Datasets:

Hendrik-a

/

arxiv

like 0

we assess the validity of various exchange - correlation functionals for computing the structural , vibrational , dielectric , and thermodynamical properties of materials in the framework of density - functional perturbation theory ( dfpt ) . we consider five generalized - gradient approximation ( gga ) functionals ( pbe , pbesol , wc , am05 , and htbs ) as well as the local density approximation ( lda ) functional . we investigate a wide variety of materials including a semiconductor ( silicon ) , a metal ( copper ) , and various insulators ( sio-quartz and stishovite , zrsio zircon , and mgo periclase ) . for the structural properties , we find that pbesol and wc are the closest to the experiments and am05 performs only slightly worse . all three functionals actually improve over lda and pbe in contrast with htbs , which is shown to fail dramatically for-quartz . for the vibrational and thermodynamical properties , lda performs surprisingly very good . in the majority of the test cases , it outperforms pbe significantly and also the wc , pbesol and am05 functionals though by a smaller margin ( and to the detriment of structural parameters ) . on the other hand , htbs performs also poorly for vibrational quantities . for the dielectric properties , none of the functionals can be put forward . they all ( i ) fail to reproduce the electronic dielectric constant due to the well - known band gap problem and ( ii ) tend to overestimate the oscillator strengths ( and hence the static dielectric constant ) .

1309.4805

the lattice - dynamical behavior ( `` the jiggling and wiggling of atoms '' , as feynman poetically said @xcite ) determines many of the physical properties of solids : infrared , raman , and neutron - diffraction spectra ; specific heats , thermal expansion , and heat conduction ; phenomena related to the electron - phonon interaction such as the resistivity of metals , superconductivity , and the temperature dependence of optical spectra ... it is thus essential to be able to model accurately the lattice dynamics of materials to understand and predict many of their properties . in the last three decades , theoretical condensed - matter physics and computational materials science have made considerable progress . nowadays , many materials properties can be computed using _ ab initio _ quantum - mechanical techniques . their only input information is the chemical composition and crystal structure of the material . the predictive power of _ ab initio _ computations has even started to be used to make materials design and predictions . @xcite in the specific case of lattice - dynamical properties , a large number of _ ab initio _ calculations based on the linear - response theory of lattice vibrations have been made possible over the past twenty years by the achievements of density - functional theory @xcite ( dft ) and by the development of density - functional perturbation theory @xcite ( dfpt ) . thanks to these theoretical and algorithmic advances , it is nowadays possible to obtain phonon dispersions on a fine grid of wave vectors covering the entire brillouin zone , which can directly be compared with neutron - diffraction data . from the phonon frequencies , many of the above physical properties can be computed . within the framework of dft , the many - body problem of interacting electrons in a static external potential is cast into a tractable problem of non - interacting electrons moving in an effective potential . the latter includes the external potential and the effects of the coulomb interactions between the electrons , i.e. the so - called hartree term , describing the electron - electron repulsion , and the exchange and correlation ( xc ) interactions , which includes all the many - body interactions . modeling the xc interactions is the main difficulty of dft . the simplest approximation is the local - density approximation @xcite ( lda ) , which is based upon exact exchange energy for a uniform electron gas and only requires the density at each point in space . a slightly more elaborate approach consists in using both the density and its gradient at each point in space in the so - called generalized gradient approximation ( gga ) . the most commonly used gga functional is the one proposed by perdew , burke , and ernzherhof ( pbe ) . @xcite it almost always overestimates the lattice constants of solids , while lda consistently underestimates the volume . in both cases , the typical errors amount to 1 - 2% of the lattice parameters . a series of alternative gga functionals have recently been proposed to overcome this problem , as well as to obtain accurate surface energies for solids . one such functional is the am05 functional , @xcite which combines the lda functional for bulk - like regions with a local airy approximation @xcite for surface - like regions . in contrast , the pbesol @xcite and wu and cohen ( wc ) @xcite functionals can be seen as revised versions of pbe specifically adapted for solids . these three functional ( am05 , pbesol , and wc ) yield indeed lattice constants that are in excellent agreement with experiments . @xcite more recently , the htbs functional @xcite has been proposed as an efficient compromise leading to good results on challenging problems requiring accuracy for both solids and molecules ( e.g. , the co adsorption on noble metals surfaces ) . it is obtained by mixing two functionals : one performing well for atomization energies of molecules ( the rpbe @xcite functional ) and one achieving very good accuracy for lattice constants in solids ( wc ) . these alternative functionals , which mainly focus on improving the modeling of the lattice constants , bond lengths and surface energies of solids , often compromise the accuracy for bulk total energies . it is therefore of interest to study how they perform when it comes to predicting properties governed by lattice dynamics ( such as phonon frequencies , dielectric constants , and thermal properties ) and that should be affected by both the accuracy in structural parameters and energy . it should be mentioned that the performance of pbe has already been tested for the phonon frequencies of a few selected solids . @xcite a comparison of lda , pbe , and an ad - hoc functional ( mixing 50% of the previous ones ) has been proposed for copper focusing on the theoretical debye - waller factors . @xcite more recently , the phonon dispersions of face - centered cubic metals ( cu , ag , au , ni , pd , pt , rh , and ir ) have also been investigated using the wc and pbesol functionals . @xcite in this paper , we consider five gga functionals ( pbe , pbesol , wc , am05 , and htbs ) as well as the lda to compute such properties using dfpt . we investigate a wide variety of materials including a semiconductor ( silicon ) , a metal ( copper ) , and various insulators ( sio@xmath0 @xmath1-quartz and stishovite , zrsio@xmath2 zircon , and mgo periclase ) . this allows us to assess the validity of the different xc functionals not only for structural properties ( as commonly reported in the literature ) but also for vibrational , thermodynamical , and dielectric properties . this paper is organized as follows . section [ sec : techn ] is devoted to the technical details of our calculations . our main results are then presented in sec . [ sec : reslt ] : we present successively ( [ sec : struc ] ) the structural properties , ( [ sec : phfrq ] ) the phonon frequencies at @xmath3 , ( [ sec : phbst ] ) the phonon dispersion curves , ( [ sec : phfrqvol ] ) the volume dependence of the phonon frequencies , ( [ sec : therm ] ) the thermodynamical properties , and ( [ sec : dielc ] ) the dielectric properties . based on these results , a few general statements and considerations are given in sec . [ sec : discuss ] . finally , our most important findings are summarized in sec . [ sec : recap ] . the _ ab initio _ calculations were performed using the abinit package @xcite combined with the libxc library . @xcite total energies were computed using dft , and phonon frequencies using dfpt . the interaction between the ions and valence electrons was described using the troullier - martins norm - conserving pseudopotentials generated by the atomic pseudopotentials engine . @xcite the reference configurations and core radii of our pseudopotentials are reported in table [ tab : psp ] . the exchange - correlation energy is evaluated within lda using perdew - wang s parametrization @xcite and gga using the parameterization of pbe , @xcite pbesol , @xcite am05 , @xcite wc , @xcite and htbs . @xcite the planewave kinetic energy cutoff was set to 10 ha for silicon , 30 ha for zircon , @xmath1-quartz , and stishovite , 40 ha for copper , 33 ha for periclase . the brillouin zone ( bz ) integrations were performed within monkhorst - pack ( mp ) scheme using @xmath4 grids for silicon , zircon , @xmath1-quartz , stishovite , periclase and using @xmath5 grids for copper . in order to deal with the possible convergence problems for copper , gaussian smearing technique was employed with the smearing parameter set equal to 0.01 hartree . . atomic valence configurations , core radii ( @xmath6 in a.u . for the different channels ) , and local channel ( see text ) of the pseudopotentials . [ tab : psp ] [ cols=">,>,^,^,^,^",options="header " , ] for a statistical analysis of the results obtained with the different xc functionals , we use the mean relative error ( mre , in % ) and the mean absolute relative error ( mare , in % ) . we have studied in table [ tab : struc ] the structural parameters ( lattice constants but also bond lengths and angles ) for the different compounds and functionals . ry9.1mmy9.1mmy9.1mmy9.1mmy9.1mmy9.1mmy9.1mmrx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmx9.1 mm & & & & & & & & & & & & & & & @xmath7 & 5.382 & 5.467 & 5.422 & 5.431 & 5.420 & 5.446 & 5.431 & @xmath8(si - si ) & 2.331 & 2.367 & 2.348 & 2.352 & 2.347 & 2.358 & 2.352 @xmath7 & 4.866 & & 4.959 & & 4.975 & & 4.916 & @xmath8(si - o ) & 1.600 & 1.624 & 1.614 & 1.614 & 1.613 & 1.620 & 1.605 @xmath9 & 5.361 & & 5.443 & 5.510 & 5.458 & & 5.406 & & 1.605 & 1.627 & 1.618 & 1.617 & 1.616 & 1.620 & 1.614 @xmath10 & 0.4666&0.4790&0.4725&&0.4748&&0.4697 & @xmath11(si - o - si ) & 142.2 & & 144.9 & & 146.1 & & 143.8 @xmath12 & 0.4128&0.4155&0.4140&0.4166&0.4149&0.4186&0.4135 & @xmath11(o - si - o ) & 108.5 & 109.4 & 109.0 & 109.5 & 109.1 & 110.0 & 108.8 @xmath13 & & & 0.2628&&&&0.2669 & & 109.3 & 108.2 & 108.6 & 108.3 & 108.6 & 108.2 & 109.0 @xmath14 & & & & & & & 0.1191 & & 109.4 & 108.9 & 109.2 & 108.9 & 109.2 & 108.2 & 109.2 & & 110.6 & 110.5 & 110.5 & 110.4 & 110.4 & 110.3 & 110.5 @xmath7 & 4.155 & 4.246 & 4.199 & 4.206 & 4.198 & 4.217 & 4.180 & @xmath8(si - o ) & 1.753 & 1.777 & 1.765 & 1.765 & 1.765 & 1.771 & 1.758 @xmath9 & 2.654 & 2.694 & 2.676 & 2.675 & 2.676 & 2.685 & 2.667 & & 1.792 & 1.844 & 1.818 & 1.822 & 1.818 & 1.827 & 1.810 @xmath10 & 0.3057&0.3070&0.3062&0.3064&0.3062&0.3060&0.3062 & @xmath11(si - o - si ) & 98.4 & 98.6 & 98.6 & 98.5 & 98.6 & 98.6 & 98.7 & & 130.8 & 130.7 & 130.7 & 130.7 & 130.7 & 130.7 & 130.7 @xmath7 & 6.575 & 6.697 & 6.631 & 6.642 & 6.629 & 6.661 & 6.610 & @xmath8(zr - o ) & 2.120 & 2.162 & 2.138 & 2.143 & 2.138 & 2.148 & 2.124 @xmath9 & 5.900 & 6.007 & 5.949 & 5.956 & 5.949 & 5.972 & 6.001 & & 2.252 & 2.288 & 2.268 & 2.267 & 2.270 & 2.272 & 2.287 @xmath10 & & & & & & & 0.0646 & @xmath8(si - o ) & 1.612 & 1.639 & 1.627 & 1.628 & 1.626 & 1.633 & 1.627 @xmath15 & 0.1953&0.1951&0.1949&0.1947&0.1951&0.1945&0.1967 & @xmath11(zr - o - si ) & 98.9 & 99.0 & 99.0 & 98.9 & 99.0 & 98.9 & 98.7 & & 150.0 & 149.8 & 149.8 & 149.7 & 149.9 & 149.7 & 150.6 & @xmath11(zr - o - zr ) & 111.1 & 111.2 & 111.2 & 111.3 & 111.2 & 111.4 & 110.7 & @xmath11(o - si - o ) & 97.2 & 97.1 & 97.1 & 97.0 & 97.1 & 97.0 & 97.8 & & 115.9 & 116.0 & 116.0 & 116.0 & 116.0 & 116.0 & 115.6 @xmath7 & 4.130 & 4.231 & 4.194 & 4.194 & 4.192 & 4.238 & 4.212 & @xmath8(mg - o ) & 2.065 & 2.115 & 2.097 & 2.097 & 2.096 & 2.119 & 2.106 @xmath7 & & 3.661 & 3.596 & 3.595 & 3.599 & 3.599 & 3.615 & @xmath8(cu - cu ) & 2.512 & 2.589 & 2.543 & 2.542 & 2.545 & 2.545 & 2.556 & -0.79 & 1.31 & 0.27 & 0.99 & 0.43 & 2.31 & & 1.35 & 2.15 & 0.82 & 2.32 & 1.17 & 5.28 & the space group of silicon is @xmath16 . the si atoms are located at @xmath17 and @xmath18 occupied the @xmath19 wyckoff sites . @xcite in table [ tab : struc ] , the optimal lattice constant of silicon calculated using different xc functionals are summarized and compared to the experimental value @xmath20 . @xcite the unit cell of @xmath1-quartz is trigonal ( space group @xmath21 ) . it contains nine atoms : the si atoms are located at @xmath22 on the @xmath23 wyckoff sites , while the o atoms are located at @xmath24 on the @xmath25 wyckoff sites . hence , four internal coordinates @xmath10 , @xmath12 , @xmath13 , @xmath14 are required , besides the two lattice constants @xmath7 and @xmath9 , in order to completely determine the structure . the theoretical value calculated here using the different xc functionals are reported in table [ tab : struc ] . a comparison with the experimental values of ref . is also provided . stishovite has the rutile structure which is tetragonal ( space group @xmath26 ) . the positions of the si atoms are imposed by symmetry : they occupy the @xmath27 wyckoff sites located at @xmath28 and @xmath29 . the o atoms are located at @xmath30 , and @xmath31 on the @xmath32 wyckoff sites . the structure of stishovite is thus completely defined by the lattice constants @xmath7 and @xmath9 and a single internal coordinate @xmath10 . in table [ tab : struc ] , our calculated results are compared with experimental values . @xcite zircon has a conventional unit cell which is body - centered tetragonal ( space group @xmath33 ) . the positions of zr and si atoms are imposed by symmetry : they are located at @xmath34 and @xmath35 on the @xmath36 and @xmath37 wyckoff sites , respectively . the o atoms occupy the @xmath38 wyckoff sites @xmath39 , where @xmath10 and @xmath15 are internal parameters . hence , the structure of zircon is completely determined by the two lattice constants @xmath7 and @xmath9 and the two internal coordinates @xmath10 and @xmath15 . table [ tab : struc ] summarizes our theoretical results for zircon together with the measurements of ref . . the space group of periclase is @xmath40 . the o atom occupies the @xmath36 wyckoff site @xmath28 and the mg atom is located at @xmath29 occupying the @xmath37 wyckoff sites . in table [ tab : struc ] , the optimal lattice constant of periclase calculated using different xc functionals are summarized and compared to the experimental value @xmath41 . @xcite the space group of copper is @xmath40 . the cu atom is located at @xmath28 occupied the @xmath36 wyckoff sites . the equilibrium lattice constants for copper are summarized in table [ tab : struc ] . in table [ tab : struc ] , we observe the well known general tendency of lda and pbe to respectively underestimate and overestimate lattice constants . pbesol and wc are the closest to the experimental lattice parameters . am05 performs only slightly worse than these functionals because of its larger errors for the @xmath1-quartz lattice parameter . these results are in agreement with a larger data set showing that wc , pbesol and am05 perform similarly and better than lda and pbe for lattice constants . @xcite on the other hand , the htbs functional is not performing better than lda and pbe , and definitely does not reach the accuracy of the wc functional . the most dramatic failure of htbs happens with @xmath1-quartz which can not be stabilized and relaxes to @xmath42-quartz ( or high - quartz ) . the latter is the stable polymorph of silica at higher temperature : at normal pressures , silica undergoes a structural transition from @xmath1- to @xmath42-quartz above @xmath43500@xmath44c . in fact , these two forms of quartz are very similar , the relative positions of the atoms being only slightly shifted . the si ( resp . o ) atoms move from the 3a ( resp . 6c ) and wyckoff sites located at @xmath22 [ resp . @xmath24 ] in the @xmath1 phase to the 3c ( resp . 6i ) wyckoff sites located at @xmath46 [ resp . @xmath47 in the @xmath42 phase . the number of internal coordinates is thus reduced from four ( @xmath10 , @xmath12 , @xmath13 , @xmath14 ) to one ( @xmath13 ) . the @xmath1- to @xmath42-quartz transition is displacive : the nearest neighbors of each atom remain the same , without breaking any of the chemical bonds . the relative positions of the atoms within the sio@xmath2 tetrahedra remain almost identical and the tetrahedra are not distorted very much . instead , the sio@xmath2 tetrahedra get slightly twisted in a way that causes the unit cell which is trigonal ( space group @xmath21 ) in @xmath1-quartz to become hexagonal ( space group @xmath48 ) in @xmath42-quartz . energy for various quartz structures along the path from @xmath1- to @xmath42-quartz for the different xc functionals : lda in solid black with circles , pbe in solid red ( gray ) with circles , pbesol in dashed black with squares , am05 in dashed red ( gray ) with squares , wc in dotted black with triangles , and htbs in dotted red ( gray ) with triangles . htbs fails to predict @xmath1-quartz as the stable phase . ] in fig . [ fig : trans ] , we report the energy for various quartz structures ( obtained by a linear interpolation of the internal coordinates and the lattice constants ) along the path from @xmath1- to @xmath42-quartz for the different xc functionals . the @xmath1-quartz is found to be the stable phase with all functionals but htbs , while @xmath42-quartz is a local minimum . with the htbs functional , it is the @xmath42-quartz phase which is stable while the @xmath1-quartz phase is unstable . we would like to point out that most of the studies evaluating the new gga functionals @xcite were carried out on highly - symmetric solids with few structural degrees of freedom ( often only one lattice constant ) . the failure of htbs for @xmath1-quartz shows that more complex structures ( with internal parameters ) should also be used to assess the performance of xc functionals in modeling structural parameters . we first compute the phonon frequencies at the @xmath3 point of the brillouin zone for the different materials using the different xc functionals . the calculations are performed using the corresponding theoretical equilibrium geometry ( lattice parameters and internal coordinates ) rather than the experimental one . as baroni _ et al . _ , @xcite we think that this is the most consistent choice when comparing with experimental data at low temperature . furthermore , our aim is to evaluate the performance of the different xc functionals for predicting properties governed by lattice dynamics and it is more sensible to do so assuming that the experimental geometry is unknown . indeed , when this is the case , it is mandatory to choose the theoretical equilibrium geometry for further calculations . note that , due to the weak volume dependence of the grneisen parameter , it is straightforward to relate results at the equilibrium geometry with those e.g. , at the experimental geometry . this will be discussed in more detail in sec . [ sec : phfrqvol ] below . in @xmath1-quartz , stishovite , zircon , and periclase , the nonvanishing components of the born effective charge tensors @xcite make it necessary to properly include the dipole - dipole interaction in the calculation of the interatomic force constants . in particular , the dipole - dipole contribution is found to be responsible for the splitting between the longitudinal and transverse optic ( lo and to , respectively ) modes e@xmath49 ( perpendicular to @xmath9 ) and a@xmath49 or a@xmath50 ( parallel to @xmath9 ) . for silicon , the following irreducible representations of optical and acoustical zone - center modes can be obtained from group theory : @xmath51 our results for silicon are reported in table [ tab : phfrq_silicon ] and compared with experimental data . @xcite lrx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmx9.1 mm mode & lda & pbe & pbesol & am05 & wc & htbs & expt.t@xmath52 & 514 & 502 & 509 & 508 & 509 & 506 & 506&1.53&-0.80&0.54&0.58&0.37&0.05 & they are in line with the conclusions of ref . regarding the evolution of the phonon frequencies with the lattice constant . lda ( resp . pbe ) underestimates ( overestimates ) the lattice constants and overestimates ( resp . underestimates ) the phonon frequency . the other xc functionals produce lattice constants and phonon frequencies in excellent agreement with experiment . in fact , the results are very good ( absolute relative error smaller than 2% ) for all the xc functionals including lda and pbe . thus , this system is not really meaningful on its own to discriminate among them . for @xmath1-quartz , the theoretical group analysis predicts the following irreducible representations of optical and acoustical zone - center modes : @xmath53 table [ tab : phfrq_quartz ] summarizes our results for @xmath1-quartz , and compares with experimental results at @xmath54k . @xcite note that the results for the htbs functional refer to the @xmath42-quartz . lrx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmx9.1 mm mode & lda & pbe & pbesol & am05 & wc & htbs & expt.raman a@xmath55(1 ) & 215 & & & & & & 219a@xmath55(2 ) & 344 & 343 & 341 & 346 & 344 & 352 & 358a@xmath55(3 ) & 454 & & & & & & 469a@xmath55(4 ) & 1069&1043&1055&1062&1058&1063&1082infrared a@xmath56(to1 ) & 351 & 358 & 350 & 366 & 355 & & 361a@xmath56(lo1 ) & 374 & 377 & 372 & 383 & 376 & 396 & 385a@xmath56(to2 ) & 486 & & & & & & 499a@xmath56(lo2 ) & 540 & & & & & & 553a@xmath56(to3 ) & 764 & 742 & 752 & 752 & 753 & 751 & 778a@xmath56(lo3 ) & 781 & & 762 & 756 & 761 & & 791a@xmath56(to4 ) & 1063&1029&1044&1049&1047&1047&1072a@xmath56(lo4 ) & 1226&1196&1207&1215&1211&1215&1230e@xmath57(to1 ) & 129 & & & & & & 133e@xmath57(lo1 ) & 129 & & & & & & 133e@xmath57(to2 ) & 258 & & & & & & 269e@xmath57(lo2 ) & 259 & & & & & & 269e@xmath57(to3 ) & 381 & 377 & 376 & 379 & 378 & 384 & 394e@xmath57(lo3 ) & 391 & 383 & 384 & 385 & 386 & 384 & 402e@xmath57(to4 ) & 441 & & & & & & 453e@xmath57(lo4 ) & 498 & & & & & & 512e@xmath57(to5 ) & 685 & & 667 & & 666 & & 698e@xmath57(lo5 ) & 688 & & 669 & & 668 & & 701e@xmath57(to6 ) & 786 & & 767 & & 766 & & 799e@xmath57(lo6 ) & 799 & & 779 & & 778 & & 812e@xmath57(to7 ) & 1054&1025&1038&1044&1041&1044&1066e@xmath57(lo7 ) & 1217&1188&1200&1207&1204&1209&1227e@xmath57(to8 ) & 1140&1126&1130&1144&1136&1152&1158e@xmath57(lo8 ) & 1137&1125&1128&1143&1134&1152&1155&-2.22&-6.92&-5.10&-6.61&-5.23&-8.79 & & 2.22 & 6.92 & 5.10 & 6.71 & 5.23 & 9.68 & in contrast with silicon , it is not obvious to correlate the performance for phonon frequencies with the structural properties . however , the relative errors vary much more significantly for the different xc functionals . lda shows the smallest relative errors , which are actually two to three times smaller than for the other xc functionals . lda , pbe , pbesol , and wc are found to systematically underestimate the phonon frequencies . it is almost systematically the case for am05 . htbs leads to the worst results since they refer to the @xmath42-quartz . for stishovite , the following irreducible representations of optical and acoustical zone - center modes can be obtained from group theory : @xmath58 in table [ tab : phfrq_stisho ] , we show the calculated and experimental values @xcite for the frequency of each mode at the @xmath3 point of stishovite . lrx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmx9.1 mm mode & lda & pbe & pbesol & am05 & wc & htbs & expt.raman a@xmath59 & 748 & 715 & 734 & 732 & 732 & 727 & 751b@xmath59 & & & & & & & 234b@xmath60 & 945 & & 923 & 920 & 924 & 916 & 964e@xmath61 & 578 & & 565 & 564 & 564 & 560 & 586infrared a@xmath50(to ) & 643 & 619 & 632 & 633 & 630 & 629 & 650a@xmath50(lo ) & & & & & & & 950e@xmath57(to1 ) & 465 & & & & & & 470e@xmath57(lo1 ) & 563 & & 545 & 543 & 546 & 539 & 565e@xmath57(to2 ) & 577 & & 555 & & 555 & & 580e@xmath57(lo2 ) & 701 & 672 & 684 & 684 & 686 & 680 & 700e@xmath57(to3 ) & 819 & 791 & 804 & 805 & 803 & 800 & 820e@xmath57(lo3 ) & 1032 & 989&1011&1009&1011&1004&1020silent b@xmath52(1 ) & 383 & 367 & 374 & 374 & 374 & 372 & b@xmath52(2 ) & 749 & 718 & 734 & 732 & 734 & 728 & a@xmath60 & 600 & 573 & 581 & 582 & 584 & 577 & & -0.11&-4.72&-2.48&-2.59&-2.49&-3.14 & & 2.03 & 5.82 & 3.85 & 3.95 & 3.86 & 4.42 & while pbesol and wc provide the best structural properties , it is again lda which leads to the smallest relative error on the phonon frequencies . it is worth mentioning that all the alternative gga functionals ( pbesol , am05 , wc , and htbs ) improve compared to pbe , pbesol and wc providing the best results ( though worse than lda ) . none of the functionals present a systematic trend ( under- or overestimation ) in the errors . for zircon , the theoretical group analysis provides the following irreducible representations of optical and acoustical zone - center modes : @xmath62 table [ tab : phfrq_zircon ] summarizes our results for zircon and compares with experimental results . @xcite the dipole - dipole contribution is found to be responsible for the splitting between the longitudinal optical ( lo ) and transverse optic ( to ) modes @xmath63 and @xmath64 at the @xmath3 point . lrx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmx9.1 mm mode & lda & pbe & pbesol & am05 & wc & htbs & expt.raman a@xmath59(1 ) & 437&&425&423&426&422&439a@xmath59(2 ) & 960&&940&936&941&933&974b@xmath59(1 ) & 222&&212&210&213&209&214b@xmath59(2 ) & 388&&378&375&379&375&393b@xmath59(3 ) & 626&603&612&610&614&607 & b@xmath59(4 ) & 1002&&981&977&981&975&1008b@xmath60 & 256&256&253&256&254&254&266e@xmath61(1 ) & 193&191&&192&191&191&201e@xmath61(2 ) & 224&&218&218&218&217&225e@xmath61(3 ) & 362&&343&&345&&357e@xmath61(4 ) & 535&521&525&527&526&524&547e@xmath61(5 ) & 910&867&889&886&890&884 & infrared a@xmath50(to1 ) & 345&&330&327&331&327&338a@xmath50(lo1 ) & 471&&460&459&460&457&480a@xmath50(to2 ) & 596&&582&580&584&&608a@xmath50(lo2 ) & 640&&624&621&626&618&647a@xmath50(to3 ) & 967&&946&942&946&940&989a@xmath50(lo3 ) & 1084&&1061&1059&1063&1055&1108e@xmath57(to1 ) & 279&&&&&&287e@xmath57(lo1 ) & 338&&&&&&352e@xmath57(to2 ) & 376&&&&&&389e@xmath57(lo2 ) & 416&&404&400&405&&419e@xmath57(to3 ) & 422&413&414&415&416&413&430e@xmath57(lo3 ) & 463&&452&452&452&449&471e@xmath57(to4 ) & 856&&&&&&885e@xmath57(lo4 ) & 1017&&994&990&996&987&1035silent b@xmath52 & 124&135&127&132&128&134 & a@xmath60 & 243&242&240&242&241&241 & a@xmath52 & 391&388&387&389&388&389 & b@xmath50(1 ) & 562&541&551&550&551&547 & b@xmath50(2 ) & 933&892&913&908&914&905 & & -1.49&-6.36&-4.08&-4.45&-3.84&-4.82 & & 2.09 & 6.36 & 4.08 & 4.45 & 3.84 & 4.82 & relative error with respect to the experimental frequencies of zircon for the different xc functionals : lda ( black circles ) , pbe [ red ( gray ) circles ] , pbesol ( black squares ) , am05 [ red ( gray ) squares ] , wc ( black triangles ) , and htbs [ red ( gray ) triangles ] . ] as for stishovite , there is no clear correlation between the performance with respect to the prediction of the structural properties and the quality of the phonon frequencies . in the case of zircon , all the gga functionals tend to underestimate the phonon frequencies , whereas there is no systematic trend for lda . once again , pbe is performing the worst , while pbesol and wc are the best ggas . these conclusions are clearly seen in fig . [ fig : phfrq_compar ] , which shows the relative errors for each phonon mode as a function of the experimental frequency . a similar graph ( not shown here for sake of brevity ) is obtained for both sio@xmath0 phases ( @xmath1-quartz and stishovite ) . it is also interesting to note that all functionals tend to underestimate the experimental data at high frequency . for periclase , the following irreducible representations of optical and acoustical zone - center modes can be obtained from group theory : @xmath65 table [ tab : phfrq_periclase ] shows our results for periclase and experimental values . @xcite lrx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmx9.1 mm mode & lda & pbe & pbesol & am05 & wc & htbs & expt.infrared t@xmath52(to1 ) & & & 391 & 390 & 391 & & 396t@xmath52(lo1 ) & 727 & 686 & 698 & 699 & 700 & 684 & 710 & 4.23&-4.59&-1.48&-1.53&-1.34&-5.11 & & 4.23 & 4.59 & 1.48 & 1.53 & 1.34 & 5.11 & in periclase , the worst results are obtained with htbs . lda and pbe perform slightly better ( with a similar accuracy ) , the best performances being obtained with pbesol , wc , and am05 in contrast with the observations for the other oxides ( quartz , stishovite , and zircon ) . note that the lda systematically overestimates while all the ggas underestimate the phonon frequencies . for copper , all the phonon frequencies at the @xmath3 point are zero . there are only three acoustic since there is one atom per unit cell . an analysis of the fundamental frequencies at other selected points of the brillouin zone will be given in sec . [ sec : phbst ] . globally , lda is the functional that performs the best among all , while pbesol and wc are the best among the ggas . the worst result is obtained with htbs for the a@xmath55(1 ) mode of quartz ( 29 @xmath66 instead of 219 @xmath66 ) . this tremendous error comes from the wrong prediction of the stable structure with this functional ( @xmath42- instead of @xmath1-quartz ) . even after the exclusion of the quartz data , the htbs functional still performs significantly worse than lda , wc , pbesol and am05 . note also that there is no systematic trend ( under- or overestimation ) in the relative errors obtained with the different functionals . we also compute the phonon bandstructure and density of states ( dos ) of the different materials using the different xc functionals . our results are presented in figs . [ fig : ph - silico]-[fig : ph - copper ] . available experimental data are also reported for comparison . the agreement is qualitatively very good though some significant discrepancies are present ( especially at higher frequencies ) independently of the xc functional . given the variations existing in the experimental data and the typical error bars , it did not seem meaningful to us to define a quantitative measure of the agreement between theory and experiment in order to discriminate between the different xc functionals . phonon bandstructure and density of states of silicon computed using the different xc functionals : lda in solid black , pbe in solid red ( gray ) , pbesol in dashed black , am05 in dashed red ( gray ) , wc in dotted black , and htbs in dotted red ( gray ) . the experimental data from refs . are also reported as blue ( black ) circles . [ fig : ph - silico ] ] phonon bandstructure and density of states of quartz computed using the different xc functionals : lda in solid black , pbe in solid red ( gray ) , pbesol in dashed black , am05 in dashed red ( gray ) , wc in dotted black , and htbs in dotted red ( gray ) . the experimental data from refs . are also reported as blue ( black ) circles . ] phonon bandstructure and density of states of stishovite computed using the different xc functionals : lda in solid black , pbe in solid red ( gray ) , pbesol in dashed black , am05 in dashed red ( gray ) , wc in dotted black , and htbs in dotted red ( gray ) . the experimental data from ref . are also reported as blue ( black ) circles . ] phonon bandstructure and density of states of zircon computed using the different xc functionals : lda in solid black , pbe in solid red ( gray ) , pbesol in dashed black , am05 in dashed red ( gray ) , wc in dotted black , and htbs in dotted red ( gray ) . the experimental data from refs . are also reported as blue ( black ) circles . ] phonon bandstructure and density of states of periclase computed using the different xc functionals : lda in solid black , pbe in solid red ( gray ) , pbesol in dashed black , am05 in dashed red ( gray ) , wc in dotted black , and htbs in dotted red ( gray ) . the experimental data from ref . are also reported as blue ( black ) circles . ] phonon bandstructure and density of states of copper computed using the different xc functionals : lda in solid black , pbe in solid red ( gray ) , pbesol in dashed black , am05 in dashed red ( gray ) , wc in dotted black , and htbs in dotted red ( gray ) . the experimental data ( with error bars ) from ref . are also reported as blue ( black ) circles . ] for the oxides ( @xmath1-quartz , stishovite , zircon , and periclase ) , it is reasonable to say that the lda seems to perform slightly better than all the gga functionals . in contrast , for copper , lda shows the largest discrepancies while pbe seems to perform the best . this can also be seen in table [ tab : phfrq_copper ] where the fundamental frequencies at selected points of the brillouin zone have been reported . we would like to point out that our results for copper are shifted up by 5 - 7% compared to those of ref . in which the wc and pbesol frequencies match the experimental data better than lda or pbe . there are several differences between the two calculations which may explain this discrepancy . in ref . , the phonon frequencies are computed at the equilibrium lattice constant at @xmath67=296 k , while we use the one at @xmath67=0 . this probably accounts for 1 - 2% difference , as observed experimentally @xcite and as discussed below . furthermore , the calculations of ref . are based on the projector augmented wave ( paw ) method , while ours rely on the use of norm - conserving pseudopotentials . it is interesting to point out that despite the equilibrium lattice constants at @xmath67=0 k are in excellent agreement ( at most 0.5% discrepancy ) , the phonon frequencies may show important differences . this emphasizes the importance of assessing the accuracy of pseudopotentials by comparing with all - electron results not only for structural but also for dynamical properties . lrx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmx9.1 mm mode & lda & pbe & pbesol & am05 & wc & htbs & expt.x@xmath68 & & 168&&&&176&169x@xmath69 & & 243&&&&&240w@xmath70 & & 162&&&&&163w@xmath71 & & 205&&&&&203l@xmath68 & & 110&&&117&117&112l@xmath69 & & 248&&&&&247 & 12.50&-0.06 & 8.21 & 9.38 & 6.99 & 5.88 & & 12.50 & 0.94 & 8.21 & 9.38 & 6.99 & 5.88 & when looking at the phonon dos in figs . [ fig : ph - silico]-[fig : ph - copper ] , it looks as if it were simply shifted when changing the xc functional ( with lda and pbe at the upper and lower ends and all the other functionals falling in the middle almost on top of one another ) . this is confirmed by computing the different moments of the phonon dos : the average frequency @xmath72 ( 1@xmath73 moment ) is much more dependent on the xc functional than all higher moments . in table [ tab : fre - moment ] , we show the average frequency @xmath72 and the standard deviation @xmath74 ( 2@xmath75 moment ) extracted from the phonon dos computed at the theoretical lattice constant of the different materials using the different xc functionals . higher moments ( not reported for sake of brevity ) do not vary much with the xc functional . rrrx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmr & & & lda & pbe & pbesol&am05 & wc & htbs & silicon & @xmath72 & 330 & 324 & 327 & 327 & 326 & 325 & & & @xmath74 & 146 & 139 & 145 & 145 & 144 & 144 & & @xmath1-quartz & @xmath72 & 557 & 539 & 545 & 546 & 544 & 541 & & & @xmath74 & 359 & 352 & 356 & 357 & 360 & 361 & & stishovite & @xmath72 & 580 & 552 & 566 & 566 & 564 & 562 & & & @xmath74 & 207 & 201 & 204 & 204 & 204 & 203 & & zircon & @xmath72 & 480 & 458 & 468 & 470 & 467 & 465 & & & @xmath74 & 294 & 281 & 288 & 288 & 287 & 286 & & periclase & @xmath72 & 413 & 383 & 392 & 393 & 392 & 379 & & & @xmath74 & 118 & 113 & 114 & 114 & 114 & 112 & & copper & @xmath72 & 183 & 163 & 176 & 175 & 178 & 173 & & & @xmath74 & 55 & 49 & 53 & 52 & 53 & 52 & the different xc functionals lead to different lattice constants and the latter influence the computed frequency . here , we study the volume dependence of the average frequency ( @xmath72 ) for the different xc functionals . we can distinguish an indirect effect through the equilibrium lattice constant from the direct effect ( for a fixed lattice constant ) . indeed , the average frequency changes quite strongly with the lattice constants as illustrated in fig . [ fig : vol - dep ] for periclase and copper : an increase of the lattice constant by 0.2 can induce a decrease of @xmath72 by 80 - 100 @xmath66 . in contrast , the direct effect is rather small at most 15 @xmath66 between the two extreme results ( produced by lda and pbe ) . to illustrate this point , let us consider periclase . in table [ tab : fre - moment ] , we have @xmath72=413 @xmath66 using lda and 383 @xmath66 using pbe . this looks like a decrease by 30 @xmath66 . however , these values are calculated for two different lattice constants @xmath7=4.13 for lda and @xmath7=4.23 for pbe . if the functional were changed from lda to pbe while keeping the lda equilibrium lattice constant , the average frequency would actually increase by 13 @xmath66 ( from 413 to 426 @xmath66 ) . this is the direct effect of the xc functional . whereas , the indirect effect ( due to the change from lda to pbe equilibrium lattice constant ) produces a decrease by 43 @xmath66 . the lattice ( or volume ) dependence of the average frequency is measured by the corresponding grneisen parameter @xmath76 which is also reported in fig . [ fig : vol - dep ] . interestingly enough , all functionals produce very comparable grneisen parameters . for periclase , @[email protected] - 1.57 is comparable to the measured values ranging from 1.49 @xcite to 1.60 . @xcite for copper , @[email protected] - 2.34 is in very good agreement with the reported experimental value of @xmath77@xmath782 @xcite and previous theoretical results . @xcite this weak variation with the xc functional might be exploited as follows . if the average frequencies were known for various xc functionals at a given lattice constant value , computing the volume dependence of the frequency for just one of those functionals would be sufficient to provide a reasonable extrapolation scheme for @xmath79 for any of the xc functionals . variation of the average frequency @xmath72 ( in @xmath66 ) of ( a ) periclase and ( b ) copper as a function of the lattice constant @xmath7 ( in ) computed using the different xc functionals : lda in solid black , pbe in solid red ( gray ) , pbesol in dashed black , am05 in dashed red ( gray ) , wc in dotted black , and htbs in dotted red ( gray ) . the vertical lines indicate the experimental lattice constant taken from refs . , and the two extreme pbe and lda lattice constants for periclase and copper , respectively . ] the vibrational entropy of a solid can be obtained from its phonon distribution . this quantity is of importance for the _ ab initio _ assessment of solid phase stability . the entropy @xmath80 at temperature @xmath67 is first computed within the harmonic approximation @xcite for the different materials in their equilibrium structural configuration . in table [ tab : entropy ] , the entropy at 298k calculated using the various xc functionals is compared with the experimental measurements . @xcite all the results are in good agreement with experiments apart from those for zircon . @xcite we note that the lda functional presents the best results . am05 , pbesol and wc are the best among the ggas , while htbs is the worst . in any case , the absolute error on the contribution of the vibrational entropic term to the free energy ( ts ) is around 2 to 3 mev / atom . this is much smaller than the typical errors on the energy term.@xcite ry9.1mmy9.1mmy9.1mmy9.1mmy9.1mmy9.1mml & & & & & & & & 19.41 & 19.56 & 19.67 & 19.61 & 19.67 & 19.71 & 18.81 - 18.80 & 42.40 & 44.32 & 43.76 & 43.86 & 44.48 & 44.95 & 41.44 - 43.40 & 24.56 & & 25.72 & 25.66 & 25.81 & 26.03 & 24.94 & & & & & & & 84.50 & 25.71 & 28.59 & 27.59 & 27.52 & 27.63 & 28.96 & 26.95 - 26.90 & & 32.94 & 31.06 & 31.34 & 30.85 & 31.57 & 33.15 - 33.10&-0.83 & 5.68 & 2.83 & 2.79 & 3.18 & 4.77 & & 4.52 & 5.87 & 4.90 & 4.58 & 5.47 & 6.34 & in the following , we take periclase and copper as examples to study how xc functionals impact other thermal properties . first , the free energy @xmath81 [ @xmath82 at temperature @xmath67 and lattice constant @xmath7 [ volume @xmath83 are computed within the quasi - harmonic approximation . @xcite by fitting the results for the free energy to the thrird - order birch - murnaghan equation of state,@xcite we can obtain the variation in temperature of the equilibrium lattice constants @xmath84 [ volumes @xmath85 , the bulk modulus @xmath86 , and the pressure derivative of the bulk modulus @xmath87 . the linear expansion @xmath88 is then given by @xmath89 where @xmath90=298 k is the reference temperature . finally , the coefficient of linear expansion is obtained as @xmath91 figs . [ fig : therm_pericl ] and [ fig : therm_copper ] show the results for ( a ) @xmath88 ( expressed as a percentage ) , ( b ) @xmath92 , and ( c ) @xmath93 using various xc functionals . variation of ( a ) the linear thermal expansion @xmath94 , ( b ) the coefficient of linear thermal expansion @xmath1 , and ( c ) the bulk modulus @xmath86 of periclase as a function of the temperature computed using the different xc functionals : lda in solid black , pbe in solid red ( gray ) , pbesol in dashed black , am05 in dashed red ( gray ) , wc in dotted black , and htbs in dotted red ( gray ) . relevant experimental data are also reported . the labels correspond to the first author and year of publication of refs . . ] variation of ( a ) the linear thermal expansion @xmath94 , ( b ) the coefficient of linear thermal expansion @xmath1 , and ( c ) the bulk modulus @xmath86 of copper as a function of the temperature computed using the different xc functionals : lda in solid black , pbe in solid red ( gray ) , pbesol in dashed black , am05 in dashed red ( gray ) , wc in dotted black , and htbs in dotted red ( gray ) . relevant experimental data are reported . the labels correspond to the first author and year of publication of refs . . ] for periclase , the anharmonic effects are very important . thus , empirical corrections are necessary to obtain good thermodynamical properties over a wide temperature range . @xcite here , we did not go through this special procedure that would not allow to discriminate the different xc functionals . so , the calculations of @xmath94 , @xmath1 , and @xmath86 were limited to temperatures lower than 1000 k. ideally , it would be better to focus on even smaller temperatures but very few data are available . at low temperature , lda seems to produce the best results . in contrast , like for the frequency at @xmath3 , the htbs results fall outside the lda - pbe range on the pbe side , leading to the worst agreement with experiments . for copper , lda ( resp . pbe ) underestimates ( resp . overestimates ) the expansion ( @xmath94 and @xmath1 ) while all the other ggas show an excellent agreement with the experimental results . for the bulk modulus , pbe underestimates compared to the measured data , while all other functionals overestimate . htbs performs the best and lda is the worst with @xmath435% and @xmath4320% discrepancy , respectively . all other ggas ( including pbe ) show an error of @xmath437 - 8% compared to experiments . finally , we turn to the calculation of the electronic ( @xmath95 ) and static ( @xmath96 ) permittivity tensors . the electronic ( ion - clamped ) dielectric permittivity tensor is related to the second - order derivatives of the energy with respect to the macroscopic electric field . the static dielectric permittivity tensor also includes the polarization coming from the ionic displacements . it is obtained by adding the contributions of different modes to @xmath95 as follows @xmath97 where @xmath98 is the volume of the primitive unit cell and @xmath99 is the mode - oscillator strength . the electronic ( @xmath100 ) and static ( @xmath101 ) permittivity tensors , as well as the contributions of the individual modes @xmath102 , are reported in table [ tab : dielc ] for @xmath1-quartz , stishovite , zircon , and periclase . due to the symmetry of the different crystalline systems , they have two different components parallel ( @xmath103 ) and perpendicular ( @xmath104 ) to the @xmath9 axis for the first three materials [ @xmath1-quartz ( trigonal ) , stishovite and zircon ( tetragonal ) ] , while the tensor is isotropic for periclase ( cubic ) . in stishovite and zircon ( resp . @xmath1-quartz ) , the phonon mode contributions to @xmath105 come from the three ir - active a@xmath50 ( resp . a@xmath0 ) modes , while the contributions to @xmath106 come from the four ir - active e@xmath57 ( resp . e ) modes . in periclase , the contribution arises from the t@xmath52 mode . for @xmath1-quartz , the experimental values ( at 298k ) are taken from ref . , except for the two lowest e modes which are from ref . . for stishovite , the measurements are those of ref . . for zircon , the reported experimental values are from ref . for the parallel components ( values at 295 k ) , and from ref . for the perpendicular components . for periclase , the measurements are those of ref . . llx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmx4mmlx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmx9.1mmx9.1mml & & lda & pbe & pbesol & am05 & wc & htbs & expt . & & & lda & pbe & pbesol & am05 & wc & htbs & expt . & & @xmath107 & 2.50 & 2.41 & 2.46 & 2.39 & 2.43 & 2.31 & 2.42 & & @xmath108 & 2.47 & 2.38 & 2.43 & 2.36 & 2.41 & 2.29 & 2.39&@xmath109 & & 0.68 & & 0.65 & 0.69 & & 0.67 & & @xmath110 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 & 0.02 & 0.00&@xmath111 & & & & & & & 0.65 & & @xmath112 & & & 0.05 & & & & 0.05&@xmath113 & & & & & & & 0.11 & & @xmath114 & & 0.32 & & & & & 0.33&@xmath115 & & & & & & & 0.66 & & @xmath116 & 0.84 & & & & & & 0.83 & & & & & & & & & & @xmath117 & & & 0.02 & & 0.02 & & 0.02 & & & & & & & & & & @xmath118 & & 0.11 & 0.11 & 0.11 & 0.11 & & 0.11 & & & & & & & & & & @xmath119 & & & & & & & 0.65 & & & & & & & & & & @xmath120 & 0.01 & & 0.01 & & 0.01 & & 0.01 & @xmath105 & 4.85 & 4.74 & 4.79 & 4.68 & 4.74 & 4.55 & 4.64 & & @xmath106 & 4.64 & 4.52 & 4.57 & 4.46 & 4.52 & 4.35 & 4.43 & @xmath107 & 3.38 & 3.44 & 3.43 & 3.42 & 3.42 & 3.43 & 3.33 & & @xmath108 & 3.30 & 3.35 & 3.35 & 3.34 & 3.34 & 3.34 & 3.24&@xmath109 & 5.59 & 5.77 & 5.63 & 5.60 & 5.69 & 5.62 & & & @xmath110 & 6.85 & 9.01 & 7.49 & 7.69 & 7.58 & 7.67 & & & & & & & & & & & @xmath112 & 0.45 & 0.09 & 0.38 & 0.30 & 0.35 & 0.34 & & & & & & & & & & & @xmath114 & 0.79 & 0.74 & 0.80 & 0.78 & 0.79 & 0.79 & & @xmath105 & 8.97 & 9.21 & 9.06 & 9.02 & 9.11 & 9.05 & & & @xmath106&11.40&13.19&12.02&12.12&12.06&12.14 & & @xmath107 & & & & & & & 3.80 & & @xmath108 & & & & & & & 3.50&@xmath109 & 6.03 & & & & & & 5.75 & & @xmath110 & 5.69 & & & & & & 5.70&@xmath111 & & & & & & & 0.36 & & @xmath112 & & & & & & & 0.60&@xmath113 & & & & & & & 0.78 & & @xmath114 & & & & & & & 0.15 & & & & & & & & & & @xmath116 & & & & & & & 1.20 & @xmath105&&&&&&&10.69 & & @xmath106&&&&&&&11.25 & @xmath100 & 3.04 & & & 3.09 & 3.09 & & 2.95&@xmath121 & & & & & & & 6.25 & @xmath101 & & & & & & & 9.20 & & & @xmath100 & 6.59 & 6.59 & 7.27 & 6.10 & 6.68 & 5.42 & & & @xmath100 & 6.59 & 6.83 & 7.27 & 6.81 & 6.68 & 7.91 & @xmath101 & 7.62 & 13.70 & 9.86 & 9.88 & 9.01 & 10.05 & & & @xmath101 & 7.62 & 13.70 & 9.86 & 9.88 & 9.01 & 11.55 the theoretical values of @xmath100 are larger than the experimental ones by about 68% , as often found in dft . the only exception is @xmath1-quartz for which the values obtained with the different ggas are much closer to the experimental values and tend to slightly underestimate it ( note that htbs results actually refer to @xmath42-quartz ) . in fact , the overestimation of the electronic dielectric constant can be related to the band - gap problem of dft . and , since the different xc functionals used here have not been designed to produce a correct band gap , there is not a significant difference among the different results . the contributions of each individual mode @xmath122 depend on the volume ( @xmath123 ) , the oscillator strength ( @xmath124 ) and the frequency of the ir - active mode ( @xmath125 ) . as a result , the effects of the different xc functionals are closely connected to previous discussions of the structural properties and the frequencies at the @xmath3 point . the largest differences ( in absolute values @xcite ) from one functional to another obviously appear in the low frequency modes for which the contributions can be quite significant ( since @xmath125 ) . for such modes , the errors with respect to experiments can also be quite large . for instance , the error on @xmath110 for zircon is about 40% when using pbe . it is also the contribution for which there is the largest difference between xc functionals ( 42% from lda which is very close the experimental value to pbe ) . the difference in @xmath122 is the combination of the differences in each of its components ( @xmath98 , @xmath126 , and @xmath127 ) . in this case , the most important differences originate from oscillator strength @xmath126 ( 22% ) and @xmath127 ( 12% , which also leads to 22% discrepancy when taken to the square ) . these are slightly compensated by the difference in @xmath98 ( 2% ) . it is worth mentioning that the differences in the oscillator strength are due essentially to differences in the atomic displacements ( related to the second derivatives of the total energy with respect to atomic displacements , just like the frequencies ) . indeed , we found that the xc functional has a very little impact ( at most 1 - 2% ) on the born effective charges ( connected to the mixed second - order derivative of the energy with respect to atomic displacements and macroscopic electric field ) . and , since our results are in excellent agreement with previous calculations , the born effective charges are not reported here for sake of brevity . however , we would like to point out that it is known that dft tends to overestimate the born effective charges . @xcite and , as a result , part of the discrepancy of @xmath122 with respect to experiments can be explained by the too large oscillator strengths . finally , we turn to the static dielectric constant @xmath101 . the theoretical values are larger than the experimental ones by about 8 - 14% . this is directly related to previous discussion about @xmath100 and @xmath122 . the best results are obtained within lda , while pbe and htbs lead to the worst results . the other three functionals ( pbesol , am05 , and wc ) have a similar accuracy which is slightly worse than for lda . in view of our results on different functionals , materials and dfpt properties , a few general statements and considerations can be made . for all oxides and semiconductors , lda performs surprisingly well in most of the cases . as the lda lattice parameter is strongly underestimated and the lattice parameter has a large role in determining the frequencies ( see sec . [ sec : phfrqvol ] ) , the success of lda must be due to some cancellation of errors . accurate vibrational properties are obtained with a significantly underestimated lattice parameter . this implies that using the experimental lattice parameter with lda will degrade the quality of the dfpt results . interestingly , pbe does not benefit from a similar effect and , as a result , it is inaccurate in both the lattice parameters and the frequencies . similar conclusions were found when investigating the theoretical debye - waller factors @xcite , leading the proposal of an ad - hoc functional obtained by mixing 50% of lda and 50% of pbe . the optimal functionals providing both accurate structural factors and reasonably accurate ( i.e. , at least better than pbe ) frequencies are the wc , pbesol and am05 . finally , htbs is not to be recommended for dfpt as both structural parameters and frequencies are not accurate . if one uses a functional such as pbesol or wc ( that give very good structural parameters ) , phonon frequencies are usually accurate within maximum 5% , vibrational entropies within maximum 6% , and static dielectric constants overestimate experiments systematically by at most 10% . these results show that while there are differences between functionals , dfpt computations are accurate enough to lead to a reasonable predictive power . on the other hand , the only metal in our data set ( copper ) shows a very different behavior with the best functional being pbe and the worst being lda , pbesol , wc and am05 . in view of previous work on metal phonon frequencies @xcite and debye - waller factors @xcite with dfpt this discrepancy might originate from the pseudopotential . in this paper , the validity of various exchange - correlation functionals ( lda , pbe , pbesol , wc , am05 , and htbs ) has been investigated for computing the structural , vibrational , dielectric , and thermodynamical properties of various materials ( silicon , sio@xmath0 @xmath1-quartz and stishovite , zrsio@xmath2 zircon , mgo periclase , and copper ) in the framework of dfpt . for the structural properties , pbesol and wc are found to provide the results closest to the experiments and am05 performs only slightly worse . these three functionals constitute an improvement over lda and pbe in contrast with htbs which has been shown to be really problematic for @xmath1-quartz . for the vibrational and thermodynamical properties , lda performs surprisingly very well . in the majority of cases , it outperforms pbe significantly and yields slightly better results than the wc , pbesol , and am05 functionals . for the latter functionals , this slight improvement is however detrimental for the structural parameters . on the other hand , htbs performs also poorly for vibrational quantities . for the dielectric properties , all the functionals fail to reproduce the electronic dielectric constant due to the well - known band gap problem . they all tend to overestimate the oscillator strengths and hence the static dielectric constant . overall , lda provides very good performances in dfpt computations provided the lda lattice parameter is used . if accurate lattice parameters are also sought for , the pbesol , wc and am05 provide the most accurate structural results while still improving in terms of vibrational quantities compared to pbe . l.h . and a.z . acknowledge support from the funds for creative research groups of china under grant 11021101 , the national basic research program of china under grant 2011cb309703 , and the national center for mathematics and interdisciplinary sciences of chinese academy of sciences . is grateful for the funding from the national science foundation of china ( grants 11071265 and 11171232 ) and from the program for innovation research in central university of finance and economics . g .- m.r . would like to thank the fund for scientific research ( f.r.s .- fnrs ) for financial sponsorship . m.j.t.o . thankfully acknowledges financial support from the portuguese fct ( contract # sfrh / bpd/44608/2008 ) . m.a.l.m . is grateful to the french anr for funding through the project anr-12-bs04 - 0001 - 02 . j.j.r . and f.d.v . would like to thank the doe for financial support through the grants de - fg03 - 97er45623 ( j.j.r . ) and de - fg02 - 03er15476 ( f.d.v . ) . the computations were carried out on lssc3 cluster in the state key laboratory of scientific and engineering computing , chinese academy of sciences . for high frequency modes , the relative errors with respect to experiments and the relative difference from one functional to another can be enormous due to the small absolute values of the contributions . this is the case of @xmath120 in @xmath1-quartz , for example .

for decades , a wide variety of observations spanning the radio through optical and on to the x - ray have attempted to uncover signs of type ia supernovae ( sne ia ) interacting with a circumstellar medium ( csm ) . the goal of these studies is to constrain the nature of the hypothesized sn ia mass - donor companion . a continuous csm is typically assumed when interpreting observations of interaction . however , while such models have been successfully applied to core - collapse sne , the assumption of continuity may not be accurate for sne ia , as shells of csm could be formed by pre - supernova eruptions ( novae ) . in this work , we model the interaction of sne with a spherical , low density , finite - extent csm and create a suite of synthetic radio synchrotron light curves . we find that csm shells produce sharply peaked light curves , and identify a fiducial set of models that all obey a common evolution and can be used to generate radio light curves for interaction with an arbitrary shell . the relations obeyed by the fiducial models can be used to deduce csm properties from radio observations ; we demonstrate this by applying them to the non - detections of sn 2011fe and sn 2014j . finally , we explore a multiple shell csm configuration and describe its more complicated dynamics and resultant radio light curves .

1604.00019

a type ia supernova ( sn ia ) is the explosion of a carbon - oxygen white dwarf @xcite . it is hypothesized that in order to explode , these stars need to gain mass from a companion star via some mass transfer mechanism , be it roche lobe overflow , winds , merger , or something more exotic . the zoo of possible mass transfer mechanisms results in a similar plethora of companion candidates from white dwarfs to red giants . to date , there has never been a direct detection of a normal sn ia companion @xcite ; but there are indirect methods of constraining the companion identity . one method is to search for signs of the ejecta impacting a circumstellar medium ( csm ) produced during mass transfer , pre - supernonova outbursts , or stellar winds . the ejecta - csm interaction radiates across a broad range of the spectrum , from x - ray to radio . the csm composition and structure will reflect the nature of the progenitor system . we refer to any supernova with csm interaction as an `` isn '' . observations of isne are often interpreted through the @xcite self - similar model of shock evolution . this model has successfully been used to fit radio light curves of core - collapse radio supernovae ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , supporting the hypothesis that the signal is from interaction of the sn ejecta with a wind . the key to the model s success in that case is that the csm is extensive and close to the stellar surface , so the shock evolves to the self - similar solution before observations begin . in the case of isne ia , the density profile of the csm is unconstrained by both observations and theory . observations so far have been used to derive upper limits on the density of _ continuous _ csm @xcite . however , sn ia progenitor models come in numerous flavors , many of which do not predict a wind - profile csm or even a continuous csm . for instance , a dense csm shell close to the progenitor white dwarf has been considered to explain high velocity features in sn ia spectra @xcite . several progenitors models suggest that nova outbursts will sweep csm into a shell of small radial extent ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? it is the purpose of this work to characterize the radio light curves of supernovae interacting with csm in a shell . we restrict our studies to the case of low - density csm such that the interaction is in the adiabatic , optically thin limit . we also assume , unless stated otherwise , that there is negligible csm interaction prior to interaction with the csm shell . these assumptions are consistent with the fact that the majority of sne ia show no signs of interaction : if present , csm must be of low density . the paper is organized as follows . in the next section we describe our model assumptions and guide the reader through our fiducial model set , which is defined by the ratio of the csm and ejecta densities at the initial contact point , motivated by @xcite . in [ sec : shell_lcs ] we present our calculated radio synchrotron light curves for the fiducial set , demonstrate their similarity , and provide a system for paramterizing these light curves such that csm properties could be easily deduced from a well - observed light curve . then in [ sec : high_vel_ej ] we relax the assumption that the csm and ejecta densities start in the ratio that defines the fiducial set , and show that the fiducial model relations can still be used to approximate csm properties even for these models . with that in mind , in [ sec : obscomp ] we use our fiducial model light curves to analyze the probability of detecting shells in recent radio observations , given that there are gaps between observations . finally , in [ sec : multishell ] we illustrate that a system with two shells will , upon collision with the second shell , produce a light curve that is significantly different from the single - shell light curves . the appendices provide the derivations of the radiation equations used in this work . in this section we briefly describe the hydrodynamic code used for this study and describe the `` fiducial model set '' . to model the shell interactions , we used the 1d spherically symmetric lagrangian hydrodynamics code of @xcite . the code solves the euler equations using a standard staggered - grid von neumann - richtmyer technique , and includes an artificial viscosity term to damp out post - shock oscillations ( see e.g. * ? ? ? although this code can treat radiative transport via monte carlo methods , we did not exercise the radiation capabilities in this paper . instead our calculations are purely hydrodynamical with an adiabatic index of an ideal gas , @xmath0 , as radiative loses are small for the csm densities considered here . figure [ fig : paramspace ] shows that our models indeed do not trap radiation and do not cool , based on the equations provided in appendix [ app : tscales ] . the exact number of resolution elements ( `` zones '' ) used in these calculations varied by model but was of order @xmath1 . as described below , the ejecta density profile has an `` inner '' and `` outer '' region ; the outer ejecta interacts with the csm and is important for the hydrodynamics . we asserted that the outer ejecta and csm zones had the same mass resolution . in most models we include a region of `` vacuum '' outside the csm shell , and in some models we include zones of inner ejecta . these regions act as buffers to ensure that the shock front does not reach the grid boundary and create numerical artifacts . the vacuum region is constant density at least a factor of @xmath2 less dense than the rest of the model , and the inner ejecta is much denser than the rest ; therefore resolution matching is infeasible in these regions and they are assigned @xmath3 zones each when present . the hydrodynamics code outputs files containing tables of gas properties at user - defined , linearly - spaced time intervals . we call these output files `` snapshots '' , and they represent only a subset of the numerical time steps taken by the code . note that , in general , snapshots are taken at different times since explosion for each model . therefore we linearly interpolate if we wish to know a model property at a specific time , introducing a small interpolation error . and [ subapp : ssa ] ) , depicted in the @xmath4 plane , where @xmath5 is the csm density and @xmath6 is the initial contact discontinuity radius for a fiducial model ( equation [ eqn : rcss ] ) . note that models of different shell fractional width ( equation [ eqn : f ] ) occupy the same position in this space . lines of constant impact time , @xmath7 ( orange ) , and maximum ejecta velocity , @xmath8 ( equation [ eqn : vejss ] ; green ) , are derived from equation [ eqn : rcss ] . shaded regions show where our assumptions would be violated : when the shock region is optically thick to electron scattering and traps photons ( @xmath9 ) , the shocked gas cools ( @xmath10 ) , the csm impacts dense inner ejecta instead of outer ejecta layers ( @xmath11 , equation [ eqn : v_t ] ) , or the light curve is altered by synchrotron self - absorption ( ssa ; @xmath12 for @xmath13 k and @xmath14 , equation [ eqn : tau_ssa ] ) . , width=326 ] previous work on csm interaction has used the self - similar formalism of @xcite ( hereafter , c82 ) . in that model , freely expanding ejecta described by a power law density profile @xmath15 runs into stationary csm with density @xmath16 . the notations of these expressions are as in c82 : @xmath17 and @xmath18 are scaling parameters , @xmath19 is the time since explosion , @xmath20 the distance from the supernova center , and @xmath21 and @xmath22 are the power law indices describing the interacting media . the self - similar evolution allows various quantities , including shock position and velocity , to be simply calculated at any time . this formalism has been applied to various core - collapse supernova data with favorable results , and for this reason , as well as its ease of use , the c82 framework is prevalent in studies of isne . the well - deserved popularity of this model has motivated its use as a foundation for the `` fiducial model set '' of this study . following c82 , we adopt power law density profiles for both the ejecta and the csm . for the ejecta , we use a broken power law profile @xcite , which provides a reasonable approximation to the structure found in hydrodynamical models of sne . for sn ia models , @xcite suggest @xmath23 interior to the transition velocity ( @xmath24 ) and @xmath25 exterior . we are primarily interested in the outer ejecta ( @xmath26 ) , which has a density profile @xmath27 v_t = ( 1.014 \times 10^{4}~ \kms)~ ( e_{51}/m_\mathrm{c } ) , \label{eqn : v_t } \end{aligned}\ ] ] where @xmath28 is ejecta mass in units of the chandrasekhar mass and @xmath29 is the explosion energy ( in this work , we take @xmath30 and @xmath31 ) . we assume that the power - law profile of the ejecta extends to a maximum velocity , @xmath32 , above which the density drops off precipitously . the ejecta is initially freely expanding with uniform temperature @xmath33 k ; the value of the preshock ejecta temperature does not play a significant role in the evolution of the shocked gas . the csm is taken to be confined to a shell with an inner radius , @xmath34 , and width @xmath35 . we define the _ fractional width _ , @xmath36 , as @xmath37 in this work a `` thin '' shell is a shell with @xmath38 and a `` thick '' shell has @xmath39 . for example , @xcite predict @xmath40 for the shells swept up by nova outbursts . in our calculations , we assume that there is no csm interior to @xmath34 . the csm at the start of our simulation ( the time of impact ) is constant density ( @xmath41 ) , constant velocity ( @xmath42 ) , and isothermal ( @xmath43 k ) . the chosen value of @xmath44 is arbitrary and unimportant as long as it is much less than the ejecta velocities . observations of variable narrow absorption features in sn ia spectra demonstrate the possibility of moderate velocities , such as @xmath45 in ptf 11kx @xcite . the ejecta impacts the shell at time @xmath7 after explosion . in our set of fiducial models , we set initial conditions such that the ratio of @xmath5 to @xmath46 at the contact discontinuity is a fixed value . physically , our motivation is that the interaction typically becomes most prominent once @xmath47 . the value we use is @xmath48 so that the initial contact discontinuity radius @xmath34 corresponds to the c82 contact discontinuity radius at the time of impact ( equation 3 of c82 ) @xmath49}^{1/(n - s ) } \\ & = & { \left ( } 0.041 \mej v_t^7 \rp^{0.1}~ \timp^{0.7}~ { \rho_\mathrm{csm}}^{-0.1 } \label{eqn : rcss } \\ & = & ( 5.850\times10^{14}~\cm)~ { \left ( } \frac{\timp}{\dayunit } \rp^{0.7 } \rho_{{\mathrm{csm}},-18}^{-0.1 } , \end{aligned}\ ] ] where @xmath50 ( for @xmath51 ) and @xmath52 ; because of this connection to the c82 self - similar solution , we use the subscript `` @xmath53 '' in this work to note when this constraint on the initial conditions is in place . although our calculations do not assume a self - similar structure , restricting our initial setup in this way turns out to produce a family of light curve models that are amenable to simple parameterization . we explore calculations with different initial conditions in [ sec : high_vel_ej ] . equation [ eqn : rcss ] shows that , within the fiducial model set , choosing @xmath34 and @xmath5 fixes the impact time @xmath7 , and hence the maximum ejecta velocity @xmath54 . figure [ fig : paramspace ] illustrates the values of @xmath6 and @xmath5 shown in this work , and shows that the fiducial model set obeys the assumptions of transparency and lack of radiative cooling ( quantified in [ app : tscales ] ) . figure [ fig : hydro_ev ] shows the evolution of the energy and mass density profiles for a supernova impacting a typical @xmath55 shell model . immediately before impact ( panel a ) , the csm has low energy density , a flat density profile , and is moving slowly . the interaction creates a forward shock moving into the csm , and a reverse shock propagating backwards into the ejecta . the forward shock eventually reaches the outer edge of the shell at a time we call shock breakout " . at this time ( panel b ) , the dynamics have not yet reached a self - similar state ; the internal energy density of the gas , @xmath56 , is high and the csm has been accelerated to nearly the ejecta speed . shock breakout accelerates the csm such that shortly afterward ( panel c ) , the mass and energy densities have dropped drastically and the outermost csm has reached speeds over @xmath57 . at a time 1.41 times the impact time ( panel d ) , the velocity profile has approximately returned to the @xmath58 of free expansion . the mass and energy density profiles will thereafter decrease according to adiabatic free expansion . the impact radius constraint @xmath59 ( equation [ eqn : rcss ] ) defines a fiducial set of single - shell interaction models . as we will see in the next section , these models generate a family of light curves from which it is possible to deduce csm properties . radio emission is a tell - tale sign of interaction and commonly used to study astronomical shocks . here we present radio synchrotron light curves of the models described in the previous section , using the equations given in appendix [ subapp : emissivity ] . these models are restricted to low densities where synchrotron self - absorption is not important ( figure [ fig : paramspace ] ) . calculating light curves requires that we parameterize the fraction of postshock energy in magnetic fields , @xmath60 , and in non - thermal relativistic electrons , @xmath61 . we assume constant values typically used in the literature , @xmath62 , but note that these quantities are a main source of uncertainty in predicting the radio emission . calculating x - ray and optical line signatures of our models is an exciting possibility for future work , but here we wish to focus on illustrating the radio emission produced by shell interaction and how it differs from the commonly applied self - similar model . we assume that the csm is of solar composition and that the conditions in the shocked ejecta are such that only the fully - ionized shocked csm ( temperature exceeding @xmath63 k ) contributes to the synchrotron emission @xcite . representative light curves are shown in figure [ fig : phys_lc ] for interactions with shells of various widths , impact radii , and csm densities . the light curves initially rise as the shock moves through the shell and increases the volume of shocked material ( the emission region ) . the light curves reach a sharp peak , then rapidly decline . the peak occurs at shock breakout , which we define as the moment the radius of the forward shock front reaches the outer radius of the csm . the rapid decline can be attributed to the plummeting internal energy of the gas as the shell suddenly accelerates ( figure [ fig : hydro_ev ] , equation [ eqn : jnu_synch_3p ] ) . changing only the shell width changes the time of shock breakout and therefore wider shells in which breakout necessarily occurs later produce broader , brighter light curves . thinner shells follow the rise of the thicker shells until the time of shock breakout , as expected since the shock evolution should be the same while inside the shell . in the left panel of figure [ fig : phys_lc ] we show how including the contribution of the reverse shocked ejecta affects the radio light curves . we use the same radiation parameters for the ejecta as for the csm ( except we assume metal - rich gas so @xmath64 in the calculation of electron density ; see appendix [ subapp : emissivity ] ) . we find that under these assumptions the reverse shock contributes negligibly ( about 10% ) to the light curve before shock breakout but becomes dominant at later times , when it has a much higher energy density than the csm ( see panel d of figure [ fig : hydro_ev ] ) . however , we stress that it is unclear whether the conditions in the postshock ejecta are the same as in the postshock csm ; emission from the postshock ejecta could be dimmer than we have calculated here if , for instance , this region has a lower value of @xmath65 . because of the uncertainty in how to treat the ejecta relative to the csm , in this work we do not include the ejecta when calculating the radio emission unless we explicitly state otherwise . the distance of the shell from the supernova primarily affects the impact time , as one may expect , and has a small effect on the peak luminosity reached ; the density of the shell primarily affects the luminosity and has a small effect on the impact time . when fractional width is held constant as is done when we vary impact radius and shell density , we see that the light curves have the same shape and are simply shifted around in @xmath66 space , as can be seen in the middle and right panels of figure [ fig : phys_lc ] . in figure [ fig : radio_lc_samp ] we show the light curves of our fiducial suite of numerical models ( having different fractional widths , csm densities , and impact times ) normalized by their impact time and early - time luminosity . this view confirms that changing the impact radius and csm density really does just shift the light curves in log - space and details of the light curve shape are governed by @xmath36 alone . , and luminosity normalized to the luminosity at @xmath67 ( @xmath68 ) , when all light curves are rising . curves are color coded by shell fractional width , @xmath36 = @xmath69 , with logarithmic spacing between @xmath55 and @xmath70 . multiple models of varying @xmath5 and @xmath7 are shown for @xmath71 ( red , cyan , and orange , respectively ) to demonstrate that the normalized light curves are nearly identical . reconstructed light curves based on the empirical parameterization described in [ sec : shell_lcs ] are shown for @xmath72 ( black dashed lines ) , and provide an excellent approximation to the hydrodynamical model light curves . the slopes predicted for self - similar evolution ( @xmath73 , appendix [ subapp : selfsimlc_calc ] ) and for free expansion with adiabatic energy loss ( @xmath74 , appendix [ subapp : adiabaticlc ] ) are shown as dotted lines with arbitrary normalization ; they do not capture the shape of these light curves . , width=336 ] for reference , we show in figure [ fig : radio_lc_samp ] the slope of the radio light curve predicted by the self - similar model ( equation [ eqn : lss_tscale ] ) . while at later times the shell light curves approach the shallow slope of the self - similar model , at most times this slope does not describe the light curve . we also show the decay expected after peak from a simple analytic model that accounts for adiabatic losses in freely expansion gas ( equation [ eqn : ladiab ] ) . the analytic curve is shallower than the decline from our models , as the models accelerate and expand faster than free expansion immediately after shock breakout . the similarity of the fiducial model set light curves is rooted in fact that the models all have the same value of @xmath75 , the initial ratio between the pre - shock csm and ejecta density at the contact discontinuity ( equation [ eqn : rcss ] ) . our next step is to quantitatively define the family of light curves so that any light curve can be cheaply reconstructed for a given @xmath36 and peak luminosity @xmath76 . first , we quantify how the fractional width of the shell , @xmath36 , affects the timescale for the light curve to reach peak , @xmath77 . the peak occurs when the forward shock radius ( @xmath78 ) is equal to the outer shell radius , @xmath79 r_{c,0}$ ] ( see equation [ eqn : f ] ) . in the c82 self - similar solution , @xmath80 and @xmath81 for our choices of @xmath82 and @xmath41 this predicts that the time of peak is related to @xmath36 through @xmath83 if instead the shock is assumed to move at constant velocity ( as is true in the early evolution before self - similarity is reached ) @xmath84 so @xmath85 . figure [ fig : radio_tpeak ] shows @xmath86 versus @xmath87 for our suite of numerical calculations , which we find are well fit with the power law @xmath88 unsurprisingly , the exponent @xmath89 lies between the self - similar ( @xmath90 ) and free - expansion ( @xmath91 ) predictions . ( blue ) normalized to impact time versus the fractional width of the shell , @xmath36 ( equation [ eqn : f ] ) for our fiducial model suite ( open circles ) . broader shells ( i.e. , higher @xmath36 ) have broader light curves ( as illustrated in figure [ fig : radio_lc_samp ] ) . this trend is well fit by a power law ( dashed lines ) , which forms the basis for the parameterization of a fiducial model light curve . fit parameters for the power law of equation [ eqn : plawfit ] are given in table [ tab : fits ] . , width=326 ] we find empirically that the peak luminosity depends on @xmath34 and @xmath5 in the way predicted by self - similar evolution ( equation [ eqn : lss ] ) but with a different normalization than predicted by that relation . for the @xmath39 light curves the best fit peak luminosity is @xmath92 where @xmath93 is the scaled initial contact discontinuity radius . the peak luminosity of thinner shells which as we see in figure [ fig : phys_lc ] peak at a lower luminosity because shock breakout occurs sooner can be found with a parameterization of the light curve rise , given below . to fit the light curve rise , we note that all light curves follow the same rise shape up until the time of peak . a function of the form @xmath94 fits the rise of @xmath70 light curves well , with @xmath95\ ] ] so using equation [ eqn : empirl - wide ] the rise of fiducial model light curves is @xmath96 ~. \label{eqn : asyml } \end{aligned}\ ] ] if desired , the peak luminosity of _ any _ fiducial model can be found by substituting equation [ eqn : plaw_peak ] in to equation [ eqn : asyml ] @xmath97 \label{eqn : empirl } \end{aligned}\ ] ] to describe the light curve decline , we choose four characteristic times along the fall : the time to fall a factor of @xmath98 and @xmath99 of @xmath100 . as demonstrated in figure [ fig : radio_tpeak ] for the @xmath101 , case , these characteristic times follow a power law in @xmath87 , @xmath102 with fit parameters @xmath103 and @xmath104 given in table [ tab : fits ] . between characteristic points , we interpolate light curves linearly in @xmath105 space ( figure [ fig : radio_lc_samp ] ) . lcc[ht ! ] @xmath106 & 1.11 & 1.28 + @xmath107 & 1.16 & 1.39 + @xmath108 & 1.22 & 1.52 + @xmath109 & 1.32 & 1.62 + @xmath110 & 1.48 & 1.70 putting the parameterized rise and fall expressions together , one can create an approximate light curve for any peak luminosity and shell width ; the reconstruction will be a good approximation to the full numerical calculation , as shown in figure [ fig : radio_lc_samp ] . finally , as mentioned previously , the decline of the light curve is modified when one includes a contribution from the reverse shock ( figure [ fig : phys_lc ] ) . empirically , we find that the light curve at times later than the characteristic point @xmath111 declines like @xmath112 , which when combined with the equation for time to decline to @xmath111 ( table [ tab : fits ] ) yields @xmath113 where the subscript `` f+r '' indicates the luminosity of the light curve including both the forward and reverse shock components while @xmath100 is as given in equation [ eqn : empirl ] . we illustrate this fit in the left panel of figure [ fig : phys_lc ] . ; @xmath114 , light thin ) to `` high velocity ejecta '' models ( [ sec : high_vel_ej ] ; @xmath115 , @xmath116 , dark thick ) . the csm shell widths shown are : @xmath55 ( green , narrowest light curves ) , @xmath117 ( brown ) , and @xmath70 ( purple , broadest light curves ) . all models have @xmath118 . decreasing the ejecta density relative to csm ( i.e. increasing maximum velocity ) lowers the peak luminosity and causes the light curve to peak at an earlier time since explosion . however , the effect is negligible for thick shells , and even for thinner shells the fiducial relations can estimate shell properties . [ fig : radio_highv ] , width=326 ] in our simulations we must choose an initial radius ( equivalently , velocity ) at which to truncate the ejecta , @xmath34 = @xmath119 . the fiducial model set is defined by @xmath120 ( equation [ eqn : rcss ] ) , so the maximum velocity is @xmath121 with @xmath122 . early time spectra of normal sne ia show ejecta with maximum velocities @xmath123 @xcite , and we take @xmath124 as an upper limit . one could constrain oneself to fiducial models in this velocity range , but as illustrated in figure [ fig : paramspace ] that is a narrow band of models . considering that the ejecta velocity is one of the most important constraints on sn ia explosion models , we wanted to explore the effect of ejecta with @xmath125 on our light curves . we call these `` high - velocity models '' , although they could also be thought of as `` super-@xmath75 '' models , as they have a higher initial csm to ejecta density than the fiducial model value @xmath126 . note that these models are not plotted on figure [ fig : paramspace ] , to avoid confusion regarding the reference lines that apply only to fiducial models . figure [ fig : hydro_highv ] shows the hydrodynamic evolution for a high - velocity model analogue of the model shown in figure [ fig : hydro_ev ] . the fiducial model of figure [ fig : hydro_ev ] was initialized with @xmath127 and @xmath128 , which implies @xmath129 and @xmath130 days . the high - velocity model figure of [ fig : hydro_highv ] has the same csm properties but @xmath115 , thus it has @xmath131 days ( @xmath132 ) . the high - velocity model has a lower initial ram pressure , since at a fixed radius @xmath133 ( equation [ eqn : rhoej ] ) and @xmath134 ; therefore the forward shock is slower and the light curve takes a few hours longer to reach peak ( time of shock breakout ; figure [ fig : hydro_highv ] , panel b ) . example light curves of high - velocity models with varying @xmath36 are presented in figure [ fig : radio_highv ] , along with their fiducial model counterparts . including the higher velocity ejecta leads to light curves that are broader , dimmer , and peak earlier . as discussed , the effect on breadth is because the shock is slower so it takes longer to cross the shell . the dimming is also due to the lower shock speed ( equations [ eqn : u_rhovsq ] and [ eqn : jnu_synch_3p ] ) . that the light curves peak earlier relative to explosion time is simply due to the earlier impact time . the offsets in peak luminosity and time of peak are more pronounced in thinner shell models , both in absolute and relative terms . therefore we decided to apply the fiducial model relations of [ sec : shell_lcs ] to see how well they recover shell properties from the @xmath55 high - velocity model . if the @xmath115 light curve had been observed , one would calculate that @xmath135 ( equation [ eqn : plaw_peak ] ) , @xmath136 , and @xmath137 ( equations [ eqn : rcss ] and [ eqn : empirl ] ) all correct to within a factor of a few . overall , we conclude that one can still apply the fiducial model relations to an observed light curve to reasonably estimate csm properties even if the derived parameters yield a @xmath8 that is below the observed ejecta velocity , @xmath138 . therefore , in the next section , we will take the liberty of applying the fiducial relations to the analysis of isne ia radio non - detections . in this section we use the radio observations of sn 2011fe from ( * ? ? ? * hereafter csm12 ) and of sn 2014j from ( * ? ? ? * hereafter pt14 ) to illustrate how our models can be used to interpret radio nondetections in sne ia . as detailed below , we do this by calculating a suite of parameterized light curves relevant to the observations , determining if each would be observed , then converting these observation counts into observation likelihoods . the analysis begins by defining the set of observations . csm12 observed sn 2011fe in two bands with effective frequency @xmath139 ghz at times after explosion @xmath140 , with luminosity limits @xmath141 for each observation . to interpret these observational constraints , we generate @xmath142 light curves corresponding to different interaction models by log - uniform randomly distributed sampling of peak luminosity , shell width , and impact time with values in the range @xmath143 , @xmath144 , and @xmath145 . for each light curve , we then determine if it was observed , i.e. if it was above the csm12 luminosity limit for any single observation . we restrict ourselves this way since if the shell had been observed it would have affected the observation schedule ; but using our models one could interpret stacked data sets as well . finally , we bin the light curves in @xmath36 and @xmath76 and in each bin calculate the probability that it would have been observed ( the ratio of the number of light curves observed to the number sampled ) . the probability of the light curve appearing in any single observation is shown in figure [ fig : p_obs ] . ) with peak luminosity @xmath146 from a shell of fractional width @xmath36 ( equation [ eqn : f ] ) for a set of light curves that have log - uniform randomly assigned values of @xmath147 and @xmath148 . for a given @xmath36 and initial shell radius , peak luminosity can be converted into csm density using equation [ eqn : empirl ] as is done in figure [ fig : p_obsthin ] for sn 2014j.,width=326 ] ) that any single radio observation of @xcite would detect a thin csm shell ( @xmath55 ) of a given mass density ( @xmath5 ) around sn 2014j if impact occurred during their observation window , 8.2 - 35.0 days . , width=326 ] csm12 attempted to estimate , without using explicit shell light curve models , the probability of detecting a specific csm shell . the shell they considered had fractional width @xmath149 , mass @xmath150 ( from a wind of @xmath151 ) , and sat at @xmath152 cm ; the authors estimated a 30% chance of detecting the shell . the preshocked csm density is @xmath153^{-1}\ ] ] where @xmath154 and @xmath155 . we estimate that the shell described by csm12 , which has @xmath156 , and @xmath157 would have a peak luminosity ( equation [ eqn : empirl ] ) @xmath158 , and would be undetected in their observations . for reference we note that at our assumed distance and width , a shell would need to have @xmath159 to have @xmath160 chance of detection . for sn 2014j , we will demonstrate how one can instead analyze the probability of observing a shell of a given density , rather than peak luminosity . the observations of sn 2014j by @xcite span the range 8.2 - 35.0 days . using the observation times , frequencies , and limits from pt14 table 1 , we perform the same random sampling test as for sn 2011fe described above , but with a peak luminosity range @xmath161 at 1.55 ghz ( their lowest frequency band ) and impact times between 2 and 35 days . isolating the @xmath55 bin , we can convert luminosity to density and analyze the probability that the authors would have observed an impact with a shell of a given @xmath5 that occurred during their observations , illustrated in figure [ fig : p_obsthin ] . the same shell considered for csm12 but at @xmath162 , to account for the later observation time , would again escape detection at @xmath163 . as we have shown here , by using the parameterized light curves and fiducial model relations one can easily quantify the probability of detecting shells spanning a range of @xmath36 and @xmath5 for a given set of observations . alternatively , one could constrain for instance the ejecta velocity and @xmath36 , and use equation [ eqn : vejss ] to analyze the likelihood of detecting shells of varying radii and densities . we have demonstrated that the light curves of ejecta impacting a single shell form a family of solutions that can be straightforwardly applied to the interpretation of radio observations . in these simulations , we assume that there is negligible csm interaction prior to the interaction being considered - i.e. the space between the supernova progenitor and the csm shell is empty . however , it is natural to wonder how it would look if one shell collision was followed by another : could we simply add the light curves , preserving their convenient properties ? in this section we perform this experiment and show that the light curves from multiple shell collisions differ dramatically from a simple addition of two single - shell light curves . imagine a system with two thin ( @xmath55 ) csm shells : shell 1 at @xmath164 cm with @xmath165 and shell 2 at @xmath166 cm with @xmath167 . the ejecta impacts shell 1 at 5 days in a self - similar model , and if shell 1 were not present the ejecta would impact shell 2 at 50 days ; therefore , models of the _ single - shell _ impact with these shells are called `` 1sh5 '' ( shell 1 ) and `` 1sh50 '' ( shell 2 ) . the double - shell model ( `` 2sh27 '' ) is the same as 1sh5 until shell 1 impacts shell 2 on day 27 . note that the earlier impact time ( 27 days versus 50 days ) is due to the fact that this second interaction starts not when the _ ejecta _ impacts shell 2 but rather when _ shell 1 _ does , which happens earlier since shell 1 is exterior to the ejecta . the initial condition of 2sh27 is the day-27 state of 1sh5 for @xmath168 plus the initial state of 1sh50 for @xmath169 , where @xmath170 is the contact discontinuity radius between the shells . the number of zones in the 1sh50 csm was adjusted to match the mass resolution of 1sh5 . figure [ fig:2sh_synchlc ] shows that the 2sh27 light curve is dramatically different than those of 1sh5 and 1sh50 . first , the 2sh27 light curve is strikingly flat and broad , maintaining a luminosity similar to the peak luminosity of 1sh50 for nearly twenty days . second , the light curve has an elbow at @xmath171 days ( time c ; corresponding to panel c of figure [ fig:2sh_toon ] ) . note that in this figure , the 1sh5 light curve would in fact be subject to synchrotron self - absorption , an effect that we neglect here because it is not the interesting part of the evolution . .,width=326 ] since our 1-d simulations do not have mixing , we can calculate the luminosity contribution from the shell 1 and shell 2 zones separately ( which zones belong to each shell is known by construction , as illustrated in figure [ fig:2sh_toon ] ) . when separated into the contribution from each shell , we see that the light curve is dominated by the signal from shell 1 . this is simply because of the chosen frequency of @xmath172 ghz ; in our model the ratio of cyclotron frequency in shell 2 to shell 1 is @xmath173 , thus @xmath174 . the shell 2 light curve also has an elbow ( time e ; @xmath175 days ) that occurs later than the shell 1 elbow . finally , both the shell 1 and shell 2 light curves start their final decline at @xmath176 days ( time f ) . this complicated light curve is caused by `` sloshing '' of shocks in shell 1 which re - energize the gas for @xmath177 days , the dynamical time ( appendix [ app : tscales ] ) this `` sloshing '' can be seen in snapshots of the simulation at critical times ( figure [ fig:2sh_toon ] ) . remember that shell 1 dominates the signal in the following narrative of the interaction : ( a ) _ impact . _ shell 1 is much less dense than either the ejecta or shell 2 , and it is approaching shell 2 at high speeds . ( b - c ) _ shock moves leftward through shell 1 . _ when shell 1 hits the `` wall '' of shell 2 , a shock forms . it energizes shell 2 in a way reminiscent of single - shell evolution , breaking out quickly . ( d ) _ shock forms at ejecta - shell 1 boundary and re - energizes shell 1 . _ because of the new shock , @xmath56 increases and the light curve has another small rise . ( e ) _ shock forms at the shell 1-shell 2 boundary , re - energizing both shells . _ we see that when the shock moving through shell 1 ( seen in panel d ) hits shell 2 , a new shock forms . the shell 2 light curve begins a dramatic second rise and shell 1 is able to maintain its luminosity . ( f ) _ after a dynamical time the system has had time to adjust and the light curve begins to decline . _ this panel shows a snapshot just before the decline of both light curves . minor shocks continue to slosh inside shell 1 between the ejecta and shell 2 , causing the more gradual light curve fall compared to single - shell models . this is just one possible configuration of a double - shell system and its evolution probably does not generalize to the whole class of such collisions ; however , it illustrates that the evolution of these systems is distinct from single - shell model evolution and so produces a radically different light curve . note that higher - dimensional simulations may be required for investigating this class further because the significance of the `` shock slosh '' could be diminished by turbulence in order to quantitatively interpret the radio light curves and spectra of isne , one needs to understand the evolution of the shocked gas . one popular model for shock evolution is the ( * ? ? ? * c82 ) one - dimensional self - similar interaction model , which assumes a continuous csm . here we explored moving beyond these models to the case of csm shells ( e.g. due to novae ) in the low - density case where the shocked csm does not cool and is optically thin . this removes the self - similar model assumption that the interaction has had time to reach self - similarity ( i.e. that the csm is close to the progenitor star or is very extended ) . we created a suite of models to explore the parameter space of shell properties , focusing on shell density , thickness , and time of impact . in our `` fiducial '' models , we enforce an initial csm - to - ejecta density ratio of @xmath178 at the contact discontinuity ( motivated by c82 ; equation [ eqn : rcss ] ) . if the shell were infinite , these models would evolve to the c82 self - similar solution ( [ sec : mods ] ) ; however , because of the finite extent of the shells , self - similarity is never reached in our models . we calculated synchrotron radio light curves for these models , since this emission traces the evolution of the shocked region only . we found a similar behavior among the radio light curves of fiducial models they peak at a predictable time ( equation [ eqn : plaw_peak ] ) , have the same shape ( equation [ eqn : asyml ] ) , and peak at a luminosity that can be calculated from shell characteristics ( equation [ eqn : empirl ] ) . therefore these light curves can be parameterized in a way that allows csm properties to be inferred readily from an observed light curve ( [ sec : shell_lcs ] ) . in fact we find that shell properties can be approximated even if there is a higher initial density ratio ( i.e. faster , lower density ejecta colliding with the shell ) than is assumed in the fiducial set ( [ sec : high_vel_ej ] ) . we then showed how one can use the fiducial models to better interpret radio non - detections of sne ia by applying our work to the radio observations of sn 2011fe and sn 2014j ( [ sec : obscomp ] ) . we presented a two - shell model to illustrate that our single shell light curves do not apply in that case , as a double shell light curve is much longer lived because of an effect that we call `` shock sloshing '' that occurs in the first shell ( [ sec : multishell ] ) . shock sloshing produces elbows in the light curve that may be an observable signature of this behavior ; exploring the diversity in multiple shell collisions is an enticing direction for future work . our fiducial model set is a new tool for studying the interaction of sne ia with a csm , with the goal of constraining the supernova progenitor system . with the fiducial set , one can easily make observational predictions and analyze radio observations . future work will focus on calculations of the radiation transport in the optical and x - ray , to show how data at those wavelengths can reinforce radio constraints on supernovae interaction . the authors thank the anonymous referee for improving the clarity of this manuscript . c.e.h . is supported by the department of energy computational science graduate fellowship . we simulate the interaction between ejecta and csm under the assumptions that ( 1 ) the gas is optically thin to electron scattering , so photons can free - stream through it and ( 2 ) the gas does not radiatively cool . to investigate the legitimacy of these assumptions , we can compare the timescale for dynamical changes ( @xmath179 ) to the diffusion time ( @xmath180 ) and the cooling time ( @xmath181 ) in the forward shock region . the assumption that photons can free - stream requires that the diffusion time be much less than the dynamical time . the diffusion time for electron scattering through a shocked region of width @xmath183 and optical depth @xmath184 is @xmath185 assuming that all species are fully ionized within the shocked region , the electron number density ( @xmath186 ) and ion number density ( @xmath187 ) are related to the mass density ( @xmath188 ) , atomic number ( @xmath189 ) , atomic mass ( @xmath190 ) , and proton mass ( @xmath191 ) through @xmath192 and @xmath193 , so the optical depth to electron scattering is @xmath194 with @xmath195 the density of the gas in the shocked region per @xmath196 and @xmath197 the width of the shocked region per @xmath198 . for the estimates here , we assume the gas is hydrogen dominated so @xmath199 . assuming a typical value for the shock width @xmath200 and that the postshock density is related to the initial csm density as @xmath201 , @xmath202 the assumption that the shocked gas does not cool requires that the cooling time be much longer than the dynamical time . the cooling time for any radiative process is given by @xmath203 where @xmath56 is the gas internal energy density ( which we have assumed is that of a monatomic ideal gas ) and @xmath204 is the angle - averaged , frequency - integrated emissivity of a radiative process . for free - free emission of thermal electrons , the angle - integrated emissivity ( corresponding to @xmath205 in the above equation ) is @xmath206 in cgs units @xcite . choosing a value of 1.2 for the gaunt factor @xmath207 gives an accuracy of about 20% . using @xmath208 as above and again assuming the shocked material has a density @xmath209 and @xmath210 , @xmath211 where @xmath212 . the strong shock jump conditions for an ideal gas link the temperature and shock velocity as @xmath213 \label{eqn : u_rhovsq } \ ; \rightarrow \ ; t_9 & \approx & 9 \mu v_{s,9}^2 \label{eqn : tfromv } \end{aligned}\ ] ] where here @xmath214=5/3 is the adiabatic index and @xmath215 is the shock velocity per @xmath216 . for our models we expect that @xmath217 is similar to the ejecta velocity so @xmath218 and thus the cooling timescale is much longer than the dynamical timescale . for the fiducial model set , we can use equations [ eqn : rcss ] and [ eqn : tfromv ] to recast the diffusion and cooling timescales in terms of physical parameters . figure [ fig : paramspace ] illustrates these constraints and shows that our models obey them . to calculate radio synchrotron light curves , we assume a fraction @xmath219 of electrons are accelerated into a non - thermal power law distribution @xmath220 ( with @xmath221 the electron lorentz factor ) with a total number density @xmath222 . data for stripped - envelope core - collapse supernovae favors @xmath223 @xcite . the normalization @xmath224 is given by @xmath225 if a fraction @xmath226 of the shock energy went into accelerating the electrons ( energy density @xmath227 ) , then @xmath228 which we can combine with equation [ eqn : c ] to show that @xmath229 for the special case @xmath230 which we assume in this work , @xmath231 c & = & \frac{1}{2 } \nent { \left ( } \frac{u_e}{\nent m_e c^2}{\right)}^2 \label{eqn : c_3p } \end{aligned}\ ] ] assuming isotropic radiation , we can derive the synchrotron emissivity of these electrons as the power density per frequency per steradian @xmath232 where @xmath233 is an electron s power output in synchrotron radiation @xmath234 with @xmath235 the magnetic field energy density ( @xmath236 ) and @xmath237 . we parameterize the magnetic field energy density as @xmath238 , as is common practice . the frequency of synchrotron radiation is related to the cyclotron frequency @xmath239 as @xmath240 thus @xmath241 using equations [ eqn : psyn][eqn : dgdn ] in equation [ eqn : j_syn_orig ] , @xmath242 where here we have assumed that @xmath243 , with @xmath244 the speed of light ; to relax this assumption one would multiply the above by @xmath245 . this is a slightly simpler version of the anisotropic equation given in @xcite , and the two are equal to within a factor of order unity if one takes the pitch angle terms in the latter equation to be their angle - averaged value @xmath246 . for @xmath230 , we can use equation [ eqn : c_3p ] and our parameterizations @xmath238 , @xmath247 to write this as @xmath248 assuming fully ionized hydrogen ( @xmath249 with @xmath250 as described for equation [ eqn : tau_es ] ) , this gives @xmath251 & & \times { \left ( } \frac{\rho}{10^{-18}~\gccunit } \rp^{-1 } { \left ( } \frac{\ugas}{10 ^ 2~\mathrm{erg~cm^{-3 } } } \rp^3~ { \left ( } \frac{\nu}{\mathrm{ghz } } \rp^{-1 } \label{eqn : physj_p3 } \end{aligned}\ ] ] for our simulations , we assume the csm is solar composition and that only regions with @xmath252 k radiate ( and that these regions are fully ionized ; recall that for times of interest the radiating regions have @xmath253 k ) . the radiating region extends from the ejecta - csm contact radius to the forward shock radius , which is identified by an algorithm that finds the sharp break in the @xmath56 profile ( see , e.g. figure [ fig : hydro_ev ] ) . we calculate the luminosity by calculating the emissivity in each radiating zone ( zone defined in [ sec : mods ] ) scaled by its peak value ( at @xmath254 ) , then multiplying by the volume of the zone and @xmath255 steradians , as is appropriate in the optically thin limit . for a quantitative discussion of our optically thin assumption as it applies to synchrotron self - absorption , see [ subapp : ssa ] . as discussed in the main text , for a few models we consider the emission from the reverse shock region . this is calculated in the same manner except that the composition is assumed to dominated by species heavier than hydrogen , in which case @xmath256 . next we use the synchrotron emissivity to calculate the luminosity predicted from the c82 self - similar evolution . in the case of constant - density csm , we can use our model parameters ( see [ sec : mods ] , particularly equation [ eqn : rhoej ] ) in equation 3 of c82 to derive that a self - similar shock has speed @xmath257 where @xmath78 is the radius of the forward shock front , @xmath217 is the shock velocity , and as usual @xmath5 is the preshock csm density and @xmath258 is the radius of the contact discontinuity . substituting this into equation [ eqn : u_rhovsq ] , @xmath259 thus the emissivity ( equation [ eqn : jnu_synch_3p ] , with @xmath260 ) , assuming fully ionized hydrogen gas , evolves like @xmath261 and the emitting volume increases like @xmath262 which together produce @xmath263 for our model parameters of @xmath30 , @xmath264 , @xmath265 , @xmath266 where @xmath267 is the current contact discontinuity radius and @xmath52 . from equation [ eqn : rcss ] , we see that @xmath268 so the scaling of luminosity with time is @xmath269 after several dynamical times ( equation [ eqn : tdyn ] ) have passed since shock breakout , the system should have relaxed into a state of free expansion , with @xmath270 . in this state , we expect adiabatic evolution of the gas , @xmath271 this produces a deline in the emissivity @xmath272 so in this state of free expansion we have @xmath273 this explains why the decline in the radio light curves is so steep after shock breakout . the relativistic electrons that create synchrotron emission can also absorb synchrotron radiation in a process known as synchrotron self - absorption ( ssa ) . from @xcite , the extinction coefficient for ssa is @xmath274 where @xmath275 is the gamma function and @xmath276 is from @xmath277 and is thus related to the @xmath224 of equation [ eqn : c ] by @xmath278 . we have used @xmath230 in the second expression . as previously mentioned , the @xmath279 term is a pitch angle term that is attached to the magnetic field amplitude @xmath280 and thus carried through the derivation . in our situation , we expect both the magnetic fields and the electron velocities to be randomly oriented and therefore when using synchrotron equations from @xcite , we adopt the angle - averaged value @xmath281 . thus with some substitution we can write this as @xmath282 & = & ( 4.74\times10^{-7}~\mathrm{cm^{-1 } } ) { \left ( } \frac{\epse}{0.1 } \rp^2 { \left ( } \frac{\epsb}{0.1 } \rp^{5/4}~\nonumber\\[8pt ] & & \times { \left ( } \frac{\rho}{10^{-18}~\gccunit } \rp^{-1 } { \left ( } \frac{\ugas}{10 ^ 2~\mathrm{erg~cm^{-3 } } } \rp^{13/4 } { \left ( } \frac{\nu}{\mathrm{ghz } } \rp^{-7/2 } \label{eqn : alssa } \end{aligned}\ ] ] where @xmath188 is the mass density of the shocked gas and @xmath56 its internal energy density . we can estimate the optical depth of a model by assuming @xmath283 is uniform in the shocked gas . with @xmath284 and @xmath285 the energy density is @xmath286 so @xmath287 where @xmath288 . assuming the shocked region has a width @xmath289 then @xmath290 the optical depth of a model varies with time because of these dependencies on temperature and radius , but we show the line of @xmath291 in figure [ fig : paramspace ] assuming @xmath292 and @xmath293 for reference . when fiducial models are optically thin such that the synchrotron luminosity is set by the emissivity , their light curves obey scaling relations that reflect the similar hydrodynamic evolution of the models as described in [ sec : shell_lcs ] . as the models become more optically thick , the light curve is set by the source function @xmath294 , which from equations [ eqn : physj_p3 ] and [ eqn : alssa ] we calculate to be @xmath295 & & \times { \left ( } \frac{u_b}{10~\mathrm{erg~cm^{-3 } } } \rp^{-1/4}~ { \left ( } \frac{\nu}{\mathrm{ghz } } \rp^{-5/2}~. \end{aligned}\ ] ] , y. , kulkarni , s. r. , howell , d. a. , gal - yam , a. , kasliwal , m. m. , valenti , s. , johansson , j. , amanullah , r. , goobar , a. , sollerman , j. , taddia , f. , horesh , a. , sagiv , i. , cenko , s. b. , nugent , p. e. , arcavi , i. , surace , j. , woniak , p. r. , moody , d. i. , rebbapragada , u. d. , bue , b. d. , & gehrels , n. 2015 , , 521 , 328 , b. , howell , d. a. , cenko , s. b. , silverman , j. m. , nugent , p. e. , sullivan , m. , ben - ami , s. , bildsten , l. , bolte , m. , endl , m. , filippenko , a. v. , gnat , o. , horesh , a. , hsiao , e. , kasliwal , m. m. , kirkman , d. , maguire , k. , marcy , g. w. , moore , k. , pan , y. , parrent , j. t. , podsiadlowski , p. , quimby , r. m. , sternberg , a. , suzuki , n. , tytler , d. r. , xu , d. , bloom , j. s. , gal - yam , a. , hook , i. m. , kulkarni , s. r. , law , n. m. , ofek , e. o. , polishook , d. , & poznanski , d. 2012 , science , 337 , 942 , j. m. , alberdi , a. , ros , e. , diamond , p. , shapiro , i. i. , guirado , j. c. , jones , d. l. , mantovani , f. , prez - torres , m. a. , preston , r. a. , schilizzi , r. t. , sramek , r. a. , trigilio , c. , van dyk , s. d. , weiler , k. w. , & whitney , a. r. 1997 , , 486 , l31 , r. , soderberg , a. m. , chomiuk , l. , chevalier , r. , hurley , k. , milisavljevic , d. , foley , r. j. , hughes , j. p. , slane , p. , fransson , c. , moe , m. , barthelmy , s. , boynton , w. , briggs , m. , connaughton , v. , costa , e. , cummings , j. , del monte , e. , enos , h. , fellows , c. , feroci , m. , fukazawa , y. , gehrels , n. , goldsten , j. , golovin , d. , hanabata , y. , harshman , k. , krimm , h. , litvak , m. l. , makishima , k. , marisaldi , m. , mitrofanov , i. g. , murakami , t. , ohno , m. , palmer , d. m. , sanin , a. b. , starr , r. , svinkin , d. , takahashi , t. , tashiro , m. , terada , y. , & yamaoka , k. 2012 , , 751 , 134 , p. e. , sullivan , m. , cenko , s. b. , thomas , r. c. , kasen , d. , howell , d. a. , bersier , d. , bloom , j. s. , kulkarni , s. r. , kandrashoff , m. t. , filippenko , a. v. , silverman , j. m. , marcy , g. w. , howard , a. w. , isaacson , h. t. , maguire , k. , suzuki , n. , tarlton , j. e. , pan , y .- c . , bildsten , l. , fulton , b. j. , parrent , j. t. , sand , d. , podsiadlowski , p. , bianco , f. b. , dilday , b. , graham , m. l. , lyman , j. , james , p. , kasliwal , m. m. , law , n. m. , quimby , r. m. , hook , i. m. , walker , e. s. , mazzali , p. , pian , e. , ofek , e. o. , gal - yam , a. , & poznanski , d. 2011 , , 480 , 344 , j. t. , howell , d. a. , friesen , b. , thomas , r. c. , fesen , r. a. , milisavljevic , d. , bianco , f. b. , dilday , b. , nugent , p. , baron , e. , arcavi , i. , ben - ami , s. , bersier , d. , bildsten , l. , bloom , j. , cao , y. , cenko , s. b. , filippenko , a. v. , gal - yam , a. , kasliwal , m. m. , konidaris , n. , kulkarni , s. r. , law , n. m. , levitan , d. , maguire , k. , mazzali , p. a. , ofek , e. o. , pan , y. , polishook , d. , poznanski , d. , quimby , r. m. , silverman , j. m. , sternberg , a. , sullivan , m. , walker , e. s. , xu , d. , buton , c. , & pereira , r. 2012 , , 752 , l26 , m. a. , lundqvist , p. , beswick , r. j. , bjrnsson , c. i. , muxlow , t. w. b. , paragi , z. , ryder , s. , alberdi , a. , fransson , c. , marcaide , j. m. , mart - vidal , i. , ros , e. , argo , m. k. , & guirado , j. c. 2014 , , 792 , 38 , j. m. , nugent , p. e. , gal - yam , a. , sullivan , m. , howell , d. a. , filippenko , a. v. , arcavi , i. , ben - ami , s. , bloom , j. s. , cenko , s. b. , cao , y. , chornock , r. , clubb , k. i. , coil , a. l. , foley , r. j. , graham , m. l. , griffith , c. v. , horesh , a. , kasliwal , m. m. , kulkarni , s. r. , leonard , d. c. , li , w. , matheson , t. , miller , a. a. , modjaz , m. , ofek , e. o. , pan , y .- c . , perley , d. a. , poznanski , d. , quimby , r. m. , steele , t. n. , sternberg , a. , xu , d. , & yaron , o. 2013 , , 207 , 3 , a. m. , berger , e. , page , k. l. , schady , p. , parrent , j. , pooley , d. , wang , x .- y . , ofek , e. o. , cucchiara , a. , rau , a. , waxman , e. , simon , j. d. , bock , d. c .- j . , milne , p. a. , page , m. j. , barentine , j. c. , barthelmy , s. d. , beardmore , a. p. , bietenholz , m. f. , brown , p. , burrows , a. , burrows , d. n. , byrngelson , g. , cenko , s. b. , chandra , p. , cummings , j. r. , fox , d. b. , gal - yam , a. , gehrels , n. , immler , s. , kasliwal , m. , kong , a. k. h. , krimm , h. a. , kulkarni , s. r. , maccarone , t. j. , mszros , p. , nakar , e. , obrien , p. t. , overzier , r. a. , de pasquale , m. , racusin , j. , rea , n. , & york , d. g. 2008 , , 453 , 469

using the irac images from the _ spitzer c2d _ program , we have made a survey of mid - infrared outflows in the ophiuchi molecular cloud . extended objects that have prominent emission in irac channel 2 ( 4.5 ) compared to irac channel 1 ( 3.6 ) and stand out as green objects in the three - color images ( 3.6 in blue , 4.5 in green , 8.0 in red ) are identified as mid - infrared outflows . as a result , we detected 13 new outflows in the ophiuchi molecular cloud that have not been previously observed in optical or near - infrared . in addition , at the positions of previously observed hh objects or near - infrared emission , we detected 31 mid - infrared outflows , among which seven correspond to previously observed hh objects and 30 to near - infrared emission . most of the mid - infrared outflows detected in the ophiuchi cloud are concentrated in the l1688 dense core region . in combination with the survey results for young stellar objects ( ysos ) and millimeter and sub - millimeter sources , the distribution of mid - infrared outflows in the ophiuchi molecular complex hints a propagation of star formation in the cloud in the direction from the northwest to the southeast as suggested by previous studies of the region .

0910.0540

mass outflows play an essential role in the process of star formation and have been found to be ubiquitous in various stages of star formation @xcite . the specific angular momentum of a star - forming molecular core is about four orders of magnitude times higher than that of a typical t tauri star @xcite . one way to transfer the excessive angular momentum from the star - forming cores so that the circumstellar material can be accreted onto the central stars is through mass outflows @xcite . mass outflows from ysos have been observed at different wavelengths . in the visual , high velocity jets and hh objects with a typical velocity of 100 - 300 km s@xmath1 trace material ejected by the star or shocked surrounding medium @xcite . near - infrared molecular hydrogen emission features , which usually have velocities of several tens km s@xmath1 , trace h@xmath2 gas in the jets or in the entrained surrounding medium @xcite . in the millimeter wavelength , high velocity co gas with velocity in the range from a few to ten km s@xmath1 have been observed @xcite . the high velocity co gas probes the swept or entrained medium . the different velocity ranges and excitation conditions for hh objects , near - infrared molecular hydrogen emission , and co outflows indicate the transfer of energy and momentum from the high velocity material to the surrounding medium . shocked gas in outflows has abundant atomic and molecular hydrogen line emission in the mid - infrared . hydrodynamic simulations of outflows from ysos by @xcite have shown that outflows have strong h@xmath2 line emission in all the irac four bands . in fact , mid - infrared observations with irac aboard of _ spitzer _ have discovered a lot of new outflows @xcite and plenty new features of previously known outflows @xcite , showing that _ spitzer _ irac imaging is a powerful tool in the survey of outflows , particular for those that are deeply embedded in clouds due to the much less extinction in the mid - infrared compared to the optical and near - infrared wavelengths . the _ spitzer _ legacy program _ c2d _ performed a complete imaging of five nearby star - forming regions with the irac and mips instruments @xcite , which provides a valuable data set to search for mid - infrared outflows in these regions . in this paper we present the results of a mid - infrared outflow survey of the @xmath0 ophiuchi molecular cloud based on the _ c2d _ archive images . the @xmath0 ophiuchi molecular cloud is one of the closest star - forming regions at a distance of about 130 pc @xcite . it is intermediate in star formation activity compared to the isolated star formation in the taurus cloud and the rich clusters in the orion cloud . the large scale structure of the @xmath0 ophiuchi cloud has been revealed by extensive molecular @xmath3co line mapping @xcite . millimeter and submillimeter surveys of the central part of the cloud have revealed the clumpy structure of the cloud and have found about 55 starless cores with masses in the range of 0.02 to 6.3 m@xmath4 @xcite . the cloud has been surveyed for young stars at wavelengths from x - ray @xcite , visual @xcite , near - infrared @xcite , mid- and far - infrared @xcite , to millimeter and submillimeter @xcite . based on these surveys , @xcite compiled a list of 316 young stars in l1688 , the main cloud of the @xmath0 ophiuchi complex . about 46 hh objects @xcite , 119 h@xmath2 emission features @xcite , 16 high velocity co outflows @xcite have been observed in the @xmath0 ophiuchi complex . for a summary of the @xmath0 ophiuchi complex we refer to the recent review by @xcite the irac observations of @xmath0 ophiuchi were conducted on 2004 march 28 - 30 and 2005 september 15 and 18 . these observations are part of the spitzer legacy program _ c2d _ @xcite and the program i d is 177 . fig . [ fig1 ] shows the three - color image of the area mapped with irac which is about 8.0 deg@xmath5 . this area is included in the region mapped with mips ( see fig.1 of @xcite ) . each region in fig . [ fig1 ] was individually mapped in two epochs with an exposure time of 12 seconds ( 24 seconds in total ) and one map was made in the high dynamic range ( hdr ) mode , which added a short exposure time of 0.6 seconds @xcite . six off - cloud fields were observed in order to sample the background source counts . we used the final mosaic images and source catalogs of the fourth delivery of irac and mips data . the data were processed with _ spitzer _ science center pipeline s13 and calibrated by _ c2d _ team ( details can be found in the delivery document ) . in the final mosaic images bright source artifacts have been removed . however , some instrumental effects still survive . for example , optical ghosts produced by internal reflections within the filters ( labelled with gh " in the subsequent figures ) and stray lights produced by illuminating sources off the edges of the detector arrays ( labelled in the subsequent figures with stray light " ) remain in the final mosaic images ( see the irac data handbook for details ) . a catalog of young stellar object candidates in the @xmath0 ophiuchi , ysoc , is also delivered by the _ c2d _ team . @xcite improved over this catalog . they removed some suspect sources , added known sources and data obtained at other wavelengths , and calculated additional quantities such as bolometric temperature and bolometric luminosity . their list makes a relatively pure yso candidate sample which includes 292 ysos . we used their list of ysos in this paper . two schemes for yso classification were used in @xcite . one is the spectral index that was obtained using a least - squares fit to all photometry between 2 and 24 , and the criteria for classification are from @xcite . the other scheme used the calculation of bolometric temperature @xcite . using the calculated spectral indices or bolometric temperatures , 292 ysos in the cloud can be classified into 4 classes : 27 in class i or class 0 , 44 in the flat - spectrum class , 179 in class ii , and 42 in class iii . note especially that there are three class 0 sources in ophiuchus , iras 16293 - 2422 , mms 126 , and vla 1623 - 243 , which were distinguished from class i sources on the basis of their bolometric temperatures . to analyze the distribution of detected outflows in the @xmath0 ophiuchi molecular complex , we used the @xmath3co 1 - 0 ( 110.201ghz ) emission data from the fcrao 14 m telescope at new salem , ma , u.s.a . the data were obtained during 2002 - 2005 with the on - the - fly ( otf ) mapping method @xcite and are part of the _ complete _ survey of star - forming regions . these @xmath3co observations cover 10.0 deg@xmath5 of the @xmath0 ophiuchi molecular complex and the fcrao half power beam width ( hpbw ) measures 44 . the @xmath3co integrated intensity maps that have been re - gridded to 23were used . in our @xmath3co emission contour maps , the units of integrated intensity is antenna temperature ( t@xmath6 ) in kelvin . irac channels cover a relatively broad wavelength range from 3.6 to 8.0 . in this wavelength range there are many h@xmath2 lines and atomic and ionic fine - structure lines . @xcite have shown near- and mid - infrared spectra of the brightest h@xmath2 emission peak of the orion omc-1 outflow which are taken with iso - sws . the spectra revealed a wealth of molecular lines of h@xmath2 , co , and h@xmath2o , and a number of h i recombination lines , atomic and ionic fine - structure lines . the irac 4.5 band contains strong h@xmath2 lines ( @xmath7 , s(9 , 10 , 11 ) ) and co bandheads ( @xmath8 ) @xcite . like the near - infrared h@xmath9 emission at 2.12 , the emission of these lines can be excited by the interaction of protostellar outflows with the ambient medium . @xcite presented synthetic _ irac band maps from hydrodynamic simulations of processing protostellar jets and they predicted that the integrated h@xmath9 line luminosity is greatest in irac band 2 ( 4.5 ) , about an order of magnitude higher than those in other irac bands ( see table 3 in @xcite ) . observations by _ spitzer _ irac have in general confirmed this prediction @xcite . according to @xcite , the remaining emission in the _ spitzer _ irac band 2 image after the subtraction of the emission in band 1 is mainly from shocked h@xmath9 and co gas . using the three - color images constructed from irac bands 1 , 2 , and 4 . , @xcite compiled a catalog of outflow candidates with the _ spitzer _ galactic legacy infrared mid - plane survey extraordinaire ( glimpse ) data . they identified the extended green objects ( egos ) in the three - color images as outflow candidates . in this paper , we use the difference image constructed from irac channel 2 ( 4.5 ) minus channel 1 ( 3.6 ) to search for mid - infrared outflows . we use the three - color ( 3.6 in blue , 4.5 in green , 8.0 in red ) images to confirm that the identified outflows are egos . images covering the common sky area of the @xmath0 ophiuchi cloud in the four irac bands were trimmed from the downloaded _ c2d _ archive mosaic images . the trimmed images were then rotated so that the image north direction is top and the east is to the left . for the subtraction of emission in band 1 from that in band 2 , we selected 80 bright stars in the images that are roughly evenly distributed over the sky and are not saturated in either band 1 or band 2 . the average flux ratio in band 1 to band 2 of these 80 stars is used to adjust the flux level of stars in band 1 to that in band 2 . the resultant difference images were visually inspected to search for mid - infrared outflows . we selected extended sources that have significant excess at 4.5 as mid - infrared outflows . toward the core of l1688 , several surveys for hh objects and h@xmath9 emission have been carried out @xcite and 46 hh objects , including components of hh objects , and 119 h@xmath9 emission features have been reported . we examined whether there are mid - infrared counterparts of these hh objects and h@xmath9 emission features at their corresponding positions in the irac images . we discovered 13 irac mid - infrared outflows . their coordinates and figure reference are list in table [ tab1 ] . the likely exciting sources for these outflows and their lada classes are given in table [ tab2 ] . note that iras 16293 - 2422 is classified as a class 0 source based on its bolometric temperature @xcite . we also identified 31 mid - infrared counterparts of previously known hh objects and h@xmath9 2.12 emission features . the coordinates of these mid - infrared counterparts of the previously known outflows , together with the corresponding hh objects and h@xmath2 2.12 emission features , are presented in table [ tab3 ] . in this paper we designate all the detected irac mid - infrared outflows as egos . the egos are numbered by order in right ascension . if the overall morphology of a group of knots suggests physical association among them , the knots are considered as parts of one ego and the individual knots are distinguished with additional letters to the ego numbers . on the other hand , if the knots in a region are relatively separated from each other and the relationship among them is not clear , each knot is treated as an individual ego . apparent sub - structures in a knot are labelled with additional numerals , such as ego 21b1 . the locations of all the mid - infrared outflows identified in the @xmath0 ophiuchi cloud are shown in fig . [ fig2 ] . the grey scale image in fig . [ fig2 ] is the irac channel 2 ( 4.5 ) image . the outflows newly discovered in this work are marked with circles and the counterparts of known flows are labelled with pluses . two yso aggregates in the region , l1689- and l1709-aggregate @xcite , are marked with open squares . the identified mid - infrared outflows are located mainly in three regions , i.e. , the l1688 core , the iras 16293 - 2422 region , and the l1709 yso aggregate region . the images of the newly discovered irac mid - infrared outflows are shown in figs . [ fig3]-[fig17 ] , and the images of mid - infrared counterparts of hh objects and h@xmath2 emission features are presented in figs . [ fig18]-[fig40 ] . figures 3 - 4 show the region of objects ego 04 and ego 12 . the distance between the two objects is about 3 . each object consists of two knots . ego 04 and ego 12 are only detected in irac channels 1 ( 3.6 ) and 2 ( 4.5 ) . it is brighter in channel 2 ( 4.5 ) , therefore , appears as an ego in the three - color image ( fig . [ fig4 ] ) . there are three ysos to the southeast of ego 04 and ego 12 , among which we suggest bklt j162624 - 241616 ( class ii in @xcite ) as the exciting source of the ego 04 and ego 12 outflow , as this source is also detected in the mips 24 image . the distance from bklt j162624 - 241616 to ego 04 is 47 and the distance to ego 12 is 18 . outflow ego 33 ( figs . 5 and 6 ) is a diffuse nebula and is visible in irac channels 2 ( 4.5 ) , 3 ( 5.8 ) , and 4 ( 8.0 ) . mid - infrared counterparts of the known near - infrared outflow in this region , three h@xmath9 emission features in @xcite , are also detected in the irac images . we designated these mid - infrared counterparts as ego 34a - c ( see table [ tab3 ] ) . ego 33 appears green and yellow in the three - color image ( fig . [ fig6 ] ) , which indicates that ego 33 may have some emission from dust or pah . there are two ysos in this region , mms126 ( class 0 in @xcite ) and bklt j162816 - 243657 ( class ii in @xcite ) . bklt j162816 - 243657 is associated with 1.3 mm and mid - infrared emission @xcite . these ysos are visible in all irac four channels and in the mips 24 image . we suggest bklt j162816 - 243657 as the driving source of ego 33 on the basis that mms126 has close association with the known outflow detected in near - infrared and in co ( 3 - 2 ) emission @xcite . ego 33 is unlikely to be associated with mms 126 because ego 33 is not coincident with the axis of the known co outflow . the distance from bklt j162816 - 243657 to ego 33 is about 19 . both ego 35 and ego 36 are complexes of emission filaments ( figs . 7 and 8) . their components are list in table [ tab1 ] . egos 35 and 36 are visible in irac channels 1 - 3 and are brightest in channel 2 ( 4.5 ) . the nearby ysos are shown in fig . bklt j162908 - 241549 ( class ii in @xcite ) is the nearest yso to egos 35 and 36 . however , the driving source of egos 35 and 36 should be located in the direction of the line connecting egos 35 and 36 if there is some association between egos 35 and 36 . other ysos in this region are located to the northeast or southwest of egos 35 and 36 , all roughly aligned with egos 35 and 36 . the morphology of ego 35 and ego 36 is bow - shock - like , with wings extending to the southwest , which suggests that the outflow source should be located to the southwest . hh 677 is about 30 to the northeast of sr 10 and it may be driven by this source @xcite . we can see that egos 35 and 36 and hh 677 are roughly located on a line . therefore , it is possible that they are physically associated . egos 37 - 40 are located in the l1709 yso aggregate @xcite ( figs . 10 - 11 ) . ego 37 is a faint knot and ego 38 is a faint bow structure and they may have some connection . they are visible in irac channels 1 and 2 . a low - luminosity class i source in @xcite and classified as a flat spectrum source in @xcite , is located in the northeastern prolongation of the egos 37 - 38 outflow . the distance from gwayl 4 to ego 37 and ego 38 is about 36 and 26 , respectively . on the basis of the alignment we suggest gwayl 4 as the exciting source of egos 37 and 38 . egos 39 and 40 are located to the southeast and east of yso l1709 - 3 , a flat - spectrum source in @xcite . ego 39 and ego 40 are visible in all irac four channels and ego 39 is also visible in the mips 24 image . the l1709 yso aggregate is a small cluster of six ysos and it is difficult at present to identify which ysos in this cluster as the exciting sources of ego 39 and ego 40 . ego 41 consists of a series of knots , ego 41a - i , and ego 42 is a relatively isolated patch ( fig . [ fig12 ] ) . knots ego 41a - i constitute a bowlike structure facing to northeast . the bowlike structure is visible in all irac four channels while ego 42 is only visible in irac channel 2 . the symmetry axis of the bowlike structure suggests that the driving source for ego 41 is located in the ne - sw direction . sstc2d j163145.81 - 243909.0 , a class ii source @xcite , is located about 8 to the southwest of ego 41 ( see fig . [ fig2 ] ) . this source is a possible driving source of ego 41 . ego 41 is located about 52 to the southwest of the class 0 object iras 16293 - 2422 @xcite which drives two molecular outflows , one in the east - west direction and another one in the ne - sw direction @xcite . the iras 16293 - 2422 e - w molecular outflow also has its spectacular demonstration in the mid - infrared which is designated as ego 43 and , together with egos 41 and 42 , is shown in fig . [ fig13 ] . the scale of the ego 43 outflow measures about 11 . it consists of four main portions , ego 43 a - d . the details of the inner portions a - b and the outer portions c - d of this outflow are shown in figs . [ fig14 ] and [ fig15 ] , respectively . iras 16293 - 2422 is a binary with components a and b @xcite and their positions are marked in fig . [ fig14 ] with pentagrams . iras 16293 - 2422 is visible in the mips 24 image and invisible in the irac images . in fig . [ fig14 ] , the position of the submillimeter source iras 16293e @xcite , 90 to the southeast of iras 16293 - 2422 , is also marked . from fig . [ fig13 ] we can see that the symmetry axis of the ego 41 bowlike structure goes through ego 42 and iras 16293 - 2422 . this axis coincides well with the axis of the iras 16293 - 2422 ne - sw molecular outflow ( see fig . 12 in @xcite ) , suggesting that egos 41 and 42 may have physical association with this molecular outflow which is attributed to the a component of iras 16293 - 2422 @xcite . we note that the ego 41 bow shock points to source iras 16293 - 2422 . this morphology usually suggests that the driving source is located on the direction opposite to iras 16293 - 2422 . however , bow shocks are not always in a configuration of facing away from their driving sources . for example , the hh 311 bow shock @xcite and the southeastern bow shock of the v380 ori - ne flow @xcite are in a configuration facing toward their exciting sources . this configuration can be understood if the flow impacts a dense , ambient clump , around which shocked gas is streaming @xcite . so it is also possible that ego 41 is associated with iras 16293 - 2422 . ego 44 is a bow shock in the irac channel 2 minus channel 1 image ( fig . [ fig16 ] ) . it is visible in all the irac channels and fig . [ fig17 ] presents the three - color image of the region . ego 44 is located at the east outskirts of the @xmath0 ophiuchi cloud ( fig . [ fig2 ] ) . the nearest known young stellar object to ego 44 is iras 16367 - 2356 ( class ii in @xcite ) which is about 15 to the west of ego 44 . as the ego 44 bow shock faces to the east , iras 16367 - 2356 is likely the exciting source of ego 44 . extensive surveys of outflows toward the @xmath0 ophiuchi cloud , including hh objects in the optical and h@xmath2 2.12 emission in the near - infrared , have been performed . in total , 46 hh objects , including components of hh objects , and 119 h@xmath9 emission features have be detected in this region @xcite . we identified in the irac images 31 egos that correspond to the known outflows , among which seven egos correspond to known hh objects and 30 egos to h@xmath9 near - infrared emission . figs . [ fig18 ] and [ fig19 ] show the region of ego 01 which has been identified at 2.12 by @xcite . ego 01 are three diffuse nebulae in the 2.12 image while it consists of 3 faint knots in the 3.6 and 4.5 images . two ysos , gss 23 and sr 4 , are located in the nearby of ego 01 . gss 23 is classified as a weak - line t tauri star @xcite . @xcite and @xcite detected narrow h@xmath9 emission surrounding gss 23 . sr 4 is an emission - line star @xcite and @xcite detected an hh object , hh 312 , to the southeast of sr 4 . @xcite suggested that there are some connection between the h@xmath9 emission and the outflow driving by vla1623 - 243 ( see figs . [ fig24]- [ fig26 ] and discussions on egos in that region ) on the basis of their locations and morphology . 5 - 6 have shown the region of ego 34 which are counterparts of the outflow [ kgs2004 ] f05 - 04 @xcite . @xcite and @xcite observed three knots which correspond to ego 34a - c in their near - infrared observations . however the _ spitzer _ images show much more details of the outflow than the near - infrared images . in figs . [ fig5 ] and [ fig6 ] we can see extended diffuse emission to the northeast of ego 34c and to the southwest of ego 34a . the central source of the outflow , mms126 , which is identified as a low - mass class 0 object in the millimetre continuum observations by @xcite ( classified as class 0 in @xcite ) is visible in all irac bands and also in the mips 24 image . @xcite has observed a molecular co outflow from mms126 which is orientated in the ne - sw direction . [ fig20 ] and [ fig21 ] present the images of egos 02 , 06 , 08 , 15 , and 16 that corresponds to near - infrared flows [ kgs2004 ] f10 - 01 and [ kgs2004 ] 10 - 02 @xcite . @xcite detected nine h@xmath9 emission components of [ kgs2004 ] f10 - 01 . however , only four knots , egos 02 , 08 , 15 , and 16 , can be detected in the irac images . these four egos are located on the two sides of bklt j162619 - 242820 , a class ii object @xcite . @xcite proposed bklt j162619 - 242820 as the driving source of this flow . ego 06 has two components , ego 06a and 06b , corresponding to [ kgs2004 ] f10 - 02 . the driving source of ego 06 is unclear . the irac images of egos 03 , 05 , and 10 are presented in figs . [ fig22 ] and [ fig23 ] . ego 03 is the counterpart of [ gswc2003 ] 8a and ego 05 is the counterpart of [ gswc2003 ] 8d . ego 10 corresponds to [ gswc2003 ] 9 @xcite . @xcite identified [ gswc2003 ] 8a as three distinct components . in figs . [ fig22 ] and [ fig23 ] ego 03 is a diffuse patch . gss 26 , a class ii source @xcite , is the nearest yso to ego 03 and ego 05 and may be the driving source for these egos . ego 10 is a diffuse structure within which we can identify several relatively bright knots . the components of ego 10 constitute a faint bow - like structure with its symmetry axis pointing to gss 26 , suggesting that ego 10 is also driven by this source . [ fig24 ] shows the locations of egos 07 , 09 , 11 , 13 - 14 , and 17 - 18 . their previous identifications in optical and near - infrared can be found in table [ tab3 ] . @xcite observed this region and proposed that most of the emission features in this region are driven by vla1623 - 243 , a class 0 source in l1688 @xcite which is marked with a pentagram in fig . [ fig24 ] . vla1623 - 243 has been identified as the driving source of a large - scale molecular outflow which lies in the direction of nw - se @xcite . @xcite presented maps of the large - scale @xmath10co outflow and h@xmath2 near - infrared emission images of this region . [ fig25 ] shows the northwest part of the vla1623 - 243 region . egos 07 , 11 , and 13 - 14 distribute roughly as a chain in the nw - se direction . @xcite observed this region in near - infrared and obtained proper motion measurements for the knots in the vla1623 region . ego 11 corresponds to hh 313b , which is likely associated with gss 30 , a class i binary system candidate in @xcite , as the position angle of proper motion measured for hh 313b is about 243@xcite . ego 09 is a diffuse nebula on the bottom right of fig . [ fig25 ] . @xcite detected this feature in the near - infrared and suggested that it is driven by a source to the north . [ fig26 ] shows the vla1623 - 243 southeast region . we can see a chain of knots in the ego 17 and ego 18 flows . ego 17 and ego 18 correspond to near - infrared flows [ gswc2003 ] 20 and 21 , respectively @xcite . the bright mid - infrared emission knots are visible at the irac 3.6 , 4.5 , and 5.8 images while the faint ones are only visible in the 3.6 and 4.5 images . the irac images of egos 19 , 21 , and 25 are shown in figs . [ fig27 ] and [ fig28 ] . ego 19 is a bow - like structure which corresponds to [ gswc2003 ] 24a in the near - infrared @xcite and an hh object candidate in the optical @xcite . @xcite suggested elias 2 - 26 ( class ii in @xcite ) as the driving source of [ gswc2003 ] 24a . the distance from elias 2 - 26 to ego 19 is about 22 . ego 21a is a faint patch and egos 21b1 and 21b2 are elongated knots . egos 21b1 and 21b2 correspond to hh 79a1 and a2 in the optical image @xcite . the relationship between ego 21a and ego 21b is uncertain and they could be excited by different exciting sources @xcite . ego 25 consists of two bright knots on the top left of figs . [ fig27 ] and [ fig28 ] . they correspond to [ gswc2003 ] 25a and 25b in the near - infrared images @xcite and hh 711 in the optical @xcite . @xcite suggested three possible driving sources for this outflow , including bklt j162658 - 241836 , vssg 24 , and sr 21 . in contrast , @xcite proposed that vssg 3 is the exciting source of hh 711 . [ fig29 ] and [ fig30 ] show the region of egos 20 and 22 - 23 . ego 20 is a faint knot on the bottom right of figs . [ fig29 ] and [ fig30 ] . ego 20 corresponds to a known hh object , hh 314 @xcite . ego 22 is a diffuse knot which corresponds to [ kgs2004 ] f04 - 01 in the near - infrared and hh 673 @xcite in the optical . ego 23 has a bow shock shape and coincides to [ kgs2004 ] f04 - 01b . there are three ysos in this region , bklt j162641 - 244015 ( class ii in @xcite ) , bbrcg10 ( class ii in @xcite ) , and bbrcg19 ( class i , @xcite ) . @xcite proposed bklt j162641 - 244015 as the driving source of hh 314 . [ fig31 ] and [ fig32 ] show the region of ego 24 , which consists of components egos 24a - d . these components correspond to four emission knots in the near - infrared flow [ kgs2004 ] f09 - 01 . in this region @xcite detected seven h@xmath2 emission knots , f09 - 01a - g . ego 24a has a bow shock shape and ego 24d is a bright knot while ego 24b - c are faint diffuse nebulae . ego 24a and ego 24c may belong to the same bow shock structure . @xcite suggested that bbrcg 24 ( class ii in @xcite ) is the driving source of the emission feature corresponding to ego 24d and that wl 16 , a early type embedded star @xcite located about 3 south of ego 24a , might be responsible for the emission features corresponding to ego 24a - c . @xcite suggested another possibility that f09 - 01a - g belong to the same flow . [ fig33 ] and [ fig34 ] show the images of ego 26 that coincides to [ kgs2004 ] f03 - 02 @xcite in the near - infrared . ego 26 also corresponds to an hh object , hh 224s @xcite . bklt j162722 - 244807 , a class ii source in @xcite , is the nearest protostar to this ego . however , considering the association between [ kgs2004 ] f03 - 02 and other flows in this region , neither @xcite nor @xcite suggest this object as the exciting source of the emission feature corresponding to ego 26 . @xcite detected several [ sii ] emission features named hh 224n , hh 224nw1 and hh 224 nw2 to the northwest of hh 224s . they believed that hh 224s belongs to a large flow driven by an embedded source located to the northwest of ego 26 . @xcite suggested gy 193 , a yso located in the northwest about 7 away from hh 224s , as the likely exciting source . @xcite detected the h@xmath9 emission counterpart of hh 224s , but they did not detect the counterparts of hh 224n and hh 224nw . therefore , they suggested em*sr 24 , which is about 8 to the northwest of [ kgs2004 ] f03 - 02 , as the exciting source based on the fact that [ kgs2004 ] f03 - 02 , [ kgs2004 ] f03 - 01 , and em*sr 24 are aligned along the northwest - southeast direction . in the irac images , only the counterpart of hh 224s , i.e. , ego 26 , is detected . as ego 26a and ego 26b form a wide - cavity lobe which faces away from bklt j162722 - 244807 , we propose bklt j162722 - 244807 to be the exciting source of ego 26 . [ fig35 ] and [ fig36 ] show the region of egos 27 - 29 , and ego 31 is shown in figs . [ fig37 ] and [ fig38 ] . egos 27 - 29 all are faint knots . @xcite proposed ylw 15 ( class i , @xcite ) to be the likely driving source of the near - infrared emission feature corresponding to ego 29 . there are abundant ysos in this region and the nearby ysos are marked in figs . [ fig35 ] and [ fig37 ] . figs . [ fig39 ] and [ fig40 ] show the region of ego 30 and ego 32 . @xcite detected eight h@xmath2 emission knots in this region , f08 - 01a - h . some of the non - detected near - infrared emission knots are located at the position of the stray light marked in fig . [ fig39 ] or near the bright star ylw52 . the non - detection of these knots may be due to these effects . ego 30 , corresponding to [ kgs2004 ] f08 - 01a , is a bright arc - like structure . ego 32a - b are two faint knots that coincide to [ kgs2004 ] f08 - 01 g and h. @xcite proposed that ego 30 arises from the interaction of two different jets emanating from two class i protostars , ylw52 ( see fig . [ fig39 ] ) and ylw15 , which is about 10 to the southwest of ego 30 . the distribution of mid - infrared outflows in the @xmath0 ophiuchi molecular cloud , together with that of hh objects @xcite and molecular co outflows @xcite , is shown in fig . [ fig41 ] . the grey scale image in fig . [ fig41 ] is the irac 4.5 emission and the contours give @xmath3co 1 - 0 emission in the region . it can be seen that mid - infrared outflows , hh objects , and molecular co outflows roughly have the same spatial distribution and they are in general concentrated in regions with high gas density , particularly in the region of the l1688 dense core . however , difference in the distribution of mid - infrared outflows , hh objects , and molecular co outflows are apparent in fig . [ fig41 ] . first , a substantial number of hh objects , 17 out of 46 that are detected in the @xmath0 ophiuchi molecular cloud @xcite , are located in regions of @xmath3co 1 - 0 emission less than 2k , while none of mid - infrared outflows and molecular co outflows are detected in these regions of relatively low gas density . this result may indicate that the excitation of mid - infrared and co molecular outflows require more dense surrounding medium than hh objects . secondly , we can see that in the core of the l1688 cloud mid - infrared outflows are distributed in the northern and southern parts of the core while there are several co molecular outflows detected in the middle of the core . using outflows as a tracer of current star formation , we can see from fig . [ fig41 ] that the core of the l1688 cloud is much more active in current star formation than the core of the l1689 cloud . the distribution of yso candidates in the @xmath0 ophiuchi molecular cloud is shown in fig . [ fig42 ] . ysos in this region are mainly concentrated in the l1688 cloud core . we note that abundant ysos are located in the region to the northwest of the l1688 cloud core though there is little dense gas in this region . in fact this part of the @xmath0 ophiuchi molecular cloud hosts four class i and two flat - spectrum sources . however , no mid - infrared outflows are detected in this region ( see fig . [ fig42 ] ) . the non - detection of mid - infrared outflows in this region may be due to that little dense gas remains in this region . several extensive surveys have been made toward the @xmath0 ophiuchi molecular cloud for millimeter sources ( mmss ) and sub - millimeter sources ( smms ) @xcite . the mmss and smms detected in these surveys are shown in fig . [ fig43 ] . it can be seen that the detected mmss and smms are exclusively distributed in the l1688 cloud core and the l1689 cloud core . the l1689 cloud core , hosting about 33 mmss and smms , is comparable to the l1688 cloud core , which hosts about 31 mmss and smms . however , we have seen that the l1688 cloud core hosts much more outflows and ysos than the l1689 cloud core , which may mean that the mmss and smms in the l1689 cloud core are in general in earlier evolutionary stages than those in the l1688 cloud cores and that they are mainly prestellar cores . the distributions of ouflows , ysos , and mmss and smms suggest a global star formation history in the @xmath0 ophiuchi molecular cloud which may be that the star formation in this region took place first in the northwestern part of the @xmath0 ophiuchi molecular cloud and most dense gas in this region has been dispersed . the l1688 cloud core is now undergoing active star formation while most of the dense cores in the l1689 cloud are still at the prestellar stages . this result is consistent with the suggestion that star formation in ophiuchus is triggered by ionisation fronts and winds from the upper scorpius ob association which is located to the west of the @xmath0 ophiuchi cloud @xcite . in summary , we detected 13 new mid - infrared outflows in the @xmath0 ophiuchi molecular cloud that have not been detected previously at other wavelengths . in addition , 31 mid - infrared outflows which correspond to previously detected hh objects or near - infrared emission are detected as well . seven of these mid - infrared outflow features correspond to previously detected hh objects and 30 to near - infrared emission . most of the detected mid - infrared outflows are concentrated in the dense core of the l1688 cloud , with only twelve of which locating outside of this core . in combination with the distribution of ysos and mmss and smms in the @xmath0 ophiuchi molecular cloud , the distribution of mid - infrared outflows hints a propagation of star formation in the @xmath0 ophiuchi molecular complex in the direction from northwest to southeast . we wish to thank s. b. zhang for idl plotting code which has been uploaded to the web site . this work is based on observations made with the _ spizer space telescope _ , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa . this work was based on data taken by the c2d _ spizer space telescope _ legacy programs . this work has made use of the simbad database operated at cds , strasbourg , france . we acknowledge the support by nsfc grants 10733030 and 10621303 . allen , l. e. , myers , p. c. , di francesco , j. , mathieu , r. , chen , h. , & young , e. 2002 , , 56 , 993 alves de oliveira , c. , & casali , m. 2008 , , 485 , 155 andr , p. , martn - pintado , j. , despois , d. , & montmerle , t. 1993 , , 236 , 180 andr , p. , & montmerle , t. 1994 , , 420 , 837 andr , p. , ward - thompson , d. , & barsony , m. 1993 , , 406 , 122 andrews , s. m. , & williams , j. p. 2007 , , 671 , 1800 arce , h. g. , et al . 2007 , protostars and planets v , ed . b. reipurth , d. jewitt , & k. keil , ( tucson , az : univ . arizona press ) , 245 bally , j. , reipurth , b. , & davis , c. j. 2007 , protostars and planets v , ed . b. reipurth , d. jewitt , & k. keil ( tucson , az : univ . arizona press ) , 215 barsony , m. , burton , m. g. , russell , a. p. g. , carlstrom , j. e. , & garden , r. 1989 , , 346 , l93 barsony , m. , kenyon , s. j. , lada , e. a. , & teuben , p. j. 1997 , , 112 , 109 bary , j. s. , weintraub , d. a. , & kastner , j. h. 2002 , , 576 , l73 bitner , m. a. , et al . 2008,,688,1326 bodenheime , p. 1995 , , 33 , 199 bontemps , s. 1996 , ph.d . thesis , univ . paris , xi bontemps , s. , et al . 2001 , , 372 , 173b bouvier , j. , & appenzeller , i. 1992 , , 92 , 481 bussmann , r. s. , wong , t. w. , hedden , a. s. , kulesa , c. a. , & walker , c. k. 2007 , , 657 , l33 caratti o garatti , a. , giannini , t. , nisini , b. , & lorenzetti , d. 2006 , , 449 , 1077 castets , a. , ceccarelli , c. , loinard , l. , caux , e. , & lefloch , b. 2001 , , 375 , 40 chen , x. p. , launhardt , r. , & henning , th . 2007 , , 475 , 277 chen , h. , myers , p. c. , ladd , e. f. , & wood , d. o. s. 1995 , , 445 , 377 cyganowski , c. j. , et al . 2008 , , 136 , 2391 davis , c. j. , dent , w. r. f. , matthews , h. e. , coulson , i. m. , & mccaughrean , m. j. 2000 , , 318 , 952 dent , w. r. f. , matthews , h. e. , & walther , d. m. 1995 , , 277 , 193 dunham , m. m. , crapsi , a. , evans , n. j. , ii , bourke , t. l. , huard , t. l. , myers , p. c. , & kauffmann , j. 2008 , , 179 , 249 eislffel , j. , mundt , r. , ray , t. p. , & rodrguez , l. f. 2000 , in protostars and planets iv , ed . v. mannings , a. p. boss , & s. s. russell ( tucson : univ . arizona press ) , 815 elias , j. h. 1978 , , 224 , 453 evans , n. j. , ii , et al . 2003 , , 115 , 965 evans , n. j. , ii , et al . 2009 , , 181 , 321 gmez , m. , stark , d. p. , whitney , b. a. , & churchwell , e. 2003 , , 126 , 863 gmez , m. , whitney , b. a. , & wood , k. 1998 , , 115 , 2018 grasdalen , g. l. , strom , k. m. , & strom , s. e. 1973 , , 184 , l53 greene , t. p. , wilking , b. a. , andr , p. , young , e. t. , & lada , c. j. 1994 , , 434 , 614 greene , t. p. , & young , e. t. 1992 , , 395 , 516 grosso , n. , alves , j. , neuhuser , r. , & montmerle , t. 2001 , , 380 , l1 hirano , n. , mikami , h. , umemoto , t. , yamamoto , s. , & taniguchi , y. 2001 , , 547 , 899 imanishi , k. , koyama , k. , & tsuboi , y. 2001 , , 557 , 747 johnstone , d. , francesco , j. d. , & kirk , h. 2004 , , 611 , l45 johnstone , d. , wilson , c. d. , moriarty - schieven , g. , joncas , g. , smith , g. , gregersen , e. , fich , m. 2000 , , 545 , 327 jrgensen , j. k. , et al . 2006 , , 645 , 1246 khanzadyan , t. , gredel , r. , smith , m. d. , & stanke , t. 2004 , , 426 , 171 lis , d. c. , gerin , m. , phillips , t. g. , & motte , f. 2002 , , 569 , 322 loren , r. b. 1989a , , 338 , 902 loren , r. b. 1989b , , 338 , 925 loren , r. b. , & wootten , a. 1986 , , 306 , 142 lucas , p. w. , et al . 2008 , , 391 , 136 lynds , b. t. 1962 , , 7 , 1l mern , b. , et al . 2008 , , 177 , 551 montmerle , t. , koch - miramond , l. , falgarone , e. , & grindlay , j. e. 1983 , , 269 , 182 motte , f. , andr , p. , & neri , r. 1998 , , 336 , 150 mundy , l. g. , wootten , a. , wilking , b. a. , blake , g. a. , & sargent , a. i. 1992 , , 385 , 306 noriega - crespo , a. , et al . 2004 , , 154 , 352 nutter , d. , ward - thompson , d. , & andr , p. 2006 , , 368 , 1833n ozawa , h. , grosso , n. , & montmerle , t. 2005 , , 429 , 963 padgett , d. l. , et al . 2008 , , 672,1013 phelps , r. l. , & barsony , m. 2004 , , 127 , 420 preibisch , th . , guenther , e. , zinnecker , h. , sterzik , m. , frink , s. , & rser , s. 1998 , , 333 , 619 reach , w. t. , et al . 2006 , , 131 , 1479 reipurth , b. , & bally , j. 2001 , , 39 , 403 reipurth , b. , bally , j. , & devine , d. 1997 , , 114 , 2708r ridge , n. a. , et al . 2006 , , 131 , 2921r rosenthal , d. , bertoldi , f. , & drapatz , s. 2000 , , 356 , 705 shang , h. , li , z .- y . , & hirano , n. 2007 , protostars and planets v , ed . b. reipurth , d. jewitt , & k. keil ( tucson , az : univ . arizona press ) , 261 smith , m. d. , gredel , r. , khanzadyan , t. , & stanke , t. 2005 , , 76 , 247 smith , h. a. , hora , j. l. , marengo , m. , & pipher , j. l. 2006 , , 645 , 1264 smith , m. d. & rosen , a. 2005 , , 357 , 1370 stanke , t. , smith , m. d. , gredel , r. , & khanzadyan , t. 2006 , , 447 , 609 stark , r. , et al . 2004 , , 608 , 341 struve , o. , & rudkjobing , m. 1949 , , 109 , 92 teixeira , p. s. , mccoey , c. , fich , m. , & lada , c. j. 2008 , , 384 , 71 visser , a. e. , richer , j. s. , & chandler , c. j. 2002,,124,2756 vrba , f. j. , strom , k. m. , strom , s. e. , & grasdalen , g. l. 1975 , , 197 , 77 wilking , b. a. , bontemps , s. , schuler , r. e. , greene , t. p. , & andr , p. 2001 , , 551 , 357 wilking , b. a. , gagn , m. , & allen , l. e. 2008 , astronomical society of the pacific , handbook of star forming regions vol . ii wilking , b. a. , lada , c. j. , & young , e. t. 1989 , , 340 , 823 wilking , b. a. , meyer , m. r. , robinson , j. g. , & greene , t. p. 2005 , , 130 , 1733 wilking , b. a. , schwartz , r. d. , & blackwell , j. h. 1987 , , 94 , 106 wilking , b. a. , schwartz , r. d. , fanetti , t. m. , & friel , e. d. 1997 , , 109 , 549 wilson , c. d. , et al . 1999 , , 513 , l139 wootten , a. 1989 , , 337 , 858 wootten , a. , & loren , r. b. 1987 , , 317 , 220 wu , j. , wang , m. , yang , j. , deng , l. , & chen , j. 2002 , , 123 , 1986 wu , y. , wei , y. , zhao , m. , shi , y. , yu , w. , qin , s. , & huang , m. 2004 , , 426 , 503 young , e. t , lada , c. j. , & wilking , b. a. 1986 , , 304 , l45 young , k. e. , et al . 2006 , , 644 , 326 de zeeuw , p. t. , hoogerwerf , r. , & de bruijne , j. h. j. 1999 , , 117 , 354 lccc ego 04 & 16:26:10.1 & -24:12:40 & 3,4 + ego 12 & 16:26:18.2 & -24:15:07 & 3,4 + ego 33 & 16:28:16.2 & -24:35:03 & 5,6 + ego 35a & 16:28:58.9 & -24:17:54 & 7,8,9 + ego 35b & 16:28:59.3 & -24:17:51 & 7,8,9 + ego 35c & 16:28:59.7 & -24:17:46 & 7,8,9 + ego 35d & 16:28:59.9 & -24:18:18 & 7,8,9 + ego 35e & 16:29:00.7 & -24:18:02 & 7,8,9 + ego 35f & 16:29:00.7 & -24:17:49 & 7,8,9 + ego 35 g & 16:29:01.9 & -24:18:24 & 7,8,9 + ego 35h & 16:29:02.3 & -24:18:28 & 7,8,9 + ego 35i & 16:29:02.5 & -24:17:56 & 7,8,9 + ego 35j & 16:29:03.1 & -24:18:23 & 7,8,9 + ego 36a & 16:29:13.4 & -24:16:59 & 7,8,9 + ego 36b & 16:29:13.5 & -24:16:25 & 7,8,9 + ego 36c & 16:29:14.4 & -24:16:39 & 7,8,9 + ego 36d & 16:29:14.4 & -24:17:17 & 7,8,9 + ego 36e & 16:29:14.5 & -24:17:03 & 7,8,9 + ego 36f & 16:29:15.2 & -24:16:32 & 7,8,9 + ego 36 g & 16:29:15.4 & -24:16:11 & 7,8,9 + ego 36h & 16:29:17.2 & -24:16:07 & 7,8,9 + ego 37 & 16:31:22.0 & -24:03:17 & 10,11 + ego 38 & 16:31:25.8 & -24:02:42 & 10,11 + ego 39 & 16:31:36.0 & -24:03:51 & 10,11 + ego 40 & 16:31:37.2 & -24:03:30 & 10,11 + ego 41a & 16:32:01.0 & -24:31:59 & 12,13 + ego 41b & 16:32:01.2 & -24:31:54 & 12,13 + ego 41c & 16:32:02.2 & -24:31:42 & 12,13 + ego 41d & 16:32:03.2 & -24:32:27 & 12,13 + ego 41e & 16:32:03.2 & -24:32:19 & 12,13 + ego 41f & 16:32:03.3 & -24:31:39 & 12,13 + ego 41 g & 16:32:04.2 & -24:31:55 & 12,13 + ego 41h & 16:32:04.2 & -24:31:35 & 12,13 + ego 41i & 16:32:04.7 & -24:31:40 & 12,13 + ego 42 & 16:32:11.8 & -24:30:33 & 12,13 + ego 43a1 & 16:32:19.3 & -24:28:31 & 13,15 + ego 43a2 & 16:32:19.5 & -24:28:11 & 13,15 + ego 43a3 & 16:32:21.1 & -24:28:30 & 13,15 + ego 43a4 & 16:32:21.5 & -24:28:38 & 13,15 + ego 43a5 & 16:32:22.3 & -24:28:32 & 13,15 + ego 43a6 & 16:32:23.6 & -24:28:35 & 13,15 + ego 43b1 & 16:32:27.0 & -24:28:19 & 13,15 + ego 43b2 & 16:32:27.8 & -24:28:36 & 13,15 + ego 43b3 & 16:32:28.6 & -24:28:31 & 13,15 + ego 43b4 & 16:32:29.2 & -24:28:16 & 13,15 + ego 43b5 & 16:32:30.1 & -24:28:28 & 13,15 + ego 43b6 & 16:32:33.8 & -24:29:04 & 13,15 + ego 43b7 & 16:32:34.7 & -24:28:33 & 13,15 + ego 43b8 & 16:32:35.4 & -24:29:04 & 13,15 + ego 43b9 & 16:32:36.1 & -24:28:28 & 13,15 + ego 43b10 & 16:32:36.5 & -24:28:32 & 13,15 + ego 43b11 & 16:32:36.6 & -24:28:56 & 13,15 + ego 43c1 & 16:32:44.7 & -24:29:28 & 13,14 + ego 43c2 & 16:32:45.7 & -24:29:09 & 13,14 + ego 43c3 & 16:32:47.1 & -24:29:09 & 13,14 + ego 43c4 & 16:32:47.2 & -24:28:49 & 13,14 + ego 43c5 & 16:32:48.9 & -24:29:13 & 13,14 + ego 43d1 & 16:32:55.9 & -24:28:13 & 13,14 + ego 43d2 & 16:32:57.5 & -24:28:24 & 13,14 + ego 43d3 & 16:32:58.4 & -24:28:39 & 13,14 + ego 43d4 & 16:33:01.3 & -24:28:54 & 13,14 + ego 43d5 & 16:33:02.9 & -24:28:53 & 13,14 + ego 43d6 & 16:33:04.2 & -24:28:40 & 13,14 + ego 43d7 & 16:33:07.2 & -24:28:27 & 13,14 + ego 44 & 16:40:50.6 & -24:01:21 & 16,17 + llcr ego 04 & bklt j162624 - 241616 & ii & 4.74 + ego 12 & bklt j162624 - 241616 & ii & 1.78 + ego 33 & bklt j162816 - 243657 & ii & 1.88 + ego 35 & sr 10 , sr 20w , bklt j162908 - 241549 , sr 20 , bklt j162923 - 241359 & ii , ii , ii , iii , iii & 2.84 , 16.94 + ego 36 & sr 10 , sr 20w , bklt j162908 - 241549 , sr 20 , bklt j162923 - 241359 & ii , ii , ii , iii , iii & 1.70 , 20.58 + ego 37 & gwayl 4 & flat & 3.6 + ego 38 & gwayl 4 & flat & 2.55 + ego 39 & l1709 3 & flat & 0.60 + ego 40 & l1709 3 & flat & 0.70 + ego 41 & sstc2d j163145.81 - 243909.0 , iras 16293 - 2422 & ii , 0 & 8.00 , 5.50 + ego 42 & sstc2d j163145.81 - 243909.0 , iras 16293 - 2422 & ii , 0 & 10.36 , 3.00 + ego 43 & iras 16293 - 2422 & 0 & 11.35 + ego 44 & iras 16367 - 2356 & ii & 14.90 + . lcclr ego 01a & 16:25:49.8 & -24:22:12 & a & 18,19 + ego 01b & 16:25:50.4 & -24:22:22 & a & 18,19 + ego 01c & 16:25:51.3 & -24:22:16 & a & 18,19 + ego 02 & 16:26:03.1 & -24:29:33 & [ kgs2004 ] f10 - 01b & 20,21 + ego 03 & 16:26:06.5 & -24:21:35 & [ gswc2003 ] 8a & 22,23 + ego 05 & 16:26:11.1 & -24:20:33 & [ gswc2003 ] 8d & 22,23 + ego 06a & 16:26:12.7 & -24:31:13 & [ kgs2004 ] f10 - 02a & 20,21 + ego 06b & 16:26:13.5 & -24:31:10 & [ kgs2004 ] f10 - 02b & 20,21 + ego 07a & 16:26:15.9 & -24:22:41 & [ gswc2003 ] 11a & 24,25 + ego 07b & 16:26:16.4 & -24:22:45 & [ gswc2003 ] 11b & 24,25 + ego 08 & 16:26:16.6 & -24:28:35 & [ kgs2004 ] f10 - 01e & 20,21 + ego 09 & 16:26:16.6 & -24:25:39 & [ kgs2004 ] f10 - 03 & 24,25 + ego 10 & 16:26:17.3 & -24:18:05 & [ gswc2003 ] 9 & 22,23 + ego 11 & 16:26:17.5 & -24:23:13 & hh 313b & 24,25 + ego 13 & 16:26:19.0 & -24:23:05 & hh 313a , vla 1623a , [ dmw95 ] h5 & 24,25 + ego 14a & 16:26:21.4 & -24:23:31 & [ gswc2003 ] 14f & 24,25 + ego 14b & 16:26:21.8 & -24:23:37 & [ gswc2003 ] 14e & 24,25 + ego 14c & 16:26:21.9 & -24:23:26 & [ gswc2003 ] 14 g & 24,25 + ego 15 & 16:26:25.3 & -24:27:34 & [ kgs2004 ] f10 - 01h & 20,21 + ego 16 & 16:26:28.4 & -24:27:05 & [ kgs2004 ] f10 - 01i & 20,21 + ego 17a & 16:26:31.1 & -24:25:14 & [ gswc2003 ] 20a , [ kgs2004 ] f10 - 04d , [ dmw95 ] h3 & 24,26 + ego 17b & 16:26:31.7 & -24:25:17 & [ gswc2003 ] 20b , [ kgs2004 ] f10 - 04e & 24,26 + ego 17c & 16:26:32.3 & -24:25:19 & [ gswc2003 ] 20c , [ kgs2004 ] f10 - 04f & 24,26 + ego 17d & 16:26:32.9 & -24:25:22 & [ gswc2003 ] 20d , [ kgs2004 ] f10 - 04 g , [ dmw95 ] h2 & 24,26 + ego 17e & 16:26:33.7 & -24:25:29 & [ gswc2003 ] 20f , [ kgs2004 ] f10 - 04h , [ dmw95 ] h1 & 24,26 + ego 18a & 16:26:35.7 & -24:25:27 & [ gswc2003 ] 21a , [ kgs2004 ] f10 - 04i & 24,26 + ego 18b & 16:26:36.6 & -24:25:39 & [ gswc2003 ] 21b , [ kgs2004 ] f10 - 04j & 24,26 + ego 18c & 16:26:37.5 & -24:25:40 & [ gswc2003 ] 21c , [ kgs2004 ] f10 - 04k & 24,26 + ego 18d & 16:26:38.3 & -24:25:46 & [ gswc2003 ] 21d & 24,26 + ego 18e & 16:26:39.2 & -24:25:54 & [ kgs2004 ] f10 - 04 m & 24,26 + ego 18f & 16:26:40.2 & -24:26:07 & new & 24,26 + ego 19 & 16:26:38.2 & -24:18:29 & [ gswc2003 ] 24a & 27,28 + ego 20 & 16:26:39.0 & -24:40:53 & hh 314 & 29,30 + ego 21a & 16:26:44.4 & -24:20:07 & hh 79b & 27,28 + ego 21b1 & 16:26:46.4 & -24:20:15 & hh 79a1 & 27,28 + ego 21b2 & 16:26:47.2 & -24:20:12 & hh 79a2 & 27,28 + ego 22 & 16:26:47.5 & -24:38:16 & hh 673 , [ kgs2004 ] f04 - 01a & 29,30 + ego 23 & 16:26:59.0 & -24:37:03 & [ kgs2004 ] f04 - 01b & 29,30 + ego 24a & 16:27:08.1 & -24:33:05 & [ kgs2004 ] f09 - 01a , [ gswc2003 ] 3f & 31,32 + ego 24b & 16:27:08.4 & -24:33:32 & [ kgs2004 ] f09 - 01d , [ gswc2003 ] 3d & 31,32 + ego 24c & 16:27:08.6 & -24:33:17 & [ kgs2004 ] f09 - 01c , & 31,32 + ego 24d & 16:27:09.5 & -24:33:41 & [ kgs2004 ] f09 - 01e , [ gswc2003 ] 3b & 31,32 + ego 25a & 16:27:10.3 & -24:17:45 & [ gswc2003 ] 25a , hh 711 & 27,28 + ego 25b & 16:27:16.4 & -24:43:56 & [ gswc2003 ] 25b & 27,28 + ego 26a & 16:27:22.3 & -24:49:14 & [ kgs2004 ] f03 - 02a & 33,34 + ego 26b & 16:27:22.8 & -24:48:53 & [ kgs2004 ] f03 - 02b , hh 224s & 33,34 + ego 27 & 16:27:28.2 & -24:37:56 & [ kgs2004 ] f04 - 06 & 35,36 + ego 28 & 16:27:39.6 & -24:41:50 & [ kgs2004 ] f05 - 01 & 35,36,37,38 + ego 29a & 16:27:39.8 & -24:38:09 & [ kgs2004 ] f04 - 03e , [ gswc2003 ] 7a & 35,36 + ego 29b & 16:27:40.2 & -24:37:60 & [ kgs2004 ] f04 - 03f , [ gswc2003 ] 7b & 35,36 + ego 30 & 16:27:43.2 & -24:31:51 & [ kgs2004 ] f08 - 01a & 39,40 + ego 31 & 16:27:50.4 & -24:40:08 & [ kgs2004 ] f05 - 03 & 37,38 + ego 32a & 16:28:03.9 & -24:30:54 & [ kgs2004 ] f08 - 01 g & 39,40 + ego 32b & 16:28:05.2 & -24:30:40 & [ kgs2004 ] f08 - 01h & 39,40 + ego 34a & 16:28:17.7 & -24:38:09 & [ kgs2004 ] f05 - 04a & 5,6 + ego 34b & 16:28:19.4 & -24:37:09 & [ kgs2004 ] f05 - 04b & 5,6 + ego 34c & 16:28:23.1 & -24:36:10 & [ kgs2004 ] f05 - 04c & 5,6 +

we present model independent high resolution amplitude analyses of cern data on at 17.2 gev / c for dipion masses 580 - 1080 mev at small momentum transfers on polarized target ( analysis i ) and unpolarized target ( analysis ii ) . the results for- and-wave transversity amplitudes from the analysis i are used in a model independent determination of helicity amplitudes . all three analyses provide similar evidence for mixing in-and-wave amplitudes . in analyses i and ii the-wave amplitudes peak at mass while the-wave amplitudes dip at mass . in both analyses the-wave amplitudes are nearly in phase with longitudinal amplitudes and nearly out of phase with transverse amplitudes indicating resonant phases of also in the-wave for both target nucleon transversities . the-wave single flip helicity amplitudes and the relative phases show the same pattern . while the non - flip helicity amplitudes are small below 960 mev , there is a sudden rise above this mass that may reflect the presence of resonance . we show that the apparent dependence of width on helicity in amplitudes and results from the interference of transversity amplitudes with definite dipion helicities and not from violation of rotational / lorentz symmetry . this suggests that the production mechanism and the mechanism responsible for mixing have independent dynamical origins .

1205.6381

following the discovery in 1961 of @xmath19 meson in @xmath20 reactions @xcite , the measurements of forward - backward asymmetry in @xmath21 suggested the existence of a rho - like resonance in the @xmath1-wave , later referred to as @xmath22 scalar meson @xcite . it was expected that @xmath22 would show up prominently in @xmath23 production where @xmath5 does not contribute . the measurements of this reaction at cern in 1972 found no evidence for a rho - like @xmath22 @xcite . furthermore , in 1973 , pennington and protopopescu used analyticity and unitarity constraints on partial wave amplitudes in @xmath24 scattering ( roy equations ) to show that a narrow @xmath22 resonance can not contribute to @xmath25 scattering @xcite . from these facts it was concluded that @xmath22 does not exist in @xmath0 and in 1974 particle data group dropped this state from its listings . in 1978 lutz and rybicki showed that almost complete amplitude analysis of reactions @xmath26 is possible from measurements in a single experiment on a transversely polarized target @xcite . cern - munich - cracow group measured @xmath27 on transversely polarized proton target in a high statistics experiment at 17.2 gev / c at low momentum transfers @xmath28 ( gev / c)@xmath29 @xcite and at high momentum transfers @xmath30 ( gev / c)@xmath29 @xcite . saclay group measured @xmath31 on transversely polarized deuteron target at 5.98 and 11.85 gev / c at larger momentum transfers @xmath32 ( gev / c)@xmath29 @xcite . measurements of @xmath27 on transversely polarized proton target at 1.78 gev / c at low momentum transfers @xmath28 ( gev / c)@xmath29 were made at itep @xcite . the cern measurements of @xmath27 and @xmath31 on polarized targets reopened the question of the existence of @xmath22 scalar meson . evidence for a narrow @xmath22 was found in amplitude analyses of @xmath27 at 17.2 gev / c @xcite and in @xmath31 at 5.98 and 11.85 @xcite . clear evidence for @xmath22 emerged from later and more precise amplitude analyses of both reactions @xcite . another evidence for @xmath22 comes from the amplitude analysis of the itep data at 1.78 gev / c @xcite . in 2001 , e852 collaboration at bnl reported high statistics measurements @xmath23 at 18.3 gev / c @xcite . the data revealed large differences in the @xmath1-wave intensities in @xmath33 and @xmath34 production . there was no evidence for a rho - like @xmath22 resonance in the @xmath34 @xmath1-wave . the bnl data presented anew the puzzle of the @xmath22 resonance . the cern and bnl data are both high quality data that can not be used to exclude one another . we must accept that they are both correct and that the apparent contradictions between them are telling us something new and important . the puzzle of @xmath22 resonance has become a unique opportunity to learn new physics . the first hints on the solution of the puzzle emerged from the amplitude analysis of cern measurements of @xmath0 in ref . this study extended the dipion mass range of 600 - 900 mev of previous analyses @xcite to 580 - 1080 mev to include the @xmath7 resonance . the @xmath1-wave transversity ampltudes @xmath8 with target nucleon trasversity @xmath35 were fitted using a parametrization with breit - wigner amplitudes for both resonances and complex backgoudnds . the best fit to all moduli of @xmath1-wave transversity amplitudes gives @xmath36 mev and @xmath37 mev . these values are very close to resonance parameters of @xmath5 with @xmath38 mev and @xmath39 mev . both solutions for the @xmath2-wave amplitude with zero dipion helicity @xmath6 showed an unexpected dip at @xmath40 980 mev , the mass of @xmath7 resonance . in 2001 , gale and collaborators investigated @xmath41 mixing in lepton pair production at rhic arising from the violation of lorentz symmetry by vacuum but conserved by the interaction lagrangian , and found large measurable effects @xcite . their work suggested that the dip in @xmath6 arises from @xmath3 mixing in the @xmath2-wave . similar mixing in the @xmath1-wave then allows to identify @xmath22 with @xmath5 which immediately explains why no @xmath22 was ever found in @xmath23 production . the @xmath3 mixing implies an apparent violation of rotational and lorentz symmetry which appeared to be supported by the observation of large differences in @xmath5 widths in the moduli of @xmath2-wave amplitues @xmath16 and @xmath17 with transverse dipion helicities . in ref . @xcite we used the results of the fits to the transversity amplitudes to calculate @xmath1-wave helicity non - flip and flip amplitudes @xmath42 and @xmath43 assuming the residues of @xmath22 were real and positive in both transversity amplitudes and there was no additional relative phase apart from the fits . the results suggested that @xmath22 dominates @xmath42 while it is suppressed in @xmath43 , and therefore that it dominates in @xmath44 and is suppressed in @xmath24 scattering . however , the results for @xmath43 disagreed with established @xmath25 scattering amplitudes . in over 100 later fits with different input values of complex residues of @xmath22 and other fitting parameters we obtained mostly convergent fits . they all had the same @xmath45 and values of resonance parameters of @xmath22 and @xmath7 identical to the values from our original fits . however , the corresponding aplitudes @xmath42 and @xmath43 were widely different with many of them unphysical solutions . we must conclude that the presence of a rho - like resonance in @xmath1-wave transversity amplitudes is not related to @xmath44 or @xmath24 scattering . it stands on its own and lends support to the hypothesis of @xmath3 mixing . in this work we present our results of a high resolution amplitude analysis of cern data on polarized target ( analysis i ) and of a new model independent amplitude analysis of cern data on unpolarized target ( analysis ii ) . we trace the presence of @xmath5 in the @xmath1-wave transversity amplitudes in both analyses to the data component @xmath46 in the equation @xmath47 . the data @xmath46 have a large @xmath5 peak that is not balanced out by @xmath5 peak in the @xmath2-wave amplitude @xmath48 . the relative phases between @xmath1-wave amplitudes and @xmath2-wave amplitudes @xmath9 and @xmath11 are near @xmath49 and @xmath50 , respectively in both analyses . since the @xmath2-wave amplitudes have @xmath5 breit - wigner phase , the @xmath1-wave amplitudes must have nearly the same resonant phase . next we outline a model independent method to determine the @xmath1- and @xmath2-wave helicity amplitudes from the knowledge of transversity amplitudes and present new results on helicity amplitudes @xmath51 and @xmath52 , @xmath53 . the four solutions for the single flip amplitude @xmath13 show a clear @xmath5 peak and the phase of @xmath43 is again nearly equal to the phase of @xmath2-wave amplitude @xmath54 . there is a remarkable similarity and concordance between the analyses i , ii and the helicity amplitudes . finally we show that the large differences in the widths of @xmath16 and @xmath17 amplitudes arise from the interference of amplitudes of opposite dipion helicities @xmath55 , and that the width of @xmath5 resonance does not depend on its helicity in both analyses i and ii . this suggests that the production process conserves rotational / lorentz symmetry . this means that the mechanism for the @xmath56 mixing and the production mechanism have different dynamical origins . while there is no reason to assume that the production mechanism is governed by anything other than the @xmath1-matrix , the @xmath3 mixing must originate in a new spin mixing interaction . while the production mechanism conserves lorentz symmetry , the spin mixing mechanism spontaneously breaks this symmetry in the measured amplitudes that combine both mechanisms . we address these issues in two related papers @xcite . the paper is organized as follows . in section ii . we define the @xmath1-and @xmath2-wave subsystem and the reduced transversity amplitudes used in all our analyses . in sections iii . and iv . we present the analytical solutions for transversity amplitudes with polarized and unpolarized target data which define the analyses i and ii , respectively . evidence for the @xmath3 mixing from analyses i and ii is presented in section v. in section vi . we outline our method of the model independent determination of @xmath1-and @xmath2-wave helicity amplitudes and present evidence for @xmath3 mixing from amplitudes @xmath13 and phases @xmath14 . in section vii . we use analyses i and ii to demonstrate the helicity independence of the @xmath5 width and thus lorentz symmetry of the production mechanism . the paper closes with our conclusions and outlook in section viii . in the appendix we show that the measurements of @xmath57 and @xmath58 determine uniquely the signs of relative phases in the analyses i and ii . the final state density matrix in @xmath0 has a general form @xcite @xmath59 where @xmath60 describe the direction of @xmath61 in the dipion centre - of - mass system and @xmath62 and @xmath63 are polarization vectors of the target and recoil nucleon , respectively . the most feasible experiments are measurements on unpolarized or polarized targets leaving the recoil nucleon polarization vector unobserved . such measurements provide information on the reduced final state density matrix @xmath64 given by @xcite @xmath65 where the polarization components of @xmath64 are expressed in terms of density matrix elements @xmath66 for k = u ( unpolarized target ) and k = y ( transversely polarized taget normal to the scattering plane ) , and @xmath67 for k = x ( transversely polarized target in the scattering plane ) and k = z ( longitudinally polarized target ) . in ( 2.3 ) and ( 2.4 ) @xmath68 and @xmath69 for @xmath70 and the factor @xmath71 for @xmath72 and @xmath73 for @xmath74 . in a given region of dipion mass @xmath75 and momentum transfer @xmath76 only amplitudes with @xmath77 contribute and all sums in ( 2.3 ) and ( 2.4 ) are finite . from the measured intensity @xmath78 the density matrix elements are determined in small @xmath79 bins using maximum likelihood method @xcite . amplitude analyses of measurements on polarized targets are best performed using nucleon transversity amplitudes with definite @xmath76-channel naturality @xmath80 and @xmath81 corresponding to unnatural and natural exchange , respectively @xcite . @xmath82 and @xmath83 are dimeson spin and helicity and @xmath12 is the nucleon transversity for target spin `` up '' or `` down '' relative to the scattering plane . in @xmath21 the amplitudes @xmath84 exchange @xmath85 and @xmath86 quantum numbers in the @xmath76-channel while the amplitudes @xmath81 exchange @xmath87 quantum numbers . general expressions for density matrix elements in terms of transversity amplitudes were tabulated by lutz and rybicki @xcite and are reproduced in the ref . @xcite in our notation . for @xmath1- and @xmath2-wave transversity amplitudes we shall use a simplified notation @xmath88 where the amplitudes @xmath89 are defined as @xmath90 @xmath91 amplitude analysis of the @xmath1- and @xmath2-wave subsystem of the reduced density matrix determines reduced transversity amplitudes defined as follows @xmath92 @xmath93 @xmath94 @xmath95 where @xmath96 is the phase of the amplitude @xmath88 . the reduced transversity amplitudes are related to transversity amplitudes by phase factors @xmath97 for unnatural exchange amplitudes @xmath98 and @xmath99 for natural exchange amplitude @xmath100 . in ( 2.7 ) and ( 2.8 ) @xmath101 is the arbitrary absolute phase and @xmath102 is the relative phase between @xmath1-wave amplitudes of opposite transversity . ccc @xmath103&@xmath104&@xmath105 + @xmath106&@xmath107&@xmath108 + @xmath109&@xmath110&@xmath111 + @xmath112&@xmath113 & @xmath114 + @xmath115&@xmath116 & @xmath117 + @xmath118&@xmath119&@xmath120 + @xmath121&@xmath122&@xmath123 + @xmath124&@xmath125&@xmath126 + @xmath103&@xmath127&@xmath128 + @xmath129&@xmath130&@xmath131 + @xmath132&@xmath133&@xmath134 + @xmath135&@xmath136&@xmath137 + [ table i. ] in the table i. we present @xmath1- and @xmath2-wave density matrix elements expressed in terms of the reduced transversity amplitudes . because of the angular properties of @xmath138 , the elements @xmath139 , @xmath140 and @xmath141 are not independent but appear as two independent combinations in the measured angular distributions ( 2.3 ) @xmath142 at any dipion mass the entire @xmath1- and @xmath2-wave subsystem can be solved analytically for the transversity amplitudes @xmath88 up to an overall phase selected to be @xmath101 - withount any measurements of recoil nucleon polarization . the observables @xmath143 and @xmath144 determine the amplitudes @xmath145 and @xmath146 . the observables @xmath147 and @xmath148 determine the phases of amplitudes @xmath100 and @xmath149 ( appendix ) . the relative phase @xmath150 is determined analytically by the process of conversion of transversity amplitudes into helicity amplitudes ( section vii . ) . the amplitude analysis yields two solutions for @xmath151 and two independent solutions for @xmath152 which give rise to four different @xmath1-and @xmath2-wave final state density matrices @xmath153 . the measured observables @xmath156 and @xmath157 organize themselves into two groups involving amplitudes of opposite transversity . using the expressions in table i. and @xmath158 , the two groups are @xmath159 for reduced transversity amplitudes with transversity @xmath160 , and @xmath161 . the relative phases are defined as in ( 2.6 ) . for dipion masses where @xmath1- and @xmath2-wave dominate , @xmath162 and @xmath163 are traces @xmath164 where @xmath165 is target spin asymmetry . the 6 equations ( 2.11 ) and ( 2.12 ) each involve 7 unknowns for 4 moduli and 3 cosines , and are not solvable . the missing equation in each group is supplied not by the data but by phase relations @xmath166 these conditions lead to non - linear relations between the cosines @xmath167 substituting for the cosines from ( 3.1 ) we get @xmath168 from ( 3.1 ) we have three equations for moduli @xmath169 @xmath170 where @xmath171 and @xmath172\\ \nonumber c & = & { 1 \over { 2}}(a_1+a_2)(a_2+a_3)+a_4 ^ 2+a_6 ^ 2 - 3a_5 ^ 2\\ \nonumber d & = & ( a_1+a_2)a_5 ^ 2-{1 \over { 2}}(a_2+a_3)a_4 ^ 2 - 2a_4a_5a_6\end{aligned}\ ] ] from ( 3.2 ) we obtain equations for moduli @xmath173 similar to ( 3.7 ) and from ( 3.5 ) a cubic equation for @xmath174 with observables @xmath175 replacing @xmath176 in the coefficients ( 3.9 ) . the cubic equations for @xmath48 and @xmath177 are just another form of the phase conditions ( 3.4 ) . the equations can be solved analyticaly . to solve for @xmath178 we write @xmath179 and require that the cubic equation ( 3.8 ) transforms to the form @xmath180 this is accomplished with @xmath181 and @xmath182 next we define quantities @xmath183 there are three categories of solutions of cubic equation ( 3.10 ) @xcite . their numerical study using a monte carlo method shows there are two physical solutions that have an analytical form @xmath184 where @xmath185 , @xmath186 and @xmath187 . it follows from ( 3.7 ) that the two physical solutions for the moduli @xmath188 , @xmath189 and @xmath190 , @xmath191 , have a similar analytical form . similarly , there are two independent solutions for the moduli @xmath192 , @xmath193 with @xmath186 and @xmath194 . for each set of solutions for moduli @xmath195 , @xmath191 we calculate from ( 3.1 ) the cosines @xmath196 with similar equations for @xmath197 , @xmath198 , @xmath199 . from the phase condition @xmath200 we obtain 3 equations for sines @xmath201 the system is solvable provided that sign(@xmath202)=sign(@xmath203 ) . with @xmath204 where @xmath205 , @xmath206 and @xmath207 can be calculated from @xmath208 and @xmath209 , respectively . this procedure allows us to determine the phases @xmath210 , @xmath211 and @xmath212 . with @xmath213 where @xmath214 , we obtain for each set of solutions for the moduli @xmath215 , @xmath193 the phases @xmath216 , @xmath217 , @xmath218 . the amplitude analyses assumed positive root for both sines @xmath219 and @xmath220 . such procedure is uniqe provided that the phases @xmath221 and @xmath222 do not change signs . a change of sign would manifest itself as a double zero in the phases @xmath221 and @xmath222 calculated from the positive roots for the sines . we found no evidence of a double zero in the phases @xmath221 and @xmath222 so calculated , and apart from the sign ambiguity , these phases are uniquely determined . the phases @xmath221 and @xmath222 are closely related to the positivity of the reduced density matrices @xmath223 for pure initial states @xmath224 by a relation @xmath225 . the fact that these phases do not change signs implies that the eigenvalues of the density matrices @xmath226 are all non - zero . for negative roots in ( 3.17 ) and ( 3.18 ) all phases change signs . there are four combinations of the signs @xmath227 and @xmath228 . the phases with opposite signs correspond to complex conjugate reduced transversity amplitudes . for any given solution of moduli @xmath229 there is a four - fold ambiguity in the phases of the reduced transversity amplitudes @xmath230 @xmath231 @xmath232 in the appendix we show that this four - fold sign ambiguity can be resolved into a single set of phases for each set of moduli @xmath233 in the process of determination of the phases of natural exchange amplitudes from the measurements of @xmath147 and @xmath148 on polarized targets . we now show that a model independent amplitude analysis is possible with unpolarized target data @xmath143 . perhaps more appropriately , the analysis should be called `` intensity analysis '' . first we define partial wave intensities @xmath234 , @xmath89 . using the expressions in the table i we can define a new set of 6 equations @xmath235 where for @xmath236 the interference terms are @xmath237 the system of equations ( 4.1 ) can be solved if the correlations @xmath238 satisfy a cosine condition similar to ( 3.5 ) . to see if such is the case we used ( 4.2 ) to expresse the correlations in terms of transversity amplitudes . upon substitution into the cosine condition we obtained a non - linear constraint connecting moduli and cosines of relative phases of all transversity amplitudes @xmath239 and @xmath240 . this condition was tested numerically for all four combinations of the amplitudes @xmath241 by every single monte carlo sampling of the data error volume that produced a physical solution . in each case the constraint was satisfied exactly as an identity . we concluded that the system ( 4.1 ) is analytically solvable using the method described in the previous section .. to prove the analytical solvability in a simpler way we set in ( 3.1 ) and ( 3.2 ) all observables @xmath242 . then ( 4.1 ) reads @xmath243 where the amplitudes @xmath244 are analytic continuation of the amplitudes @xmath245 to @xmath242 . we obtain a similar set of equations from ( 3.2 ) with @xmath246 for amplitudes @xmath247 which are analytical continuation of the amplitudes @xmath248 to @xmath242 . since @xmath249 we must have @xmath250 the equations ( 4.1 ) and ( 4.3 ) coincide when @xmath251 since the amplitudes @xmath252 and @xmath253 are complex valued functions , they satisfy the phase conditions ( 3.4 ) and cosine conditions ( 3.5 ) . this means that the equations ( 4.1 ) are analyticaly solvable . helicity nonflip and single flip amplitudes @xmath254 and @xmath255 are related to transversity amplitudes by the relations @xcite @xmath256 where @xmath257 . then at @xmath242 we have @xmath258 . thus @xmath259 and @xmath260 . we thus find another useful relation for physical intensities in addition to ( 4.5 ) @xmath261 there are two solutions for the single flip helicity amplitude @xmath262 extrapolated to @xmath242 which correspond to the two analytical solutions of the equations ( 4.1 ) satisfying the conditions ( 4.5 ) @xmath263 and @xmath264 are solutions of the equations ( 4.1 ) that are not related to the extrapolation of the amplitudes to @xmath242 and thus do not meet the conditions ( 4.5 ) . the reason for this is the fact that @xmath265 while @xmath266 . the amplitude analysis determines also the relative phases between single flip amplitudes @xmath262 . to determine @xmath267 we find from ( 4.6 ) the interference term @xmath268 where @xmath269 are real , @xmath270 and @xmath271 . at @xmath242 we have @xmath272 , @xmath273 and @xmath274 . then ( 4.10 ) reads @xmath275 with @xmath276 we find for both solutions 1 , 2 @xmath277 where @xmath12 . note how our results on the single flip amplitudes @xmath262 depend on our finding in section vi . that the relative phases @xmath278 . we have performed high resolution amplitude analyses of cern data on @xmath21 at 17.2 gev / c at @xmath279 ( gev / c@xmath280 and in dipion mass range @xmath281 mev with polarized target data @xmath156 , @xmath157 @xcite ( analysis i ) as well as with only unpolarized target data @xmath156 @xcite ( analysis ii ) . the two analyses were performed using the same code but in the analysis ii we set @xmath242 . the analysis with polarized target data used 1 million of monte carlo samplings of error volumes of measured density matrix elements . not all values of density matrix elements correspond to physical amplitudes . the procedure produces frequency distribution of physical solutions for the moduli and phases and for various observables calculated from the solutions for amplitudes . the average values of the distributions for the moduli add up exactly to @xmath158 and those for the phases satisfy cosine conditions ( 3.5 ) . these averaged values can thus be interpreted as the measured amplitudes . the range of distributions determines the asymmetric error bars on the amplitudes and calculated observables . analysis i using 5 million monte carlo samplings produced virtually identical averaged values and slightly larger errors , indicating the analysis is stable . the analysis with unpolarized target data used 200 000 monte carlo samplings . the results from both analyses are shown in figures 1 - 7 and 11 - 13 . the figures show data points with errors from both analyses . and @xmath282 from analyses i and ii , respectively.,width=453,height=396 ] -wave amplitudes @xmath283 and @xmath284 from analyses i and ii ( with line).,width=453,height=396 ] -wave amplitudes @xmath285 and @xmath286 from analyses i and ii ( with line).,width=453,height=396 ] to understand the origin of the @xmath3 mixing in the @xmath1-wave amplitudes we recall from ( 3.7 ) the relation @xmath287 in the analysis i and a similar equation for the amplitudes @xmath288 in terms of @xmath282 in the analysis ii . we show that the presence of @xmath5 in the @xmath1-wave amplitudes in both analyses arises from the dominant presence of this resonance in the data components @xmath46 and @xmath282 . the data components were calculated from those monte carlo samplings of density matrix elements in the error volume of the data for which physical solutions for the amplitudes were found . the results for @xmath46 and @xmath282 are shown in figure 1 . the figure shows pronounced @xmath5 peak for target spin `` down '' component and a suppression of @xmath5 in the target spin `` up '' component . there are pronounced dips at 970 mev and 1030 mev in target spin `` down '' and `` up '' components , respectively , corresponding to the @xmath7 resonance . in contrast , the components @xmath289 are equal for both target transversities , peak at @xmath5 mass and dip at 980 mev . and @xmath290 for @xmath1- and @xmath2-wave amplitudes from analyses i and ii ( with line).,width=453,height=396 ] figure 2 shows the two solutions for the @xmath2-wave amplitudes @xmath291 and @xmath6 from the analysis i and the amplitudes @xmath292 from the analysis ii . remarkably , in both analyses the two solutions of the two cubic equations reproduce closely the general features of the data components in the figure 1 . in the analysis i the @xmath5 peak is enhanced in @xmath6 and suppressed in @xmath291 . figure 3 shows the two solutions for the @xmath1-wave amplitudes @xmath293 and @xmath4 from the analysis i and amplitudes @xmath294 from the analysis ii . the rho - like resonance and @xmath7 are clearly resolved in both solutions for @xmath4 in analysis i. this pronounced peak at 770 mev with a width at half - height of @xmath295 mev has its origin in the @xmath5 peak of the data component @xmath46 which peak survives the subtraction of @xmath296 . in the analysis i the @xmath5 peak is enhanced in @xmath4 and suppressed in @xmath293 . the analysis ii shows a clear @xmath5 peak in solution 1 and a subdued @xmath5 peak in solution 2 in the amplitude @xmath297 . in both analyses the amplitudes @xmath4 and @xmath297 show a shoulder at @xmath298 mev in solution 1 and a secondary peak at the same mass in solution 2 . in ref . @xcite we have shown in fits to the moduli of @xmath1-wave amplitudes that these structures arise from the interference of @xmath5 and @xmath7 breit - wigner amplitudes with a complex backgroud . in figure 4 we present the partial wave intensities @xmath299 from the analysis i and @xmath300 from the analysis ii for the solutions @xmath301 . the intensities from the two analyses are virtually identical , although they come from two different calculations with differenet sets and numbers of physical monte carlo samplings . this result confirms the validity of our analysis with unpolarized target data . both analyses show the expected dominant @xmath5 peak in the solutions @xmath302 and @xmath303 . the crucial result however is the observation that both analyses show a clear albeit subdued @xmath5 peak in the solutions @xmath304 and @xmath305 . this means that an exact amplitude analysis of unpolarized target data provides evidence for the presence of @xmath5 in both single flip amplitudes @xmath306 and @xmath307 . the observation of @xmath3 mixing in the @xmath1-wave amplitudes brings up the question of @xmath3 mixing in the @xmath2-wave amplitudes . figure 2 shows a clear dip at 970 mev in the amplitude @xmath6 followed by a sharp rise . the dip occurs at the mass of scalar resonance @xmath7 and is more pronounced in the solution 2 . figure 11 shows unexpected resonant structure ( a bump ) at the @xmath7 mass in the solution 1 of the amplitude @xmath308 . these findings are a strong indication for @xmath3 mixing in the @xmath2-wave amplitudes . since @xmath7 resonance is produced by unnatural exchange we expect @xmath7 to be suppressed in natural exchange amplitudes @xmath17 as evidenced by the data in figure 12 . figures 5 , 6 and 7 show the mass dependence of the phases @xmath309 and @xmath310 , @xmath311 from the analysis i. as the figures show , these phases are continous and nearly constant functions . from this fact we can understand why the phases @xmath312 can not change sign . from ( 3.16 ) we see that a change of signs of @xmath221 or @xmath313 results in the change of sign of all phases of the same transversity . this would result in large unphysical discontinuities in the phases @xmath314 and @xmath315 . and @xmath316 from analyses i and ii ( with line).,width=453,height=396 ] and @xmath317 from analyses i and ii ( with line).,width=453,height=396 ] and @xmath318 from analyses i and ii ( with line).,width=453,height=396 ] figures 5 , 6 and 7 also show the mass dependence of the phases @xmath319 , @xmath320 , @xmath311 from the analysis ii . there is a remarkable agreement between the analyses i and ii except for solution 1 above 900 mev in phases @xmath321 , @xmath322 , @xmath323 . while the phases @xmath309 and @xmath324 are similar , the phases @xmath319 and @xmath325 are equal . due to the similarity of phases @xmath221 and @xmath326 we may conclude that not only @xmath327 but also @xmath328 for both solutions 1 and 2 . independent evidence for the presence of @xmath5 in the @xmath1-wave amplitudes comes from the relative phases @xmath329 and @xmath330 from both analyses . for target spin `` up '' the phase @xmath221 is constant and small , reflecting the similarity of @xmath293 and @xmath291 in both solutions 1 and 2 . for target spin `` down '' the phase @xmath222 is near zero in the solution 1 and small and slowly varying in solution 2 below 1000 mev . these results indicate that the amplitudes @xmath8 and @xmath9 are nearly in phase . since the amplitudes @xmath9 resonate at @xmath5 , so must the amplitudes @xmath8 . this conclusion is in agreement with the phases @xmath331 , @xmath332 from both analyses . these phases are approximately constant below 1000 mev and indicate that the amplitudes @xmath8 and @xmath11 are nearly 180@xmath333 out of phase . since amplitudes @xmath11 resonate at @xmath5 , so must the amplitudes @xmath8 . the figure 7 shows that the relative phases of the amplitudes @xmath9 and @xmath11 are nearly constant and nearly 180@xmath333 out of phase , as expected from these two @xmath2-wave amplitudes . amplitude analysis of the complete @xmath1- and @xmath2-wave subsystem of the reduced density matrix yields four sets of solutions for reduced transversity amplitudes @xmath334 . for unnatural exchange amplitudes @xmath98 the reduced transversity amplitudes are related to transversity amplitudes by phase factors @xmath335 where @xmath336 is the arbitrary absolute phase and @xmath337 is the relative phase between @xmath1-wave amplitudes of opposite transversity . recall from ( 2.6 ) that @xmath338 are real and positive amplitudes , the phases of @xmath339 and @xmath340 are @xmath210 and @xmath341 , and the phases of @xmath342 and @xmath343 are @xmath212 and @xmath344 , respectively . helicity amplitudes @xmath345 with definite @xmath76-channel naturality were defined and related to transversity amplitudes of definite @xmath76-channel naturality in ref . @xcite for any dipion spin @xmath82 and helicity @xmath83 . due to the @xmath2-parity conservation only helicity nonflip and helicity flip amplitudes @xmath346 and @xmath347 are independent . here @xmath53 is nucleon helicity flip @xmath348 . the @xmath1- and @xmath2-wave helicity amplitudes @xmath349 are related to the transversity amplitudes @xmath88 by relations @xcite @xmath350 in terms of reduced transversity amplitudes we can write for the unnatural exchange amplitudes @xmath351 each set of solutions comes with a fourfold sign ambiguity in the phases of the amplitudes @xmath352 where @xmath353 are the signs of relative phases @xmath210 and @xmath341 . from ( 3.16 ) we see that the change of sign of @xmath210 ( @xmath341 ) results in the change of sign of the phase of amplitude @xmath342 ( @xmath343 ) , or complex conjugation of amplitudes @xmath354 ( @xmath355 ) . we can thus write the four sets of phases for each @xmath356 @xmath357 as shown in the appendix , the measurements with planar target polarization can select uniquely one of the four solutions for the phases . since the transversity amplitudes @xmath358 and @xmath359 are also complex conjugated , we get for helicity amplitudes @xmath360 since we do not get new solutions with the phase sets @xmath361 and @xmath362 numerical calculations were done only for the sets with phases @xmath363 and @xmath364 . we now look at the bilinear terms @xmath365 , @xmath98 and @xmath366,@xmath367 . for the sake of brevity we shall omit in the following the indices @xmath368 and @xmath363 , @xmath364 . using ( 6.4 ) or ( 6.6 ) we obtain @xmath369 where @xmath370 , @xmath371 and @xmath372 is partial wave intensity . note that @xmath373 as both @xmath1 and @xmath374 are real . for the bilinear terms @xmath366 we obtain @xmath375 the real part reads @xmath376 @xmath377 the imaginary part reads @xmath378 @xmath379 it is apparent from ( 6.4 ) and ( 6.6 ) that the knowledge of @xmath150 allows to determine the helicity amplitudes up to an absolute phase for each solution set @xmath356 and phase set @xmath363 and @xmath364 . as we show in the next section , the phase @xmath150 can be determined analytically from the consistency condition @xmath380 where the terms are given by ( 6.7),(6.9 ) and ( 6.10 ) . in order to make use of the consistency condition ( 6.11 ) to determine @xmath150 , we first rewrite the terms ( 6.8 ) , ( 6.10 ) and ( 6.11 ) in a simplified form . to this end we define @xmath381 then @xmath382 note that @xmath383 , @xmath384 and @xmath385 . we can write @xmath386 where @xmath387 next we require that @xmath380 the l.h.s . of ( 6.16 ) reads @xmath388 @xmath389 the r.h.s . of ( 6.16 ) reads @xmath390 @xmath391 subtracting ( 6.16 ) with @xmath392 from ( 6.16 ) with @xmath393 and using ( 6.15 ) we obtain equation linear in @xmath394 and @xmath395 @xmath396 @xmath397 which can be cast in the form @xmath398 using @xmath399 we find @xmath400 where @xmath401 . using ( 6.7 ) , ( 6.9 ) and ( 6.10 ) it is straightforward to verify that the consistency condition ( 6.16 ) reduces to identity in the following two cases @xmath402 the first case leads to two solutions @xmath403 and @xmath404 . the second case leads to another two solutions @xmath405 and @xmath406 . the monte carlo amplitude analysis of reduced transversity amplitudes and related observables was carried out by a computer code a. the mean values of moduli and phases of the reduced transversity amplitudes satisfy strict normalization and phase conditions , respectively , and thus represent the true measured amplitudes at the @xmath76-bin average of 0.067 ( gev / c)@xmath29 . these mean values were used as an input in an exploratory computer code b to calculate @xmath150 using ( 6.21 ) and helicity amplitudes @xmath407 , @xmath98 using @xmath150 from ( 6.21 ) , ( 6.22 ) and ( 6.23 ) . errors on @xmath150 from ( 6.21 ) and on helicity amplitudes were not calculated due to high non - linearity of the equations . and @xmath408 for @xmath409 . the solutions for the phase signs + + and + -are equal.,width=453,height=396 ] and @xmath410 for @xmath411 . the solutions for the phase signs + + and + - are different.,width=453,height=396 ] for @xmath411 . the solutions for the phase signs + + and + - are different.,width=453,height=396 ] equations ( 6.21 ) were used with amplitudes @xmath412 to calculate @xmath413 for each set @xmath363 and @xmath364 of signs of phases . using these values of @xmath413 as well as the values corresponding to the solutions ( 6.22 ) and ( 6.23 ) , moduli @xmath414 and interference terms @xmath415 were calculated to determine cosines and sines of the relative phases @xmath416 since the calculations of @xmath150 , moduli and the interference terms are all entirely independent , the selfconsistency of the data and calculations was checked using the following three tests on the relative phases . the first test are trigonometric identities @xmath417 the second test are phase conditions @xmath418 the third test are cosine conditions equivalent to phase conditions @xmath419 all conditions are satisfied identically within the single precision calculation used by the code for all combinations of solutions @xmath409 for both sets @xmath363 and @xmath364 of signs of phases . for instance , typical deviation from 1 of the trigonometric identities ( 6.25 ) are of order @xmath420 . to select a physical solution we used two criteria : ( 1 ) the single flip amplitudes must dominate non - flip amplitudes due to the pion exchange dominance at small @xmath76 . ( 2 ) the amplitudes @xmath54 and @xmath421 must show a clear @xmath5 peak . the computer codes a and b were merged in a computer code c to calulate errors on helicity amplitudes using a physical solution for @xmath150 found in the exploratory code b. each monte carlo selection from the data error volume that yields a physical solution of transversity amplitudes also yields the corresponding physical solution of helicity amplitudes , enabling code c to calculate their average value and error range in each mass bin in the same way as for the transversity amplitudes . the transversity and helicity amplitudes with errors thus represent identical experimental data . the values of @xmath422 and @xmath423 calculated from ( 6.21 ) show random variations as a function of dipion mass for all solution sets @xmath409 and signs of phases . all amplitudes exhibit the same random behaviour and do not show the required resonant breit - wigner behaviour at @xmath5 resonance . moreover , the non - flip amplitude @xmath424 has magnitude comparable to or larger than the flip amplitude @xmath54 . these solutions with @xmath425 and @xmath426 are rejected as unphysical . the two values for @xmath427 from ( 6.22 ) correspond to @xmath428 and @xmath429 for all solution sets @xmath409 . the solutions with @xmath430 are excluded because they require that the non - flip amplitude @xmath424 is larger than the flip amplitude @xmath54 . because @xmath431 and the amplitudes @xmath42 and @xmath43 do not depend on @xmath395 , both solutions @xmath432 give @xmath433 where @xmath434 is @xmath1-wave intensity . as a result , both values @xmath427 are unphysical for any solution set @xmath409 and signs of phases @xmath363 and @xmath364 . the solutions with @xmath435 ( @xmath405 ) from ( 6.23 ) require that @xmath436 and @xmath437 for any solution set @xmath409 and signs of phases . the magnitudes of @xmath43 are small compared to magnitudes of @xmath42 , in contradiction with the pion exchange dominance . these solutions are thus excluded as unphysical solutions . in the solutions with @xmath276 ( @xmath438 ) the pion exchange dominance is observed in all amplitudes and @xmath439 and @xmath440 show a well defined @xmath5 peak . the resulting amplitudes thus represent a unique solution to both helicity and transversity amplitudes , up to signs of the phases @xmath221 and @xmath326 to be resolved by the measurements of @xmath147 and @xmath441 . the solution @xmath276 was used in the computer code c to calulate average values and errors of helicity amplitudes in a monte carlo analysis of the data . figure 8 shows the moduli of @xmath1-wave helicity amplitudes . the solutions for the phase signs @xmath363 and @xmath364 are equal . all solutions for @xmath442 provide evidence for @xmath3 mixing in the @xmath1-wave helicity flip amplitudes . the @xmath5 peak stands out in in solutions @xmath443 and @xmath444 which resemble the solution 1 of @xmath4 . the peak is clearly visible in the solutions @xmath445 and @xmath446 which are similar to solution 2 of @xmath4 . the secondary structure in the region of 900 - 1000 mev is due to @xmath7 . figure 9 shows the moduli of helicity amplitudes @xmath424 and @xmath54 . the solutions @xmath447 and @xmath448 ( not shown ) are similar to solutions @xmath449 and @xmath450 , respectively . the helicity flip amplitudes @xmath439 dominate the small non - flip amlitudes with the @xmath5 peak below 980 mev . above 980 mev @xmath439 decreases rapidly while there is a sudden rise in @xmath15 . these structures are associated with @xmath7 resonance and support the suggestion of @xmath56 mixing in the helicity non - flip amplitude @xmath424 . the exception is the solution @xmath451 where the rise is observed in @xmath439 . figure 10 shows the relative phase @xmath14 which provide an independent evidence for @xmath56 mixing in the @xmath1-wave . again , the solutions @xmath447 and @xmath448 ( not shown ) are similar to solutions @xmath449 and @xmath450 , respectively . in all solutions the phase is small and nearly constant below 960 mev . since @xmath54 resonates near @xmath5 with a breit - wigner phase @xmath452 , the phase of the amplitude @xmath43 must be nearly the same resonant phase . the sudden rise of the phase @xmath14 in the @xmath7 mass region is due to the presence @xmath7 resonance in the amplitude @xmath43 . the non - flip amplitudes @xmath453 are very small and nearly constant . as a result we find @xmath454 for all @xmath409 . using ( 4.9 ) this means that @xmath455 there is also a clear similarity between the phase @xmath14 and the phases @xmath221 and @xmath326 which implies that @xmath456 in equations ( 6.28 ) and ( 6.29 ) the solutions @xmath191 on the l.h.s . correspond to solutions @xmath411 on the r.h.s . , respectively . the extrapolation of @xmath43 to @xmath242 obtained in the analysis of unpolarazied target data carries nearly as much information as the exact amplitude from analysis of polarized target data . the residues of the single flip helicity amplitudes @xmath43 and @xmath54 at the pion pole @xmath458 in the unphysical region are related to the @xmath1-wave and @xmath2-wave partial wave amplitudes @xmath459 and @xmath460 in @xmath461 scattering . analyticity and unitarity of @xmath25 scattering amplitudes ( roy equations ) exclude a rho - like resonance in the @xmath1-wave @xcite . past extrapolations of unpolarized target data to the pion pole produced a residue of @xmath462 with a broad structure in dipion mass below 900 mev without any evidence for a rho - like resonance while the residue of @xmath463 showed the @xmath5-like peak . the question arises whether the extrapolation of our single flip amplitudes @xmath464 or @xmath465 and @xmath466 to the pion pole preserves the @xmath5 peak or smears it out . the broad shapes of moduli @xmath467 , @xmath445 and @xmath446 are all very similar and strongly suggest that an extrapolation of these solutions to the pion pole will show a broad structure consistent with previous extrapolations of unpolarized data . in this regard our results agree with the previous @xmath25 scattering data analyses . roy equations only inform us that we should use the extrapolation of @xmath464 and not that of @xmath468 , or extrapolations of @xmath465 and @xmath466 and not those of @xmath469 and @xmath470 , to calculate @xmath25 scattering amplitudes . roy equations have nothing to say about the physical amplitudes and intensities in @xmath0 scattering process itself . roy equations thus can not be used to select and reject solutions for physical intensities and amplitudes in the @xmath0 scattering , and so deny any physical meaning to the multiple solutions for amplitudes which may have - and actually do have - such a meaning apart from and outside of the @xmath25 scattering theory . it is generally expected that the position and the width of the @xmath5 peak as observed in the spin averaged cross - section will be faithfully reproduced on the level of spin amplitudes . the cern data on polarized target show that this is not the case . from figures 2 , 11 and 12 we see that the @xmath5 production is suppressed in all target spin `` up '' amplitudes while the target spin `` down '' spectra dominate the @xmath5 production . the width at half - height of the peaks in the longitudinal spectra @xmath291 and @xmath6 is the expected @xmath471 mev . however , the @xmath5 width shows different values in different transverse spectra . in the spin `` down '' spectra the @xmath5 width is narrower at @xmath472 mev in @xmath308 but wider at @xmath473 mev in @xmath474 . these values are reversed in the spin `` up '' spectra with @xmath473 mev and @xmath472 mev in amplitudes @xmath475 and @xmath476 , respectively . such large variations in the @xmath5 width are entirely unexpected and appear anomalous . the rotational symmetry of strong interactions requires that the width and mass of a resonance do not depend on its helicity . it also prevents the mixing of scalar and vector resonances in the same partial wave with a definite spin @xmath82 . if rotational symmetry was broken in pion production then the widths of @xmath5 in the @xmath2-wave amplitudes @xmath9 , @xmath11 and @xmath477 could differ and @xmath3 mixing could occur in both @xmath1- and @xmath2-wave amplitudes . but how realistic is the hypothesis that rotational invariance is violated in pion production ? the breaking of rotational symmetry implies a breaking of lorentz symmetry . in 2001 , gale and collaborators examined vector - scalar mixing of @xmath478 resonances in dilepton production in hot and dense matter . the medium - induced breaking of rotational and lorentz symmetry leads to mixing of different spin states even when the interaction lagrangian respects all required symmetry properties @xcite . their work inspired the idea to test rotational and lorentz symmetry in cern data on @xmath21 on polarized target . to test the rotational symmetry we need information on transversity amplitudes @xmath479 with definite dipion helicity @xmath480 . the transverse amplitudes @xmath11 and @xmath477 are a mix of transverse amplitudes with dipion helicities @xmath83=+1 and -1 , and thus are not suitable to test the rotational symmetry . the required amplitudes with transverse helicities @xmath481 and @xmath482 are related to the amplitudes @xmath11 and @xmath477 @xcite @xmath483 their partial wave intensities and polarizations can be calculated from the data on polarized target @xmath484 for amplitudes with zero helicity we have longitudinal amplitudes @xmath9 . it is convenient to relabel them as @xmath485 . their moduli squared shown in figure 2 are then the longitudinal intensities @xmath486 -wave amplitudes @xmath16 and @xmath487 from analyses i and ii ( with line).,width=453,height=396 ] -wave amplitudes @xmath17 and @xmath488 from analyses i and ii ( with line).,width=453,height=396 ] and @xmath489 from analyses i and ii.,width=453,height=396 ] and @xmath490 from analyses i and ii.,width=453,height=396 ] figures 13 and 14 show the transverse intensities @xmath491 and @xmath492 for two solutions ( 1,1 ) and ( 2,2 ) . the intensities show a clear peak with the same width at half - height of @xmath40 150 mev for both target spins in both solutions . this transverse @xmath5 width is exactly the same as the longitudinal width at half - height in the intensities @xmath493 and @xmath494 . this indicates that the @xmath5 poles in all helicity amplitudes @xmath495 , @xmath496 have the same width @xmath40 150 mev . to explain the large differences in the @xmath5 witdth observed in the spectra @xmath308 and @xmath474 we note from ( 7.2 ) @xmath497 is negative and broad in the @xmath5 mass region . as the result of the opposite signs of @xmath498 in ( 7.4 ) , the @xmath5 width is narrower in @xmath308 and wider in @xmath474 . the calculated polarization @xmath499 is positive and broad around the @xmath5 mass and therefore it has an opposite effect on their widths , as observed . thus the apparent difference in resonance widths in the amplitudes @xmath16 and @xmath17 is entirely due to the interference of the amplitudes @xmath500 and @xmath501 . figures 11 and 12 show also the results for amplitudes @xmath487 and @xmath488 from the analysis ii . it is remarkable that in this analysis the resonance widths at half - height in both amplitudes and in both solutions 1,2 are the same at @xmath471 mev . from ( 7.4 ) this implies @xmath502 , as observed in figures 11 and 12 . the same widths are therefore observed also in both solutions ( 1,1 ) and ( 2,2 ) for the transverse intensities @xmath503 in figures 13 and 14 . we conclude from these tests in the analyses i and ii that the production mechanism conserves rotational / lorentz symmetry . in contrast , the spin mixing mechanism responsible for @xmath3 mixing spontaneosly violates this symmetry in the measured amplitudes . while the production mechanism is governed by the @xmath1-matrix dynamics of the standard model , the spin mixing mechanism must be governed by a different dynamics outside of the standard model . we have presented experimental evidence for @xmath3 mixing in @xmath1- and @xmath2-wave transversity amplitudes in @xmath0 from both polarized and unpolarized target data measured at cern . a model independent determination of @xmath1-and @xmath2-wave helicity amplitudes confirms the evidence from transversity amplitudes and specifies the relative phase of amplitudes with opposite transversities . all three analyses yield mutually consistent results . we show that the measurements of @xmath147 and @xmath148 resolve the sign ambiguity of relative phases . as a result , there are four sets of transversity amplitudes and four sets of helicity amplitudes . the principal difference between the analyses i. and ii . is the spin dependence of the moduli @xmath504 . in the analysis i. the spin @xmath505 moduli @xmath506 are suppressed compared to the spin @xmath507 moduli @xmath508 while in the analysis ii . these moduli are equal @xmath509 . the suppression of @xmath506 in the analysis i. is a part of a more complex spin dependence of the amplitudes @xmath504 which show oscillations as a function of momentum transfer @xmath76 in the @xmath5 mass region @xcite . the amplitude analyses i and ii provide the first direct experimental evidence for the independence of the width of a resonance on its helicity expected theoretically from the rotational symmetry of strong interactions . since the interactions in the standard model do not violate lorentz symmetry then the observed @xmath510 mixing must arise from a different kind of interaction involved in the pion creation process . in two related papers @xcite we propose to identify this new kind of interaction with a non - unitary interaction of the produced @xmath1-matrix final state @xmath511 with a quantum environment @xmath512 . the four sets of transversity amplitudes and the four sets of helicity amplitudes provide two equivalent descriptions of this non - unitary interaction for dipion masses below 1080 mev . the @xmath1- and @xmath2-wave density matrix elements @xmath147 and @xmath148 involve interferences between reduced unnatural and natural exchange amplitudes of opposite transversity shown in table i .. first we shall show that the elements @xmath515 and @xmath516 determine the phases @xmath517 and @xmath518 of the natural exchange amplitudes @xmath100 and @xmath519 , respectively , with a two - fold ambiguity for each set of moduli @xmath520 . from the table i. and ( 2.6 ) we have @xmath521 @xmath522 where @xmath523 . from ( 1 ) we can solve for @xmath524 and @xmath525 @xmath526 substituting into @xmath527 we find @xmath528 where @xmath529 substituting into @xmath530 yields a quadratic equation for @xmath531 with two solutions @xmath532 where @xmath533 from ( 4 ) we can now calculate @xmath534 @xmath535 and from ( 3 ) we obtain @xmath536 where @xmath537 note that for each set of moduli @xmath538 , @xmath409 the two solutions for the phases @xmath539 depend @xmath356 and the sign @xmath540 in ( 4 ) but do not depend the sign ambiguities @xmath541 of amplitudes @xmath230 in ( 3.19 ) . as a result , the amplitudes @xmath542 and @xmath543 do not depend on signs @xmath541 . we shall now show that for each set of the moduli @xmath356 only one solution @xmath542 and @xmath543 is consistent with the remaining equations from the table i. and that this solution resolves the four - fold sign ambiguity in relative phases of unnatural exchange amplitudes @xmath544 and @xmath545 . to this end we introduce a convenient notation for real and imaginary parts of reduced transversity amplitudes @xmath546 @xmath547 where the phases @xmath548 and @xmath549 for amplitudes @xmath230 are given by ( 3.19 ) . the equations for the remaining density matrix elements from the table i. then take the form @xmath550 @xmath551 the equations ( 8) and ( 9 ) are four linear equations for four unknowns @xmath552 with four solutions @xmath553 , @xmath554 , @xmath555 and @xmath556 corresponding to @xmath557 and @xmath558 . one of these four solutions of ( 8) and ( 9 ) must be equal to one of the solutions of ( 1 ) . to be specific , let us suppose that this physical solution for @xmath559 is the solution with the positive sign @xmath560 in ( 4 ) . it corresponds to some specific signs @xmath561 and @xmath562 so let us label all amplitudes in this solution @xmath563 . we now show that the amplitudes @xmath564 for the other solution with @xmath565 can not be a solution of ( 8) and ( 9 ) for any choice of signs @xmath227 and @xmath566 of the phases in ( 3.19 ) . to prove this statetement let us suppose the contrary and assume that the amplitudes @xmath564 satisfy ( 8) and ( 9 ) for some amplitudes @xmath567 , @xmath230 from ( 3.19 ) corresponding to particular values of signs @xmath568 and @xmath569 . since the amplitudes @xmath567 , @xmath230 differ from the amplitudes @xmath570 , @xmath230 only in the sign of phases , their real parts are the same and the imaginary parts differ at most by sign @xmath571 where @xmath572 . in the next step we subtract the set of equations with amplitudes @xmath564 from the set with amplitudes @xmath573 and get a homogeneous set of equations @xmath574 @xmath575 we now write ( 6 ) in the form @xmath576 and calculate differences and sums @xmath577 @xmath578 we can write similar expressions for amplitudes @xmath579 and @xmath580 by replacing the paramerters @xmath581 in ( 14 ) with parameters @xmath582 corresponding to the form ( 13 ) of the equations ( 5 ) and ( 4 ) for @xmath534 and @xmath531 . we now examine the equations ( 11 ) and ( 12 ) for each possible choice of @xmath83 and @xmath583 . the case @xmath584 is excluded as the amplitudes @xmath564 can not satisfy the same system as the amplitudes @xmath573 . for the case with @xmath585 the equations ( 11 ) and ( 12 ) take the form @xmath586 the terms in the parentheses are all different . in particular , the large differences between the moduli @xmath587 and @xmath588 mean large differences in the parentheses for @xmath589 at least one of which must be non - zero . since @xmath589 has been measured in the cern experiments on polarized targets and it is non - zero , the equations ( 15 ) and ( 16 ) can not be satisfied and this case is excluded . for the case @xmath590 and @xmath591 the equations ( 11 ) and ( 12 ) read @xmath592 combining the first two and the last two equations we obtain two equations of interest @xmath593 the terms in the parentheses are all different . the parentheses for amplitudes @xmath594 and amplitudes @xmath595 are not proportional to each other since the phases of @xmath594 and @xmath595 are 180@xmath596 out of phase , as seen in fig . 7 . moreover , as we have discussed in section vii . , the moduli of these amplitudes have different @xmath5 widths and structures around @xmath7 . the equations ( 17 ) and ( 18 ) thus can not be satisfied and this case is excluded . the analysis of the case @xmath597 and @xmath598 is similar with the same conclusion . the solution @xmath599 selects a unique set of phases of amplitudes @xmath600 and @xmath601 corresponding to @xmath561 and @xmath562 . the equations ( 8) and ( 9 ) change when the phases of these amplitudes change sign . the change results in a different solution for amplitudes @xmath602 which is not compatible with the data in ( 1 ) . we conclude that the measurements of @xmath1- and @xmath2-wave density matrix elements @xmath147 and @xmath148 unambiguously select a unique solution for @xmath1- and @xmath2-wave reduced transversity amplitudes . for any set @xmath356 of solutions for the moduli there exists only one solution for the phases of natural exchange amplitudes and only one set of signs of phases of unnatural exchange amplitudes in ( 3.19 ) . the corresponding solution of equations ( 8) and ( 9 ) for @xmath552 is the only solution that satisfies the conditions @xmath603 and @xmath604 . k. rybicki , _ data tables of spin moments measured in @xmath0 on polarized target at 17.2 gev / c for dipiom masses 580 - 1600 mev and four - momentum transfers @xmath621=0.01 - 0.20 @xmath622 _ , private communication , 1996 .

early - type galaxies often contain a hot x - ray emitting interstellar medium ( 3 - 8 k ) with an apparent radiative cooling time much less than a hubble time . if unopposed by a heating mechanism , the gas will radiatively cool to temperatures 10 k at a rate proportional to/ , typically 0.03 - 1 m yr . we can test if gas is cooling through the 3 k range by observing the ovi doublet , whose luminosity is proportional to the cooling rate . here we report on a study of an unbiased sample of 24 galaxies , obtaining _ far ultraviolet spectroscopic explorer _ spectra to complement the x - ray data of _ rosat _ and _ chandra_. the ovi line emission was detected in about 40% of the galaxies and at a luminosity level similar to the prediction from the cooling flow model . there is a correlation between and , although there is significant dispersion about the relationship , where the ovi is brighter or dimmer than expected by a factor of three or more . if the cooling flow picture is to be retained , this dispersion requires that cooling flows be time - dependent , as might occur by the activity of an agn . however , of detected objects , those with the highest or lowest values of/ are not systematically hot or cool , as one might predict from agn heating .

astro-ph0511284

the cumulative stellar mass loss from the stars in an early - type galaxy is typically 0.1 - 3 m@xmath6 yr@xmath7 ( e.g. , @xcite ) , which , integrated over a hubble time , is comparable to the gaseous mass of a spiral galaxy . the absence of a massive disk of cool atomic and molecular gas in early - type galaxies indicates a different life cycle for the gas , which may be divided into distinct stages . ignoring accretion onto a galaxy , the first stage is mass loss from the stars , a process that can be measured by detecting the infrared signature associated with the stellar winds of red giants . the measurement of this process successfully detects the infrared emission near 10 @xmath11 m and yields a value for the mass loss rate that is approximately the value predicted from theoretical stellar evolution models @xcite . the stellar ejecta will not remain in orbit around its star because the expanding ejecta will eventually collide with the ejecta from other stars , undergoing shocks that convert their random orbital motion to thermal energy ( e.g. , @xcite ) . this process heats the gas to 10@xmath12 -10@xmath13 k , and if there were no additional heating or cooling , the gas would be bound to the galaxy and have the same spatial distribution as the stars . however , the radiative cooling time for the gas is less than a hubble time , so the gas will evolve with time , and this is the basis of the cooling flow model . in the absence of a heating mechanism , radiative cooling drains energy from the gas most rapidly at small radii , causing a loss of buoyancy and a subsequent inflow of gas . then , the rate at which gas cools and flows inward ( @xmath10 ) is proportional to two observed quantities , the energy loss rate ( @xmath14 ) divided by the thermal energy per gram ( @xmath15 ) , or @xmath10 @xmath16 @xmath14/@xmath17 , typically 0.03 - 3 m@xmath6 yr@xmath7 for ellipticals ( e.g. , @xcite ) . the location at which the gas cools ( to 10@xmath3 k or cooler ) depends upon a free parameter in this _ standard _ cooling flow picture . picture does not include heating effects by supernovae , which will make profound changes , since the net cooling rate is given by @xmath10 @xmath16 ( @xmath14-_h_)/@xmath17 , where _ h _ is a heating component . in principle , the rate at which cool gas is produced can be reduced to zero . if one uses the recent values for the supernova rate in early - type galaxies ( @xcite ; a factor of three lower than the older tammann and sandage rates ; @xcite ) , the characteristic temperature of the ism would be 1@xmath010@xmath13 k , which is about the escape temperature for a typical elliptical galaxy ( @xmath18 @xmath2 0.3@xmath19 ) . more detailed hydrodynamic calculations @xcite show that galactic winds ( or partial galactic winds ) will play an important role in the evolution of the hot gas even if the supernova heating rate is lower than that implied by @xcite . a galactic wind carries away nearly all of its energy rather than radiating it , substantially lowering the x - ray luminosity . not only is a galactic wind important for understanding the x - ray emission and the ism in the galaxy , it has implications for the surrounding intergalactic medium , as this is the primary way that it becomes polluted with metals . the x - ray emission surveys of early - type galaxies show that there can be a wide range in the mass and luminosity of x - ray emitting gas for galaxies of similar optical properties ( e.g. , @xcite ) . this variation , along with the trend of rapidly decreasing @xmath14 with decreasing optical luminosity @xmath18 , can be understood if galactic winds play a role in some systems ( other systems may have accretion from their surrounding region ; review by @xcite ) . therefore , the prediction is that the systems with large x - ray gas masses and short cooling times ( high @xmath14 ) should have cooling flows while the x - ray poor systems have partial or total galactic winds , so the cooling flow will be weak or absent in those systems . because the heating rate is not well known , it is difficult to determine the net cooling rate through x - ray luminosity measurements . alternatively , if gas is cooling to lower temperatures , emission lines will be produced that are indicative of those lower temperatures . the best lines for this test are from ovi because this ionization state dominates the total radiative cooling as the gas passes through the 2 - 4@xmath010@xmath8 k range @xcite . the primary cooling lines come from the doublet at @xmath20@xmath201032 , 1038 , which is accessible with the _ far ultraviolet spectroscopic explorer _ ( _ fuse _ ) . calculations show that there is a linear relationship between the line luminosity and the cooling rate , which is insensitive to the metallicity of the gas or whether the gas is out of collisional ionization equilibrium @xcite . at 10@xmath21 k , the radiative cooling rate per unit volume is significantly higher than at the temperature of the x - ray emitting gas ( 10@xmath12 - 10@xmath13 k ) , so the relative contribution of a heating mechanism is greatly reduced and can be ignored . therefore , the luminosity of the ovi doublet should be a direct measure of the gas cooling rate , @xmath22 . previously , we reported upon ovi observations of two early - type galaxies , ngc 1404 and ngc 4636 @xcite , two of the x - ray luminous galaxies widely believed to host cooling flows . for ngc 1404 , ovi was not detected and the upper limit is several times below the standard cooling flow prediction , based on _ rosat _ data . however , ovi emission was detected from ngc 4636 , and the luminosity of these lines corresponds to a cooling rate of 0.4 m@xmath6 yr@xmath7 . this is less than the total rate from the cooling flow model of 2 m@xmath6 yr@xmath7 , but the _ fuse _ aperture ( a 30@xmath23 square aperture , or an effective radius of about 17@xmath23 ) only takes in a part of the galaxy . correcting for the flux that falls outside the aperture is a model - dependent procedure , but if one uses a model with distributed mass drop out ( _ q_=1 from @xcite ) , the corrected ovi luminosity approximately equals the cooling flow prediction . following on this work , we began a ovi emission line survey of an unbiased sample of nearby early - type galaxies . the basic observations define the emission line characteristics of the sample , and permit us to test a few predictions of the model . one would expect that the x - ray faint galaxies would be very weak ovi emitters , if most of the thermal energy is being carried away in galactic winds . secondly , the galaxies with significant hot gas masses should usually possess ovi emission . we defined a sample of optically selected early - type galaxies in order to be unbiased with respect to x - ray properties . in previous work with _ rosat _ observations @xcite , we developed a complete optically selected sample of early - type galaxies , which contains the optically brightest elliptical galaxies ( by total apparent magnitude ) in the @xcite sample that do not have agns pointing at us , are not at low galactic latitude , and excluding dwarf galaxies and cd systems . this sample of 33 galaxies have been studied extensively in the optical region and were observed at x - ray energies with _ rosat_. most of the objects in the sample were observed more recently with _ chandra_. this sample was accepted as a program on _ fuse _ , and nearly all were observed , with only one exception due to the pointing constraints following the reaction wheel failure . the redshift range of the sample is -250 km s@xmath7 to 1950 km s@xmath7 , and the apparent magnitude range is 9.3 - 11.0 ( in b ) ; for completeness , we included m87 . the optical properties and exposure times are given in table 1 , where we list the galaxy name , absolute blue magnitude , r@xmath24 , one - dimensional stellar velocity dispersion , extinction in b , distance ( in km s@xmath7 and mpc , total exposure time , and night exposure time ( boldface denotes the data used ) . the raw _ fuse _ data were processed using the pipeline calfuse 2.2.1 , 2.4 , and 3.0 , usually with very little change between the data sets . we inspected each of the data sets to determine whether using the night - only data was superior to using the total data set ( night plus day data ) . the decisions depended partly on the redshift of the galaxy since the weaker ovi line at @xmath201037.62 can be redshifted into any of three airglow lines at 1039.2 , 1040.8 , and 1041.6 at about 450 km s@xmath7 , 920 km s@xmath7 , and 1150 km s@xmath7 . these airglow lines are about 100 km s@xmath7 wide and can shift relative to the source depending upon the placement in the slit . the ratio of these airglow lines , and relative to the stronger one at 1027.3 , allows us to estimate the degree to which contamination occurs . also , there is a dead spot in the detector that covers part of the spectrum near 1043 and this sometimes produces artifacts in the spectrum ( false absorption or emission features ) , so we inspect this wavelength region closely in the event that an emission line is found nearby ( the weaker ovi lines occurs here for redshifts of 1500 - 1600 km s@xmath7 ) . there are four detectors that cover this wavelength region , two lif channels and two sic channels , but the effective area is significantly better for the lif channels . also , the lif1a channel is significantly better than the lif2b channel and we find that adding this second channel generally reduces the s / n , so we usually rely upon the lif1a data for the region covering the ovi lines . we inspect the other channels for confirmation of features . the other important emission line in the _ fuse _ band is from ciii @xmath20977.02 , which is covered by the two sic channels , with the sic2a being the best . sometimes there is an offset in the wavelength scale , and when this is an issue , we mention it and use the strong atomic absorption line , cii @xmath201036.34 to realign the spectrum . the spectra from early - type galaxies have several components that need to be modeled or recognized when trying to determine ovi line strengths . given the redshift range of our objects , the ovi lines fall in the region 1030 - 1045 , but these lines are superimposed upon the continuum of the host galaxy , plus there is absorption by milky way gas , and there are airglow lines ( discussed above ) , all of which are taken into account when analyzing line strengths . the underlying stellar spectrum of the host galaxy was modeled in this spectral region by @xcite in order to analyze _ hopkins ultraviolet telescope _ ( _ hut _ ) data , which have a resolution of about 3 , significantly worse than _ fuse_. when we compare their model spectrum to the stellar continuum from ngc 1399 , the galaxy with the brightest continuum , we find that the model reproduces the general properties of the spectrum . however , there are aspects of the spectrum that depart from the model , and since the s / n is good for ngc 1399 , we choose to use it as the template spectrum for other objects . there are important absorption lines , produced by the milky way , superimposed upon the observed spectra of these external galaxies . the strongest atomic line in the spectral range 1028 - 1053 is cii @xmath201036.34 , which is always optically thick . this is followed in line strength by oi @xmath201039.23 , ari @xmath201048.22 , and the galactic ovi lines . in addition to atomic absorption lines , there can be absorption by molecular lines , although the strength of this component can vary greatly , from being absent ( at our typical s / n ) to being optically thick . the lines occur in bands , one occurring in the 1036 - 1039 region and another that occurs in the 1048.5 - 1052 region . the three strongest lines in the group at 1036 - 1039 are centered at 1036.54 , 1037.15 , and 1038.16 , the first of which can blend with the atomic line of cii @xmath201036.34 . these atomic and molecular lines can absorb part or all of an ovi emission line from the external galaxy if the redshift coincidence is unfortunate . we show the location of these absorption components in the figures for the spectra , but we do not determine their absorption equivalent widths , as this generally does not bear upon the analysis . to extract line fluxes or upper limits , we defined the continuum for about 5 on either side of the line , along the shape of the stellar continuum ; over this narrow range , the continuum is nearly a simple slanted line . to determine the line flux , we integrated over the line , but to determine the fwhm , a gaussian was fit . the location of the continuum is one of the main contributors to the uncertainty of the line flux , so the continuum location was varied ( @xmath25 1@xmath26 ) to determine the errors in the line flux and line width . when the line flux was less than 3@xmath26 , we consider it a non - detection and quote an upper limit of 3@xmath26 . the individual objects are discussed , beginning with ngc 1399 , which was used as the stellar template . extracted fluxes or upper limits to the ovi lines are given in table 2 , and for the detections , possible detections , and most upper limits , the spectra are shown ( some non - detections are largely noise so we do not show their spectra ) . for uniformity , and to avoid the strong geocoronal ly@xmath27 airglow line , we show the 1028 - 1053 region in our figures of all objects . the spectra were binned by five pixels to a bin size of 0.034 ( 9.8 km s@xmath7 at the ovi lines ) and then smoothed by five , seven , or 11 for the purposes of producing the figures shown below ; line fluxes were extracted from the unsmoothed data . the reddenings quoted below are from @xcite , the galacitc hi column density is from @xcite , and other galaxy properties ( e.g. , redshift ) are taken from the _ nasa / ipac extragalactic data base_. ngc 1399 : this galaxy has the brightest stellar continuum , typically an order of magnitude brighter than other systems . as with other galaxies in the fornax cluster ( ngc 1316 , ngc 1404 ) , the galactic hi column is low ( 1.31@xmath010@xmath28 @xmath29 ) as is the extinction , and there is no evidence for galactic h@xmath30 absorption , although galactic atomic absorption is found ( figure 1 ) . the lack of h@xmath30 absorption simplifies the determination of the location of the continuum and we see no evidence for ovi emission . due to the absence of ovi emission and the strength of the continuum , we use the stellar continuum as the standard template for the other systems , after filling in the galactic absorption lines ( for guidance , we used the model by @xcite ) . ngc 1316 : this is also known as fornax a , a radio galaxy in the fornax cluster , and one of the optically most luminous galaxies in the cluster . it is also a liner with weak low ionization emission lines , extended ionized and neutral gas , and some evidence of dust . the system has a low galactic extinction and low galactic hi column ( 1.9@xmath010@xmath28 @xmath29 ) , so the absence of detectable galactic h@xmath30 is expected . the _ fuse _ lif1a spectrum ( figure 2 ) shows emission from the strong line of ovi ( @xmath201032 ) . there is weaker emission from the ovi ( @xmath201038 ) line , also seen in the lif2b channel . the sharp feature that is coincident with ly@xmath27 for ngc 1316 and the ovi ( @xmath201032 ) line from the milky way is not confirmed in the lif2b channel and may not be real ( also , the higher series ly lines are not seen ) . in addition , there is emission from the ciii ( @xmath20977 ) line at the redshift of ngc 1316 ( sic2a and sic1a channels ) . the two ovi lines and the ciii line have the same fwhm of about 1.5 ( 440 km sec@xmath7 ) , which is comparable to the fwhm that one would infer from the one - dimensional velocity dispersion of the stars in ngc 1316 ( 223 km s@xmath7 , which would lead to a fwhm of 525 km s@xmath7 ) . the line ratio of the two ovi lines is [email protected] , consistent with the value of 2 expected for optically thin gas . ngc 1395 : galactic absorption is present but there is little h@xmath30 absorption ( about 10@xmath31 @xmath29 ) and there is no evidence for galactic ovi emission , which would be difficult to detect in this short observation . the strong ovi line from ngc 1395 would lie at 1037.84 , a fairly clean part of the spectrum ; there is only an upper limit to this line ( figure 3 ) . ngc 1404 : like the other fornax galaxies , h@xmath30 absorption is nearly absent , and in this observation , using night data , there airglow lines are nearly absent as well ( figure 4 ) ; the usual galactic atomic absorption lines are apparent ( presented and discussed in @xcite ) . there is no emission from either ovi line at the redshift of ngc 1404 . ngc 1407 : there is a moderate galactic hi column ( 5.42@xmath010@xmath28 @xmath29 ) and extinction , and the galactic h@xmath30 lines are prominent , along with the atomic absorption lines . the stronger ovi line from ngc 1407 ( 1038.1 ) corresponds to the location of a galactic h@xmath30 line , although not one of the strongest ones , but there is no apparent emission , nor is there emission at the location of the weaker ovi line ( figure 5 ) . ngc 1549 : there is very little galactic absorption or extinction ( 1.57@xmath010@xmath28 @xmath29 , and there is little evidence of h@xmath30 absorption ( figure 6 ) . absorption by galactic atomic gas is present and the cii @xmath201036 line ( 0.6 wide ) lies in the middle of the redshifted ovi @xmath201032 line , although emission near this region is seen in the lif2b channel as well . the weaker line , ovi @xmath201038 , is in an uncontaminated region and it is detected with a fwhm of about 1.2 . the flux is elevated just to the blue side of the cii @xmath201036 line , which may be the unabsorbed part of the ovi @xmath201032 line . the line width is about 0.7 of the fwhm of the stellar velocity dispersion . there is no emission from ciii @xmath20977 . there is a narrow emission line at the location of the galactic ovi line . ngc 3115 : galactic molecular absorption lines are strong , but the redshifted ovi lines lie in uncontaminated parts of the spectrum . there is no emission from ovi ( figure 7 ) . ngc 3379 : the redshift of this galaxy places the ovi lines in a part of the spectrum uncontaminated by galactic absorption . no ovi emission is found ( figure 8) . ngc 3585 : this fairly isolated e7/s0 galaxy lies in a loose group and has moderate galactic reddening and an hi column of 5.6@xmath010@xmath28 @xmath29 . there were problems that occurred when this spectrum was being obtained , and of the 16 ksec exposure , only 3 ksec was acceptable , nearly all during the day ( figure 9 ) . consequently , the airglow lines are strong . absorption by galactic atomic and molecular gas is nearly absent , which is surprising given the column density of the gas . due to the absence of the strong galactic c ii , ar i and fe ii ( 1063.18 ) lines , we do not believe this to be a reliable spectrum . this object may be reobserved at a future date but this data only provides a poorly constrained upper limit . ngc 3607 : the galactic hi column and extinction are low toward this system ( 1.52@xmath010@xmath28 @xmath29 ) and there are no strong h@xmath30 absorption lines that can be identified . the only clear galactic atomic absorption line is the cii @xmath201036 feature , and the strong ovi line would be redshifted to 1035.2 , where there appear to be an emission feature ( figure 10 ) . this feature also occurs in the lower s / n lif2b channel , which does not add much to the s / n of this result , but provides some consistency . however , the emission from the weaker ovi line is not present , although since this line is only half the strength of the other ovi line , it s absence does not lead to an inconsistency with the presence of the other line . the center of the ovi line is at 1035.3 , a redshift of 979 km sec@xmath7 ( and a fwhm of [email protected] , or 150 km sec@xmath7 ) , which is similar to the redshift of the galaxy , 935 km sec@xmath7 . in addition to the weak ovi line , the ciii @xmath20977 line is also detected at a somewhat higher level of significance and at exactly the redshift of the galaxy . it has a width of [email protected] , which is consistent with the width of the ovi line . a line width of 0.6 corresponds to 180 km sec@xmath7 , which is narrower than the fwhm that would be inferred from the stellar velocity dispersion ( 248 km sec@xmath7 , or a fwhm of 584 km sec@xmath7 ) . ngc 3923 : this galaxy has significant extinction and a moderately large hi column ( 6.3@xmath010@xmath28 @xmath29 ) , although the galactic h@xmath30 lines are not especially strong . no emission lines are seen ( figure 11 ) . the weaker ovi line is in an uncontaminated part of the spectrum , while the stronger ovi line would coincide with a weak galactic h@xmath30 absorption line and is adjacent to strong airglow lines . ngc 4125 : the stellar continuum is very weak in this galaxy , despite one of the longest observations in the sample ( figure 12 ) . the continuum is so poorly defined that no galactic absorption lines can be identified . the night - only data ( 40% of the total exposure time ) set yields stricter upper limits to the ovi emission . no emission features from any lines are detected , despite this galaxy being listed as a liner and classified as e6 pec . ngc 4374 : this elliptical in the virgo cluster has radio lobes and this is classified as a liner . despite being at high galactic latitude ( _ b _ = 74@xmath32 ) , it has moderate absorption and the _ fuse _ spectrum ( figure 13 ) shows clear evidence for galactic h@xmath30 absorption . the day spectrum had particularly strong airglow lines , so we used the night - only spectrum , which had only 3.7 ksec , yet it reveals a strong ovi @xmath201032 at the galaxy redshift ( also present at the same level in the day+night spectrum ) ; the red side of the line may be partly absorbed by the galactic cii @xmath201036 line . the weaker ovi line lies between two airglow lines and there is an instrumental feature near 1044 , but a line half of the strength of the ovi @xmath201032 line is consistent with the level of the spectrum above that expected from the stars . the line width of 130 km sec@xmath7 ( fwhm ) is only about one - fifth the velocity dispersion of the stars , an equivalent fwhm of 674 km sec@xmath7 , although we can not rule out that the line has a broader base . ngc 4406 : also known as m86 , this lies in the central part of the cluster , 1.2@xmath33 west of m87 , and with a modest hi column ( 2.6@xmath010@xmath28 @xmath29 ) . it has a nearly radial orbit , as its redshift ( -244 km s@xmath7 ) differs from the systemic velocity of the cluster by more than the cluster velocity dispersion . it has an elongated x - ray distribution that has been interpreted as stripping of the gas within ngc 4406 by the ambient cluster medium . the stellar continuum is very faint so that it is difficult to detect any galactic absorption features ( figure 14 ) . the stronger ovi line would be shifted to 1031.1 , and a possible emission feature in this low s / n spectrum occurs at that location . although this is the broadest and most significant emission feature , it is less than a 3@xmath26 detection . the weaker ovi line is not detected , nor are any other emission lines . ngc 4472 : known as m49 , this lies in the center of the southern grouping of the virgo cluster . it has a modest galactic hi column and extinction ( 1.65@xmath010@xmath28 @xmath29 ) , but it has clear galactic h@xmath30 absorption . there is an indication of a ovi @xmath201032 line associated with ngc 4472 , although if present , it is redshifted relative to the systemic velocity of the galaxy by about 0.4 ( 120 km sec@xmath7 ; figure 15 ) . there is a minor peak at the location of the redshifted ovi @xmath201038 line in this night - only spectrum . we regard the ovi line emission as a possible detection . the line width of fwhm = 120 km sec@xmath7 is about 18% of the stellar velocity dispersion expressed as a fwhm , 676 km sec@xmath7 . there is no indication of emission from the cii @xmath20977 line . ngc 4486 : this central galaxy of the virgo cluster ( m87 ) has low extinction but measurable lines of galactic atomic and molecular gas . unfortunately , the galactic cii @xmath201036 line ( 1036.3 ) is coincident with the redshifted ovi @xmath201032 line ( at 1036.4 ) and there are galactic h@xmath30 lines to the red side of the galactic cii line , so much of a redshifted ovi @xmath201032 line would be absorbed ( figure 16 ) . nevertheless , to the blue side of the galactic cii line , the continuum rises well above the expected stellar continuum ( in both the lif1a and lif2b spectra ) , suggestive of the blue side of a wide ovi @xmath201032 line from m87 . if this is the case , the line width would needs to be about 1.5 ( fwhm ) and at least 1.2 . if it is 1.5 , we estimate a line strength of 7@xmath010@xmath34 erg @xmath29 sec@xmath7 , which is consistent with the upper limit to the weaker ovi line of 5@xmath010@xmath34 erg @xmath29 sec@xmath7 . the ciii @xmath20977 line is detected , with a line width of 1.4 , although it could be wider since the red side is absorbed by galactic h@xmath35 lines ; ignoring absorption by these lines , the line flux is 1.5@xmath010@xmath36 erg @xmath29 sec@xmath7 , stronger than the ovi line . if the h@xmath30 absorption produces the decrease in the red side of the line , the line flux and line width would be about 30% larger . ngc 4494 : the redshifted ovi @xmath201032 line ( 1036.6 ) lies in the galactic cii absorption line ( 1036.34 ) , but there is no emission line on either side of the cii line ( figure 17 ) . there is no evidence for the redshifted ovi @xmath201038 line either , although it falls on an instrumental feature in the lif1a channel . in the less sensitive lif2b channel , there is no indication of either ovi line , and there is no instrumental feature near the weaker line . ngc 4552 : this galaxy ( m89 ) also lies in the virgo cluster , but has slightly more extinction than most other members ( 2.56@xmath010@xmath28 @xmath29 ) and shows clear evidence of galactic molecular hydrogen absorption . although galactic atomic and molecular absorption compromises our ability to detect the redshifted ovi @xmath201038 line , the redshifted ovi @xmath201032 line is in a part of the spectrum that is uncontaminated by galactic absorption or airglow ( figure 18 ) . there is an emission feature near the redshift of the systemic velocity of the galaxy ( for the ovi @xmath201032 line ) , although the center of the emission feature is redshifted relative to the systemic velocity by about 0.5 ( 150 km sec@xmath7 ; the stellar velocity dispersion is 260 km sec@xmath7 , or an equivalent fwhm of 615 km sec@xmath7 , so the line shift is well within the velocity dispersion value ) . the ovi @xmath201032 line is visible on both the lif1a and lif2b channels . ngc 4621 : this e5 galaxy ( m59 ) lies about 3@xmath33 west of m87 in the virgo cluster , it has about the average extinction for the virgo cluster ( 2.2@xmath010@xmath28 @xmath29 ) , but it clearly shows galactic h@xmath30 absorption along with the usual galactic atomic absorption lines . however , the redshift of the galaxy places both ovi lines in uncontaminated regions and the spectrum is of moderately good s / n ( figure 19 ) . there is no evidence for emission lines of either ovi line or of ciii . ngc 4636 : this virgo cluster galaxy was presented by @xcite , who reported the detection of both ovi lines . the width of the emission lines but not the line strength seems to depend upon the version of the pipeline used ; these results employ pipeline version 2.4 ( figure 20 ) . the ciii line may be detected , although the centroid is 120 km sec@xmath7 ( 0.4 ) blueward of the expected line center ( the line would have a fwhm of 0.5 and a flux of 5@xmath010@xmath34 erg @xmath29 sec@xmath7 , detected at about the 2@xmath26 level ) . ngc 4649 : the redshifted ovi @xmath201032 line would be shifted ( 1035.8 ) just to the blue side of the galactic cii @xmath201036 line ( 1036.3 ) , but there is no evidence for ovi emission above the level of the rather strong stellar continuum , which is similar to the continuum of ngc 1399 ( figure 21 ) . there are moderately strong galactic h@xmath30 lines and atomic absorption lines present , although these do not contaminate the locations of the ovi lines . ngc 4697 : the redshift of this galaxy would place the redshifted ovi @xmath201032 line ( 1036.2 ) into the strong galactic cii absorption line ( 1036.34 ) . there is no strong line to the blue side of the cii line ( figure 22 ) . ngc 5846 : this is the optically most luminous galaxy in the center of its group and it has a moderate amount of extinction ( 4.24@xmath010@xmath28 @xmath29 ) , with evidence of absorption from galactic molecular hydrogen and atomic lines . this is one of the more distant objects , so the continuum is fairly faint but is well - fit by the stellar continuum of ngc 1399 ( figure 23 ) . both the strong and weak redshifted ovi lines appear to be present , but not in the expected 2:1 ratio ( for ovi @xmath201032 to ovi @xmath201038 ) . the ovi @xmath201032 line is broken up by galactic molecular absorption features , so it could be significantly stronger . we can not fit the h@xmath30 parameters well enough to make an accurate absorption correction for the ovi @xmath201032 line . the usually weaker ovi @xmath201038 line ( redshifted to 1043.6 ) falls close to a feature in the detector . whereas this feature is stationary and well - known , we have seen instances of unusual absorption or emission near it , and this may contribute to the strength of this feature . the spectra in the lif2b channel , while poorer , confirms the presence of the ovi @xmath201032 line but not the ovi @xmath201038 line . consequently , we estimate the ovi luminosity from the ovi @xmath201032 line . there is emission from the ciii @xmath20977 line at a flux of 7.7@xmath010@xmath36 erg @xmath29 sec@xmath7 ( 15% error ) , with a line width of [email protected] and a line center of 982.87 ( at the systemic redshift , it would have been at 982.61 ) . ic 1459 : this galaxy has among the lowest galactic hi column in the sample and very little reddening ( 1.19@xmath010@xmath28 @xmath29 ) , although it has the usual galactic atomic absorption lines that can be seen on the stellar continuum ( figure 24 ) . the airglow lines are strong during the daytime observing , so although only 25% of the observing time was at night ( only 2.5 ksec ) , the night - time data are better for the analysis of the ovi region , while the total data set is better for analysis in the ciii region . there is a broad emission line coincident with the strong redshifted ovi line , which is relatively uncontaminated by galactic absorption features ( also present in the lif2b channel ) . the weaker ovi line is not detected at a statistically significant level , but as it is half the strength of the ovi @xmath201032 line , this is not inconsistent with a detection of the ovi @xmath201032 line . the ciii @xmath20977 line is also detected with a flux that is poorly determined because it is unclear where to mark the boundaries of the line . if we use the single peak , to the blue of line center , the line width is 0.8 ( line flux of 1.3@xmath010@xmath36 erg @xmath29 sec@xmath7 ) and a line center at 982.0 ( expected to be at 982.5 ) . if we use the weaker red part of the line as well , the line center is exactly that expected for the recession velocity , the line width becomes 2.1 ( at a line flux of 2.1@xmath010@xmath36 erg @xmath29 sec@xmath7 ) . given the uncertainties associated with each of the lines , it is difficult to compare the ovi and ciii lines . this galaxy has a strong radio source and is known to be a liner . an important goal of the program is to test whether the x - ray cooling rate is a good predictor of the true cooling rate , which would be reflected in the ovi line luminosity , presumed to be a better measure of the net cooling rate . there certainly is not a one - to - one relationship between the x - ray value of @xmath22 ( @xmath10 ) and the ovi value of @xmath22 ( @xmath9 ) , such as for the galaxies ngc 1399 and ngc 1404 , two of the most x - ray luminous sources but with no detectable ovi emission . however , there are statistical correlations between the x - ray and ovi cooling rates . for the analysis of our sample , we have summarized the results in table 3 , where we give a variety of x - ray and optical properties , along with the ovi results . the measure of a detection is given by either upper limits , detections , or possible detections . the detections are lines whose strength exceeds 3@xmath26 and where the signal appears clearly on the spectrum . possible detections are generally near 3@xmath26 threshold or where it has been difficult to establish the presence of the line unambiguously . there are a variety of issues , such as contamination from galactic absorption line , airglow features , or uncertainties in the level of the underlying stellar continuum . for the purposes of scoring the detections in a simple non - parametic fashion , we assign 0 to upper limits , 0.5 to possible detections , and 1.0 to detections . the conversion of the ovi luminosity to a cooling rate ( @xmath9 ) has been calculated by @xcite and by @xcite , and here we use the conversion l(1032 ) = 9@xmath010@xmath37 @xmath9 erg / sec , where @xmath9 is in m@xmath6 yr@xmath7 ( as used in bmi ) . this is probably accurate to 30% , where the uncertainty derives from issues such as whether the cooling gas is isobaric or isochoric . this is most likely an uncertainty in the absolute calibration and is unlikely to cause an additional random scatter of 30% . the quantity @xmath9 is fairly insensitive to the metallicity provided that the metals remain the primary cooling agent ( edgar and chevalier 1986 ) . in the temperature range where ovi is the most prominent ion ( near 10@xmath21 k ) , this requirement is satisfied for metallicities greater than 10@xmath38 of the solar value ( see @xcite ) . we wish to compare the ovi cooling rate to the x - ray cooling rate within the _ fuse _ aperture , and this depends somewhat on whether gas cools primarily in the center or in a distributed manner throughout the galaxy . effectively , the x - ray cooling rate is @xmath39 , where @xmath40 is the net cooling rate ( the x - ray luminosity , if there is no opposing heat source ) and e is the specific thermal energy of the gas . the factor @xmath41 represents the model correction due to fluid effects , such as gravitational reheating as gas flows inward or a sink of mass as gas cools out of the flow , which can be extracted from model calculations , such as @xcite . we discussed the differences between the predictions in the various models previously ( bmi ) and here we use the value @xmath41 = 0.4 , corresponding to the _ q _ = 1 model . in this _ q _ = 1 model , gas loss is distributed through the galaxy in the sense that hot gas is converted locally to cold gas at a rate inversely proportional to the instantaneous cooling time @xmath42 as given by @xmath43 . there is a second consideration in calculating @xmath10 , having to do with the _ fuse _ aperture , a square 30@xmath23 on a side , or an effective radius of 17@xmath23 . this is smaller than the effective optical radius ( r@xmath44 , typically 30 - 60@xmath23 ) or of the extent of the x - ray emission , which is similar to the optical size . if we were to adopt the cooling flow model without mass drop out , all of the cooling would occur within the _ fuse _ aperture ( although this makes an x - ray surface brightness profile that is too sharply peaked in the center ) . if there is mass drop - out with radius , then we need to correct for only enclosing part of the cooling gas in the _ fuse _ aperture . as described in bmi , this is accomplished by multiplying the total @xmath10 by the fraction of the x - ray luminosity projected into the _ fuse _ aperture . here we use the value of @xmath10 for the distributed model ( _ q=1 _ ) , corrected to the size of the _ fuse _ aperture . if we had used the model with no mass drop - out ( _ q=0 _ ) , the median predicted value of @xmath10 hardly changes , although for the most x - ray luminous galaxies in the sample ( often the largest ) , the predicted @xmath10 could be a factor of two higher . the various values of @xmath22 and other relevant derived quantities , including the x - ray luminosity , @xmath4 , and the x - ray temperature , @xmath5 are given in table 3 . there is a complete set of _ rosat _ x - ray fluxes and luminosities for this sample , so we begin by comparing these to the ovi data . the nominal prediction was that there would be a connection between @xmath10 and the detection of ovi emission , but we examined other relationships as well . some of those relationships investigated were between ovi and @xmath4 , @xmath45 , and @xmath5 , where no strong correlations were found and no strong relationships were predicted . however , there seems to be a correlation between the ovi and @xmath10 , as seen in figure 25 . we see that none of the six galaxies with the lowest values of @xmath10 have any ovi emission , yet the significance of this correlation is difficult to quantify . we would like to use the kaplan - meier estimator for the analysis of censored data @xcite , but an underlying assumption is that the data are censored randomly . here , the upper limits ( censored data ) are not randomly distributed , but preferentially occur at the lower values of @xmath10 . also , we have introduced `` possible '' detections , which is difficult to incorporate in statistical schemes . given these challenges , we can divide the sample into three bins by @xmath10 , using our scoring for detections , possible detections , and upper limits . we find that for galaxies with the lowest @xmath10 values , 1/8 have ovi , while 4.5/8 of the highest @xmath10 objects have ovi ( and 3.5/8 of the intermediate objects have ovi ) . using poisson statistics , the joint probability that of the lowest @xmath10 objects , 1/8 ( or fewer ) have ovi while 4.5/8 ( or more ) of the highest @xmath10 objects have ovi would occur by chance 2.5% of the time ( 97.5% confidence level ) . the correlation is probably a bit stronger than this value since none of the lowest six @xmath10 object have either an ovi detection or possible detection . for a kendall s @xmath46 test or a spearman s @xmath47 test for the whole sample ( treating upper limits and detections equally ) , the significance improves to 98 - 99.5% confidence , with the higher significance if the very poor upper limit object , ngc 3585 is eliminated from the sample . conservatively , we conclude that the correlation exists at the 98% confidence level when using the _ rosat _ data . a correlation is suggested between the ovi detectability and the radio luminosity , where the lowest third ( in @xmath48 ) has 1.5/8 ovi detections , the middle third has 3/8 and the most luminous third has 4.5/8 ovi detections . when we divide the sample in halves instead of thirds , for the low radio luminosity group , only 1.5/12 have ovi detections while 7.5/12 of the radio luminous objects have ovi emission . this correlation is significant at the 93% confidence level ( using poisson statistics ) , so more objects would be required to confirm this result . if the correlation is real , it could be caused if the cooling gas eventually feeds the radio source , although there is not a good correlation between @xmath10 and @xmath48 . alternatively , it might be possible for the agn to heat the gas and produce ovi emission , and four of the radio luminous objects have ciii emission as well , although this could be consistent with cooling gas also . the _ chandra _ data offer a number of advantages in determining x - ray properties , relative to the _ rosat _ data . first , the point sources can be removed from the observations , and a correction can be made for the unresolved point source contribution , yielding a more accurate measure of the properties of the x - ray emitting gas @xcite . the most important observational quantities used to calculate @xmath10 are @xmath4 and @xmath5 ( @xmath49 ) , with the surface brightness distribution , usually described by a @xmath27 model , being of less importance . chandra _ data not only provide better values for @xmath4 , but also for @xmath5 and for the @xmath27 model as well . however , the values of @xmath5 are typically in the range of 0.3 - 0.8 kev , so it is important to have observations that cover this low energy range adequately , and this is provided by the acis - s chip , which has sensitivity to 0.3 kev . the other detector used for imaging , acis - i , is three times less sensitive in the 0.5 - 1 kev range , it is insensitive below 0.5 kev , and it was subject to radiation damage during the early part of the mission . for these reasons , we do not include acis - i data in this program for galaxies where _ rosat _ data indicated that @xmath5 @xmath50 0.5 kev . this affects two objects , ngc 1395 and ngc 3607 , which are excluded from this analysis , making 22 objects with useful _ chandra _ and _ fuse _ data . this sample includes acis - i data for ngc 4486 ( m87 ) , a hotter galaxy , and we used the values given in the work of @xcite . using the _ chandra _ data , the relationship between @xmath9 and @xmath10 is little changed ( figure 26 ) . the objects with the lowest values of @xmath10 are smaller in the _ chandra _ data due to the removal of the x - ray point sources . the significance of the correlation is a bit lower for the _ chandra _ data , mainly because of having two fewer objects ( 95% confidence of a correlation ) . the nominal cooling flow prediction is that there should be a relationship between @xmath9 and @xmath10 , but the relationship should be @xmath9 @xmath51 @xmath10 . this is nearly the case for the _ rosat _ data and is almost the case for the _ chandra _ data , save @xmath10 appears to be about 20 - 30% too large , which is within the uncertainties in the model predictions . the more important discrepancy with the models is the very large scatter of @xmath9 for a fixed @xmath10 ( the dispersion of the relationship ) . this occurs because there are several x - ray bright galaxies without ovi emission , such as ngc 4649 ( the upper limit on @xmath9 is 5 times less than the prediction ) , ngc 1399 ( 2.6 time less than the prediction ) and ngc 3923 ( 2 times below the prediction ) . these differences are much greater than the statistical uncertainties in the data . also , there are some galaxies in which ovi is surprisingly bright : ic 1459 lies an order of magnitude above the predicted @xmath9 , ncg 1549 is five times above the prediction ; and ngc 1316 is four times above the predicted value . even if one were to suggest that an agn were responsible for the high values of ovi , we are still left with the objects that are much fainter than predicted . this broad dispersion about the relationship could be caused if cooling flows varied in time , as has been suggested for the cluster case ( e.g. , @xcite ) . heating by an agn could reheat the gas , choking off the cooling flow . if that were the case , one might expect that , in the moderate and high @xmath10 objects without ovi emission , the gas temperature would be above the velocity dispersion temperature , indicating that net heating is occurring . similarly , the ovi emitters , at similar values of @xmath10 , should have temperatures at or below the stellar velocity dispersion temperature . however , we find no such trend in the data supporting this picture nor is there any correlation between the detection of ovi and @xmath17 @xmath26@xmath38 for the entire sample . also , one might expect a division by radio properties , assuming that this is a proxy for agn activity . yet the objects furthest above the line , @xmath9 = @xmath10 , are not obviously different in their radio properties than those furthest below that line . therefore , within the context of the cooling flow picture , time variation in the flow rate is necessary to explain the ovi detection properties , yet we do not find telltale signs that should accompany intermittent heating . another possibility is that the ovi emission is not a result of cooling gas , but a result of gas being heated , such as by thermal conduction . however , there is a very large difference in the emission line strength between gas that is radiatively cooling and gas that is being conductively heated . this difference , which is typically a factor of 10@xmath52 - 10@xmath53 ( canizares et al . 1993 ) , means that to achieve the same line flux for conductively heated gas , the mass flux rate would have to be typically 10@xmath54 - 10@xmath52 m@xmath6 yr@xmath7 instead of a cooling rate of 0.1 - 1 m@xmath6 yr@xmath7 . if this cold gaseous reservoir is hi or h@xmath30 , the mass would have to be less than typical upper limits for the galaxies in this sample , about 10@xmath55 m@xmath6 , with some galaxies having significantly stricter upper limits @xcite . at a conduction rate of 10@xmath56 m@xmath6 yr@xmath7 , this would lead to a lifetime for the cold gas reservoir of 10@xmath1 yr , so the gas would have to be replenished every 10@xmath1 yr by the amount of gas found in galaxies about one - tenth the mass of the milky way . the orbital interaction time would be about 10@xmath57 yr , so we should see many of these galaxies passing through the early - type galaxy , which is not the case . consequently , it seems unlikely that conductive heating of gas could produce the lines that we observe . a final issue is whether the gas is undergoing simple radiative cooling or whether turbulent mixing layers play a role @xcite . in this case , the mixing would be between the hot ambient medium and gas that has already cooled , and this process can cause gas to spend less time at a given temperature compared to the pure radiative model . this has been suggested as a process within cluster cooling flows in order to remain consistent with the discrepancy between the observed and predicted strength of the x - ray ovii line within the cooling flow model ( fabian et al . 2001 ) . a similar discrepancy exists for ngc 4636 , where the ovii line is weaker than expected for gas cooling at the rate derived from the _ fuse _ data ( xu et al . if turbulent mixing causes gas to effectively jump over the ovii temperature region ( @xmath58 10@xmath1 k ) , it could solve this issue , but it also makes the prediction that there will be emission from the civ @xmath20 1550 line at a luminosity greater than the ovi doublet . unfortunately , there is presently no instrument that can measure the civ @xmath20 1550 line with the required sensitivity . the use of ovi emission has given a new insight into the properties of the hot gas in early - type galaxies , and this study highlights the need for further investigations . the greatest need is for a substantial increase in sensitivity , since most of the galaxies were not detected and even the detections are typically at the 3 - 5@xmath26 level . for objects with short exposure times ( @xmath50 5 ksec ) , it should be possible to double or even triple the s / n with moderate investments of observing time ( this applies to six objects in the sample , four of which are detections or possible detections ) . the other objects generally have exposure times of 5 - 10 ksec , requiring an additional 15 - 30 ksec per object ( generally of night data ) to double the s / n . for the objects with upper limits , improving the s / n by 2 - 3 will not make these objects secure detections even if there is a weak feature present . improved studies of these objects , or of more distant sources will require a new instrument with at least an order of magnitude greater sensitivity . this instrumental goal should be achievable as _ fuse _ is a rather small instrument by the standards of any space - based optical telescope ( or even most x - ray telescopes ) . we would like to thank the _ fuse _ team for their assistance in the collection and reduction of these data , and in particular , b - g andersson for his patience in dealing with many data reduction issues . also , we wish to acknowledge the advice of renato dupke and bill mathews . this research has made use of the nasa / ipac extragalactic database ( operated by jpl , caltech ) , the high energy astrophysics science archive , the multimission archive at space telescope , and the nasa astrophysics data system , all operated under contract with nasa . we gratefully acknowledge support by nasa through grants nag5 - 9021 , nag5 - 11483 , g01 - 2089 , go1 - 2087 , go2 - 3114 , and nag5 - 10765 . athey , a. 2003 , ph.d . , university of michigan athey , a. , bregman , j.n . , bregman , j.d . , temi , p. , and sauvage , m. 2002 , , 571 , 272 athey , a.e . , bregman , j.n . , and irwin , j.a . 2005 , , submitted bregman , j.n . , hogg , d.e . , and roberts , m.s . 1992 , , 387 , 484 bregman , j.n . , miller , e.d . , and irwin , j.a . 2001 , , 553 , l125 bregman , j.n . , snider , b.a . , grego , l. , and cox , c.v . 1998 , , 499 , 670 brown , b.a . , and bregman , j.n . 2000 , , 539 , 592 brown , b.a . , and bregman , j.n . 1998 , , 5495 , l75 . brown , t.m . , ferguson , h.c . , davidsen , a.f . , and dorman , b. 1997 , , 482 , 685 cappellaro , e. , evans , r. , and turatto , m. 1999 , , 351 , 459 david , l.p . , forman , w. , and jones , c. 1991 , , 369 , 121 dickey , j. , and lockman , f.j . 1990 , , 28 , 215 edgar , r.j . , and chevalier , r.a . 1986 , 310 , l27 faber , s.m . , 1989 , , 69 , 763 isobe , t. , feigelson , e. d. , and nelson , p. i. 1986 , , 306 , 490 kaiser , c.r . , and binney , j. 2003 , , 338 , 837 knapp , g.r . , gunn , j.e . , and wynn - williams , c.g . 1992 , , 399 , 76 mathews , w.g . , and brighenti , f. 2003 , , 41 , 191 matsushita , k. , belsole , e. , finoguenov , a. , bhringer , h. 2002 , , 386 , 77 osullivan , e. , forbes , d.a . , and ponman , t.j . 2001 , , 328 , 461 osullivan , e. , ponman , t.j . , and collins , r.s . 2003 , , 340 , 1375 pellegrini , s. , and ciotti , l. 1998 , , 433 , 444 roberts , m.s . , hogg , d.e . , bregman , j.n . , forman , w.r . , and jones , c. 1991 , , 75 , 751 sarazin , c.l . and ashe , g.a . 1989 , , 345 , 22 schlegel , d.j . , finkbeiner , d.p . , and davis , m. 1998 , , 500 , 525 slavin , j.d . , shull , j.m . , and begelman , m.c . 1993 , , 407 , 83 sutherland , r.s . , and dopita , m.a . 1993 , , 88 , 253 van den bergh , s. , and tammann , g.a . 1991 , , 29 , 363 voit , g.m . , donahue , m. , and slavin , j.d . 1994 , , 95 , 87 rrrrrrrrrr ngc 1316 & 1760 & 11.34 & 80.1 & 174 & 0.09 & 1422 & 20.3 & 18861 & * 9166 + ngc 1395 & 1717 & 11.02 & 45.1 & 258 & 0.10 & 1990 & 28.4 & 16573 & * 9875 + ngc 1399 & 1425 & 10.55 & 42.4 & 310 & 0.06 & 1422 & 20.3 & 16523 & * 10956 + ngc 1404 & 1947 & 10.89 & 26.7 & 225 & 0.05 & 1422 & 20.3 & * 7751 & 6363 + ngc 1407 & 1779 & 10.57 & 72.0 & 285 & 0.30 & 1990 & 28.4 & 16282 & * 10287 + ngc 1549 & 1220 & 10.58 & 47.4 & 205 & 0.05 & 1213 & 17.3 & 17900 & * 11868 + ngc 3115 & 720 & 9.95 & 32.3 & 266 & 0.21 & 1021 & 14.6 & 11561 & * 8080 + ngc 3379 & 911 & 10.43 & 35.2 & 201 & 0.11 & 857 & 12.2 & 4002 & * 2565 + ngc 3585 & 1399 & 10.53 & 38.0 & 220 & 0.28 & 1177 & 16.8 & * 2998 * & * 2998 + ngc 3607 & 935 & 10.53 & 65.5 & 248 & 0.09 & 1991 & 28.4 & 12918 & * 9131 + ngc 3923 & 1788 & 10.52 & 52.2 & 216 & 0.36 & 1583 & 22.6 & * 15667 & 10420 + ngc 4125 & 1356 & 10.57 & 58.4 & 229 & 0.08 & 1986 & 28.4 & 25129 & * 9366 + ngc 4374 & 1060 & 10.13 & 54.5 & 287 & 0.17 & 1333 & 19.0 & 6444 & * 3692 + ngc 4406 & -244 & 9.87 & 101.7 & 250 & 0.12 & 1333 & 19.0 & 5378 & * 4248 + ngc 4472 & 997 & 9.32 & 103.6 & 287 & 0.10 & 1333 & 19.0 & 5381 & * 3732 + ngc 4486 & 1307 & 9.52 & 103.6 & 361 & 0.10 & 1333 & 19.0 & 3612 & * 3612 + ngc 4494 & 1351 & 10.69 & 45.1 & 124 & 0.09 & 695 & 9.9 & 18616 & * 12679 + ngc 4552 & 340 & 10.84 & 30.0 & 261 & 0.18 & 1333 & 19.0 & * 11013 & 8623 + ngc 4621 & 410 & 10.65 & 45.5 & 240 & 0.10 & 1333 & 19.0 & * 6747 & 3413 + ngc 4636 & 938 & 10.20 & 101.1 & 191 & 0.12 & 1333 & 19.0 & 6457 & * 5137 + ngc 4649 & 1117 & 9.77 & 73.4 & 341 & 0.11 & 1333 & 19.0 & 15771 & * 7192 + ngc 4697 & 1241 & 10.03 & 73.5 & 165 & 0.13 & 794 & 11.3 & 11712 & * 9446 + ngc 5846 & 1714 & 10.67 & 82.6 & 278 & 0.24 & 2336 & 33.4 & 9636 & * 7739 + ic 1459 & 1691 & 10.88 & 34.0 & 308 & 0.07 & 2225 & 31.8 & 10358 & * 2563 + * * * * * * * * * * * * * * * * * * * * * * * * rrrrrrrrr ngc 1316 & 1037.99 & 1043.71 & y & 1.1e-14 & 2.0e-15 & 1.5 & 0.2 & y + ngc 1395 & 1037.84 & 1043.56 & ul & 1.5e-15 & & & & n + ngc 1399 & 1036.84 & 1042.55 & ul & 3.0e-15 & & & & n + ngc 1404 & 1038.63 & 1044.36 & ul & 3.0e-15 & & & & n + ngc 1407 & 1038.05 & 1043.78 & ul & 7.0e-16 & & & & n + ngc 1549 & 1036.13 & 1041.84 & y & 4.6e-15 & 8.0e-16 & 1.2 & 0.2 & n + ngc 3115 & 1034.41 & 1040.11 & ul & 1.5e-15 & & & & n + ngc 3379 & 1035.07 & 1040.77 & ul & 2.5e-15 & & & & n + ngc 3585 & 1036.75 & 1042.46 & ul & 1.0e-14 & & & & n + ngc 3607 & 1035.15 & 1040.86 & y & 1.6e-15 & 7.0e-16 & 0.5 & 0.1 & y + ngc 3923 & 1038.08 & 1043.81 & ul & 1.0e-15 & & & & n + ngc 4125 & 1036.60 & 1042.31 & ul & 8.0e-16 & & & & n + ngc 4374 & 1035.58 & 1041.29 & y & 4.0e-15 & 6.0e-16 & & & n + ngc 4406 & 1031.09 & 1036.78 & p & 3.8e-15 & 1.5e-15 & 1 & 0.2 & n + ngc 4472 & 1035.36 & 1041.07 & p & 3.2e-15 & 1.0e-15 & 0.4 & 0.1 & n + ngc 4486 & 1036.43 & 1042.14 & p & 7.0e-15 & 3.0e-15 & & & p + ngc 4494 & 1036.58 & 1042.30 & ul & 1.0e-15 & & & & n + ngc 4552 & 1033.10 & 1038.80 & p & 3.0e-15 & 1.0e-15 & 1 & 0.2 & n + ngc 4621 & 1033.34 & 1039.04 & ul & 3.0e-15 & & & & n + ngc 4636 & 1035.16 & 1040.87 & y & 3.9e-15 & 6.0e-16 & 1 & 0.2 & p + ngc 4649 & 1035.77 & 1041.49 & ul & 9.0e-16 & & & & n + ngc 4697 & 1036.20 & 1041.92 & ul & 1.4e-15 & & & & n + ngc 5846 & 1037.83 & 1043.55 & y & 3.0e-15 & 1.0e-15 & 0.6 & 0.2 & y + ic 1459 & 1037.75 & 1043.47 & y & 9.5e-15 & 4.0e-15 & 1.5 & 0.5 & y + rrrrrrrrrrr ngc 1316 & y & 38.85 & 0.78 & 40.79 & 0.35 & 0.15 & 40.77 & 0.58 & 0.20 & 120 + ngc 1395 & ul & 38.29 & 0.21 & 40.75 & 0.44 & 0.18 & & & & ul + ngc 1399 & ul & 38.24 & 0.19 & 41.15 & 0.94 & 0.22 & 41.16 & 0.77 & 0.48 & 2.5 + ngc 1404 & ul & 38.23 & 0.19 & 40.98 & 0.56 & 0.32 & 40.54 & 0.63 & 0.13 & 0.0039 + ngc 1407 & ul & 38.20 & 0.18 & 41.05 & 0.91 & 0.12 & 41.00 & 0.68 & 0.22 & 0.088 + ngc 1549 & y & 38.29 & 0.21 & 39.75 & 0.19 & 0.040 & 39.70 & 0.29 & 0.041 & ul + ngc 3115 & ul & 37.84 & 0.08 & 39.45 & 0.45 & 0.011 & 38.88 & 0.61 & 0.004 & 0.0012 + ngc 3379 & ul & 37.78 & 0.07 & 39.49 & 0.26 & 0.019 & 39.15 & 0.61 & 0.006 & 0.0024 + ngc 3585 & ul & 38.87 & 0.83 & 39.55 & 0.31 & 0.018 & 39.59 & 0.53 & 0.020 & ul + ngc 3607 & y & 38.30 & 0.22 & 40.53 & 0.37 & 0.095 & & & & 0.0069 + ngc 3923 & ul & 38.23 & 0.19 & 40.61 & 0.55 & 0.092 & 40.63 & 0.29 & 0.37 & ul + ngc 4125 & ul & 37.99 & 0.11 & 40.72 & 0.33 & 0.18 & 40.46 & 0.37 & 0.11 & 0.0025 + ngc 4374 & y & 38.46 & 0.32 & 40.80 & 0.70 & 0.11 & 40.72 & 0.50 & 0.16 & 2.95 + ngc 4406 & p & 38.37 & 0.26 & 41.72 & 0.67 & 0.53 & 41.35 & 0.67 & 0.24 & ul + ngc 4472 & p & 38.26 & 0.20 & 41.48 & 0.94 & 0.21 & 41.26 & 0.72 & 0.28 & 0.22 + ngc 4486 & p & 38.60 & 0.44 & 42.00 & 0.80 & 0.84 & 42.40 & 1.70 & 0.25 & 38.5 + ngc 4494 & ul & 37.19 & 0.02 & 38.99 & 0.30 & 0.005 & 38.78 & 0.48 & 0.004 & 0.0035 + ngc 4552 & p & 38.33 & 0.24 & 40.63 & 0.43 & 0.18 & 40.39 & 0.55 & 0.13 & 0.1 + ngc 4621 & ul & 38.24 & 0.19 & 39.50 & 0.63 & 0.007 & 39.25 & 0.63 & 0.007 & ul + ngc 4636 & y & 38.38 & 0.27 & 41.52 & 0.72 & 0.31 & 41.26 & 0.68 & 0.24 & 0.079 + ngc 4649 & ul & 37.73 & 0.06 & 41.19 & 0.82 & 0.18 & 41.09 & 0.76 & 0.32 & 0.029 + ngc 4697 & ul & 37.50 & 0.03 & 39.84 & 0.30 & 0.007 & 39.30 & 0.60 & 0.003 & ul + ngc 5846 & y & 38.90 & 0.87 & 41.72 & 0.73 & 0.61 & 41.55 & 0.63 & 0.53 & 0.021 + ic 1459 & y & 39.14 & 1.55 & 40.90 & 0.6 & 0.22 & 40.30 & 0.59 & 0.11 & 0.97 +

massive datasets in the gigabyte and terabyte range combined with the availability of increasingly sophisticated statistical tools yield analyses at the boundary of what is computationally feasible . compromising in the face of this computational burden by partitioning the dataset into more tractable sizes results in stratified analyses , removed from the context that justified the initial data collection . in a bayesian framework , these stratified analyses generate intermediate realizations , often compared using point estimates that fail to account for the variability within and correlation between the distributions these realizations approximate . however , although the initial concession to stratify generally precludes the more sensible analysis using a single joint hierarchical model , we can circumvent this outcome and capitalize on the intermediate realizations by extending the dynamic iterative reweighting mcmc algorithm . in doing so , we reuse the available realizations by reweighting them with importance weights , recycling them into a now tractable joint hierarchical model . we apply this technique to intermediate realizations generated from stratified analyses of 687 influenza a genomes spanning 13 years allowing us to revisit hypotheses regarding the evolutionary history of influenza within a hierarchical statistical framework . , .

1101.0959

influenza a continues to evade eradication resulting in ongoing economic and human cost . yearly epidemics are responsible for 36,000 deaths on average in the united states . three times in the past century global pandemics , including the infamous spanish influenza of 1918 , resulted in catastrophic mortality rates [ @xcite ] . influenza epidemiologists believe a future influenza pandemic is an imminent threat [ @xcite ] . nearly 400 documented transfers [ @xcite ] of the highly virulent and potentially pandemic [ @xcite ] h5n1 strain of avian flu to humans in addition to the recent development of h1n1 swine flu [ @xcite ] bolster the threat . the increasingly relevant necessity of preventing future influenza pandemics requires a clear understanding of the evolutionary mechanisms of influenza as it is the key to vaccine development [ @xcite ] . influenza a is a negative single - stranded rna virus composed of 8 segments that total approximately 13 kb in length and encode 11 proteins . the three largest segments encode polymerases pb1 , pb1-f2 , pb2 and pa all of which are involved in rna transcription and replication . the next three segments code for the two surface glycoproteins haemagglutinin ( ha ) and neuraminidase ( na ) as well as the nucleoprotein ( np ) . the two smallest segments encode the matrix proteins m1 and m2 and the nonstructural proteins ns1 and ns2 [ @xcite , @xcite ] . influenza a research typically focuses on the epitope - rich ha and na segments because they exhibit strong evolutionary selective pressure due to their direct interaction with the host immune system and are the primary determinants of the antigenic variation of influenza [ @xcite ] . the 16 ha and 9 na glycoproteins found in the avian reservoir , referred to as h1 to h16 and n1 to n9 , respectively , characterize and name the subtypes of influenza a [ @xcite ] . the evolutionary history of influenza a involves the interaction of a number of mechanisms including mutation and reassortment . approximately one random sequence mutation every replication cycle combined with the selective pressure on the surface glycoproteins results in an accumulation of point mutations on the ha and na segments termed antigenic drift . the influenza genome also evolves through reassortment in which two subtypes coinfect a single host cell and exchange segments . this exchange of genetic material can lead to an antigenic shift or the creation of a new , potentially lethal , subtype . reassortment between a virus in the avian reservoir and human influenza a resulted in the subtypes responsible for the asian and hong kong influenza pandemics [ @xcite ] and the current swine flu pandemic derives from a triple reassortment event [ @xcite ] . as critically important events in influenza evolution occur at the genome - level , complete genome analysis yields scientific insight that single segments can not afford . for example , @xcite clarify a perplexing question in the evolutionary dynamics of ha by considering the varying histories of each segment . analyzing 156 complete h3n2 viruses over a five - year time span from 1999 to 2004 , @xcite ( @xcite ) discover that while the fujian - variant ha segment has been co - circulating since at least 2000 , the variant only rises to dominance in 2002 after other segments within the influenza genome reassort and provide a synergistic background . this important reassortment event is only understood by studying the influenza genome in its entirety . a more recent study by @xcite emphasizes the importance of incorporating model parameter uncertainty in drawing conclusions about influenza evolution through a bayesian analysis of a truly massive dataset . @xcite compile 687 h3n2 influenza a full genomes sampled from new york over a twelve - year period . @xcite address a host of biologically and clinically relevant questions including : ( 1 ) are reassortment events coincident with shifts in ha antigenicity ? ( 2 ) do certain segments maintain greater genetic diversity ? ( 3 ) are the genetic histories of certain segments correlated ? however , due to a dearth of bayesian massive data techniques , computational constraints force @xcite to partition the data by stratifying on segment , using the data inefficiently , and drawing ad hoc conclusions about potential correlation . this current study rectifies the stratified analyses by fully capitalizing on the hierarchical nature of the influenza data and making formal inference after modeling the complete data in a single bayesian analysis . despite optimized algorithms for missing data integration [ @xcite ] , phylogenetic analysis of dna sequences is lengthy and computationally intensive . massive data measure in the gigabyte to terabyte range [ @xcite ] and are increasingly common [ @xcite , @xcite ] . this pervasiveness is particularly poignant in bayesian models with missing data and especially in bayesian models for stochastic processes where the dimensionality of the missing data can far outweigh the observed data . such is the case in the evolutionary reconstruction of @xcite . one strategy pertinent to massive data inference is stratification [ @xcite , @xcite ] , often undesirable because it comes shackled with the host of difficulties arising from subgroup analysis [ @xcite , @xcite ] . this identifies the direction that @xcite originally follow as they treat each of the eight segments independently . shared computer memory and communication latency between computers limit hopes for considering a proper hierarchical model across segments simultaneously through which to share information and learn about segment - to - segment correlation . even on state - of - the - art equipment , simulating sufficient realizations from posterior distributions conditional only on the data from a single segment , or what we refer to as `` stratified distributions , '' still compels one to devote weeks of computing time per segment . this huge computational investment raises a critical point regarding a massive dataset with hierarchical structure . often researchers perform preliminary analysis stratified by the exchangeable identifiers in the data simply because the statistical tools and computational resources exist for the stratified analysis . in attempting to fit the full hierarchical analysis , the ability to reuse the results from these suboptimal analyses represents a major savings in terms of time and resources and may even be the only feasible option . to this end , we examine the dynamic iteratively reweighting mcmc algorithm ( dyirma ) [ @xcite , @xcite ] . dyirma is based on the meta - analysis technique of using summary statistics from independent studies to infer a single hierarchical model [ @xcite , @xcite ] . instead of summary statistics , however , dyirma combines realizations from independent distributions using importance sampling and markov chain monte carlo ( mcmc ) in an iterative process . importantly , we can adopt dyirma to reuse realizations from preliminary stratified analyses . this desire to not waste intermediate realizations from the stratified analyses is particularly relevant in our influenza example because the realizations themselves require massive computing resources to generate . we further extend the insight of @xcite who combine intermediate realizations from two uncorrelated distributions . our extension is necessary to accommodate correlated sequence data sampled over a span of thirteen years and allows us to entertain a much richer collection of hierarchical models motivated by the science at hand . our scientific aim in this study is to create a joint hierarchical model that addresses the questions raised by @xcite regarding the evolutionary history of influenza a. to this end , the hierarchical model must account for an unknown correlation structure between segments and allow for a flexible time - course in the model response , for which we exploit gibbs variable selection ( gvs ) [ @xcite , @xcite ] to estimate a nonparametric response . the influenza a example illustrates that dyirma is a particularly flexible and valuable approach that reuses realizations via reweighting from computationally expensive distributions in a hierarchical framework . this widely applicable technique can be used to jointly model other independently generated , but in truth correlated , massive datasets . as a preview , the paper continues as follows : section [ sec : stratified ] describes the generation of realizations from the stratified analyses , section [ sec : sampling ] introduces the basic framework used to estimate genealogies . section [ sec : recycle ] relates the machinery necessary to combine these realizations to estimate the joint hierarchical model along with computational concerns . section [ sec : model ] reviews the hierarchical model proposed , prior distributions , mcmc sampling concerns , and various modeling extensions . sections [ sec : theresults ] and [ sec : discuss ] present results and conclude with a discussion . rambaut et al . ( @xcite ) compile aligned sequence data for coding regions of each of the eight segments of the influenza a genome from the influenza genome sequencing project ncbi database [ @xcite ] . these alignments contain all 687 h3n2 influenza a genomes available over the 12 influenza seasons between 1993 and 2005 . season 2002 yields no sequences as it was predominantly an h1n1 season . from these data , @xcite are most interested in estimating and formally comparing the times to most recent common ancestor ( tmrca ) of all the sequences sampled within each season for each segment . tmrca can be thought of as a measure of genetic diversity because evolutionarily distant present - day sequences converge to a genealogy with longer branch lengths and consequently a larger tmrca . to keep computation manageable , @xcite are forced to partition these data into independent blocks by segment . as the virus evolves through time , samples from different seasons are highly interrelated through their shared history . standard phylogenetic software packages account for this correlation . on the other hand , joint models across segments are less developed [ @xcite ] . consequently , initial analyses consider the segments independently , clouding conclusions about segment segment interactions important to influenza a evolution . rambaut s et al . ( @xcite ) analyses provide samples from the intermediate distributions of tmrcas given sequences from each individual segment . we are interested in the interaction of the evolutionary dynamics of influenza a segments over time and we use as our starting point realizations from these analyses stratified on segment . we let @xmath0 be the tmrca for segment @xmath1 and season @xmath2 and @xmath3 the sequence data for segment @xmath1 . from each of the stratified analyses , we tabulate samples @xmath4 or @xmath5 for all @xmath6 where @xmath7 indexes the mcmc sample , @xmath8 is the total number of mcmc samples , @xmath9 indexes the segment , and @xmath10 indexes the season . next , let @xmath11 be the matrix constructed from @xmath12 columns @xmath13 or @xmath14 rows @xmath15 , where @xmath13 is an @xmath16 vector with all samples of tmrca at iteration @xmath17 for season @xmath2 and @xmath15 is a @xmath18 vector for segment @xmath1 with all samples of tmrca at iteration @xmath17 for all @xmath12 seasons . these @xmath8 matrices @xmath11 are the intermediate samples from the stratified distributions provided by @xcite that we propose to recycle into a hierarchical model . to describe the construction of this stratified distribution , we first introduce some nomenclature . in brief , let @xmath19 be the genealogy composed of a bifurcating acyclic graph ( commonly called a topology ) that describes the relatedness of a set of sequences and a vector of edge weights for the edges in this topology . edges reflect the passage of time between bifurcation events and are also called branch lengths . estimates of @xmath20 different tmrca , @xmath21 , embed in @xmath19 because @xmath22 where @xmath23 is a deterministic mapping@xmath24 . here , @xmath21 represents a @xmath25 vector containing the tmrcas of interest and we use this general vector as the starting point before building up to @xmath26 . in order to describe @xmath23 , a brief introduction to the coalescent process is warranted [ @xcite ] . in the isochronous case , there are @xmath27 sequences sampled at the same time @xmath28 where @xmath29 represents the present - day . formation of a genealogy begins by randomly selecting two lineages at time @xmath30 , @xmath31 where @xmath32 indexes the coalescent event and @xmath33 is the total number of coalescent events . proceeding back in time , the inter - coalescent time between the @xmath32th and @xmath34th event is @xmath35 where @xmath36 are independent exponential random variables . let @xmath37 be the set of all @xmath27 taxa ; then the @xmath38 independent intercoalescent intervals yield an estimate @xmath39 , the tmrca of set @xmath37 . since each inter - coalescent interval is distributed as an exponential , the summation is a convolution of exponential distributions [ @xcite ] . we can estimate @xmath40 for any subset of taxa . let @xmath41 be a subset of the taxa , @xmath42 represent the first event in this subset which for the isochronous case occurs at @xmath43 , and @xmath44 the last so that @xmath45 . then @xmath46 is calculated similarly by identifying @xmath47 , the subtree formed within @xmath19 from the tips at @xmath28 and proceeding back in time to the last coalescent event occurring at @xmath48 for all taxa in @xmath49 . we then calculate @xmath50 . finally , as a nomenclature device , let @xmath51 be the set of all times of coalescent events identified in @xmath49 , @xmath52 , and let @xmath53 be the complement . then the genealogy can be decomposed into two disjoint sets of coalescent times as @xmath54 , a device that will prove useful in section [ sec : reweigh ] . one particularly interesting subset @xmath55 arises in the case of heterochronous data , namely @xmath56 or tmrca for all sequences sampled in a given season @xmath2 . because we have influenza a sequences sampled over time , our data are commonly called `` heterochronous . '' we know the exact date of sampling for the influenza a sequences and can extend the coalescent to reflect this additional information . elaborating on the description of the coalescent process above , the heterochronous case has two events of interest , coalescent and sampling , both of which can occur at multiple times . if there are @xmath57 sampling times , there are now a total of @xmath58 intervals , so in the heterochronous case , @xmath59 . let @xmath60 represent the times of coalescent or sampling events where @xmath28 is the most recent chronological sample and as before , the inter - event intervals are @xmath35 . for @xmath55 identify the earliest event @xmath42 and the last event @xmath44 and again we have @xmath61 . we can now fully describe the intermediate realizations available from @xcite . recall that our research goals require we extract estimates of @xmath62 for a specific influenza season @xmath2 where @xmath10 indexes the @xmath12 distinct seasons from a @xmath63 for segment @xmath9 . we accomplish this by allowing the subset @xmath64 to be the set of all taxa for a given influenza season @xmath2 . then @xmath65 is the summary statistic of interest for segment @xmath1 and season @xmath2 . finally , let @xmath66 contain all estimates of tmrca for a given segment and let @xmath67 be all estimates of tmrca for a given year . refer to figure [ fig : tree ] for a simple example . and extends back into the past until @xmath68 or the @xmath69 for all five sequences . inferred ancestral nodes between the samples are represented by circles . the black circle is the most recent common ancestor for both of the sequences sampled in the 1993 influenza season . we calculate @xmath70 by isolating the relevant branches , represented here by the purple subtree , and summing the inter - coalescent intervals , @xmath71 . similarly , the gray circle is the most recent common ancestor of all three sequences sampled in the 1994 influenza season and @xmath72 . ] although we have outlined how to generate our summary statistics of interest given a genealogy , further description is necessary regarding sampling from the distribution of the unknown genealogy @xmath63 conditional on the data @xmath3 for segment @xmath1 . this posterior distribution can be represented as @xmath73 where @xmath74 is the likelihood of the sequence data given the genealogy and other phylogenetic parameters @xmath75 that model sequence change over time and @xmath76 , @xmath77 , and @xmath78 are the prior distributions for the genealogy and phylogenetic parameters . the above decomposition specifies a marginal prior distribution on @xmath63 because @xmath79 . we take note of this marginal distribution because we wish to ultimately replace @xmath80 by a joint prior distribution @xmath81 in our hierarchical model . from @xcite , @xmath82 derives from a semiparametric relaxation of the coalescent process parameterized in terms of time - varying effective population size vector @xmath83 that follows a piecewise constant multiple - changepoint - process hyperprior distribution [ drummond et al . ( @xcite , @xcite ) ] . effective population size is meant to reflect amount of genetic diversity rather than census count [ @xcite ] and can be thought of as the average number of unique individuals that actually contribute genes to subsequent generations . investigators generate samples from @xmath84 using mcmc in the bayesian software beast [ drummond et al . ( @xcite , @xcite ) ] for each segment independently . due to the prominence the prior distribution @xmath82 plays in the iterative reweighting scheme , it is outlined in some detail as follows , where we drop the subscript @xmath1 for clarity . recall that @xmath85 are the times of events going into the past and @xmath86 are the inter - coalescent intervals . let @xmath87 be the number of lineages that exist in @xmath19 during a given inter - coalescent interval . we want to generate a sequence of effective population sizes of length @xmath88 where @xmath89 indexed by @xmath90 with time similarly partitioned into @xmath91 . essentially , we want to partition @xmath92 into @xmath88 groups @xmath93 where @xmath94 is constant between @xmath95 and @xmath96 . in the heterochronous case , the number of lineages can increase ( for a sampling event ) or decrease ( for a coalescent event ) so there are two events of interest that can change the number of lineages . these events are distinguished by the indicator function @xmath97 which indicates that @xmath32 is a coalescent event . @xcite specify the following heterochronous semiparametric coalescent prior distribution ( again , ignoring dependence on @xmath1 for clarity ) : @xmath98 where the function @xmath99 maps from the larger number of @xmath38 events to the @xmath88 groups , or in other words @xmath100 if @xmath30 is between @xmath101 and @xmath96 . finally , to complete the of the prior distribution , the first effective population size follows a scale - invariant prior distribution , @xmath102 [ @xcite ] and the remaining @xmath103 effective population sizes are distributed as exponential with scale parameter equal to the previous effective population size , @xmath104 @xmath105 . we make no claims about the appropriateness of this prior distribution choice . however , since the mean and variance of this prior distribution on @xmath106 grow with @xmath107 and lead to some difficulty later , we point out that @xcite provide a stable alternative with the joint skyride prior distribution on @xmath108 . now we have a foundation for understanding how genealogies are sampled and summarized using @xmath62 and what realizations from the intermediate distributions of the stratified analyses we have available . we wish to point out the benefits of the hierarchical model whose estimation we describe in the next section . these benefits include shrinkage estimators , a framework for statistical inference that accounts for correlation between strata , and models based on all available data . we now delve into how these independently generated estimates are combined into a joint statistical model reusing the preliminary realizations . in addition to the stratified realizations , the process of reweighting the stratified analysis samples of @xmath109 to generate the joint hierarchical posterior distribution through dyirma requires @xmath82 and the ability to evaluate the marginal prior distribution @xmath110 . we first present the dyirma machinery and then comment on how we extend it to accommodate this dataset . we save computational concerns regarding the calculation of @xmath110 for section [ sec : compute ] . recall that we employ the following decomposition of genealogy @xmath111 which allows us to relate the summary statistic @xmath109 to @xmath63 as @xmath112 the symbol @xmath113 identifies the stratified analyses and for notational consistency can be viewed as the forthcoming hierarchical model parameters , @xmath114 fixed at an arbitrary value . the subscript @xmath51 is dropped in the following equations to simplify notation . we have intermediate realizations of the multivariate vector of @xmath115 under the individual models generated during the initial stratified analysis . we combine these realizations into a single joint posterior distribution conditional on all of the sequence data , @xmath116 where @xmath117 and @xmath118 , by preferentially weighting samples that have a high likelihood under the hierarchical model relative to the probability of the prior distribution in the individual models . we make the following assumptions of conditional independence , namely that given @xmath21 the sequence data are independent of the parameters in the hierarchical model , or @xmath119 . we also assume that given the hierarchical parameters , the tmrcas @xmath109 , which we treat as exchangeable units , are independent or in other words we assume @xmath120 . given these assumptions , we can then use the following relationship : @xmath121 therefore we reuse all available @xmath8 realizations from @xmath122 by identifying the following importance weights which are calculated for every sample of the stratified distributions : @xmath123 the numerator of the weights , @xmath124 , is the conditional density under the hierarchical model . these hierarchical parameters are updated during each round of gibbs sampling requiring recalculation of the weights for each iteration . the proposal density , @xmath125 , is conditional on the parameters in the stratified analyses . as the variance of the estimate in ( [ eq : ira ] ) relies on the proposal density , some thought should be given toward selecting an appropriate prior distribution during individual analysis . generally it is desirable to have a proposal density with heavier tails than the numerator of the weights [ @xcite ] . as shown by @xcite and relying on the concept of importance sampling [ @xcite ] , we can solve for the conditional distribution of the @xmath109 s under the hierarchical model , which is specified as the following : @xmath126 where @xmath127 and @xmath128 is a degenerate distribution at @xmath15 . the weighted stratified realizations are sampled during each round of gibbs updates to generate samples from the hierarchical posterior distributions of @xmath109 conditional on the current values of the parameters in the hierarchical model , @xmath129 . these samples in turn are then used to update the parameters in the hierarchical model from @xmath130 to @xmath131 which completes another iteration of sampling and contributes the last remaining piece of machinery necessary to jointly model our intermediate realizations from stratified analyses . we now describe in detail the strategy we employed to calculate @xmath110 which is necessary for the calculation of weights . we include this description because for many scientifically interesting choices of the mapping @xmath23 , such as jointly modeling tmrcas , @xmath110 remains intractable in analytic form . fortunately , standard machinery already exists to draw simulants from this distribution , namely the mcmc sampler exploited to generate the stratified samples . a single additional run of this sampler without data provides all ingredients necessary to tackle the seemingly computationally intractable joint inference . each partitioned dataset assumes identical prior distributions so a single simulation of @xmath110 suffices for the reweighting of all the stratified distributions . two successful approaches to estimating marginal distributions frommcmc samples are multivariate kernel density [ @xcite , @xcite ] and importance weighted marginal density estimation ( iwmde ) [ @xcite ] . iwmde proceeds by identifying a weighting function and sampling from a weighted ratio of the likelihood at a given value of the marginal of interest ( here @xmath132 ) and realizations of the full joint distribution . the weighting function ideally has a similar shape to the unknown conditional marginal density , a catch-22 circumvented with multivariate kernel density estimation . as a consequence of the strictly nonnegative support of @xmath109 we control for potential boundary effects during kernel density estimation by using a gamma kernel [ @xcite ] that removes the boundary effect and has the best mean integrated squared error among all nonnegative densities . for computational ease , we select the bandwidth using a multivariate adapted scott s rule - of - thumb in which the bandwidth for the @xmath133th variable , @xmath134 , is @xmath135 [ @xcite ] where @xmath136 is the univariate sample standard deviation and bandwidths are allowed to vary for each univariate kernel resulting in a multiplicative kernel [ @xcite ] . the importance of selecting an appropriate prior distribution in a bayesian framework is a topic of considerable depth [ @xcite , @xcite ] . even the less contentious tactic of selecting a noninformative prior distribution by placing a uniform distribution over the parameter space can prove to be subjective [ @xcite , @xcite ] . in the case where the likelihood function overwhelms the prior distribution , a potentially informative uniform prior distribution is rendered noninformative and specification of the prior distribution is deceptively unimportant . when @xcite specify a multiple changepoint prior distribution , the effective population sizes , @xmath137 , were further constrained to lie between 0 and 120,000 . this truncated prior distribution is well outside of the desired range of the values of @xmath109 , a fact that is inconsequential when sampling from the distribution conditional on the data . however , this creates difficulties for evaluating the prior distribution at the realized values of the distribution conditional on the stratified data as they mostly lie well outside the region of the kernel density estimate . luckily , when the maximum population size is constrained to be lower than 120,000 and coverage extends to the region of interest the kde of the prior distribution reveals a relatively flat density surface . . we want the value of the prior density used during stratified analyses for the intermediate realizations sampled in the range illustrated in gray dot and black line . we have a kde constructed with coverage from realizations of the prior distribution we generated in the range in black line . the gray dot region , or the region of interest the kde is forced to extrapolate , expands as the maximum allowable @xmath92 , @xmath138 , increases . the density with @xmath139120,000 approaches the flat line illustrated with the perhaps more reasonable @xmath140 . ] as the maximum is gradually increased to 120,000 , this surface decreases in the value of the density but remains relatively flat . this is illustrated in figure [ fig : kde ] , which shows a representative ( @xmath141 ) prior density predicted by kde conditional on the mean of the other parameters for different maximum allowable @xmath137 sizes along with realizations from the prior distribution . the range of the values for the realizations from the prior distribution we wish to evaluate are indicated in the lower range of the density and the realizations used to predict the kde are in the upper range . taken together , this indicates the region the kde is forced to extrapolate . as the addition of a constant value on the log scale does not affect the weights , a constant density equivalent to some arbitrary @xmath142 can be selected . for computational ease it is desirable for @xmath142 to be similar in range to the density under the joint hierarchical model . our methods enable us to conceive of a model that tests a greater range of hypotheses than those based on a single stratum . we revisit each of the statements of @xcite aimed at understanding the evolutionary history of influenza a with the advantages afforded by a hierarchical framework . we construct our model out of three basic elements . the first modeling element identifies seasons with a significant change in tmrca from the previous year using gvs on the timepoints . a significant increase in tmrca between timepoints suggests a reassortment event , whereas a significant decrease suggests a selective sweep . the second element introduces fixed segment effects that test whether certain segments have a higher tmrca and therefore greater genetic diversity than others across time . finally , we address correlation between the segments by exploring constrained variance matrices . to recycle random samples generated under existing stratified analyses , we implement an additional dyirma step during each round of gibbs sampling of the joint model parameters . we first build up to the biologically motivated mean structure of the model containing gvs and fixed segment effects with an independent variance matrix and then introduce extensions for modeling the variance structure . at the most basic biological level , we must identify significant changes in tmrca between influenza seasons and test if these correspond with shifts in ha antigenicity . we accomplish this with gvs as parameterized by @xcite where the outcomes of the regression are the tmrcas and season effects represent potential predictors . note that this analysis would not be possible using a single segment as there would be insufficient degrees of freedom . the goal of bayesian stochastic search variable selection is to identify the underlying generative model @xmath143 from the set of all possible models @xmath8 . if @xmath144 is the total number of possible predictors in the regression model , the model space has dimension @xmath145 , an arduous dimension from which to draw inference . @xcite bypass this task by introducing indicator variables , @xmath146 , that identify the potential predictors of the outcome variable . to clarify , if the @xmath2th predictor has a marked effect on the outcome , the posterior probability that the corresponding indicator variable is one , @xmath147 , is high . on the other hand , if the @xmath2th predictor is not critical , the posterior probability that the corresponding indicator variable is zero , @xmath148 , is high . this implies that estimates of @xmath149 clarify which timepoints correspond to significant changes in tmrca . in order to proceed , we must introduce some additional nomenclature . let @xmath150 be the @xmath151 additive design matrix for all seasons where we remove the intercept in order to avoid overparameterization . in other words , @xmath152 is a row vector where the first @xmath153 entries are 1 and the remaining @xmath154 are 0 s . then @xmath155 is made of @xmath14 identical rows of @xmath152 and stacking the @xmath12 matrices of @xmath156 generates @xmath150 . let @xmath157 equal the diagonal matrix @xmath158 and @xmath159 be the unknown season effect sizes . the flexible gvs - induced mean time - course for mrcas for all segments at season @xmath2 becomes @xmath160 and is identical for all segments @xmath1 . we select the following conjugate , independent , and noninformative priors distributions : @xmath161 and @xmath162 where @xmath163 , and @xmath164 are hyperprior constants . estimation of @xmath149 addresses questions about the evolution of influenza data over time and clarify whether shifts in ha antigenicity correlate with significant changes in tmrca . modeling the segment effect highlights the importance of jointly modeling the influenza genome in concert in order to draw meaningful inference about similarities and differences in their evolutionary histories . segment effects identify consistent differences in tmrca over time and can test the hypothesis that np has higher genetic diversity than ha . we also garner indirect information regarding the unresolved physical location of segments within the influenza a genome because we can resolve the correlation between segments and highly correlated segment histories are consistent with close proximity . multidimensional scaling ( mds ) suggests a relationship with decreasing intensity among the following three groupings : ( 1 ) \{ha , m1@xmath1652 } , ( 2 ) \{ns1@xmath1652 , np } , and ( 3 ) \{pa , pb1 , pb2 } [ @xcite ] . the na segment is not grouped with any other segments . let @xmath166 be unknown segment effects . then we model @xmath167 where in its most general form @xmath168 is assumed to be an @xmath169 unstructured ( uns ) covariance matrix . conjugate prior distributions for this portion of the model are @xmath170 for all @xmath1 and @xmath171 where @xmath172 , and @xmath173 are hyperprior constants and @xmath174 is the inverse of the hyperprior constant @xmath175 scaling matrix . when combined with the prior distributions above , drawing realizations from the conditional posterior distributions for @xmath176 , and @xmath157 helps address questions about segment effects , significant timepoints , and segment correlation simultaneously . we now specify how to draw mcmc samples from the complete model . let @xmath177 contain the unknown parameters of the hierarchical model , of which we wish to draw inference . recall that we specify the conditional distribution for @xmath109 in ( [ eq : dyirmacond ] ) , which shows how we obtain realizations from @xmath178 by reweighting samples from @xmath179 . therefore to estimate the parameters from this complete model using gibbs sampling , we use dyirma during each gibbs cycle over @xmath180 . we have stated the denominator of the weights previously as the predictions from the kde of the stratified prior distributions that do not depend on @xmath114 and hence are constant during mcmc sampling . the numerator is updated at each iteration of mcmc and is simply the density of the vector of tmrcas given the parameters in the hierarchical model , a straightforward way to determine multivariate normal . at this stage we are replacing the standard coalescent prior distribution with a normal prior distribution , a trade - off that yields straightforward interpretability as it allows us to directly test @xcite hypotheses , in addition to a simple computational implementation . for details on the other update steps for @xmath181 and a schematic of sampling refer to the supplementary material [ @xcite ] . to identify segments with similar evolutionary histories , several constraints to @xmath168 may provide more effective estimates . we explore an independent ( ind ) parameterization , such that @xmath182 , with marginal variance @xmath183 unknown that implies the evolutionary history of segments is not correlated . this specification allows inference to focus on the segment effects and has the additional advantage of substantially decreasing the number of inferred parameters . also informative is compound symmetry ( cs ) which gives a general estimate of correlation between segments and provides a model nested within uns to test for similar levels of correlation between segments . cs implies that the evolutionary histories of all segments are correlated with the same strength . finally , autoregressive first order ( ar1 ) and tridiagonal ( tri ) structures with an estimable ordering of the segments directly identifies which segments have similar evolutionary histories . the motivation for nonexchangeable structures relies on the reasoning that segments with similar evolutionary histories have higher correlation than those with dissimilar histories . for the cs model , we modify the gibbs sampling by replacing the step for @xmath168 with a metropolis hastings step . let @xmath184 where @xmath185 is the segment correlation and assume prior distributions @xmath186 and @xmath187 where @xmath188 , and @xmath189 are hyperprior constants . refer to the supplementary material [ @xcite ] for further details on this modification to sampling . segments are not exchangeable in the tri and ar1 parameterizations of the covariance matrix . in ar1 , nearest - neighbor segments in the covariance matrix have higher correlation than those further apart . in tri , the structure is more restrictive with segments more than one neighbor away from each other having no correlation . as the ordering of the segments @xmath1 is not known and of paramount scientific interest , we estimate the labeling or ordering of the segments within the covariance matrix by parameterizing a permutation vector @xmath190 where @xmath191 indexes the different possible permutations . sampling @xmath190 requires an additional metropolis hastings step . we propose @xmath192 by randomly swapping two positions in the current permutation @xmath190 . we accept this proposal with probability @xmath193 where @xmath194 is @xmath195 assuming all permutations occur with equal probability , the terms @xmath196 and @xmath197 cancel . we proceed with caution when drawing inference due to label switching [ @xcite ] . instead of focusing inference on the permutation itself , we concentrate on the posterior probability that a given pair of segments are nearest neighbors , suggesting these segments have similar evolutionary histories . our reasoning is that because only nearest neighbors are correlated , segments with similar evolutionary histories have a high posterior probability of adjacent positions within the permutation vector conditional on the correlation @xmath185 being positive . this model that includes estimation of @xmath190 ( tri - p ) clarifies which evolutionary histories of segments are strongly correlated , which in turn has implications regarding structural interaction . rambaut et al . ( @xcite ) have run stratified analyses for each of the eight influenza segments . each analysis required an exhausting 23 weeks on high - end computers to approximate the stratified distributions @xmath84 via mcmc . each mcmc chain runs for @xmath198 iterations and subsampling every @xmath199 iterations yields @xmath200 approximately independent samples from these stratified distributions . we recycle these precomputed random samples to fit our hierarchical model that corrects for the stratification using the bayesian machinery described in sections [ sec : genomic ] and [ sec : recycle ] . in particular , we implemented the dyirma gibbs sampling scheme in cross - platform java . we simulate three independent mcmc chains for each hierarchical model for @xmath201 iterations with 10% burn - in and a 10-fold thinning . each chain takes only approximately five hours to run on a mid - end desktop computer , representing a compelling and efficient alternative to fitting a joint hierarchical model starting from the sequence data . we assess the combined chains from three independent chains via several convergence criteria including trace plots , histograms , geweke s convergence diagnostic [ @xcite ] , and rhat [ @xcite ] . table [ tab : tmrca ] presents mean time - course estimates for the across - segments independent ( ind ) and tridiagonal with permutation ( tri - p ) models . an additional subscript of ind or tri - p clarifies the model for the parameter estimate . the additive parameterization of the design matrix results in parameter estimates that reflect an increase or decrease from the previous season . table [ tab : tmrca ] therefore reports both the relative posterior conditional mean of @xmath202 , @xmath203 , that reflects the relative change in tmrca ( from the previous season ) and , what we term , the absolute posterior conditional mean or posterior conditional mean of the segment effect average added to the cumulative posterior mean of @xmath202 , @xmath204 where @xmath205 . additionally , we report the posterior mean estimates for @xmath149 that reflect the posterior probability that @xmath202 is included in the model . @ld1.6d2.6d1.2@ & & & + * year * & & & + & & + 1994 & 0.143 & 0.166 & 2.24 + 1995 & 0.113 & 0.110 & 2.35 + 1996 & 0.00667 & 0.00327 & 2.35 + @xmath206 & 0.00115 & 0.0104 & 2.34 + @xmath207 & 0.0748 & -0.0665 & 2.30 + 1999 & 0.0207 & -0.0128 & 2.28 + 2000 & 0.0133 & -0.00605 & 2.27 + 2002 & 0.207 & -0.242 & 2.03 + 2003 & 1.00 & 3.02 & 5.05 + @xmath208 & 0.00852 & 0.00538 & 5.05 + @xmath209 & 1.00 & -3.86 & 1.20 + & & + 1994 & 0.0500 & 0.0568 & 2.11 + 1995 & 0.0356 & 0.0326 & 2.14 + 1996 & 0.00593 & 0.00235 & 2.14 + @xmath206 & 0.00556 & 0.00122 & 2.14 + @xmath207 & 0.00100 & -0.00511 & 2.14 + 1999 & 0.00815 & -0.00346 & 2.13 + 2000 & 0.00852 & -0.00268 & 2.13 + 2002 & 0.0437 & -0.0456 & 2.09 + 2003 & 1.00 & 2.89 & 4.96 + @xmath208 & 0.0159 & 0.0118 & 4.98 + @xmath209 & 1.00 & -3.79 & 1.18 + looking first at the ind model , an indicator estimate of 1 at season 2003 suggests that there exists a significant difference between the average segment time to mrca , @xmath40 , at influenza season 2002 and that of 2003 because the posterior probability that the inclusion of a regression parameter captures this difference approaches 1 . this decisive support of a significant jump in @xmath40 suggests an infusion of genetic diversity and is consistent with a _ reassortment event_. similarly , the posterior probability for the indicator representing the shift between seasons 2004 and 2005 also approaches 1 . however , in this instance , as the negative sign of the estimate for @xmath210 reflects , this suggests a decrease in genetic diversity which is consistent with a _ selective sweep_. looking more closely at the ind model , the 2003 shift corresponds to an increase in @xmath40 with a posterior mean regression parameter estimate of @xmath211 and 95% bayesian credible interval ( bci ) @xmath212 . the 2005 shift suggests a @xmath213 @xmath214 decrease in @xmath40 and is concomitant with the fu02-ca04 ha antigenic shift . no other indicators have posterior probabilities greater than 0.95 . similarly , the tri - p model furnishes strong support for including seasons 2003 ( @xmath215 ) and 2005 ( @xmath216 ) and shows the similar pattern of jump in @xmath40 of @xmath217 @xmath218 followed by decrease of @xmath219 @xmath220 . the increase in genetic diversity at season 2003 followed by a decrease at 2005 clearly identify themselves in the third column of table [ tab : tmrca ] that shows that the posterior conditional segment means for the seasons between 1994 and 2002 as around 2 but jumps to nearly 5 in seasons 2003 and 2004 before undergoing a decline in 2005 and decreasing to around 1.2 . figure [ fig : fitted ] reiterates this finding . in general , the tri - p indicator probabilities are closer to 0 or 1 than in the ind model . for example , the posterior probability of the season 2002 shift @xmath221 is less than that for the ind model @xmath222 . one explanation for the differences between the two models is that the parameter estimate standard errors are generally reduced for the tri - p model . 2 , and ns1@xmath1652 ) of h3n2 influenza a listed from largest to smallest . samples from the hierarchical posterior distributions are from the tridiagonal with permutations ( tri - p ) model . twelve seasons are depicted with season 2001 missing due to an h1n1 dominant season . segment distributions are staggered for clarity . the y - axis represents tmrca and ha antigenicity ( be93 , wu95 , sy97 , fu02 , and ca04 ) is indicated across the top . hierarchical posterior distributions exhibit shrinkage toward the mean relative to the stratified . ] there exists a wide number of covariance structures that are biologically interesting to explore within our framework . specifically , we also consider the unstructured ( uns ) and first order autoregressive with permutation ( ar1-p ) models . although we observe a similar pattern of increase and decrease in tmrca in 2003 and 2005 , these models that allow a high level of correlation between the segments lose identifiability of the indicator variables across time , presumably because the variation present in the observations is used to model the correlation instead of the mean structure . combined with the fact that inference focuses on the mean structure , we continue to exclusively discuss results for sparse covariance matrices such as those used in the ind and tri - p models . summarizing the posterior probability of a model that contains all of the antigenic shifts addresses our question regarding whether ha shifts are concomitant with significant changes in _ t_. antigenic shifts occur at 1997 ( be93-wu95 ) , 1998 ( wu95-sy97 ) , 2004 ( sy97-fu02 ) , and 2005 ( fu02-ca04 ) and these seasons are superscripted in table [ tab : tmrca ] . we refer to these seasons from now on as shifts 1 , 2 , 3 , and 4 , respectively , and use these shift numbers as subscripts on a given model @xmath8 to clarify which model variables are included . for example , @xmath223 describes a model where both the first and third antigenic shifts are included @xmath224 and @xmath225 . as we have already noted , the posterior probability that the fu02-ca04 shift in 2005 is included in both the ind and tri - p models approaches 1 . however , the posterior probability of this being the only antigenic shift selected is also large because @xmath226 and @xmath227 . the posterior probability of a model with all four antigenic shifts , @xmath228 , approaches zero for both tri - p and ind . this strongly suggests that ha antigenic shifts are not strictly concomitant with significant changes in tmrca from the previous season after correcting for the correlation structure of the segments . @xcite are unable to correct for the correlation structure across segments . given the limited time series data available relative to the number of antigenic shifts , more data would certainly enhance the understanding of the relationship between reassortment events , selective sweeps , and ha antigenic shifts . [email protected]@ & & + & + pb2 & 2.51 & ( 1.30 , 3.40 ) + pb1 & 2.65 & ( 1.42 , 3.49 ) + pa & 2.11 & ( 0.888 , 2.96 ) + @xmath229 & 1.37 & ( 0.161 , 2.17 ) + @xmath230 & 2.53 & ( 1.30 , 3.36 ) + na & 2.21 & ( 1.01 , 3.02 ) + m1@xmath1652 & 1.39 & ( 0.244 , 2.19 ) + ns1@xmath1652 & 1.78 & ( 0.585 , 2.62 ) + & + pb2 & 2.35 & ( 1.24 , 3.17 ) + pb1 & 2.52 & ( 1.29 , 3.37 ) + pa & 2.08 & ( 1.03 , 2.91 ) + @xmath229 & 1.54 & ( 0.381 , 2.47 ) + @xmath230 & 2.46 & ( 1.25 , 3.28 ) + na & 2.20 & ( 1.04 , 3.04 ) + m1@xmath1652 & 1.37 & ( 0.168 , 2.44 ) + ns1@xmath1652 & 1.88 & ( 0.770 , 2.79 ) + to more thoroughly unravel the evolutionary history of influenza a , we need to address the relative level of evolutionary diversity between segments . to this end we include segment - specific effects whose posterior estimates we summarize in table [ tab : fixed ] for both the ind and the tri - p models . we report both the posterior mean and the 95% bcis with the range of the tri - p intervals generally slightly reduced from the ind , perhaps indicative of the tri - p being a more appropriate covariance structure with which to model the correlated parameters of the different segments . pb1 returns the highest posterior mean estimates of @xmath40 which at season 1993 for ind is @xmath231 ( 1.30 , 3.40 ) and for tri - p is @xmath232 @xmath233 . ha and m1@xmath1652 yield the lowest posterior means of @xmath40 which , also given at season 1993 , are @xmath234 @xmath235 and @xmath236 @xmath237 respectively for ind and @xmath238 @xmath239 and @xmath240 @xmath241 respectively for tri - p . this implies that genetic diversity is maintained longer in pb1 than ha and m1@xmath1652 with the difference between these segments for @xmath40 on the order of an entire year . we go further than @xcite who simply observe the differences in the @xmath40 of different segments of influenza a by formally testing whether certain segments maintain greater diversity than others . one relationship of particular interest , explored but not formally tested by @xcite , is the comparison between np and ha , superscripted in table [ tab : fixed ] . when assuming equal prior probability of both outcomes ( 0.5 each ) , the bayes factors ( bf ) of whether the @xmath40 of np is greater than ha are 179 for the ind model and 16.1 for tri - p . this means that the posterior probability of np being greater than ha in the ind model is 0.994 and 0.941 in the tri - p , providing strong support for the hypothesis that np maintains greater genetic diversity than ha . finally , we approach teasing out the correlation between segments by estimating the posterior distribution of all possible segment order permutation within the tridiagonal covariance matrix in the tri - p model . the three groupings of \{ha , m1@xmath1652 } , \{np , ns1@xmath1652 } , and \{pa , pb1 , pb2 } originally posited by @xcite arise from ad hoc multidimensional scaling results . for notational convenience , we now refer to these segments by number and place them in alphabetical order as 1 ( ha ) , 2 ( m1@xmath1652 ) , 3 ( na ) , 4 ( ns1@xmath1652 ) , 5 ( np ) , 6 ( pa ) , 7 ( pb1 ) , and 8 ( pb2 ) . this notational device means that we refer to a model with both the \{ha , m1@xmath1652 } and \{pa , pb1 , pb2 } grouping as @xmath242 . note that there is no implied ordering of the remaining unlisted segments so that @xmath243 and @xmath244 , say , have some overlapping groupings . there are seven pairs of neighbors possible in the tridiagonal matrix because the two segments occupying the corners of the diagonal of the covariance matrix are restricted to having a single neighbor and the middle six segments can each have two neighbors . we are interested specifically in positive correlation between segments as we want to clarify which segments have similar evolutionary histories . with this purview , we focus on results conditional on the correlation @xmath185 being greater than 0.2 ; this provision occurs with approximately 0.593 posterior probability , but eases interpretation . in general , we assume there are two segments @xmath1 and @xmath245 where @xmath246 , and @xmath247 which are selected as neighbors . we summarize the posterior probability of models in which two segments are neighbors for our subset of interest , @xmath248 , in table [ tab : permute ] . the posterior probability of the strongly hypothesized \{ha , m1@xmath1652 } group is very high , @xmath249 . the posterior probability of the \{pb1 , pb2 } pairing is also high , @xmath250 . previously unidentified as a potential pairing is \{np , pb2 } , @xmath251 as well as \{np , pa } , @xmath252 which implies that the np segment might be just as strongly aligned with the \{pa , pb1 , pb2 } grouping as with ns1@xmath1652 because the posterior probability of the \{np , ns1@xmath1652 } grouping is similar , @xmath253 . @lcd1.7d1.7d1.3d1.4d1.3ll@ & & & & & & & & + * pb2 * & & 0.493^\{78 } & 0.297^\{68 } & 0.181 & 0.378 & 0.123 & 0.185 & 0.165 + * pb1 * & & & 0.284^\{67 } & 0.151 & 0.257 & 0.313 & 0.0990 & 0.160 + * pa * & & & & 0.192 & 0.313 & 0.230 & 0.173 & 0.290 + * ha * & & & & & 0.0620 & 0.154 & 0.768@xmath254 & 0.157 + * np * & & & & & & 0.345 & 0.0880 & 0.335@xmath255 + * na * & & & & & & & 0.222 & 0.283 + * m1@xmath2562 * & & & & & & & & 0.304 + * ns1@xmath2562 * & & & & & & & & + again , we formally test these hypothetical groupings against the null hypothesis that all permutations of the tridiagonal covariance matrix are equally likely and summarize the results in table [ tab : bf ] . we make these calculations unconditional on the correlation @xmath185 . notice in column 2 of table [ tab : bf ] , the posterior probability of grouping \{ha , m1@xmath1652 } is 0.624 and has a prior probability of 0.0357 leading to a significant bf of 17.5 . again , \{np , ns1@xmath1652 } has a weaker bf of 8.50 lending some doubt to the hypothesis that these two segments have similar evolutionary histories . finally , the posterior probability that all three pairings are found is 0.0741 which , given all models are thought a priori to have equal probability , has a prior probability of 0.000600 . this leads to a very decisive bf of 124 which strongly supports rambaut s et al . ( @xcite ) hypothesized groupings of segments . @ld2.4d1.4d2.4d3.6@ & & & & + posterior odds & 0.624 & 0.304 & 0.195 & 0.0741 + prior odds & 0.0357 & 0.0357 & 0.0179 & 0.000600 + bayes factor & 17.5 & 8.50 & 10.9 & 124 + we have already demonstrated the ease with which bf can be assigned to competing joint hierarchical analysis models once we have recycled the realizations from stratified analyses into a single statistical framework . however , an additional advantage is that this hierarchical posterior distribution demonstrates shrinkage toward the mean when compared against the stratified results , leading to more sensible segment - specific estimates . figure [ fig : fitted ] displays the marginal hierarchical posterior distributions superimposed on the stratified results for all eight segments over time . in figure [ fig : fitted ] we demonstrate that overall the direction of shrinkage estimates from the stratified mean ( black ) to the hierarchical mean ( color ) draws toward the grand mean for that time . shrinkage toward the mean especially tempers outliers , a phenomenon illustrated in figure [ fig : box ] . focusing on 1993 , this season emits positive outliers in pb1 and na . season 1997 has positive outliers in pb2 and pb1 and 2004 yields a negative outlier in ha and a positive outlier in np . in 1993 , coverage drastically shifts down for pb1 and na as well as shifts up for m1@xmath1652 and ns1@xmath1652 . in season 1997 , coverage greatly decreases and shifts downward for the outlier pb2 and in season 2004 , coverage shifts up for ha and shifts down for np . figure [ fig : fitted ] also illustrates further advantages of this technique , including the reduced credible intervals for the hierarchical distribution relative to the stratified . finally , especially apparent in this figure is a cautionary reminder of the wide variability present in the inappropriately independent , stratified results . for example in 1993 , the stratified estimate uncertainty of pb1 spans around 3 years whereas that for pb2 spans around 0.5 . these differences in variability mitigate when basing conclusions on the point estimates of the stratified results . therefore the advantages of reusing the stratified analyses in a joint model lie not only in the ability to assign across - segment bfs and the incorporation of highly desirable shrinkage estimators that lead to improved estimation [ @xcite ] , but also in enhanced modeling capabilities that more accurately represent the variability in the segments . increasing dataset sizes are engulfing the scientific community [ @xcite ] demanding novel approaches to statistical analysis . while the introduction of gpu programming to the statistical community promises solutions in the near - future [ @xcite ] , the daunting task of analyzing these massive datasets is currently made realistic by partitioning them into smaller , more tractable sizes . this stratification , while facilitating fast estimation , results in overparameterization and ignores the correlation between parameters across strata . additionally , stratification fails to profit from the massive amounts of data available because parameters are estimated from siloed strata , removed from the implicit context that motivated the initial data collection . ideally , given no computational constraints , related and exchangeable groups are represented by a hierarchical model . this framework efficiently pools information across groups while accounting for the correlation between them . this single unified model also makes it easy to draw dataset - wide inference . finally , hierarchical models lead to improved estimators due to shrinkage toward the mean ; this well - known phenomenon is termed stein s paradox [ @xcite ] . perhaps a more familiar approach to constructing this full hierarchical model is sequential monte carlo ( smc ) [ @xcite ] . @xcite , @xcite , and @xcite use particle filtering as a smc solution to the massive data problem , building up the posterior distribution of the complete data by incrementally introducing a small number of data into the posterior distribution using importance sampling [ @xcite ] . this form of smc is highly effective for linearly organized data such as time series observations but is inappropriate for data divided into exchangeable groups . further , particle filters do not completely recycle the preliminary analyses . our methods in this paper create a new strategy , combining the advantages of stratification , namely speed , with the statistical framework of hierarchical modeling . any hypothesis addressed in the subpar stratified model can be reused , benefiting from assigned measures of statistical certainty . our methods capitalize on the intermediate realizations from stratified analyses , recycling them into the hierarchical model by reweighting via importance sampling . from the standpoint of the evolutionary history of influenza a , @xcite are ambitious in their goal of understanding individual segments within the larger context of the complete genome . our methods enable us to revisit @xcite s conclusions with the insight afforded by a hierarchical statistical framework . we find that for these data , the tri - p model is quite sufficient for our re - examination of the biological questions . however , in some circumstances it may be necessary to allow for some small degree of correlation between segments that are not nearest neighbors . one way to accommodate this correlation is by assigning a prior inverse - wishart distribution to @xmath168 . our approach is to center the inverse - wishart distribution on the structured covariance matrix [ as in @xcite ] . the degrees of freedom of the inverse - wishart provide a tuning parameter . as the degrees of freedom go to infinity the extra correlation goes to zero and we recover our original model . the use of this model requires that we replace the gibbs sampling step for @xmath168 with a metropolis hastings step , and so adds some computational burden . the applications of reusing , recycling , and reweighting are limited only by the biological questions of interest . this flexible framework has far - reaching value into areas such as resequencing and phylogeography [ @xcite ] , in other words , situations where computational complexity forces data partitioning , preventing the more appropriate hierarchical model . from an applied statistician s perspective , this technique delivers a much needed strategy for analyzing massive datasets . we are thankful to robert weiss for modeling discussion and suggestions and to eddie holmes and andrew rambaut for generously providing the results from their analyses and feedback . interested readers may request the source - code for this project from j. a. tom . allison , d. b. , visscher , p. m. , rosa , g. j. m. and amos , c. i. ( 2009 ) . statistical genetics and statistical genomics : where biology , epistemology , statistics , and computation collide . statist . data anal . _ * 53 * 15311534 . anderson , c. ( 2008 ) . the end of theory : the data deluge makes the scientific method obsolete . _ wired _ * 16*. available at http://wired.com/science/discoveries/magazine/16-07/pb_theory[http://wired.com/science/discoveries/magazine/]http://wired.com/science/discoveries/magazine/16-07/pb_theory[16-07/pb_theory ] . boscardin , w. j. and zhang , x. ( 2004 ) . modeling the covariance and correlation matrix of repeated measures . in _ applied bayesian modeling and causal inference from incomplete data perspectives _ ( a. gelman , ed . ) 215226 . wiley , chichester . cressie , n. , olsen , a. and cook , d. ( 1997 ) . massive data sets : problems and possibilities , with application to environmental monitoring . in _ massive data sets : proceedings of a workshop _ 115119 . national academy press , washington , dc . doucet , a. , de freitas , n. and gordon , n. ( 2001 ) . an introduction to sequential monte carlo methods . in _ sequential monte carlo methods in practice _ ( a. doucet , n. de freitas and n. gordon , eds . ) 313 . springer , new york . drummond , a. j. , nicholls , g. k. , rodrigo , a. g. and solomon , w. ( 2002 ) . estimating mutation parameters , population history and genealogy simultaneously from temporally spaced sequence data . _ genetics _ * 161 * 13071320 . geweke , j. ( 1992 ) . evaluating the accuracy of sampling - based approaches to calculating posterior moments . in _ bayesian statistics , vol . 4 _ ( j. m. bernado , j. o. berger , a. p. dawid and a. f. m. smith , eds . ) 169193 . oxford univ . press , new york . ghedin , e. , sengamalay , n. a. , shumway , m. , zaborsky , j. , feldblyum , t. , subbu , v. , spiro , d. j. , sitz , j. , koo , h. , bolotov , p. , dernovoy , d. , tatusova , t. , bao , y. , st george , k. , taylor , j. , lipman , d. j. , fraser , c. m. , taubenberger , j. k. and salzberg , s. l. ( 2005 ) . large - scale sequencing of human influenza reveals the dynamic nature of viral genome evolution . _ nature _ * 427 * 11621166 . holmes , e. c. , ghedin , e. , miller , n. , taylor , j. , bao , y. , st george , k. , grenfell , b. t. , salzberg , s. l. , fraser , c. m. , lipman , d. j. and taubenberger , j. k. ( 2005 ) . whole - genome analysis of human influenza a virus reveals multiple persistent lineages and reassortment among recent h3n2 viruses . _ plos biology _ * 3 * 15791589 . lambert , d. ( 2003 ) . what use is statistics for massive data ? in _ lecture notes monograph series , crossing boundaries : statistical essays in honor of jack hall _ ( j. e. kolassa and d. oakes , eds . ) 217228 . ims , beachwood , oh . liang , l. j. , weiss , r. e. , redelings , b. and suchard , m. a. ( 2009 ) . improving phylogenetic analyses by incorporating additional information from genetic sequence databases . _ bioinformatics _ * 25 * 25302536 . rubin , d. b. ( 1988 ) . using the sir algorithm to simulate posterior distributions . in _ bayesian statistics 3 _ ( j. m. bernardo , m. h. degroot , d. v. lindley and a. f. m. smith , eds . ) 395402 . oxford univ . press . smith , g. j. d. , vijaykrishna , d. , bahl , j. , lycett , s. j. , worobey , m. , pybus , o. g. , ma , s. k. , cheung , c. l. , raghwani , j. , bhatt , s. , peiris , j. s. m. , guan , y. and rambaut , a. ( 2009 ) . origins and evolutionary genomics of the 2009 swine - origin h1n1 influenza a epidemic . _ nature _ * 459 * 11221125 . suchard , m. a. , wang , q. , chan , c. , frelinger , j. , cron , a. and west , m. ( 2010 ) . understanding gpu programming for statistical computation : studies in massively parallel massive mixtures . _ j. comput . graph . _ * 19 * 419438 . tom , j. a. , sinsheimer , j. s. and suchard , m. a. ( 2010 ) . supplement to `` reuse , recycle , reweigh : combating influenza through efficient sequential bayesian computation for massive data . '' doi : http://dx.doi.org/10.1214/10-aoas349supp[10.1214/10-aoas349supp ] . warn , d. e. , thompson , s. g. and spiegelhalter , d. j. ( 2002 ) . bayesian random effects meta - analysis of trials with binary outcomes : methods for the absolute risk difference and relative risk scales . _ * 21 * 16011623 .

the integral observatory has been performing a deep survey of the galactic central radian since 2003 , with the goal of both extracting a catalog of sources and gaining insight into the galactic diffuse emission . this paper concentrates on the estimation of the total point sources emission contribution . it is now clear that unresolved point sources contribute to the observed diffuse emission ; the increasing sensitivity of instruments with time has lead to a steady decrease in estimates of this `` diffuse emission '' . we have analysed the first year data obtained with the spectrometer and imager spi on board integral . first , a catalog of 63 hard x - ray sources detected , time - averaged , during our 2003 galactic plane survey , is derived . second , after extracting the spectra of the sources detected by spi , their combined contribution is compared to the total ( resolved and unresolved ) emission from the galactic ridge . the data analysis is complex : it requires us to split the total emission into several components , as discrete sources and diffuse emission are superimposed in spi data . the main result is that point source emission dominates in the hard x - ray / soft-ray domain , and contributes around 90 % of the total emission around 100 kev , while above 250 kev , diffuse electron - positron annihilation , through its three - photon positronium continuum with a positronium fraction 0.97 and the 511 kev electron - positron line , dominates over the sources .

astro-ph0510084

observations carried out for more than three decades indicate that the total hard x - ray / soft @xmath0-ray emission from the galaxy results from the superposition of multiple physical processes whose relative contribution depends on the energy . the spectrum of this emission is reasonably well measured and understood above @xmath1 1 mev from osse and comptel ( kinzer et al . , 1999 , strong et al . , 1994 , 2004a ) . at energies above @xmath1 100 mev , the dominant emission process is the decay of @xmath2 mesons produced in the interaction of cosmic - ray nucleons with the interstellar matter ( bertsch et al . , 1993 ) . between 1 and 70 mev , electron bremsstrahlung and inverse compton scattering are expected to dominate over discrete source emission ( sacher & schnfelder , 1984 , skibo et al . , 1993 , strong et al . 2004a ) . in the low - energy domain , the spectrum has been measured with rxte in the 10 - 35 kev energy band ( valinia & marshall , 1998 ) and with ginga in the 3 - 16 kev band ( yamasaki et al . , 1997 ) . more recently asca ( sugizaki et al . , 2001 ) observations in the 0.7 - 10 kev band , and chandra observations ( ebisawa et al . , 2001 ) in the 2 - 10 kev band , show that there is a genuine diffuse soft x - ray component , and this result is confirmed by xmm - newton measurements ( hands et al . , 2004 ) in the 2 - 10 kev band . however , in the hard x - ray / soft @xmath0-ray band ( 50 - 500 kev ) , the situation is more complicated as multiple components are believed to contribute to the total emission . these include discrete sources , the positron annihilation line , three - photon positronium continuum radiation and a soft @xmath0-ray component corresponding to the diffuse continuum emission induced by cosmic - ray interactions ( cr diffuse ) . unfortunately , measurements of the galactic diffuse emission in the hard x - ray / soft @xmath0-ray band and its interpretation are inherently difficult because of the presence of numerous hard x - ray discrete sources in the galactic plane . moreover , generally , hard x - ray / soft @xmath0-ray instruments either have large fields of view and no imaging capabilities or have imaging capability but no sensitivity to extended emission . as a result , discrimination between diffuse emission and point sources remains a difficult task . for this reason , simultaneous multiple - instrument observations had been performed , with coordinated observations of the galactic center region with osse / cgro ( kurfess et al . , 1991 ) and the imaging instrument sigma / granat ( paul et al , 1991 ) . the main results were that , in the hard x - ray / soft @xmath0-ray regime , the spatial distribution of `` diffuse emission '' is broad and relatively flat in longitude , with an extension in latitude of @xmath1 @xmath3 width , while a few discrete sources contribute at least 50@xmath4 of the total emission ( purcell et al . , 1996 ) . however sigma had a sensitivity of about 25 mcrab ( 2@xmath5 ) for a typical 24 hours observation . as a result , weak sources escaped detection in its survey , and the unresolved emission was still contaminated significantly by discrete sources . the extension of the spectral shape of the galactic diffuse emission from the hard x - ray to @xmath0-ray regime , and how much of the emission was due to discrete sources remained to be precisely determined . further attempts to derive the diffuse emission characteristics followed ( kinzer et al . 1999 , 2001 ; boggs et al . , 1999 ; valinia et al . , 2000 ) . theoretical studies were unable to explain even 50% of the galactic emission as originating in the interstellar medium ( ism ) . two main processes can lead to an interstellar soft @xmath0-ray emission . the first one is inverse compton scattering of high energy ( gev ) cosmic ray electrons on the ambient photons field ; but these electrons would also produce radio synchrotron emission in the galactic magnetic field at a level much higher than one actually observed . the second process is bremsstrahlung of a population of electrons of a few hundred kev , radiating through interactions with interstellar gas . because these electrons will lose their energy through ionization and coulomb collisions , the total power required to compensate for these energy losses is of the order of @xmath6 erg@xmath7s@xmath8 this power , comparable or higher than that of cosmic ray protons , would affect interstellar - medium ionization equilibrium and give rise to excessive dissociation of interstellar molecules . a possible interstellar process has been proposed by dogiel et al . ( 2002 ) . alternatively , a dozen point sources with intensities around 25 mcrab ( the sigma 2@xmath5 sensitivity ) could account for most of the remaining diffuse emission . this demonstrates the need for highly sensitive imaging instruments for a precise determination of the diffuse galactic emission . + actually , a new vision of the hard x ray sky is provided by integral with the detection of a significant amount of new sources and the discovery of a new class of objects : the highly absorbed sources ( dean et al . , 2005 ) may represent 20% of the total number of sources , but their contribution could not be evaluated by x - ray survey , with an obvious implication for the diffuse emission . a recent result based on isgri / ibis / integral data ( hereafter isgri ) shows that known binary sources account for the main part of the total galactic emission ( from 86% to 74% ) from 20 kev to 220 kev ( lebrun et al . , 2004 , terrier et al . 2004 ) . however the limited sensitivity of isgri , above 200 kev , does not allow an extension of this study at higher energies . early study of the diffuse continuum emission using spi / integral ( hereafter spi ) shows that diffuse emission dominates above @xmath1 200 kev ( strong et al . 2003 ) and that the ratio of diffuse emission to total emission vary from @xmath9 to @xmath10 along spi energy domain ( strong et al . 2004b ) . spi ( see next section ) has a moderate angular resolution over a large field , thus is sensitive to extended sources as well as to point sources imaging . spi can thus be used in a self consistent way to measure both the galactic diffuse and discrete source emission , avoiding the complication due to different regions observed , different instrument responses , different models assumed and diverse ways in which results are presented ( per radian , per field - of - view etc . ) when comparisons with other instruments are made . we present the catalog of the sources detected with spi as well as their global spectral emission . an early spi sources catalog can be found in bouchet et al . moreover a broadband spectral analysis including point sources and diffuse emission components ( annihilation , positronium and cr diffuse continuum ) has been performed in a self consistent way , i.e using the same instrument for both sources detection and diffuse emission estimate . we draw attention to an independent , complementary study on this topic by strong et al . the spectrometer spi ( vedrenne et al . , 2003 ) is one of the two main instruments onboard esa s integral ( international gamma - ray astrophysics laboratory ) observatory launched from baikonour , kazakhstan , on 2002 october 17 . it consists of an array of 19 actively cooled high resolution germanium ( ge ) detectors with an area of 508 @xmath11 and a thickness of 7 cm . it is surrounded by a 5-cm thick bgo shield . the detectors cover the 20 kev - 8 mev energy range with an energy resolution ranging from 2 to 8 kev as a function of energy . in addition to its spectroscopic capability , spi can image the sky with a spatial resolution of @xmath12 over a field of view of @xmath13 , thanks to its coded mask . despite such a modest angular resolution , it is possible to locate intense sources with an accuracy of few arc minutes ( dubath et al . , 2005 ) . + a complete integral orbit lasts @xmath143 days , but scientific data can not be accumulated when the instrument is crossing the radiation belts , reducing the useful observing time @xmath14 2.5 days . the instrument in - flight performance is given in roques et al . each integral orbital revolution consists of several exposures ( typically , 30 - 40 minutes pointings dithering around the target ) . due to the small number of detectors , imaging which relies on these observations in dithering mode ( jensen et al . , 2003 ) , the pointing direction varying by steps of @xmath15 within a @xmath16 square or a 7-point hexagonal pattern . the data used to perform this analysis were recorded from 2003 , march 3 to 2003 , october 19 , including in total 42 revolutions ( table 1 ) . they cover the region @xmath17 @xmath18 l @xmath18 @xmath19 , @xmath20 @xmath18 b @xmath18 @xmath21 most of these data have been obtained through the gcde ( galactic center deep exposure ) part of the integral core program ( winkler et al . , 2003 ) . data polluted by solar flares and radiation belt entries are excluded . after image analysis and cleaning , there is 2552 pointings , which represent 5.75@xmath2210@xmath23 seconds of effective observing time . the data analysis is done in 2 steps . the first one concerns the determination of the positions of the emitting sources . we have used spiros , a software delivered in the integral osa package to produce images . in a second step , we have developed a dedicated software to extract fluxes from the positions obtained with spiros . spiros v6 ( spi iterative removal of sources ) algorithm has been used to derive source positions , with the mcm background option ( mode 5 ) . spiros ( skinner & connell , 2003 ) , a part of the integral science data centre ( isdc ) package produces synthetic and simplified sky images : they consist of a limited number of sky pixels , those which contain excesses above a given threshold . an important limitation is that sources are considered as constant during the image reconstruction process . as a consequence , at low energy ( @xmath24 50 kev ) , some exposures exhibit an unacceptable @xmath25 fit of raw data to the reconstructed sky image convolved by the response matrix along with biased residuals distribution . we found this effect to be due to intensity variations of the most intense sources , which are not taken into account . to suppress the effect of the variability of the strongest sources , a special iterative scheme described in 4.2.2 has been used . spiros works both with known and unknown sources . if an input catalog is given , spiros will first build a sky model with the proposed sources then looks for a number a new ones required by the data . one must take care of the influence of the source position errors on the image generation . if the source position introduced in spiros is different from the true position ( even slightly ) , the iterative removal of sources algorithm leads to subsequent artefacts in the image . this problem may become important for strong sources or when a large number of sources has inaccurate positions . thus , whenever possible , the exact source positions should be introduced in spiros as _ a - priori _ knowledge ( input catalog ) in order to avoid these cumulative errors . but , _ a contrario _ , to use an input catalog with too many ( non - emitting ) sources also leads to unstable solution . our analysis philosophy ( see 5 ) takes into account both limitations . a complete model fitting procedure ( called `` time - model - fit '' ) based on the likelihood statistics has been developed ( see annexe a-1 ) . the input sky model can include both point sources and diffuse components . the main features of the algorithm are : + - a self determination of the background distribution on the detector plane . + - a time dependent background normalization determination + - a time dependent flux determination for each point source with its own timescale ( meaningless for diffuse components ) . the spi imaging system , although using a combination of a coded mask with a position sensitive detector , needs , due to the small number of detector pixels , a dithering scheme to increase the number of sky pixels that can be reconstructed . this results in a time modulation of the sky signal . in order not to confuse such modulation with background variations , the background versus time profile of the 19 detectors has to be evaluated . the construction of a time - dependend background model constitutes a key point of spi data analysis . analyses of `` empty - field '' observations have shown that , for any energy band , the relative count rates ( uniformity map ) of the 19 ge detectors are constant while the global amplitude ( normalization factor ) varies with time . we thus have to determined the relative count - rates of the 19 ge detectors ( background counts - ratio pattern depending on the energy band ) leaving normalization as the only free parameter for the background intensity . this is included inside the model fitting procedure as described in annexe a-2 with the background amplitude able to vary on the pointing ( @xmath12500 s ) timescale . the spi image reconstruction relies on the dithering . as a result , variability of sources has to be explicitly included in the system of equations to be solved . this is included in our model fitting procedure by means of additionnal equations ( annexe a-3 ) . for each source , the allowed variation time scale can be chosen . in practice , it has to be carefully used as the number of parameters ( unknown fluxes ) necessary to describe the sky will increase , correspondingly the significances decrease . there is also a mathematical limitation : since there are 19 independent data per pointing , it means that , whatever the number of pointings , there can not be more than 19 variable sources on the time scale of a pointing . however , most of the sources are weak enough to neglect the influence of their variability on the image reconstruction , and thus can be considered as constant . even flaring sources are too weak and/or shortliving ( e.g. sgr 1806 - 20 ) to have any effect on long period . + in this work , the `` time - model - fit '' tool has been used for a dedicated variability treatment applied only to the most intense sources , namely , 4u 1700 - 377 , sco x-1 and oao 1657 - 4154 , below 50 kev . we build for them the light curves in a one pointing time scale . the corresponding contributions in the count space are then substracted from the original data to derive a `` corrected '' data set to use as spiros input . this cleaning is done from the original data set for each iteration , as the strong sources contributions can be affected by new sources introduction . + applying this scheme to the whole data set in the 25 - 50 kev energy band improves the global reduced @xmath25 from 3.24 ( standard procedure ) to 1.33 ( table 3 ) . images are built using all our data in the following energy bands : 300 - 600 kev , 150 - 300 kev , 50 - 150 kev and 25 - 50 kev . in order to minimise bias ( see 4.1 ) , the first step of the analysis was performed without any _ a - priori _ information about known sources . the _ a priori _ knowledge is introduced progressively . for each image , spiros is parametrized in order to search for up to 30 new excesses above 2@xmath5 in addition to the input catalog built from the previous iteration . this process ( i.e. to fix found positions ) suppresses the instabilities of such an iros algorithm . moreover , such a procedure leads to a minimal sky model able to represent the data . the catalog generation process is the following : * the first image is obtained from spiros with an empty input catalog and gives a list of source candidates . * a catalog is built with : ( i ) all identified sources above 5@xmath5 with their celestial positions . ( ii ) the unidentified excesses above 5@xmath5 with the positions found . the source identification process is described 5.1 . this current catalog is completely regenerated at each iteration as significances can evolve . * we run spiros again with this catalog as input , obtain a new list of sources / excesses , and restart with step 2 * this iterative analysis continues until no other potential excess @xmath26 2@xmath5 can be found . a flow chart illustrates this process in figure 1 . + the final position catalog for a given energy band is built from the last iteration , where sources and excesses with a significance threshold fixed to 4 @xmath5 are accepted . the use of a dedicated catalog for each energy band allows us to restrict the number of free parameters to the minimum needed , thus decreasing the flux detection limit at high energy . figure 2 and 3 present the images obtained by spiros in the 25 - 50 kev and 50 - 150 kev energy bands using our corresponding catalogs in input . to obtain source identification x - ray source catalogs are used . primarly , the isgri catalogs ( bird et al . , 2004 , revnivtsev et al . , 2004 ) are used and , if nothing is found , the isdc catalog is used ( ebisawa et al . , 2003 ) . each excess is considered associated with the nearest known emitting x/@xmath0 ray sources whose distance is less than @xmath27 . this rather high value , 3 times the theoretical value for a @xmath28 detection ( dubath et al . , 2005 ) , is used because the source localisation precision is degraded when more than one source is present , especially for crowded regions . note however that this value is much less than the geometrical resolving power ( 2.6@xmath29 ) , as the dithering pattern amounts to improving the imaging system . + so , two known sources closer than @xmath27 can be retrieved depending on the statistics ( dubath et al . , 2005 ) . the galactic center region is an exception as too many sources are present in a small area . so we represent the @xmath27 region around the galactic center by only one source , namely 1e 1740.7 - 2942 . a more sophisticated scheme is used for the 25 - 50 kev band : the dataset used is divided in two parts , one constructed with odd pointing numbers the other with even pointing numbers ; then two sub - images are built . a non - identified excess is considered as source candidate if it is detected in both images ( cross validation method ) . this method drastically reduces the number of false source detections introduced by systematics , mainly due to variable sources . + the number of excesses as function of the threshold level is shown in table 2 . above 150 kev and below 5@xmath5 , the number of unidentified excesses increase rapidely . from this , empirical thresholds of 5@xmath5 and 4 @xmath5 respectively below and above 150 kev have been fixed . once the position catalogs completed , we build a complete sky model including the detected point sources plus diffuse emission morphologies ( described is 6.1 ) . + fluxes and significances for each of this sky model components are thus computed in a set of energy bands , using `` time - fit -model '' ( 4.2 ) . the fluxes of point sources and diffuse emission are assumed constant ( except for the 20 - 50 kev band , where 4u 1700 - 377 , sco x-1 and oao 1657 - 4154 vary on the time scale of one pointing ) . the background normalization is adjusted on a time scale of one pointing . + excesses above 4.5@xmath5 are kept to form the final catalogs . their fluxes and significances are computed again to be consistent ; thus the final catalogs contain some sources below 4.5@xmath5 that were previously detected above this threshold . this process is described in figure 4 . the final @xmath25 obtained with this method between the proposed sky model and the data are shown in table 3 . the use of a dedicated input catalog for each energy band allows us to optimize the signal - to - noise ratio per energy band . + these catalogs are far from exhaustive : in the computed 1-year - averaged emission , weak or short - transient sources can be completely washed out . moreover , as we looked for a minimal sky model , and given the modest spi angular resolution , only one source is necessary to represent the complex regions ( 1e 1740.7 - 2942 for instance ) . this work , based on the whole dataset , has been complemented by applying the same procedure in 3 subsets , sorting data according to the average galactic longitude of each observation ( table 1 ) : positive ( l @xmath30 ) , negative ( l @xmath31 ) and central ( @xmath32 @xmath18 l @xmath18 @xmath33 ) longitudes , that have more or less equal exposures . these new datasets allow us to add four sources in the 25 - 50 kev band catalog . these sources are labelled with an asterisk . as the final result , in the 25 - 50 kev domain the catalog contains 63 excesses above 4@xmath5 ; 59 have been identified with known hard x - ray objects in the ibis catalog , 4 are tentatively associated with x - ray sources ( labelled with * * ) . among the tentatively identified sources , the source at ( l , b)=(1.94@xmath29,-2.02@xmath29 ) has a quite high significance of 11.2@xmath5 , while its flux is only 6.3 mcrab . it could represent several weak ( eventually extended ) sources below isgri detection threshold that are integrated within the spi angular resolution . it could also results from variable sources not taken into account or from errors accumulation due , for example , to a wrong center - of - gravity of the many sources that compose our so - called ` 1e 1740.7 - 2942 ' source actually fixed at the 1e 1740.7 - 2942 position . intensive simulations are needed to understand the behaviour of the instrument in such a crowded region , and is beyond the scope of this paper . mean fluxes and significances are given in 3 energy bands : 25 - 50 kev , 50 - 150 kev,150 - 300 kev ( table 4 ) . in the 50 - 150 kev band , the catalog contains 20 excesses above 4@xmath5 ( table 5 ) ; all but one correspond to confirmed sources , the last identification is only tentative . finally , 4 known objects emit significantly above 150 kev ( table 6 ) . we will now concentrate on the relative contribution of point sources to the galatic ridge emission as a function of the energy range . in the soft @xmath0 ray regime , multiple components contribute to the total emission . these include discrete sources , positron annihilation line and three - photon positronium annihilation continuum , and a diffuse continuum resulting probably from cosmic ray interactions with the interstellar medium . spi is sensitive to both discrete sources and diffuse emission . as a consequence the discrete sources should always be extracted simultaneously with diffuse emission . however it is difficult to separate / distinguish extended and discrete sources emission if too many free parameters are to be determined : the error bars then become very large , and it may be impossible to derive meaningful information from the analysis . therefore as much _ a priori _ information as possible is included in the analysis : + - precise external source locations is used in catalogs . + - spatial morphologies derived from previous works for each diffuse component are used and fixed _ a priori_. + in order to estimate the total point source emission , it is obvious that all the discrete sources ( also weak , transient and possibly undetected ) should be included in the analysis . on the other hand , increasing the number of sources decreases the significance of the measurement . thus only the significant sources are introduced in the analysis , their number depending on the considered energy band . three spectral regions are considered : + * in the 20 - 150 kev energy range , the 63 sources of the 25 - 50 kev catalog have been introduced . * in the 150 - 300 kev energy range , the 20 sources of our 50 - 150 kev catalog is used . * above 300 kev , the 4 sources detected above 150 kev have been considered . the diffuse emission morphology in the 50 - 400 kev estimated by measurements from osse / gro ( purcell et al . , 1996 , kinzer et al . , 1999 ) and confirmed with subsequent simultaneous rxte and osse observations ( valinia et al . , 2000 ) , is broadly distributed in longitude with a @xmath34 fwhm in latitude and a @xmath35 extent in longitude . however since most of this emission at least in the low energy range is due to point sources ( lebrun et al . , 2004 ) , it can not be used to represent the cr diffuse component . a better model may be the co map ( dame et al . , 2001 ) since it is a tracer of the interstellar matter . in the absence of a better model and for simplicity , the cr diffuse continuum spatial morphology is modelled by the co map . nevertheless , this point constitutes a weakness , as finally the morphology is poorly determined . the annihilation line detected by spi has been used to study the 511 kev spatial morphology ( kndlseder et al . , 2003 ) . this component is equally well described by models that represent the stellar bulge , by halo populations or by an azimuthally symmetric gaussian with fwhm of @xmath1 @xmath36 ( kndlseder et al . , we thus modelled it with this last simpler hypothesis . + a recent study of the positronium annihilation emission with spi shows that its all - sky distribution is consistent with that of the 511 kev electron - positron annihilation ( weidenspointmer et al . , 2005 ) . in the absence of any other information , the three - photon positronium continuum is assumed to have the same spatial distribution as the annihilation line . component intensities of the sky model are extracted channel by channel with the time - model - fit algorithm described in 4.2 . then , count rates are converted into photon spectra for each component . the total point source contribution to the galactic emission is obtained by summing all the point source spectra in the central radian , excluding those from sco x-1 and cen a which are located at high galactic latitude ( b @xmath37 20 @xmath29 ) . + figure 5 shows the composite spectrum of the three major components . the summed point source emission ( upper one ) can be represented by a power law of index 2.64@xmath38 with an exponential cutoff of 504@xmath39 and a 100 kev flux of 4.36@xmath40 @xmath22 10@xmath41 photons @xmath42 s@xmath8 kev@xmath8 , for a reduced @xmath25 of 1.74 for 13 degrees of freedom ( dof ) . the cr diffuse emission spectrum is more complex . below 50 kev it clearly exhibits a soft component . in the 300 - 500 kev domain , there is clearly some cross - talk between the positron annihilation and the cr diffuse continuum components . this can be easily understood as we try to fit simultaneously two sky model distributions that have a common part in the inner galaxy where the exposure time is the highest . + the cr diffuse continuum fit above 50 kev , excluding 300 - 500 kev points ( figure 5 ) , gives a photon power law index of @xmath1 -1.8 @xmath43 and a 100 kev flux of ( 6 @xmath442 ) @xmath22 10@xmath45 photons @xmath42 s@xmath8 kev@xmath8 for a reduced @xmath25 of 0.3(8 dof ) . + in addition , a fit using only data whose pointing direction is outside the central radian , and supposed to contain insignificant electron - positron annihilation emission , gives a power index of 1.7@xmath43 above 50 kev . + these two indexes are fully compatible with the hard component expected from a ` cosmic - ray interaction ' model ( skibo et al . , 1993 ) , which can be approximated by a power law with index of @xmath1 -1.65 over the 50 kev-10 mev range . as we poorly constrain this parameter , we decided to fix the power law index to 1.65 . + finally , in the 20 - 1800 kev energy range , and excluding the 300 - 500 kev band , the cr diffuse continuum spectrum is modelled by two components to give a reduced @xmath25 of 0.5 ( 8 dof ) : the first one is a power law with a photon index 1.7 @xmath43 and a cutoff energy of 20 kev , having a 50 kev flux of ( 3.1 @xmath44 0.7 ) @xmath22 10@xmath41 photons @xmath42 s@xmath8 kev@xmath8 . the second one is a power law with a photon index fixed to 1.65 and a 100 kev flux of ( 4.8 @xmath44 2.2 ) @xmath22 10@xmath45 photons @xmath42 s@xmath8 kev@xmath8 that represents the cr diffuse above 50 kev up to several mev ( including osse and comptel results ) . from this study , the contribution of point sources to the total galactic emission can be derived : 85% in the 25 - 50 kev energy range , 90% in the 50 - 110 kev and 85% in the 110 - 140 kev range . this fraction then decreases as the positronium emission begins to be significant . + second , the 511 kev flux extracted in a 10 kev width channel is @xmath1 0.9@xmath22 10@xmath46 photons @xmath42 s@xmath8 . this value is compatible with that obtained in a completely different way by kndlseder et al . + from these data , we also extract a positronium fraction ( defined as in brown & leventhal , 1987 ) of @xmath1 0.9 as a rough estimate . in fact , the cr diffuse component in the 300 - 500 kev energy range is clearly overestimated , leading to an underestimation of the positronium flux . a more complex and adequate method has thus been proposed for this spectral region ( see next section ) . + the photon spectra extraction method applied above ( 6.2 ) presents the advantages to be spectral model independent and relatively easy to implement . its `` simplicity '' lies on approximations well - suited only for energies below @xmath1 300 kev . indeed , the spi response is averaged over all the exposures for a given component , and the energy redistribution matrix is not properly taken into account . the latter can lead to significant errors particularly in the case of photon spectra with positive slope , such as the positronium continuum ( i.e. above 300 kev ) . for these reasons a complete model fitting software has been developed , in which the emission photon spectral models of each sky components are convolved with the full spi response ( sturner et al . , 2003 ) for each pointing . this algorithm is detailed in annex b. in the other hand , this kind of algorithm allows to fix the shape of the spectra as a function of the energy , via an analytique emission model . as a consequence , the problem of `` cross - talk '' previously encountered is drastically reduced . + each photon spectral component follows an appropriate emission model whose parameters can be adjusted simultaneously to the whole set of data along with the background intensity . given the size of the dataset such an algorithm can lead to prohibitive computing time and memory space if the number of free parameters is too high . fortunately , the benefit of this method is more significant at high energy , where the source number is low . moreover , we limit the number of free parameters by fixing some of them . first , as the used energy band is limited to 300 kev- 1 mev , the slope of the power law that modelled the cr diffuse emission is fixed to 1.65 while the flux is to be fit . second , the point source emission consists of 4 discrete sources ( table 4 ) whose spectra are represented a by power law with an exponential cut - off . because the source statistics are low , the indices and normalizations of these power laws have been previously determined below 150 kev by the first method ( 6.2 ) , while the cutoff energy is to be fitted . results may slightly depend on the sources spectral shape assumed above 300 kev . + the annihilation spectrum ( positronium continuum plus line ) is then determined by the complete model fitting . a line flux of ( [email protected] ) 10@xmath46 photons @xmath42 s@xmath8 and a positronium fraction of 0.97@xmath47 are found . the 100 kev flux of the diffuse emission is 5@xmath443 @xmath22 10@xmath45 photons @xmath42 s@xmath8 kev@xmath8 , in perfect agreement with previous determinations . the reduced @xmath25 of the fit is 0.99 ( 630337 dof ) , the reduced @xmath25 for each energy band are distributed between 0.95 and 1.2 . + the result is illustrated in figure 6 . despite the spi instrument s moderate angular resolution , and benefitting from the large number of pointings , a first catalog of hard x - ray sources detected by spi / integral in the galactic plane between @xmath17 @xmath18 l @xmath18 @xmath19 , @xmath20 @xmath18 b @xmath18 @xmath48 has been derived . + this catalog contains 63 sources and gives their fluxes up to 300 kev . it has been built by means of model - fitting procedure simultaneously applied to all pointings of the whole dataset of 5.7@xmath2210@xmath23 seconds , and represents the minimal averaged sky model needed to decribe the data . thus transient sources could escape detection . moreover , due to the modest spi angular resolution , our sky model in crowded regions is simplified , with one point source representing the combined emission of all sources near 1e 1740.7 - 2942 . on the other hand , this catalog is perfectly suited to estimate the point source contribution to the galactic ridge emission . + the diffuse emission components ( continuum , 511 kev and positronium ) have been studied by fixing their spatial distributions , as a first order model , and taking into account our catalog . + the main conclusions are : * point sources contribute at least for 85 - 90% of the total galactic emission in the energy range 25 to 140 kev . this independantly confirms the isgri result that point sources largely dominate the total galactic emission ( lebrun et al . , 2004 ) at least up to 250 kev . the interstellar emission is around 15 % of the total emission up to 150 kev . the low energy part ( @xmath24 250 kev ) of the total spectrum is source dominated while the high energy part ( @xmath49 250 kev ) is diffuse dominated . * the cr diffuse emission ( 50 - 250 kev ) can be fitted with a power law joining smoothly the high energy continuum as measured by osse ( kinzer et al . , 1999 ) and comptel ( strong et al . , 1994 ) above 1 mev . * no additional spectral component is needed above the power law spectrum to model the diffuse continuum above 50 kev . the soft diffuse component reported by kinzer et al . ( 1999 ) in this region is mostly due to point sources not taken into account by this instrument . * below 50 kev , a real diffuse soft component can still exist , but may also be related to very steep / weak x - ray sources not included in the present analysis . * there is no evidence of 511 kev point source emission , in agreement with the conclusion drawn by kndlseder et al . ( 2005 ) . * a positronium fraction 0.97@xmath47 and a 511 kev flux of ( 0.93 @xmath440.15 ) 10@xmath46 photons @xmath42 s@xmath8 . this positronium fraction is compatible with the osse determination : 0.93 @xmath44 0.04 ( kinzer et al . , 2001 ) and spi / integral determinations of 0.94 @xmath44 0.06 ( churazov et al . , 2005 ) . these results constitute a first step as they are based on spi data taken during the first year of operation and will be obviously refined in the future . an alternative approach is proposed by strong et al . ( 2005 ) . the catalogs are available electronically as apj ` supplementary electronic materials ' . the spi / integral project has been completed under the responsibility and leadership of cnes . we are grateful to asi , cea , cnes , dlr , esa , inta , nasa and ostc for support . bird a. , barlow e.j . , bassani l. , et al , 2004 , apjl , 607 , l33 bertsch d.l . , dame t.m . , fichtel c.e . , et al . , 1993 , apj , 416 , 587 boggs s.e . , lin r.p . , slassi - sennou s. , et al . , 1999 , apj , 544 , 320 bouchet l. , jourdain e. , roques j.p . , et al . , proc . 5th integral workshop , esa sp-552 , p147 brown b. & leventhal m. , 1987 , apj , 319 , 637 churazov e. , sunyaev r. , sazonov s. , et al , 2005,mnras 357 , 1377 dame t.m . , hartmann d. , & thadeus p. , 2001 , apj , 547 , 792 dean a.j . , hill , a. b. , stephen j. b. , et al , 2005 , a&a , submitted dogiel v.a . , schnfelder , v. , & strong , a.w . 2002a , a&a , 382 , 730 dubath p. , knodlseder j. , skinner g.k . , et al . , 2005 , mnras , 357 , 420 ebisawa k. , meada y. , kaneda h. & yamauchi s. , 2001 , science , 293 , 1633 ebisawa k. , bourban a. , bodaghee a. , et al , 2003 , a&a , 411 , l59 hands a.d.p . , warwick r.s . , watson m.g . et helfand d.j . , 2004 , mnras , 361 , 31 jensen p.l . , clausen k. , cassi c. , , et al . , 2003 , a&a , 411 , l7 kinzer r.l . , purcell w.r . & kurfess j.d . , 1999 , apj , 515 , 215 kinzer r.l . , milnes , p.a .. , kurfess j.d . , et al . , 2001 , apj , 559 , 282 kndlseder j. , lonjou v. , jean p. , et al , 2003 , a&a , 411 , l457 kndlseder j. , pierre j. , lonjou v. , et al . , 2005 , a&a in press kurfess j.d . , johnson w.n . , kinzer r.l . , et al . , 1991 , adv space res . , 11(8 ) , 323 lebrun f. , terrier r. , bazzano a. , et al . , 2004 , nature , 428 , 293 paul j. , manbdrou p. , ballet j. , et al . , 1991 , adv space res . , 11(8 ) , 289 purcell w.r . , bouchet l. , johnson w.n . , et al . , 1996 , a&as , 120 , 389 revnivtsev m.g . , sunyaev r.a . , varshalovich d.a . , et al , 2004 , astl , 30 , 382 roques j.p . , schanne s. , von kienlin a. , et al , 2003 , a&a , 411 , l91 sacher w. , & schoenfelder v. , 1984 , apj , 279 , 817 skibo j.g , johnson , w.n . , kinzer , r.l . , et al , 1993 , a&as , 25 , 1293 skinner g. & connell p. , 2003 , a&a , 411 , l123 strong a.w . , et al . , 1994 , a&a , 292 , 82 strong a.w . , et al , 2003 , a&a , 411 , l447 strong a.w . , moskalenko i.v & reimer o , 2004a , apj , 613 , 956 strong , a. w. , et al . 2004b , proc . 5th integral workshop , esa sp-552 , p507 ; astro - ph/0405023 strong a.w . et al . , 2005 , a&a in press , astro - ph/0509290 sturner , s.j . , schrader , c.r . , weidenspointner , g. , et al . , 2003 , a&a , 411 , l81 sugizaki , m. , kazuhisa k. , kaneda h. , et al . , 2001 , apjs , 134,77 terrier , r. , et al . , 2004 , 5th integral workshop , esa sp-552 , p513 ; astro - ph/0405207 valinia a. & marshall f.e . , 1998 , apj , 505 , 134 valinia a. , kinzer r.l . & marshall f.e . , 2000 , apj , 534 , 277 vedrenne g. , roques j.p . , schonfelder v. , et al , 2003 , a&a , 411 , l63 weidenspointner g. , kndlseder j. , pierre j. , et al , 2005 , to be submitted to a%a winkler c. , gehrels n. , schonfelder v. , et al . , 2003 , a&a , 411 , l349 yamasaki n.y . , ohashi t. , takahara f. , et al , 1997 , apj , 481 , 821 cccccc 47 & 2003/03/03 - 03:32 & 2003/03/05 - 18:02 & 136.0 & -27.4 & 0.3 + 49 & 2003/03/09 - 03:00 & 2003/03/11 - 14:43 & 46.8 & -20.7 & 1.1 + 50 & 2003/03/12 - 02:49 & 2003/03/14 - 16:44 & 92.1 & -27.9 & 0.8 + 51 & 2003/03/15 - 02:34 & 2003/03/17 - 16:30 & 81.0 & -12.1 & 0.8 + 52 & 2003/03/18 - 02:19 & 2003/03/20 - 16:36 & 91.6 & -14.0 & -0.1 + 53 & 2003/03/21 - 02:04 & 2003/03/23 - 15:43 & 143.2 & -3.1 & 0.2 + 54 & 2003/03/24 - 01:50 & 2003/03/26 - 15:56 & 116.7 & -15.8 & -1.7 + 55 & 2003/03/27 - 01:35 & 2003/03/29 - 15:50 & 26.5 & -10.3 & 10.7 + 56 & 2003/03/30 - 01:22 & 2003/04/01 - 15:27 & 49.6 & 4.1 & 4.0 + 57 & 2003/04/02 - 01:10 & 2003/04/04 - 15:10 & 0.7 & -0.1 & -1.1 + 58 & 2003/04/05 - 00:59 & 2003/04/07 - 14:20 & 18.1 & -5.2 & -2.9 + 59 & 2003/04/08 - 00:49 & 2003/04/10 - 14:44 & 48.7 & 12.7 & 1.1 + 60 & 2003/04/11 - 00:34 & 2003/04/13 - 14:59 & 43.6 & -5.7 & -0.4 + 61 & 2003/04/14 - 00:20 & 2003/04/16 - 14:17 & 51.2 & -0.6 & 0.7 + 62 & 2003/04/17 - 00:06 & 2003/04/19 - 13:35 & 22.8 & 6.1 & 4.4 + 64 & 2003/04/22 - 23:37 & 2003/04/25 - 13:41 & 189.9 & 18.1 & -0.1 + 65 & 2003/04/25 - 23:20 & 2003/04/28 - 13:11 & 201.7 & 24.2 & 0.1 + 66 & 2003/04/28 - 23:06 & 2003/05/01 - 13:25 & 145.0 & 26.2 & -2.5 + 97 & 2003/07/30 - 16:59 & 2003/08/02 - 05:57 & 180.0 & -1.4 & 23.7 + 100 & 2003/08/08 - 15:18 & 2003/08/11 - 05:07 & 137.3 & -18.7 & 0.6 + 101 & 2003/08/11 - 15:57 & 2003/08/14 - 05:19 & 181.7 & -2.1 & 23.5 + 103 & 2003/08/17 - 14:39 & 2003/08/20 - 04:32 & 119.2 & -7.8 & 0.6 + 104 & 2003/08/20 - 14:25 & 2003/08/23 - 04:30 & 156.2 & -42.1 & 31.1 + 105 & 2003/08/23 - 15:06 & 2003/08/26 - 03:41 & 177.0 & -0.1 & -0.6 + 106 & 2003/08/26 - 14:51 & 2003/08/29 - 03:38 & 166.7 & -0.7 & 0.6 + 107 & 2003/08/29 - 14:35 & 2003/09/01 - 03:17 & 197.6 & 0.8 & 0.8 + 108 & 2003/09/01 - 14:22 & 2003/09/04 - 02:54 & 201.5 & 1.5 & -0.0 + 109 & 2003/09/04 - 14:07 & 2003/09/07 - 03:05 & 193.5 & 10.1 & -0.8 + 110 & 2003/09/07 - 13:52 & 2003/09/10 - 02:46 & 139.2 & 0.1 & -0.6 + 111 & 2003/09/10 - 13:39 & 2003/09/13 - 02:33 & 199.3 & 0.5 & -0.3 + 112 & 2003/09/13 - 13:26 & 2003/09/16 - 02:21 & 195.8 & 0.3 & -0.3 + 113 & 2003/09/16 - 13:14 & 2003/09/19 - 02:09 & 171.7 & -0.1 & -0.3 + 114 & 2003/09/19 - 13:07 & 2003/09/22 - 01:58 & 177.3 & -0.2 & -0.3 + 115 & 2003/09/22 - 12:55 & 2003/09/25 - 02:01 & 193.3 & 0.0 & -0.2 + 116 & 2003/09/25 - 11:47 & 2003/09/28 - 01:41 & 148.5 & -14.8 & -2.0 + 117 & 2003/09/28 - 11:35 & 2003/10/01 - 01:50 & 188.5 & 22.0 & 0.9 + 118 & 2003/10/01 - 11:21 & 2003/10/04 - 01:45 & 180.8 & -19.2 & -1.2 + 119 & 2003/10/04 - 11:08 & 2003/10/07 - 01:12 & 178.9 & -1.9 & -1.3 + 120 & 2003/10/07 - 10:55 & 2003/10/10 - 01:02 & 190.6 & 2.6 & 0.1 + 121 & 2003/10/10 - 10:45 & 2003/10/13 - 00:46 & 198.9 & 14.3 & 0.8 + 122 & 2003/10/13 - 10:35 & 2003/10/16 - 00:32 & 199.0 & 13.6 & 0.5 + 123 & 2003/10/16 - 10:25 & 2003/10/19 - 00:22 & 176.3 & 26.9 & 0.3 + lcccc 7@xmath5 & 50(47 ) & 16(15 ) & 2(2 ) & 1(1 ) + 6@xmath5 & 56(53 ) & 20(19 ) & 4(4 ) & 3(3 ) + 5@xmath5 & 62(58 ) & 24(19 ) & 4(4 ) & 4(3 ) + 4@xmath5 & 69(64 ) & 30(21 ) & 7(4 ) & 11(5 ) + lcc 300 - 600 & 1.027 & 45867 + 150 - 300 & 1.013 & 45869 + 50 - 150 & 1.125 & 45855 + 25 - 50 & 3.242 & 45809 + 25 - 50 * & 1.328 & 45806 + rlccccc 1 & cen a & -50.48 & 19.38 & 54.7 @xmath44 2.9 ( 18.9 ) & 72.7 @xmath44 6.3 ( 11.5 ) & 85.4 @xmath44 21.8 ( 3.9 ) + 2 & 4u 1516 - 569 & -37.88 & 0.04 & 18.0 @xmath44 3.0 ( 6.0 ) & 15.0 @xmath44 6.6 ( 2.3 ) & @xmath24 46.1 + 3 & 4u 1538 - 522 & -32.58 & 2.18 & 9.6 @xmath44 2.0 ( 4.9 ) & @xmath24 9.2 & @xmath24 33.5 + 4 & 4u 1608 - 522 & -29.06 & -0.86 & 17.8 @xmath44 1.6 ( 11.0 ) & 17.8 @xmath44 3.9 ( 4.6 ) & @xmath24 28.5 + 5 & 4u 1636 - 536 & -27.08 & -4.81 & 19.2 @xmath44 1.5 ( 12.8 ) & 12.5 @xmath44 3.6 ( 3.5 ) & @xmath24 26.1 + 6 & igrj16167 - 4957 * & -26.92 & 0.51 & 5.8 @xmath44 1.4 ( 4.3 ) & @xmath24 7.2 & @xmath24 26.7 + 7 & igr j16318 - 4848 & -24.37 & -0.44 & 20.9 @xmath44 1.5 ( 13.7 ) & @xmath24 7.4 & @xmath24 27.5 + 8 & 4u 1630 - 47 & -23.08 & 0.24 & 76.1 @xmath44 1.5 ( 50.1 ) & 49.8 @xmath44 3.7 ( 13.6 ) & @xmath24 26.9 + 9 & gx 339 - 4 & -21.06 & -4.32 & 8.6 @xmath44 1.2 ( 7.0 ) & 11.8 @xmath44 2.9 ( 4.1 ) & @xmath24 20.5 + 10 & igr j16418 - 4532 & -20.81 & 0.52 & 14.1 @xmath44 1.2 ( 11.4 ) & 9.6 @xmath44 2.9 ( 3.3 ) & @xmath24 21.2 + 11 & 4u 1705 - 440 & -16.67 & -2.33 & 28.3 @xmath44 1.2 ( 22.9 ) & 18.1 @xmath44 2.9 ( 6.3 ) & @xmath24 21.1 + 12 & 4u 1702 - 429 & -16.11 & -1.31 & 8.4 @xmath44 1.3 ( 6.5 ) & 6.9 @xmath44 3.1 ( 2.2 ) & @xmath24 22.5 + 13 & oao 1657 - 4154 & -15.63 & 0.32 & 83.0 @xmath44 1.7 ( 50.3 ) & 34.1 @xmath44 2.3 ( 14.7 ) & @xmath24 17.1 + 14 & 4u 1735 - 444 & -13.94 & -6.99 & 6.8 @xmath44 0.9 ( 7.3 ) & @xmath24 4.2 & @xmath24 15.1 + 15 & igr j17195 - 4100 & -13.03 & -2.12 & 11.1 @xmath44 0.8 ( 13.9 ) & 7.1 @xmath44 1.9 ( 3.7 ) & @xmath24 14.1 + 16 & 4u 1700 - 377 & -12.25 & 2.18 & 164.1 @xmath44 1.4 ( 119.7 ) & 75.8 @xmath44 2.0 ( 38.3 ) & 16.6 @xmath44 7.4 ( 2.2 ) + 17 & gx 349 + 2 & -10.88 & 2.76 & 14.2 @xmath44 1.2 ( 11.4 ) & @xmath24 6.1 & @xmath24 22.9 + 18 & igr j17252 - 3616 * & -8.49 & -0.36 & 11.8 @xmath44 1.4 ( 8.5 ) & @xmath24 3.2 & @xmath24 11.8 + 19 & 4u 1746 - 37 & -6.48 & -4.97 & 7.7 @xmath44 0.6 ( 13.0 ) & 6.7 @xmath44 1.4 ( 4.7 ) & @xmath24 10.6 + 20 & gx 354 - 0 & -5.69 & -0.15 & 23.0 @xmath44 0.6 ( 37.2 ) & 7.5 @xmath44 1.5 ( 4.9 ) & @xmath24 11.3 + 21 & grs 1724 - 308 & -3.67 & 2.31 & 20.3 @xmath44 0.5 ( 38.7 ) & 12.0 @xmath44 1.4 ( 8.9 ) & @xmath24 10.1 + 22 & 4u 1822 - 371 & -3.14 & -11.29 & 29.6 @xmath44 0.8 ( 38.1 ) & 9.6 @xmath44 1.8 ( 5.2 ) & @xmath24 13.4 + 23 & igr j17488 - 3253 & -3.03 & -2.64 & 15.4 @xmath44 0.9 ( 16.9 ) & 9.1 @xmath44 2.3 ( 4.0 ) & @xmath24 16.9 + 24 & igr j17464 - 3213 & -2.74 & -1.83 & 6.2 @xmath44 1.1 ( 5.7 ) & 9.3 @xmath44 2.7 ( 3.5 ) & @xmath24 19.6 + 25 & grs 1734 - 292 & -1.11 & 1.41 & 7.8 @xmath44 0.6 ( 12.3 ) & @xmath24 3.4 & 15.6 @xmath44 6.2 ( 2.5 ) + 26 & sco x-1 & -0.90 & 23.77 & 216.8 @xmath44 1.6 ( 136.6 ) & 20.9 @xmath44 2.3 ( 9.2 ) & 18.9 @xmath44 8.5 ( 2.2 ) + 27 & 1e 1740.7 - 2942 & -0.87 & -0.10 & 80.4 @xmath44 1.1 ( 71.8 ) & 78.9 @xmath44 2.7 ( 28.8 ) & 45.6 @xmath44 10.0 ( 4.6 ) + 28 & slx 1744 - 299 & -0.72 & -0.89 & 25.6 @xmath44 1.2 ( 20.6 ) & @xmath24 6.3 & 24.7 @xmath44 11.5 ( 2.1 ) + 29 & v2400oph & -0.15 & 8.73 & 14.2 @xmath44 0.9 ( 16.4 ) & 4.8 @xmath44 2.1 ( 2.3 ) & @xmath24 15.2 + 30 & exmsb 1709 - 232 * & 0.59 & 9.27 & 16.8 @xmath44 3.4 ( 4.9 ) & @xmath24 4.2 & @xmath24 15.2 + 31 & slx 1735 - 269 & 0.81 & 2.41 & 10.1 @xmath44 0.5 ( 18.7 ) & 3.8 @xmath44 1.4 ( 2.8 ) & @xmath24 10.0 + 32 & rx j1832 - 33 & 1.54 & -11.36 & 4.1 @xmath44 0.7 ( 5.9 ) & 3.5 @xmath44 1.7 ( 2.1 ) & @xmath24 12.1 + 33 & xte j1807 - 294 & 1.93 & -4.27 & 5.0 @xmath44 0.6 ( 9.1 ) & @xmath24 2.8 & @xmath24 10.4 + 34 & ax j1758.0 - 2818 * * & 1.94 & -2.02 & 6.3 @xmath44 0.6 ( 11.2 ) & @xmath24 3.0 & 14.9 @xmath44 5.7 ( 2.6 ) + 35 & gx 1 + 4 & 1.95 & 4.80 & 12.0 @xmath44 0.5 ( 22.2 ) & 7.3 @xmath44 1.3 ( 5.5 ) & @xmath24 10.0 + 36 & gx 3 + 1 & 2.30 & 0.80 & 10.5 @xmath44 0.6 ( 17.6 ) & @xmath24 3.1 & @xmath24 11.5 + 37 & 4u 1820 - 303 & 2.79 & -7.92 & 14.8 @xmath44 0.6 ( 25.1 ) & @xmath24 2.8 & 11.8 @xmath44 5.2 ( 2.3 ) + 38 & grs 1758 - 258 & 4.51 & -1.36 & 61.2 @xmath44 1.0 ( 59.9 ) & 76.5 @xmath44 2.5 ( 30.7 ) & 59.6 @xmath44 9.2 ( 6.5 ) + 39 & 4u 1849 - 31 & 4.97 & -14.35 & 4.3 @xmath44 1.0 ( 4.5 ) & @xmath24 4.3 & @xmath24 14.9 + 40 & gx 5 - 1 & 5.09 & -1.01 & 20.6 @xmath44 1.1 ( 19.4 ) & 6.7 @xmath44 2.6 ( 2.6 ) & 19.2 @xmath44 9.5 ( 2.0 ) + 41 & sax j1810.8 - 2609 * * & 5.20 & -3.43 & 2.5 @xmath44 0.5 ( 4.8 ) & @xmath24 2.6 & @xmath24 9.7 + 42 & igr j17597 - 2201 & 7.58 & 0.77 & 9.5 @xmath44 0.6 ( 15.4 ) & @xmath24 3.0 & @xmath24 11.2 + 43 & 4u 1745 - 203 & 7.72 & 3.81 & 4.3 @xmath44 0.6 ( 7.7 ) & @xmath24 2.7 & @xmath24 10.1 + 44 & gx 9 + 9 & 8.53 & 9.04 & 4.5 @xmath44 0.7 ( 6.8 ) & 3.3 @xmath44 1.6 ( 2.1 ) & @xmath24 11.3 + 45 & gs 1826 - 24 & 9.28 & -6.08 & 80.5 @xmath44 0.6 ( 125.9 ) & 54.0 @xmath44 1.5 ( 35.8 ) & 30.7 @xmath44 5.6 ( 5.5 ) + 46 & sgr 1806 - 20 & 10.01 & -0.21 & 5.8 @xmath44 0.7 ( 8.8 ) & @xmath24 3.2 & @xmath24 11.9 + 47 & ax j1812.2 - 1842/snr 011.2 - 00.3 * * & 11.90 & -0.16 & 3.7 @xmath44 0.7 ( 5.3 ) & 6.4 @xmath44 1.7 ( 3.8 ) & @xmath24 12.7 + 48 & pks 1830 - 211 & 12.15 & -5.72 & 4.8 @xmath44 0.7 ( 7.0 ) & 7.6 @xmath44 1.6 ( 4.6 ) & @xmath24 12.3 + 49 & gx 13 + 1 & 13.52 & 0.12 & 7.9 @xmath44 0.8 ( 10.2 ) & 4.7 @xmath44 1.9 ( 2.5 ) & @xmath24 14.0 + 50 & gx 17 + 2 & 16.44 & 1.28 & 17.1 @xmath44 0.8 ( 21.1 ) & @xmath24 3.9 & @xmath24 14.2 + 51 & m 1812 - 12 & 18.03 & 2.40 & 20.0 @xmath44 0.9 ( 22.7 ) & 16.1 @xmath44 2.1 ( 7.7 ) & @xmath24 15.3 + 52 & ax j1825.1 - 1253 * * & 18.49 & -0.14 & 3.5 @xmath44 0.9 ( 4.0 ) & @xmath24 4.3 & @xmath24 15.9 + 53 & snr 021.5 - 00.9 * & 21.51 & -0.88 & 5.4 @xmath44 0.9 ( 6.0 ) & @xmath24 4.8 & @xmath24 17.5 + 54 & igr j18325 - 0756 * & 23.71 & 0.57 & 6.2 @xmath44 0.9 ( 6.8 ) & 6.3 @xmath44 2.5 ( 2.5 ) & @xmath24 18.9 + 55 & 4u 1850 - 087 & 25.36 & -4.32 & 5.7 @xmath44 1.1 ( 5.2 ) & @xmath24 5.2 & @xmath24 19.0 + 56 & kes 73 & 27.39 & 0.01 & 5.1 @xmath44 1.1 ( 4.5 ) & 8.1 @xmath44 2.8 ( 2.9 ) & @xmath24 20.5 + 57 & xte j1855 - 026 & 31.08 & -2.09 & 17.9 @xmath44 1.4 ( 13.0 ) & 10.6 @xmath44 3.3 ( 3.3 ) & @xmath24 23.8 + 58 & 1916 - 053 & 31.35 & -8.49 & 6.4 @xmath44 1.6 ( 4.0 ) & @xmath24 7.3 & 31.9 @xmath44 13.3 ( 2.4 ) + 59 & ax j1852.6 + 0038 & 33.65 & 0.02 & 6.2 @xmath44 1.6 ( 3.9 ) & @xmath24 7.6 & @xmath24 28.0 + 60 & 4u 1901 + 03 & 37.19 & -1.25 & 22.9 @xmath44 2.2 ( 10.6 ) & @xmath24 10.4 & @xmath24 37.9 + 61 & ss 433 & 39.70 & -2.24 & 13.6 @xmath44 2.8 ( 4.8 ) & @xmath24 13.2 & @xmath24 47.6 + 62 & 4u 1909 + 07 * & 41.90 & -0.81 & 13.5 @xmath44 3.3 ( 4.2 ) & @xmath24 17.7 & @xmath24 59.4 + 63 & grs 1915 + 105 & 45.37 & -0.21 & 91.6 @xmath44 15.4 ( 6.0 ) & @xmath24 36.9 & @xmath24 97.4 + lccc cen a & -50.48 & 19.38 & 76.8 @xmath44 6.7 ( 11.5 ) + 4u 1608 - 522 & -29.06 & -0.86 & 19.5 @xmath44 3.9 ( 5.0 ) + 4u 1630 - 47 & -23.08 & 0.24 & 55.5 @xmath44 3.2 ( 17.6 ) + oao 1657 - 4154 & -15.63 & 0.32 & 37.0 @xmath44 2.0 ( 18.2 ) + 4u 1705 - 440 & -16.67 & -2.33 & 22.5 @xmath44 2.2 ( 10.2 ) + 4u 1700 - 377 & -12.25 & 2.18 & 80.2 @xmath44 1.6 ( 48.8 ) + 4u 1746 - 37 & -6.48 & -4.97 & 9.4 @xmath44 1.3 ( 7.2 ) + gx 354 - 0 & -5.69 & -0.15 & 5.3 @xmath44 1.3 ( 4.2 ) + grs 1724 - 308 & -3.67 & 2.31 & 11.1 @xmath44 1.2 ( 9.0 ) + 4u 1822 - 371 & -3.14 & -11.29 & 9.3 @xmath44 1.7 ( 5.4 ) + igr j17488 - 3253 & -3.03 & -2.64 & 10.2 @xmath44 1.9 ( 5.3 ) + igr j17464 - 3213 & -2.74 & -1.83 & 8.2 @xmath44 1.9 ( 4.4 ) + sco x-1 & -0.90 & 23.77 & 22.4 @xmath44 2.4 ( 9.3 ) + 1e 1740.7 - 2942 & -0.87 & -0.10 & 87.3 @xmath44 1.3 ( 67.1 ) + gx 1 + 4 & 1.95 & 4.80 & 6.1 @xmath44 1.2 ( 5.0 ) + grs 1758 - 258 & 4.51 & -1.36 & 82.9 @xmath44 1.2 ( 67.8 ) + gs 1826 - 24 & 9.28 & -6.08 & 57.0 @xmath44 1.5 ( 38.4 ) + ax j1812.2 - 1842/snr 011.2 - 00.3 * * & 11.90 & -0.16 & 5.7 @xmath44 1.5 ( 3.9 ) + pks 1830 - 211 & 12.15 & -5.72 & 8.0 @xmath44 1.5 ( 5.2 ) + m 1812 - 12 & 18.03 & 2.40 & 16.9 @xmath44 2.0 ( 8.6 ) + lccc -2.74 & -1.83 & igr j17464 - 3213 & 21.2 @xmath44 4.0 ( 5.3 ) + -0.87 & -0.10 & 1e 1740.7 - 2942 & 61.4 @xmath44 4.1 ( 14.8 ) + 4.51 & -1.36 & grs 1758 - 258 & 75.8 @xmath44 4.0 ( 18.9 ) + 9.28 & -6.08 & gs 1826 - 24 & 28.6 @xmath44 4.3 ( 6.6 ) + the relation between the expected data , instrument aperture response , background and the @xmath50 sources which light the detectors is @xmath51 where e(d , p , e ) and b(d , p , e ) are respectively the expected data and background in the energy bin e , pointing p and detector d. @xmath52 is the intensity of the source number j , or sky image pixel j , in the direction @xmath53 in the energy bin e. @xmath54 the instrument response of the data element ( d , p , e ) to the sky image pixel j in the direction @xmath53 . t(d , p ) is the duration exposure for the detector d and pointing p. the background term is rewritten as @xmath55 u(d , e ) is the uniformity map that fixed the relative counts ratio between the 19 ge detectors ( 4.1 ) . if the background amplitude varies on the pointing timescale , then an amplitude coefficient a(p ) is introduced to describe the variable background @xmath56 , @xmath57 for a given energy bin e and to simplify the equation representation , the sky image pixel direction @xmath53 is represent by the sky pixel number j. the equation reduces to @xmath58 for the maximum likelihood algorithm , the cash statistics ( cash , 1979 ) is used and the following equation is to be maximized with respect to the parameters i and a ( which are constrained to be positive ) : @xmath59 where n(d , p ) are the measured data ( for energy bin e ) and @xmath60 the number of pointing . the background uniformity estimation is obtained in two steps . first an uniformity map @xmath61 is assumed and the likelihood is optimized with respect to i and a to give the best estimation @xmath62 and @xmath63 . second , knowing the current best estimate of the sources intensity and current background . the background map is obtained through the minimization for each detector of @xmath64 where s(d , p ) is the total source counts contribution , b the current background and @xmath65 a multiplicative factor of order 1 of the current background map for detector d. @xmath66 the minimization with respect to @xmath65 gives @xmath67 the best uniformity map is then @xmath68 this process is repeated several times until the equivalent @xmath25 stops to decrease , the equivalent @xmath25 being defined as @xmath69 the expected counts e from a given source k counts is @xmath70 for a variable source this equation is to be expanded in several equations , corresponding to intervals whic the source is considered as constant @xmath71 finally , instead of a single intensity for source k , there is l intensities @xmath72 to be determined . the function to be minimized is @xmath73 where @xmath74 is the number of detector energy bin ( counts space ) , @xmath60 the number of pointing and @xmath75 the number of detector . n(d , p , e ) and e(d , p , e ) are respectively measured and expected counts for detector d , pointing p and detector energy bin e. the following equation relies the response matrix , sources photon spectra , background and expected counts where @xmath77 is the response ( aperture response and detector redistribution matrix ) of the data element ( d , p , e ) to 1 incident photon in the sky direction @xmath53 for the incident in the energy bin energy @xmath78.@xmath50 and @xmath79 are respectively the number of sources and energy bin in the photon space . @xmath80 the incident photon spectrum in the energy bin @xmath81 ( photon space ) for the source j. given the parameters which describes the photon spectra and thus the photon spectrum of each sources @xmath88 , the background is determined in each energy bin e. given the best current background pattern @xmath61 , the background is modeled as

we present a new approach to analyze the validation of weakly nonlinear geometric optics for entropy solutions of nonlinear hyperbolic systems of conservation laws whose eigenvalues are allowed to have constant multiplicity and corresponding characteristic fields to be linearly degenerate . the approach is based on our careful construction of more accurate auxiliary approximation to weakly nonlinear geometric optics , the properties of wave front - tracking approximate solutions , the behavior of solutions to the approximate asymptotic equations , and the standard semigroup estimates . to illustrate this approach more clearly , we focus first on the cauchy problem for the hyperbolic systems with compact support initial data of small bounded variation and establish that theestimate between the entropy solution and the geometric optics expansion function is bounded by , _ independent of _ the time variable . this implies that the simpler geometric optics expansion functions can be employed to study the behavior of general entropy solutions to hyperbolic systems of conservation laws . finally , we extend the results to the case with non - compact support initial data of bounded variation .

1205.5930

we are concerned with weakly nonlinear geometric optics for entropy solutions of the following hyperbolic system of conservation laws : @xmath2 where @xmath3 : @xmath4 is a smooth function . the jacobian matrix @xmath5 is diagonalizable at any point @xmath6 and has @xmath7 real eigenvalues such that any eigenvalue has constant multiplicity . without loss of generality , we assume that @xmath8 where @xmath9 , @xmath10 , and the corresponding left and right eigenvectors @xmath11 and @xmath12 satisfy @xmath13 here @xmath14 is the kronecker delta satisfying @xmath15 and @xmath16 when @xmath17 . when @xmath18 , the system in is strictly hyperbolic . in addition , we assume that each characteristic field is either genuinely nonlinear or linearly degenerate , that is , either of the following holds : @xmath19 a prototype of such hyperbolic systems is the full steady euler equations for supersonic ideal gases in @xmath20 with the following form for @xmath21 : @xmath22 where @xmath23 is the density , @xmath24 the fluid velocity , and @xmath25 the pressure . the notation @xmath26 denotes the tensor product of the vectors @xmath27 and @xmath28 . set @xmath29 . when @xmath30 , the vector @xmath31 can be expressed as a vector function @xmath32 of @xmath33 . then the system can be written into the form in with @xmath34 . by direct calculation , we find that the eigenvalues of this system are @xmath35 and the corresponding eigenvectors are @xmath36 , and @xmath37 . then @xmath38 \rr_{0j}\cdot\nabla\lambda_0\equiv0 \,\,\,\ , \qquad \mbox{for } \;\ ; j=1 , 2 . \end{array}\ ] ] thus , the two characteristic fields corresponding to @xmath39 are genuinely nonlinear , while the other two fields corresponding to @xmath40 are linearly degenerate . an asymptotic geometric optics expansion is of the following form : @xmath41 where @xmath42 is a constant background state in . by a formal derivation of the expansion of weakly nonlinear geometric optics for conservation laws ( _ cf . _ diperna - majda @xcite ) , the expansion is expected to be @xmath43 and the functions @xmath44 satisfy a decoupled system of scalar conservation laws : @xmath45 with the coefficients @xmath46 then the genuinely nonlinear condition from implies that @xmath47 which yields that equation is the inviscid burgers equation : @xmath48 while the linearly degenerate condition implies that @xmath49 which yields a linear equation @xmath50 we define @xmath51 then the much simpler function @xmath52 can be used to study the behavior of general entropy solutions of nonlinear hyperbolic system of conservation laws , provided that the convergence can be rigorously verified . an approach has been introduced in chen - christoforou - zhang @xcite , based on the results presented in bressan @xcite , to compare the solutions of two different systems , which requires that one of them is the standard riemann semigroup ( srs ) while the other is only a global entropy solution with bounded variation obtained by the front tracking method . let @xmath53 be a closed domain . a map @xmath54 is a * srs * generated by system if the following three conditions hold ( _ cf . _ @xcite ) : @xmath55 semigroup property : : for every @xmath56 and @xmath57 , @xmath58 @xmath55 lipschitz continuity : : there exist constants @xmath59 and @xmath60 such that , for all @xmath61 and @xmath62 @xmath63 @xmath55 consistency with the riemann solver : : for any piecewise constant initial data @xmath64 , there exists @xmath65 such that , for all @xmath66 $ ] , the trajectory @xmath67 coincides with the solution of the cauchy problem obtained by piecing together the standard solutions for the riemann problems determined by the jumps of @xmath68 . following @xcite , in this paper , the general entropy solution @xmath69 of the cauchy problem under consideration is the srs , which can be constructed by the front tracking method ( _ cf . _ @xcite ) , and the entropy solution for the corresponding asymptotic scalar equation is constructed by polygonal approximations , first introduced in dafermos @xcite , with initial data : @xmath70 we establish the @xmath71-estimate between @xmath69 and @xmath72 by using both the properties of the wave - front tracking algorithm and the standard error formula ( _ cf . _ @xcite ) : @xmath73 where @xmath74 is the lipschitz constant of the semigroup @xmath75 , and @xmath76 is any lipschitz continuous map defined on @xmath77 $ ] . one of our objectives here is to develop a new approach to provide a rigorous mathematical proof of the following theorem . [ thm : compacted initial data]*(main theorem ) . * let @xmath78 , and let @xmath79 be an arbitrary function of bounded variation with compact support . assume that each eigenvalue of the hyperbolic system in has constant multiplicity and its corresponding characteristic field is either genuinely nonlinear or linearly degenerate . consider an entropy solution @xmath80 of the cauchy problem , which is a srs , and the weakly nonlinear geometric optics expansion function @xmath81 defined by and . then there exists @xmath82 such that , for all @xmath83 and @xmath84 $ ] , @xmath85 for some @xmath86 independent of @xmath87 and @xmath88 . we remark here that this result allows the eigenvalues of the @xmath89 hyperbolic system in to have constant multiplicity and the corresponding characteristic fields to be linearly degenerate , which answers the open problem posed by majda in @xcite . in particular , for the @xmath89 system in , we obtain that the @xmath0estimate between the entropy solution and the geometric optics expansion function is bounded by @xmath1 that is _ independent of @xmath90_. the proof of theorem [ thm : compacted initial data ] is based on our following observation : for a genuinely nonlinear system with initial data of compact support , the waves of different families in the solution will be separated each other . this enables us to follow majda - rosaales @xcite and use the front tracking method . we introduce an auxiliary approximate function @xmath91 by adding higher order term of @xmath92 to as @xmath93 and its corresponding more accurate auxiliary approximate function : @xmath94 where @xmath95 means that the corresponding @xmath96th characteristic field is genuinely nonlinear , while @xmath97 means that the corresponding characteristic field is linearly degenerate and all @xmath98 together constitute the @xmath99-th characteristic field ; furthermore , @xmath100 is given in [ sec : scalar scheme ] and @xmath101 is defined in 5 . the novelty here is that the new correction terms are introduced to deal with the contact discontinuities . with this key observation , then our approach is to prove the @xmath71distance between this auxiliary function and the general entropy solution to system with the same initial data is @xmath102 , and finally to employ the @xmath71stability of solutions with respect to initial data to establish theorem [ thm : compacted initial data ] . the complete proof of theorem [ thm : compacted initial data ] will be given in [ sec : main thm ] and [ sec : compact support ] . this provides an alternative approach to deal with nonlinear geometric optics for hyperbolic systems of conservation laws in . as an example of further applications of this approach , we extend the result to the case when the initial data has non - compact support . [ thm : noncompact case ] let @xmath78 and @xmath103 . assume that each real eigenvalue of the hyperbolic system in has constant multiplicity and its corresponding characteristic field is either genuinely nonlinear or linearly degenerate . consider an entropy solution @xmath80 of the cauchy problem , which is the srs , and the weakly nonlinear geometric optics expansion @xmath81 defined by . then @xmath104 for related earlier results in this direction , we refer the reader to diperna - majda @xcite for an order of @xmath105 for the case of periodic initial data and the same estimate for initial data with compact support for @xmath106 genuinely nonlinear and strictly hyperbolic systems of conservation laws . for general strictly hyperbolic systems with some kind of periodic properties of the initial data for which the resonance phenomena occur , schochet @xcite proved the @xmath71-estimate of order @xmath107 ; and cheverry @xcite dealt with more general initial data and proved that , for all @xmath108 , @xmath109 where @xmath110 is the corresponding geometric optics expansion . both of their results allow the characteristic fields to be linearly degenerate , but require the hyperbolic system in to be strictly hyperbolic . we also refer the reader to chen - junca - rascke @xcite , cheverry @xcite , gus @xcite , hunter - majda - rosales @xcite , joly - mtivier - rauch @xcite , majda - rosales @xcite , and the references cited therein for related results . for classical results on the front tracking method and hyperbolic systems of conservation laws , see bressan @xcite and dafermos @xcite . in this section , we analyze entropy solutions of hyperbolic systems of conservation laws and front tracking algorithms for scalar equations for subsequent development . consider the riemann problem of system with the following initial data : @xmath111 u_-,\qquad x<0 , \end{array } \right.\ ] ] where @xmath112 are constant vectors . based on the results in @xcite , we have [ prop : riemann systems ] assume that @xmath32 satisfies the same assumptions as in theorem [ thm : compacted initial data ] . then , for every compact set @xmath113 , there exists @xmath65 such that , whenever @xmath114 , the riemann problem above has a unique entropy solution , which consists of @xmath115 constant states , denoted by @xmath116 , and @xmath117 elementary waves ( shock or rarefaction waves corresponding to the genuinely nonlinear characteristic fields , or contact discontinuities to the linear degenerate fields ) . moreover , there exists a unique small parameter vector @xmath118 such that @xmath119 u_i=\ppsi_i(u_{i-1 } ; \beta_i),\qquad i=0,1,\cdots , m , m+p+1,\cdots , n,\\ \end{array}\ ] ] and @xmath120 where @xmath121 , and @xmath122 are smooth functions with respect to the respective parameter vectors and satisfy @xmath123 from lemma [ prop : riemann systems ] , one can follow the approach in @xcite and @xcite to obtain the existence and stability of the unique entropy solution that is the standard riemann semigroup ( srs ) . [ prop : stability ] assume that @xmath32 satisfies the same assumptions as in theorem [ thm : compacted initial data ] . then there is a suitably small @xmath124 such that , given any @xmath125 with @xmath126 , there exists an entropy solution @xmath127 by the wave - front tracking method or the glimm scheme . the map @xmath128 satisfies that , for all @xmath129 , @xmath130 & & \|s_t\bar{u}-s_s\bar{v}\|_{l^1}\le l_1\,\|\bar{u}-\bar{v}\|_{l^1}+l_2\,|t - s| \,\quad \mbox{for some constants $ l_1 $ and $ l_2$},\label{semigroup property}\end{aligned}\ ] ] such that the solution is a srs . furthermore , for any other srs @xmath131 , defined on a domain @xmath132 , we have @xmath133 in lemma [ prop : stability ] , the domain @xmath134 is defined by the glimm functional ; see the definition in @xcite ( chapter 7 , pp . 151 ) . we adopt the following scheme and related results from @xcite ; also see @xcite for the details . for our problem , we assume that the scalar flux function @xmath135 is a convex function . for fixed integer @xmath136 , we consider the piecewise constant initial data @xmath137 taking values within the discrete set @xmath138 and define @xmath139 to be the piecewise affine function that coincides with @xmath135 at all nodes @xmath140 with @xmath99 integer : @xmath141.\ ] ] for the scheme , we consider the cauchy problem : @xmath142 with initial data @xmath137 . first we consider the riemann problem with initial data @xmath137 as in and @xmath143 to obtain the following solutions : @xmath55 case 1 . @xmath144 : : we define the increasing sequence of jump speeds as @xmath145 where @xmath146 are the jump points of @xmath147 and satisfy @xmath148 . then we can obtain an entropy solution of the above riemann problem and as follows : @xmath149 @xmath55 case 2 . @xmath150 : : in this case , we define the shock speed as @xmath151 thus , we can also obtain an entropy solution as the previous case : @xmath152 next , consider a more general cauchy problem for with piecewise constant initial data @xmath137 , taking values within the set @xmath153 . we can construct the solution by solving the corresponding riemann problems so that the total number of interactions is finite and the solution can be prolonged for all @xmath108 . for these solutions @xmath154 , we have the following properties : @xmath155 & & \|u_{\nu}(t,\cdot)-u_{\nu}(t',\cdot)\|_{l^1}\le l\ , tv(\bar{u})\ , |t - t'| \,\quad \qquad \mbox{for any}\,\ , t , t'\ge 0,\end{aligned}\ ] ] where @xmath74 is the lipschitz constant such that @xmath156.\ ] ] then , using helly s theorem , we obtain an entropy solution @xmath157 defined for all @xmath158 by compactness , with @xmath159 we remark here that , since the scalar flux function in our case in 3 6 is quadratic , the front - tracking algorithm could be applied directly . however , we adopt the piecewise - linear approximation in our analysis in 36 so that it is more convenient to compare the solutions to present our approach . in this section we compare the riemann solvers to system and the geometric optics expansion defined by with the same initial data . from now on , we denote @xmath160 and @xmath161 through the paper . first , we consider some properties of approximate solutions to the burgers equation . [ lem : burger s riemann problem ] for some @xmath162 , assume that , for fixed @xmath163 , @xmath164 is an approximate solution to the inviscid burgers equation : @xmath165 constructed by the front tracking method . at the jump point @xmath166 , suppose that @xmath167 , @xmath168 , and @xmath169 is the corresponding point in the @xmath170-coordinates to the point @xmath166 . then 1 . if @xmath171 , the slope of the discontinuity line of solutions to the inviscid burgers equation in the @xmath170coordinates is @xmath172 2 . if @xmath173 , then @xmath174 , and the slope of the discontinuity line of solutions to the inviscid burgers equation in the @xmath170coordinates is @xmath175 first , we deduce from [ sec : scalar scheme ] that the slope of the discontinuity line of solutions to the inviscid burgers equation in the @xmath176coordinates is @xmath177 from the rankine - hugoniot condition when @xmath171 , and @xmath177 from when @xmath173 using the fact that , if @xmath170 is a point on the discontinuity line with the corresponding point @xmath176 in the @xmath176-coordinates , we have @xmath178 \tau=\varepsilon t , \end{array } \right . \qquad \mbox{and } \qquad \left\ { \begin{array}{l } \dot{s}_{b(t , x)}(\sigma)=\frac{x - x_0}{t - t_0},\quad \sigma < 0,\\[2 mm ] \lambda_{b(t , x)}^{(k)}(\sigma)=\frac{x - x_0}{t - t_0},\quad \sigma > 0 . \end{array } \right.\ ] ] this completes the proof . next , we consider the riemann problem to system with the corresponding initial data : @xmath179 [ lem : genius nonlinear riemann problem ] assume that @xmath32 satisfies all the assumptions as in theorem [ thm : compacted initial data ] , the @xmath180-th characteristic field is genuinely nonlinear , and @xmath181 has a jump point as lemma [ lem : burger s riemann problem ] . let @xmath182 be the corresponding point to @xmath166 , @xmath183 , and @xmath184 , where @xmath185 are constant valued functions near @xmath166 when @xmath186 . then we have @xmath187 where @xmath14 is the kronecker delta . let @xmath188 . from the definition , if @xmath191 , we deduce from @xmath192 that @xmath193 . we differentiate both sides of the above equation with respect to @xmath194 to obtain @xmath195 from lemma [ prop : riemann systems ] , we know @xmath196 then we find @xmath197 taking the dot product both sides of with @xmath198 , we obtain @xmath199 then we have @xmath200 this completes the proof . with lemmas [ lem : burger s riemann problem][lem : genius nonlinear riemann problem ] , we have [ lem : local l1 estimates for genius nonlinear ] assume that @xmath32 satisfies the assumptions as in theorem [ thm : compacted initial data ] , the @xmath180-th characteristic field is genuinely nonlinear , @xmath181 is a piecewise constant function as lemma [ lem : burger s riemann problem ] and has a jump point at @xmath166 , @xmath182 is the corresponding point , and @xmath166 is not a jump point of @xmath201 for @xmath17 . then , for every @xmath202 with @xmath203 being the maximum of the solution for , if @xmath204 is sufficiently small , we have @xmath205 where @xmath206 and @xmath86 is a constant independent of @xmath87 and @xmath88 . first , we consider the case @xmath171 . then , from lemma [ lem : genius nonlinear riemann problem ] , we obtain that @xmath207 when @xmath87 sufficiently small . thus , the @xmath180-th wave is a shock . from lemma [ prop : riemann systems ] and the rankine - hugoniot condition , we have @xmath208 where @xmath209 is the shock speed of the @xmath180-th shock with respect to @xmath210 , and @xmath211 is defined as in lemma [ prop : riemann systems ] . therefore , if @xmath212 and @xmath213 , we have @xmath214 \quad\quad\,\,\,\,\ , = \lambda_k(u_-)+\frac{1}{2}\beta_k+o(1)\beta^2 + o(1)\sigma(\sigma+\max_j|\sigma_{\nu_-}^{(j)}|)\,\varepsilon^2 , \end{array}\ ] ] by using lemmas [ prop : riemann systems ] and [ lem : genius nonlinear riemann problem ] and the following estimate : @xmath215 using lemma [ lem : genius nonlinear riemann problem ] , we have @xmath216 then we have @xmath217 = \lambda_k(u_-)+\frac{1}{2}\sigma\varepsilon + o(1)\sigma\big(\sigma+\max_j|\sigma_{\nu-}^{(j)}|\big)\varepsilon^2 -\big(\lambda_k^0+\sigma_{\nu-}^{(k)}\varepsilon + \frac{\sigma}{2}\varepsilon\big)\\[2 mm ] = \lambda_k(u_-)-\lambda_k^0 -\frac{1}{2}\sigma_{\nu-}^{(k)}\varepsilon + o(1)\sigma\big(\sigma+\max_j|\sigma_{\nu-}^{(j)}|\big)\varepsilon^2\\[2 mm ] = \varepsilon\sum_{j\ne k}\sigma_{\nu-}^{(j ) } + o(1)\sigma\max_j\big(|\sigma_{\nu-}^{(j)}|^2\big)\varepsilon^2 + o(1)\sigma\big(\sigma+\max_j|\sigma_{\nu-}^{(j)}|\big)\varepsilon^2\\[2 mm ] = o(1)\max_j|\sigma_{\nu-}^{(j)}|\ , \varepsilon . \end{array}\ ] ] as shown in figure [ fig : nonlinear shock ] , with estimate , and @xmath218 we have @xmath219 we now consider the case @xmath173 . as for the case @xmath171 , we obtain that @xmath220 when @xmath87 sufficiently small . then the @xmath180-th wave is a rarefaction wave . from lemmas [ prop : riemann systems ] and [ lem : genius nonlinear riemann problem ] , we have @xmath221 \lambda_k(u_{k})&=&\lambda_{k}(u_+)+o(1 ) \sigma\big(\sigma+\max_j|\sigma_{\nu-}^{(j)}|\big)\varepsilon^2,\nonumber\\[2 mm ] \lambda_k(u_{k})&=&\lambda_k(u_{k-1 } ) + \beta_k + o(1)\sigma\big(\sigma+\max_j|\sigma_{\nu-}^{(j)}|\big)\varepsilon^2\nonumber\\[2 mm ] & = & \lambda_k(u_{-})+\frac{1}{2}\sigma\varepsilon + o(1)\sigma\big(\sigma+\max_j|\sigma_{\nu-}^{(j)}|\big)\varepsilon^2 . \nonumber\end{aligned}\ ] ] these yield @xmath222 @xmath223 = \lambda_k(u_-)+\frac{1}{2}\sigma\varepsilon -\lambda^0_k-\sigma_{\nu-}^{(k)}\varepsilon-\frac{\sigma}{2}\varepsilon + o(1)\sigma\big(\sigma+\max_j|\sigma_{\nu-}^{(j)}|\big)\varepsilon^2\\[2 mm ] = \lambda_k(u^0+\varepsilon\sum_{j=1}^n\sigma_{\nu-}^{(j)}r_j^0 ) -\lambda_k^0-\sigma_{\nu-}^{(k)}\varepsilon + o(1)\sigma\big(\sigma+\max_j|\sigma_{\nu-}^{(j)}|\big)\varepsilon^2\\[2 mm ] = o(1)\varepsilon\sum_{j\ne k}(\rr_j\cdot\nabla)\lambda_k^0\sigma_{\nu-}^{(j ) } + o(1)\sigma\varepsilon+o(1)\sigma\big(\sigma+\max_j|\sigma_{\nu-}^{(j)}|\big)\varepsilon^2\\[2 mm ] = o(1)\big(\sigma+\max_j|\sigma_{\nu-}^{(j)}|\big)\varepsilon , \end{array}\ ] ] and @xmath224 = \lambda_k(u_-)-\lambda^0_k-\sigma_{\nu-}^{(k)}\varepsilon -\frac{\sigma}{2}\varepsilon+o(1)\sigma\big(\sigma+\max_j|\sigma_{\nu-}^{(j)}|\big)\varepsilon^2\\[2 mm ] = \lambda_k(u^0+\varepsilon\sum_{j=1}^n\sigma_{\nu-}^{(j)}\rr_j^0)-\lambda_k^0 -\sigma_{\nu-}^{(k)}\varepsilon -\frac{\sigma}{2}\varepsilon + o(1)\sigma\big(\sigma+\max_j|\sigma_{\nu-}^{(j)}|\big)\varepsilon^2\\[2 mm ] = o(1)\big(\sigma+\max_j|\sigma_{\nu-}^{(j)}|\big)\varepsilon , \end{array}\ ] ] where we have used the above estimates and the properties of the centered rarefaction waves : @xmath225 , \end{array } \right.\ ] ] when @xmath112 are connected by the @xmath180-rarefaction curve , and @xmath226 is a smooth function for @xmath227 and @xmath228 . as shown in figure [ fig : nonlinear rarefaction ] , similar to the case @xmath171 , we have @xmath229 this completes the proof . in this case , we first show that the characteristic fields whose eigenvalue has constant multiplicity , bigger than one , must be linearly degenerate . [ lem:3.3a ] assume that @xmath32 satisfies the assumptions in theorem [ thm : compacted initial data ] , and @xmath230 for @xmath231 . then the characteristic fields of @xmath232 must be linearly degenerate . assume that @xmath233 , and the corresponding eigenvectors @xmath234 are linearly independent . performing @xmath235 to both sides of the equation @xmath236 , we have @xmath237 where @xmath238 and @xmath17 . taking the dot product on both sides of with @xmath239 from the left and using the fact that @xmath240 , we have @xmath241 which implies @xmath242 next , taking the dot product on both sides of with @xmath243 from the left and using the fact that @xmath240 , we have @xmath244 & & = \lambda_k(u ) \ll_k(u)\cdot(\rr_j(u)\cdot\nabla)\rr_k(u)+(\rr_k(u)\cdot\nabla)\lambda_k(u),\end{aligned}\ ] ] which implies @xmath245 this completes the proof . now we consider the @xmath180th - characteristic field whose eigenvalue has constant multiplicity for @xmath246 in . in this case , the corresponding scalar case is quite simple , and the equations are reduced to be @xmath247 . thus , @xmath248 is a piecewise function and @xmath249 \sigma^{(k)}_{\nu- } \qquad \mbox{if}~y\le y_0 . \end{array } \right.\ ] ] then the slope of the discontinuity line in the @xmath250coordinates is @xmath251 next , we consider the riemann problem to system with the corresponding initial data as before . we obtain [ lem : linearly degerate ] assume that @xmath32 satisfies all the assumptions in theorem [ thm : compacted initial data ] , @xmath246 as in , @xmath181 has a jump at @xmath166 , and @xmath252 . let @xmath182 be the corresponding point to @xmath166 , @xmath183 , and @xmath184 . then we have @xmath253 let @xmath188 . from the definition , @xmath254 , that is , @xmath255 if @xmath191 , we deduce from @xmath256 that @xmath193 . we differentiate both sides of with respect to @xmath194 to obtain @xmath257 . from lemma [ prop : riemann systems ] , we know @xmath258 . thus , we have @xmath259 then we take the dot product on both sides of with @xmath198 to obtain @xmath260 and then @xmath261 this completes the proof . with lemmas [ lem:3.3a][lem : linearly degerate ] , we have [ lem : local l1 estimates for linearly degenerate ] assume that @xmath32 satisfies the assumptions in theorem [ thm : compacted initial data ] , either the @xmath180-th characteristic field is linearly degenerate or its eigenvalue has constant multiplicity , @xmath181 is a piecewise constant function which satisfies @xmath252 and has a jump point at @xmath166 , @xmath182 is the corresponding point , and @xmath166 is not a jump point of @xmath201 for @xmath17 . then , for every @xmath262 with @xmath203 being the maximum of the solution to , if @xmath204 is sufficiently small , we have @xmath263 where @xmath264 is defined in lemma [ lem : local l1 estimates for genius nonlinear ] , and @xmath86 is a constant independent of @xmath204 and @xmath87 . the proof of this lemma is similar to the one for lemma [ lem : local l1 estimates for genius nonlinear ] , especially for the shock case . the only change is about the shock speed @xmath209 . this can be calculated directly as follows : @xmath265 and then @xmath266 with this , we can obtain the desired estimate . we now prove theorem [ thm : compacted initial data ] when @xmath268 $ ] , where @xmath269 is defined as follows : since @xmath270 is a compact set for each @xmath99 ( due to the compactness of the support of @xmath271 ) and the speed @xmath272 for @xmath273 , there exists @xmath274 , independent of @xmath84 $ ] , such that , for @xmath275 , the compact sets @xmath276 are disjoint , that is , for @xmath275 , @xmath277 in order to establish theorem [ thm : compacted initial data ] for @xmath267 , a technical lemma is stated here to restrict the integral interval for the semigroup estimates to be finite . [ lem : local semigroup estimate ] let @xmath278 be the semigroup generated by system , and let @xmath279 be an upper bound for all the wave speeds , where @xmath32 satisfies all the assumptions in theorem [ thm : compacted initial data ] . given any interval @xmath280 $ ] , define @xmath281,\quad t<\frac{b - a}{2\hat{\lambda}}.\ ] ] then , for every lipschitz continuous map @xmath282\mapsto\mathcal{d}$ ] , @xmath283 first , for every @xmath284 and @xmath108 , we have @xmath285 this can be obtained by using the following facts : \(i ) if two initial conditions @xmath284 coincide on @xmath286 , then @xmath287 for all @xmath288 ; \(ii ) if @xmath64 , then @xmath289}\in\mathcal{d}$ ] . with this estimate , the other steps for the proof is the same as the ones for in @xcite ( theorem 2.9 ) . now we prove theorem [ thm : compacted initial data ] for @xmath267 . we divide the proof into three steps . the constant @xmath86 below is a universal bound independent of the parameters @xmath290 in the proof . since @xmath291 , we can construct piecewise constant functions @xmath292 , which satisfy @xmath293 \|\sigma^{(j)}_{\nu}(0,\cdot)-\sigma^{(j)}(0,\cdot)\|_{\infty } + \|\sigma^{(j)}_{\nu}(0,\cdot)-\sigma^{(j)}(0,\cdot)\|_{l^1}\le\varepsilon , \end{array}\ ] ] where @xmath294 . then @xmath295 and @xmath296 in addition , the jump points of each @xmath100 are finite , and the range of those functions is contained in the set @xmath153 , for some fixed integer @xmath163 . for this initial data , we can construct an approximate solution @xmath297 by the method introduced in [ sec : scalar scheme ] . then the solution @xmath297 satisfies @xmath298 & & \|\sigma_{\nu}^{(j)}(\tau,\cdot)\|_{l^{\infty}}\le c\,\|\sigma_{\nu}^{(j)}(0,\cdot)\|_{l^{\infty } } \le c \|u^1(\cdot)\|_{l^{\infty}},\label{4.2b}\\[2 mm ] & & \|\sigma^{(j)}_{\nu}(\tau,\cdot)-\sigma^{(j)}_{\nu}(s,\cdot)\|_{l^1}\le l|\tau - s| . \label{4.2c}\end{aligned}\ ] ] then , from helly s theorem , for every @xmath299 , there exists a subsequence of functions ( still denoted as ) @xmath300 so that @xmath301 and @xmath302 by standard argument , there exists a subsequence @xmath303 such that @xmath304 pointwise , which is also in @xmath305 , at every rational time @xmath306 . define @xmath307 where @xmath308 is a rational time and @xmath309 . then , by continuity of time in , we conclude that @xmath310 in @xmath311 , and @xmath312 satisfies @xmath313 first , consider the case that @xmath314 is a jump point of one and only one function of @xmath315 , say @xmath297 . from [ sec : riemann comparison ] , we can derive the following estimate : @xmath316 where @xmath317 , when @xmath204 is so small that there is no interaction , and @xmath318 contains only the wave fronts generated at point @xmath314 . we now consider the general case . without loss of generality , assume that @xmath314 is a jump point of @xmath201 and @xmath181 . _ claim . assume that @xmath32 satisfies all the assumptions in theorem [ thm : compacted initial data ] , @xmath99 and @xmath180 are two distinct integers , @xmath166 is a jump point of and only of @xmath201 and @xmath181 . let @xmath319 and @xmath320 @xmath182 is the corresponding point to @xmath166 , @xmath321 then we have @xmath322 & & + o(1)(|\sigma_{\nu+}^{(j)}-\sigma_{\nu-}^{(j)}|+|\sigma_{\nu+}^{(k)}-\sigma_{\nu-}^{(k)}| ) { ( \max_j|\sigma_{\nu-}^{(j)}|+\max_j|\sigma_{\nu+}^{(j)}|)}\,\varepsilon^2.\end{aligned}\ ] ] _ this property can be proved by combining the properties in the proof of lemma [ lem : genius nonlinear riemann problem ] and the ideas for the proof of lemma [ lem : linearly degerate ] . we omit the details . with this , we can show @xmath323 when @xmath204 is so small that there is no interaction and @xmath318 contains only the wave fronts generated at the point @xmath314 . the proof is similar to the one in propositions [ lem : local l1 estimates for genius nonlinear ] and [ lem : local l1 estimates for linearly degenerate ] ; the difference here is that we should divide the interval into three parts : the additional part is near the @xmath99th - wave . thus , we estimate not only @xmath324 or @xmath325 as in proposition [ lem : local l1 estimates for genius nonlinear ] or [ lem : local l1 estimates for linearly degenerate ] , but also @xmath326 or @xmath327 . with and , we can derive an estimate for arbitrary finite interval @xmath318 by dividing the interval into some subintervals , every one of which contains only one jump point . since the number of these subsets is finite , we can sum together to obtain @xmath328 where @xmath329 . . for an arbitrary finite interval @xmath330 $ ] , let @xmath331 $ ] . then , from the definition in lemma [ lem : local semigroup estimate ] , we have @xmath332 . for any initial data @xmath333 , @xmath312 and @xmath334 are defined from , , and . then , following step 1 , we construct piecewise constant functions @xmath297 so that @xmath335 & & \ ; + \|u^{\varepsilon}_{\nu}(t,\cdot)-u^{\varepsilon}_{w,\nu}(t,\cdot)\|_{l^1(j)}\\[2 mm ] & = & i_1+i_2+i_3.\end{aligned}\ ] ] for @xmath336 , from lemma [ prop : stability ] , @xmath337 & \le & l\|u^0+\varepsilon u^1(\cdot)-\big(u^0+\varepsilon\sum_{j=1}^n\sigma_{\nu}^{(j)}(0,\cdot)\rr_j^0\big)\|_{l^1}\\[2 mm ] & \le & c\sum_{j=1}^n\|\sigma^{(j)}(0,\cdot)-\sigma^{(j)}_{\nu}(0,\cdot)\|_{l^1}. \end{aligned}\ ] ] hence , @xmath338 when @xmath339 . for @xmath340 , it is reduced to estimate @xmath341 . from step 1 , we can select a rational time sequence @xmath342 and a subsequence ( still denoted ) @xmath343 such that @xmath344 when @xmath345 and , for all @xmath346 , @xmath347 for a.e . @xmath348 , when @xmath349 . then @xmath350 using the lipschitz continuity , we obtain that @xmath351 when @xmath349 . for @xmath352 , from lemma [ lem : local semigroup estimate ] , we obtain @xmath353 where we have used the fact that @xmath354 for all @xmath83 . with these three estimates together , passing limit @xmath349 , we obtain that , for any arbitrary finite interval @xmath355 $ ] , @xmath356 finally , let @xmath357 and @xmath358 , we have @xmath359 therefore , for @xmath267 , we obtain @xmath360 as a direct corollary , we conclude assume that @xmath3 satisfies the assumptions in theorem [ thm : compacted initial data ] , and @xmath361 ( not necessarily with compact support ) . consider an entropy solution @xmath80 of , which is the srs , and the weakly nonlinear geometric optics expansion @xmath81 defined by , , and . then there exists @xmath82 such that , when @xmath362 $ ] , we have @xmath363 however , estimate is not really strong enough to justify the weakly nonlinear geometric optics approximation as noted in schochet @xcite . in 6 , we further develop the approach to improve into the stronger error estimate in theorem 1.2 , which is strong enough indeed to justify the approximation even for the case when @xmath361 without compact support . notice from [ sec : main thm ] that , when @xmath275 , the compact sets @xmath364 , are disjoint : @xmath365 in this section , we give a careful calculation based on this and complete the proof of theorem [ thm : compacted initial data ] . the constant @xmath86 below is the universal constant , independent of @xmath366 in the proof . we first refine expansion or by the following corresponding approximation : @xmath367 where @xmath95 means that the corresponding @xmath96th characteristic field is genuinely nonlinear , while @xmath97 means that the corresponding characteristic field is linearly degenerate and all @xmath98 together constitute the @xmath99-th characteristic field . furthermore , @xmath368 and @xmath369 is defined as follows . first set @xmath370:=\mbox{supp}_{x}\sigma^{(j)}(\varepsilon t_0,x-\lambda_j^{0}t_0)$ ] and let @xmath371 be a point at which @xmath372 has a jump . denote the set of these jump points together by @xmath373 and then define the piecewise constant function @xmath374 with the same jump points in @xmath373 . at each jump point @xmath375 , if @xmath376 and the multiplicity of the @xmath99-th characteristic field is @xmath346 , then the difference of @xmath374 at this jump point is defined by @xmath377 for the point @xmath378 or @xmath379 , we define @xmath380 . then we have the property that @xmath381 . from the proof of lemma [ prop : cs ] later , we will see that the error term satisfies that @xmath382 . now we consider the case @xmath95 . as the bootstraps in [ sec : riemann comparison ] , we have the following lemma . [ lem : gn riemann compact support ] suppose that the assumptions in lemma [ lem : genius nonlinear riemann problem ] hold . let @xmath383 with @xmath384 . then we have @xmath385 where @xmath386 , and @xmath387 is the kronecker delta . let @xmath188 . then @xmath388 thus we have the following equation : @xmath389 clearly , by the implicit function theorem , we obtain a unique solution @xmath390 of such that @xmath391 . differentiating with respect to @xmath194 and then letting @xmath191 , we have @xmath392 since @xmath393 if @xmath394 , we find that @xmath395 if @xmath396 . using this fact to calculate , we have @xmath397 which yields @xmath398 next , taking twice derivatives on both sides of with respect to @xmath194 and then letting @xmath191 , then @xmath399 thus , we have @xmath400 noting that @xmath401 , we obtain @xmath402 combining with the taylor expansion , we finally have @xmath403 this completes the proof . with this better estimate , as in [ sec : riemann comparison ] , we further have [ prop : gn riemann compact support ] suppose that the assumptions in theorem [ thm : compacted initial data ] hold . then there exists @xmath404 such that , when @xmath405 $ ] , for @xmath406 , we have @xmath407 we divide the proof into two cases . @xmath171 . in this case , by lemma [ lem : gn riemann compact support ] , @xmath207 . denote @xmath408 the k - shock speed of @xmath210 in the @xmath170coordinates and @xmath409 . then @xmath410 on the other hand , the shock speed of the burgers equation of @xmath411 in the @xmath170coordinates is @xmath412 then @xmath413 meanwhile , with the fact that @xmath414 , we deduce the core estimate for the shock case : @xmath415 then , with this estimate in hand , following the bootstrap in the proof of proposition [ lem : local l1 estimates for genius nonlinear ] step by step , using the results in lemma [ lem : gn riemann compact support ] , and performing cumbersome calculations , we have @xmath416 : @xmath173 . from the scheme for the scalar equation in [ sec : scalar scheme ] , or by lemma [ lem : burger s riemann problem ] , we conclude that the strength of discontinuity satisfies the following property : @xmath417 first , from lemma [ lem : burger s riemann problem ] , the speed of rarefaction front of the burgers equation in the @xmath170coordinates is @xmath418 on the other hand , the speed of characteristics is @xmath419 thus , by following exactly the proof of proposition [ lem : local l1 estimates for genius nonlinear ] , direct computation yields that @xmath420 combining with , we arrive at the result . next , consider the linearly degenerate case , namely , @xmath421 . by construction , all the values of @xmath374 lie on the curve @xmath422 , _ i.e. _ the integral surface of the vector - fields spanned by @xmath423 passing through @xmath424 , while @xmath425 is the speed . here @xmath374 is introduced below to replace @xmath426 . thus , @xmath374 is exactly a riemann solution . then we have [ prop : ld riemann compact support ] suppose that the assumptions in theorem [ thm : compacted initial data ] hold . then there exists @xmath404 such that , when @xmath405 $ ] , for @xmath406 , we have @xmath427 with lemmas [ prop : gn riemann compact support ] and [ prop : ld riemann compact support ] in hand , summing up the above estimates and using , we have [ prop : v ] suppose that the assumptions in theorem [ thm : compacted initial data ] hold . then @xmath428 we first use the standard error formula and then let @xmath204 small enough such that @xmath429 is the solution obtained by piecing together the standard entropy solutions of the riemann problems determined by the jumps of @xmath430 . then we can use lemmas [ prop : gn riemann compact support ] and [ prop : ld riemann compact support ] to obtain @xmath431 & \leq l\int_{t_0}^t\liminf_{h\rightarrow0 + } \frac{\|s(h)v_{\nu}^{\varepsilon}(z)-v_{\nu}^{\varepsilon}(z+h)\|_{l^1}}{h}\mbox{d}z\\ & \leq l\int_{t_0}^t\liminf_{h\rightarrow0 + } \frac{\sum_{\sigma}|s(h)v_{\nu}^{\varepsilon}(z)-v_{\nu}^{\varepsilon}(z+h)|}{h}\mbox{d}z\\ & \leq o(1)\left(\int_{t_0}^{\max\{t_0,\,\varepsilon^{-1}\}}+\int_{\max\{t_0,\,\varepsilon^{-1}\}}^t\right ) \big(\varepsilon^3\sum_{k\in n}\big(tv(\sigma^{(k)}_\nu(\varepsilon z))\big)^3 + \varepsilon^22^{-\nu}\sum_{k\in n}tv(\sigma^{(k)}_\nu)\big)\mbox{d}z,\end{aligned}\ ] ] where @xmath432 is the sum of all the jumps at @xmath433 with jump strength @xmath227 . then , passing the limit @xmath434 , we have @xmath435 where we have used the fact that , for genuinely nonlinear scalar conservation laws , if the initial data @xmath436 has compact support and satisfies @xmath437 , then the solution @xmath438 satisfies @xmath439 where the constant @xmath440 depends only on @xmath436 . this completes the proof . since the estimates for the auxiliary function have been established , we are now at the stage to consider the estimates for the geometric optic expansion function @xmath441 . [ prop : cs ] suppose that the assumptions in theorem [ thm : compacted initial data ] hold . then @xmath442 this is in fact a simple corollary of lemma [ prop : v ] . notice that @xmath443 from lemma [ lem : linearly degerate ] and the fact that the discontinuities of @xmath374 and @xmath444 share the same speed @xmath425 , we notice that , for any @xmath445 , @xmath446 and @xmath447 & & w^{(j)}_{\nu}(x-\lambda_j^0(t - t_0))=w^{(j)}_{\nu}(x).\end{aligned}\ ] ] thus , we have @xmath448 where @xmath449 means that there is a jump point of @xmath450 at which the strength is @xmath451 . thus , by lemma [ prop : stability ] , @xmath452 & \leq \|s(t - t_0)u^{\varepsilon}_{w,\nu}(t_0)-s(t - t_0)v_{\nu}^{\varepsilon}(t_0)\|_{l^1 } + \|s(t - t_0)v_{\nu}^{\varepsilon}(t_0)-v^{\varepsilon}_{\nu}(t)\|_{l^1}\\ & \quad + \|v^{\varepsilon}_{\nu}(t)-u^{\varepsilon}_{w,\nu}(t)\|_{l^1}\\[2 mm ] & \leq l \|v^{\varepsilon}_{\nu}(t_0)-u^{\varepsilon}_{w,\nu}(t_0)\|_{l^1(\mathrm{r } ) } + \|s(t - t_0)v_{\nu}^{\varepsilon}(t_0)-v^{\varepsilon}_{\nu}(t)\|_{l^1 } + \|v^{\varepsilon}_{\nu}(t)-u^{\varepsilon}_{w,\nu}(t)\|_{l^1}\\[2 mm ] & \leq \|s(t - t_0)v_{\nu}^{\varepsilon}(t_0)-v^{\varepsilon}_{\nu}(t)\|_{l^1 } + c\,\varepsilon^2\sum_{j\in n}\|\sigma^{(j)}_{\nu}(\varepsilon t_0 , \cdot-\lambda_j^0t_0)\|_{l^\infty } \|\sigma^{(j)}_{\nu}(\varepsilon t_0 , \cdot-\lambda_j^0t_0)\|_{l^1 } \\ & \quad + c\,\varepsilon^2\sum_{j\in n}\|\sigma^{(j)}_{\nu}(\varepsilon t , \cdot-\lambda_j^0t)\|_{l^\infty } \|\sigma^{(j)}_{\nu}(\varepsilon t , \cdot-\lambda_j^0t)\|_{l^1 } + c\,\varepsilon^2\sum_{j\in l}|\mbox{supp}_x\sigma_{\nu}^{(j)}|\\ & \leq \|s(t - t_0)v_{\nu}^{\varepsilon}(t_0)-v^{\varepsilon}_{\nu}(t)\|_{l^1 } + c\varepsilon^2.\end{aligned}\ ] ] then , passing the limit @xmath349 and using lemma [ prop : v ] , we have @xmath453 which completes the proof . with lemmas [ lem : gn riemann compact support][prop : cs ] , we can now complete the proof of theorem [ thm : compacted initial data ] . notice that @xmath454 & \leq \|s(t)u^{\varepsilon}_w(0)-s(t - t_0)u^{\varepsilon}_w(t_0)\|_{l^1 } + \|s(t - t_0)u^{\varepsilon}_w(t_0)-u^{\varepsilon}_w(t)\|_{l^1}\\[2 mm ] & = i_1+i_2.\end{aligned}\ ] ] using in [ sec : main thm ] , we have @xmath455 & \leq l\|s(t_0)u^{\varepsilon}_w(0)-u^{\varepsilon}_w(t_0)\|_{l^1}\\[2 mm ] & \leq c\,t_0\varepsilon^2.\end{aligned}\ ] ] by lemma [ prop : cs ] , @xmath456 therefore , we conclude @xmath457 this completes the proof . following the approach developed in 35 , we can extend the result even for @xmath458 initial data , theorem [ thm : noncompact case ] . from the properties of @xmath464 , we know that @xmath465 and , near the jump point @xmath170 , the speed of discontinuity of @xmath466 is the same as the one of @xmath467 . in order to use the standard riemann semigroup , we need to show that the error terms @xmath468 are lipschitz - continuous with respect to @xmath88 . indeed , we have @xmath469 \cdot[\mbox{total strength of all wave - fronts of } \sigma_{\nu}^{j}(\varepsilon s , \cdot)]\\ & \leq c(t - s)\end{aligned}\ ] ] for some constant @xmath440 independent on @xmath163 . then , as before , we modify expansion by the following corresponding approximation : @xmath470 [ lem : riemann noncompact ] suppose that the assumptions in theorem [ thm : noncompact case ] hold and , at point @xmath460 , @xmath471 has a jump . let @xmath472 with @xmath384 . then @xmath473 where @xmath386 , and @xmath387 is the kronecker delta . differentiating with respect to @xmath194 and letting @xmath191 , we have @xmath476 notice that @xmath477 . then we obtain from that @xmath478 which yields @xmath479 next , takeing the twice derivatives on with respect to @xmath194 and letting @xmath191 yield @xmath480 thus , we have @xmath481 noticing that @xmath482 , we have @xmath483 with all of these formulas above , by the taylor expansion , we finally have @xmath484 = & \sigma\varepsilon\delta_{kj}+o(1)\varepsilon^3\sum_i\big((\sigma_{\nu-}^{i})^2\sigma+\sigma_{\nu-}^{(i)}\sigma^2+\sigma^3\big).\end{aligned}\ ] ] it completes the proof of this lemma . [ prop : riemann noncompact ] suppose that the assumptions in theorem [ thm : noncompact case ] hold , and the jump strength of @xmath485 is @xmath486 . then there exists @xmath404 such that , when @xmath487 , we have @xmath488 & & \leq c\sum_{\lambda_j^0\neq\lambda_k^0 } |\sigma_{\nu-}^{(j)}||\sigma_k|h \varepsilon^2 + c\big(\sum_j|\sigma_{\nu-}^{(j)}|\big)^2|\sigma_k|h \varepsilon^3 + c 2^{-\nu}|\sigma_k|h \varepsilon^2.\end{aligned}\ ] ] for the rarefaction case , the proof is the same as the one in [ sec : compact support ] , which needs use the crucial property of the wave strength in the scheme that @xmath489 . thus , it suffices to consider the other two cases . consider @xmath490 for the shock . by lemma [ lem : riemann noncompact ] , denote @xmath408 the k - shock speed of @xmath210 in the @xmath170coordinates and @xmath409 . then @xmath491 & = & \lambda_k^0+\varepsilon\sigma_{\nu-}^{(k)}+\frac{\sigma_k\varepsilon}{2 } + o(1)\varepsilon\sum_{j\neq k } \sigma_{\nu-}^{(j ) } + o(1)\varepsilon^2(\sigma^2_k+\varepsilon|\sigma_k|).\end{aligned}\ ] ] on the other hand , the shock speed of the burgers equation of @xmath485 in the @xmath170coordinates is @xmath492 then @xmath493 meanwhile , with the fact that @xmath494 , we deduce the core estimate for this case : @xmath495 consider the case of contact discontinuity whose corresponding characteristic field has constant multiplicity : @xmath496 if @xmath497 . then the kth - shock speed @xmath408 of @xmath210 in the @xmath170coordinates is @xmath498 on the other hand , the shock speed of the burgers equation of @xmath499 in the @xmath170coordinates is @xmath500 then we can obtain the same core estimate for this case as for the shock case . with these estimates in hand , following the bootstrap in [ sec : compact support ] step by step , we have @xmath501 & & \leq c \sum_{\lambda_j^0\neq\lambda_k^0 } |\sigma_{\nu-}^{(j)}||\sigma_k|h \varepsilon^2 + c\,\big(\sum_{j}|\sigma_{\nu-}^{(j)}|\big)^2|\sigma_k|h\varepsilon^3.\end{aligned}\ ] ] [ lem : noncompactproperty ] suppose that the assumptions in theorem [ thm : noncompact case ] hold . then , for @xmath83 , we have @xmath502 where @xmath503 means that , at the point @xmath504 , @xmath485 has a jump with the corresponding jump strength @xmath505 . this can be achieved by first using the standard semigroup error formula , then letting @xmath204 small enough such that @xmath429 is the solution obtained by piecing together the standard entropy solutions of the riemann problems determined by the jumps of @xmath430 , and finally using lemma [ prop : riemann noncompact ] . this is a simple corollary of lemma [ lem : noncompactproperty ] . notice that @xmath508 and that @xmath509 . thus , by lemma [ prop : stability ] , we have @xmath510 & & \leq \|s(t)u^{\varepsilon}_{w,\nu}(0)-s(t)v_{\nu}^{\varepsilon}(0)\|_{l^1(\mathrm{r } ) } + \|s(t)v_{\nu}^{\varepsilon}(0)-v^{\varepsilon}_{\nu}(t)\|_{l^1(\mathrm{r})}\\[1.5 mm ] & & \quad + \|v^{\varepsilon}_{\nu}(t)-u^{\varepsilon}_{w,\nu}(t)\|_{l^1(\mathrm{r})}\\[2 mm ] & & \leq l \|v^{\varepsilon}_{\nu}(0)-u^{\varepsilon}_{w,\nu}(0)\|_{l^1(\mathrm{r } ) } + c\,(\varepsilon t+2^{-\nu}t)\varepsilon^2 + \|v^{\varepsilon}_{\nu}(t ) -u^{\varepsilon}_{w,\nu}(t)\|_{l^1(\mathrm{r})}\\ & & \quad + c\varepsilon^2 \int_{0}^{t}\sum_{k}\big(\sum_{x\in j_k(s)}\sum_{\lambda_j^0\neq\lambda_k^0 } |\sigma^{(j)}_{\nu}(\varepsilon s , x-\lambda_js)||\sigma_k(x)|\big)\mbox{d}s\\ & & \leq c\,(1+\varepsilon t+2^{-\nu}(t - t_0))\varepsilon^2 + c\varepsilon^2 \int_{0}^{t}\sum_{k}\big(\sum_{x\in j_k(s)}\sum_{\lambda_j^0\neq\lambda_k^0}|\sigma^{(j)}_{\nu}(\varepsilon s , x-\lambda_js)||\sigma_k(x)|\big)\mbox{d}s.\end{aligned}\ ] ] then , passing the limit @xmath349 , we have @xmath511 which completes the proof . it suffices to prove @xmath512 where @xmath506 . let @xmath513 . then @xmath514 since @xmath515 and @xmath516 , then , for any given @xmath517 and any @xmath518 , we have @xmath519 then , for any @xmath520 , there exists a constant @xmath521 such that , for any @xmath522 , @xmath523 . next , let @xmath524 small enough such that , for any @xmath516 , @xmath525 thus , for any @xmath524 small enough , set @xmath526 then we have @xmath527 thus , we obtain that , for any @xmath517 , @xmath528 meanwhile , let @xmath529 where the constant @xmath530 depends only on the dimension @xmath7 . then @xmath531 notice that @xmath532 then , by the dominant convergence theorem , we have @xmath533 therefore , we conclude @xmath534 the authors thank the referees for valuable suggestions and comments . the research of gui - qiang chen was supported in part by the national science foundation under grant dms-0807551 , the uk epsrc science and innovation award to the oxford centre for nonlinear pde ( ep / e035027/1 ) , the nsfc under a joint project grant 10728101 , and the royal society wolfson research merit award ( uk ) . wei xiang was supported in part by the china scholarship council no . 2008631071 while visiting the university of oxford and the doctoral program foundation of the ministry education of china . yongqian zhang was supported in part by nsfc project 11031001 , nsfc project 11121101 , and the 111 project b08018 ( china ) . cheverry , c. ( 1997 ) . justification de loptique gomtrique non linaire pour un systme de lois de conservation [ justification of nonlinear geometric optics for a system of conservation laws ] . ( in french ) . majda , a. ( 1986 ) . nonlinear goemetric optics for hyperbolic systems of conservation laws . in : _ oscillation theory , computation , and methods of compensated compactness ( minneapolis , minn . , 1985 ) _ , 115165 , i m a vol . 2 , springer : new york .

we simulate the dilute attractive fermi - hubbard model in the unitarity regime using a diagrammatic determinant monte carlo algorithm with worm - type updates . we obtain the dependence of the critical temperature on the filling factor and , by extrapolating to , determine the universal critical temperature of the continuum unitary fermi gas in units of fermi energy : . we also determine the thermodynamic functions and show how the monte carlo results can be used for accurate thermometry of a trapped unitary gas .

cond-mat0605350

in recent years , ultracold atomic systems have served as a controlled and tunable toolbox for studying many - body quantum phenomena . the continuous tunability of the interaction by feshbach resonances makes these systems ideal candidates to study the crossover from momentum space pairing in the bardeen schrieffer ( bcs ) theory to bose - einstein condensate ( bec ) of fermions bound into molecules . this bcs - bec crossover has been one of the most studied problems in recent experiments in both magnetic or optical traps @xcite and optical lattices @xcite . tuning across the feshbach resonance , one traverses the whole range of the gas parameter @xmath3 , where @xmath4 is the fermi momentum and @xmath5 is the s - wave scattering length . the regime of @xmath6 ( negative @xmath5 ) is described by bardeen schrieffer ( bcs ) theory . at @xmath7 ( positive @xmath5 ) fermions pair into bosonic molecules and form a bose - einstein condensate ( bec ) . at the microscopic scale , the bcs and bec regimes are radically different ; however , the macroscopic , and , in particular , critical behaviour is supposed to be qualitatively the same for the whole range of @xmath3 : the system undergoes the superfluid ( sf ) phase transition at a certain critical temperature . separating the bcs and bec extremes is the so - called unitarity point @xmath8 . it is worth noting that the unitarity regime is approximately realized in the inner crust of the neutron stars , where the neutron - neutron scattering length is nearly an order of magnitude larger than the mean interparticle separation @xcite . the fermi gas at unitarity is a peculiar case of a strongly interacting system with no interaction - related energy scale : the divergent scattering length and any related energy scale drop out completely . this gives rise to universality of the dilute gas properties , in the sense that the only relevant energy scale left in the system is given by the density , @xmath9 . because of this universality one obtains a unified description of such diverse systems as cold atoms in magnetic or optical traps , fermi - hubbard model in optical lattices and inner crusts of neutron stars . the theoretical description of the fermi gas in the bcs - bec crossover regime is a major challenge , since the system features no small parameter on which one could build a theory in a rigorous way . the original analytical treatments were confined to zero temperature and were based on the extension of the bcs - type many - body wave function @xcite . most of the subsequent elaborations are also of mean - field type ( with or without remedies for the effects of fluctuations ) @xcite . the accuracy and reliability of such approximations is questionable since they inevitably involve an uncontrollable approximation . numerical investigations of the unitary fermi gases are hampered by the sign problem , inherent to any monte carlo ( mc ) simulations of fermion systems @xcite . one way of avoiding the sign problem at the expense of a systematic error is the fixed - node monte carlo framework , which has been used to study the ground state @xcite . the systematic error of the fixed - node monte carlo depends on the quality of the variational ansatz for the nodal structure of the many - body wave function and is not known precisely . only in a few exceptional cases can the sign problem be avoided without incurring systematic errors . one of such cases is given by fermions with attractive contact interaction , for which a number of sign - problem - free schemes has been introduced @xcite . fortunately , this system can be tuned to the unitarity regime . still , despite a number of calculations at finite temperature @xcite , an accurate description of the finite - temperature properties of the unitary fermi gas is missing . in the present paper , we simulate the fermi - hubbard model in the unitary regime by means of a determinant diagrammatic monte carlo method . by studying the dilute limit of the model , we extract properties of the homogeneous continuum fermi gas . a brief summary of the main results has been given in ref . @xcite . here we provide a detailed description of the monte carlo scheme and methods of data analysis . we also report new results relevant to experimental realizations of the fermi - hubbard model in optical lattices and trapped fermi gases . the fermi - hubbard model is defined by the hamiltonian @xmath10 @xmath11 @xmath12 here @xmath13 is a fermion creation operator , @xmath14 , @xmath15 is the spin index , @xmath16 enumerates @xmath17 sites of the three - dimensional ( 3d ) simple cubic lattice with periodic boundary conditions , the quasimomentum @xmath18 spans the corresponding brillouin zone , @xmath19 is the single - particle tight - binding spectrum , @xmath20 and @xmath21 are the lattice spacing and the hopping amplitude , respectively , @xmath22 stands for the chemical potential and @xmath23 is the on - site attraction . without loss of generality we henceforth set @xmath20 and @xmath21 equal to unity ; the effective mass at the bottom of the band is then @xmath24 . in sec . [ sec:2body ] we will study the two - body problem at zero temperature , show how the hubbard model can be used to study the continuum unitary gas and investigate the functional structure of lattice corrections to the continuum behaviour . in sec.[sec : ddmc ] we discuss the finite - temperature diagrammatic expansion for the hubbard model ( sec . [ ssec : matsubara ] ) , and give a qualitative description of the monte carlo procedure to sum the diagrammatic series ( sec.[ssec : mc ] ) , with details of the updating procedures given in appendix [ sec : updates ] . in order to extract the thermodynamic limit properties from mc data , we use finite - size scaling analysis described in sec . [ sec : sense ] . section [ sec : thermodynamics ] gives an overview of the scaling functions describing thermodynamics of the unitary gas , and results are presented and discussed in sec . [ sec : results ] . ( heavy dot ) . small dots represent the bare hubbard interaction @xmath25 , and lines are the single - particle propagators . ] consider a quantum mechanical problem of two fermions at zero temperature described by the hamiltonian ( [ ahm:1])([ahm:3 ] ) with @xmath26 . the most straightforward way to tackle this problem is within the diagrammatic technique in the momentum - frequency representation @xcite , which , in the present case , is built on four - point vertices , @xmath25 , with two incoming ( spin-@xmath27 and spin-@xmath28 ) and two outgoing ( spin-@xmath27 and spin-@xmath28 ) ends , connected by single - particle propagators . the scattering of two particles is then described by a series of ladder diagrams @xcite shown in . ladder diagrams can be summed by introducing the vertex insertion @xmath29 , which depends on frequency @xmath30 and momentum @xmath31 . since @xmath32 is proportional to the scattering amplitude , the unitarity limit corresponds to @xmath33 . the summation depicted in leads to @xmath34 where @xmath35 is the polarization operator ( the integration is over the brillouin zone ) : @xmath36 it immediately follows from eq . ( [ dyson ] ) and ( [ polariz ] ) that the unitary limit corresponds to @xmath37 , where @xmath38 a straightforward numeric integration yields @xmath39 . in the limit of vanishing filling factor @xmath1 for the many - body problem of ( [ ahm:1])([ahm:3 ] ) the typical values of @xmath40 and @xmath41 are related to the fermi energy @xmath42 and fermi momentum @xmath43 and are small compared to the bandwidth and reciprocal lattice vector , respectively . in zero - th order with respect to @xmath0 , the lattice system is identical to the continuum one . indeed , by combining ( [ dyson ] ) , ( [ polariz ] ) and ( [ u - unitary ] ) , we get @xmath44\ ; , % \label{gamma - latt}\ ] ] and observe that for small @xmath40 and @xmath41 one can replace @xmath45 with @xmath46 and extend integration over @xmath47 to the whole momentum space with the result @xmath48 with this form of @xmath49 , and particle propagators based on the parabolic dispersion relation one recovers the continuum limit behaviour now we are in a position to treat the lattice corrections . the first correction should come from @xmath50 , not from propagators , since only in @xmath50 large momenta play a special role due to resonance in the two - particle channel . in the lowest non - vanishing order in @xmath30 and @xmath31 we have ( the summation over repeating subscripts is implied ) @xmath51 with the difference between the lattice and continuous model given by @xmath52 where @xmath53 in the limit of @xmath54 and @xmath55 , we have @xmath56 , and @xmath57 . hence , the leading lattice correction is of the form @xmath58 incidentally , eqs . ( [ gamma - a ] ) and ( [ gamma - b ] ) hint into an intriguing possibility of completely suppressing the leading - order lattice correction by tuning the single - particle spectrum @xmath59 so that @xmath60 . we did not explore this possibility in the present study . the diagrammatic technique employed in the previous section is not particularly convenient for numerical studies . in this section , we briefly review the matsubara technique and then present a monte carlo scheme of summing the resultant diagrammatic series . to construct a diagrammatic expansion for the model ( [ ahm:1])([ahm:3 ] ) we follow refs . @xcite and consider the statistical operator in the coordinate imaginary time representation . in the interaction picture we get : @xmath61 where @xmath62 is an inverse temperature , @xmath63 , and @xmath64 stands for the imaginary time ordering . expanding eq . ( [ statoper ] ) in powers of @xmath65 , one obtains for the partition function : @xmath66\ ; . % \label{z - diagr}\ ] ] expansion ( [ z - diagr ] ) generates the standard set of feynman diagrams . graphically , the diagrams are similar to those of sec.[sec:2body ] , and consist of the four - point vertices , @xmath25 , connected by the single - particle propagators @xmath67 $ ] . the @xmath41-th order of the perturbation theory is then graphically given by a set of @xmath68 possible interconnections of vertices by propagators , see . ) , lower line depicts eq . ( [ z - summed ] ) . diagram signs are shown explicitly . ] the diagrammatic expansion ( [ z - diagr ] ) is _ unsuitable _ for the direct monte carlo simulation since it has a sign problem built in : different terms in the series have different signs a closed fermion loop brings in an extra minus sign @xcite . the trick is to consider all diagrams of a given order @xmath41 with the fixed vertex configuration @xmath69 as one . this implies summation over the @xmath68 ways of connecting vertices by propagators . upon summation , eq.([z - diagr ] ) takes on the form @xcite : @xmath70 where @xmath71 and @xmath72 are the @xmath73 matrices built on the single - particle propagators : @xmath74 for equal number of spin - up and spin - down particles latexmath:[$\det \mathbf{a}^\uparrow \det \mathbf{a}^\downarrow = | \det \mathbf{a } sign of @xmath25 changes if the hole representation is used for one of the spin components . hence , this method is also applicable to the half - filled repulsive hubbard model . ] graphically , feynman diagrams in this representation are just collections of vertices , see . for future use , we define the set of all possible vertex configurations ( [ conf - p ] ) by @xmath76 , _ i.e. , _ @xmath77 . the following two - point pair correlation function will prove useful : @xmath78 @xmath79 where @xmath80 and @xmath81 are the pair annihilation and creation operators in the heisenberg picture , respectively : @xmath82 . the non - zero asymptotic value of @xmath83 as @xmath84 is proportional to the condensate density . feynman diagrams for @xmath85 are similar to those for @xmath86 , but contain two extra elements : a pair of two - point vertices with two incoming ( outgoing ) ends which represent @xmath80 ( @xmath81 ) , see . the vertex configurations for the correlation function ( [ corr ] ) slightly differ from those for the partition function ( [ conf - p ] ) by the presence of the two extra elements : the configuration space for eq . ( [ corr ] ) is @xmath87 , with @xmath88 the partially summed diagrammatic expansion for @xmath85 is similar to eq.([z - summed ] ) : @xmath89 where @xmath90 is a @xmath91 matrix which differs from eq . ( [ matrix ] ) only by that it has an extra row @xmath92 and an extra column @xmath93 such that @xmath94 and @xmath95 . ) . diamonds represent the two - point creation / annihilation operators @xmath96 and @xmath97 . ] below we deal only with equal number of spin - up and spin - down fermions and ( for the sake of brevity ) suppress the spin indices wherever possible . we also peruse the generic notation ( with superscripts ) @xmath98 for @xmath41-th order terms of the diagrammatic expansions similar to ( [ z - summed ] ) , e.g. , @xmath99 . to simplify the notation we also omit superscripts if this does not lead to ambiguity . equations ( [ z - summed ] ) and ( [ g - summed ] ) have similar general structure of a series of integrals and sums with ever increasing number of integration variables and summations . in refs . @xcite it has been shown how to arrange a numerical procedure that sums such convergent series . to this end one considers the space of all possible vertex configurations @xmath100 ( for the series ( [ z - summed ] ) @xmath101 , while for the series ( [ g - summed ] ) @xmath102 ) , with the `` weight '' @xmath98 associated with each element of the space . one then uses the metropolis principle @xcite to arrange a stochastic markov process which sequentially generates vertex configurations @xmath103 according to their weights @xmath98 , thus sampling the space @xmath100 . in the course of sampling , one also collects statistics for observables in the form of mc estimators , see sec.[ssec : estimators ] . the stochastic process consists of elementary mc updates performed on vertex configurations @xmath103 . the set of updates is problem - specific being restricted only by the requirements of ( i ) the ergodicity , _ i.e. , _ given a particular diagram @xmath103 it takes a finite number of steps to convert it into any other diagram @xmath104 , and ( ii ) detailed balance , _ i.e. , _ relative contributions of diagrams @xmath103 and @xmath104 to the statistics is given by the ratio of their weights , @xmath105 . the set of updates satisfying these requirements is not unique , the freedom is used to maximize the efficiency of simulations , as explained in detail in [ sec : updates ] . in view of close similarity between the diagrammatic expansions ( [ z - summed ] ) and ( [ g - summed ] ) it is advantageous to construct a monte carlo process which samples these two series in a single simulation . this way one has access to both diagonal , _ e.g. , _ energy , and off - diagonal properties , _ e.g. , _ the superfluid response . an efficient way of performing such concurrent simulation is provided by the worm algorithm , which was originally devised for the worldline monte carlo simulations @xcite . in the context of the diagrammatic determinant monte carlo , the generic worm algorithm principles imply the following . first , one works in the joint configuration space @xmath106 , accommodating diagrams ( [ conf - p ] ) and ( [ conf - p - corr ] ) . second , all the updates are performed _ exclusively _ in terms of the two - point vertices @xmath80 and @xmath107through their creation / annihilation , motion , and interactions " with adjacent vertices . the worm - type updating procedures are further detailed in [ sec : updates ] . within the worm - algorithm framework , the configuration spaces @xmath76 and @xmath108 are disjoint subsets of one extended configuration space . in what follows we will refer to them as @xmath86-(or `` diagonal '' ) and @xmath109-(or `` off - diagonal '' ) sectors of the configuration space . formally , the extended configuration space corresponds to the generalized partition function @xmath110 where the value of @xmath111 is arbitrary it controls the relative statistics of @xmath86- and @xmath109-sectors and the efficiency of simulation . suppose we have an observable @xmath112 , which depends on a set of variables @xmath113 , _ e.g. _ temperature and chemical potential . a mc estimator for the observable @xmath114 is an expression which , upon averaging over the sequence of mc configurations converges to the expectation value of @xmath112 . in accordance with eq . ( [ z - worm ] ) , the simplest worm - algorithm mc estimators are @xmath115 and @xmath116 their mc averages are @xmath117 and @xmath118 where @xmath119 denotes averaging over the set of stochastically generated configurations . in particular , for our purposes it will be quite useful that @xmath120 the general rules for constructing an estimator for a quantity @xmath112 specified by the diagrammatic expansion @xmath121 are standard . we adopt a convenient convention : if the actual summation in ( [ xy ] ) involves only a subset @xmath122 of vertex configurations a typical example is an expansion defined within the @xmath86-sector only then we extend summation over the entire configuration space by simply defining @xmath123 . if vertex configurations @xmath103 are sampled from the probability density @xmath124 which comes from the expansion for the generalized partition function : @xmath125 then the mc estimator for @xmath112 is derived from @xmath126 as @xmath127 in what follows , by the estimator for a quantity @xmath128 we basically mean corresponding function @xmath129 . we start with the estimator for the number density . the expectation value of the number density reads @xmath130 here @xmath131 is an arbitrary space - time point ( the system is space / time translational invariant ) and @xmath132 is one of the two spin projections ; the factor of 2 comes from the spin summation . the diagrammatic expansion of the numerator is similar to that for @xmath86 , with the diagram weight given by @xmath133 here @xmath134 , @xmath135 is a @xmath73 matrix ( [ matrix ] ) , and @xmath136 is a similar @xmath137 matrix with an extra row and a column , corresponding to the extra creation and annihilation operators in the numerator of ( [ dens - gibbs ] ) , respectively . this immediately leads to the following estimator ( [ esti-2 ] ) for the number density @xmath138 we utilize the freedom of choosing @xmath139 to suppress autocorrelations in measurements . the density measurement starts with randomly generated @xmath139 . the estimator for kinetic energy is derived similarly . one employs the coordinate - space expression for the kinetic energy in terms of hopping operators and deals with a slightly generalized version of eq . ( [ dens - gibbs ] ) : @xmath140 the rest is identical to the previous discussion up to replacement @xmath141 , since now the spatial position of the creation operator is shifted from that of the annihilation operator by the vector @xmath142 . in our case , only the nearest - neighbor correlator ( [ eps ] ) has to be computed . the estimator for the interaction energy @xmath143 is readily constructed using a generic trick of finding the expectation value of operator in terms of which the perturbative expansion is performed . consider the hamiltonian @xmath144 and observe that @xmath145 the differentiation of eq . ( [ z - diagr ] ) for @xmath146 and letting @xmath147 afterwards is straightforward since the diagram of order @xmath41 is proportional to @xmath148 . hence @xmath149 following the general treatment of ref . @xcite , one can construct an estimator for the correlation function ( [ corr ] ) . in this work we just need to sum and integrate this correlator over all its variables ( see sec . [ sec : sense ] ) : @xmath150 and the estimator for @xmath151 is particularly simple : @xmath152 the mc setup discussed in sec . [ sec : ddmc ] works in the grand canonical ensemble with external parameters @xmath153 . in order to extract the critical temperature of a continuum gas from mc data , one has to perform the two - step extrapolation . ( i ) upon taking the limit of @xmath154 one obtains @xmath155 , the critical temperature of a lattice system at a given chemical potential , and translates it into @xmath156 by extrapolating the measured filling factor to the infinite system size : @xmath157 . ( ii ) the extrapolation towards the continuum limit is then done by taking the limit of @xmath158 . the latter procedure is based on results presented in sec . [ sec:2body ] . the finite - size extrapolation is performed by considering a series of system sizes @xmath159 . at the critical point the correlation function ( [ corr ] ) decays at large distances as a power - law : @xmath160 . if one rescales the summed correlator ( [ q - rescaled ] ) according to @xmath161 the corresponding quantity is supposed to become size - independent at the critical point , i.e. the crossing of @xmath162 and @xmath163 curves can be used to obtain an estimate @xmath164 for the critical temperature @xmath155 @xcite . indeed , for temperatures in the vicinity of the critical point the correlation length diverges as @xmath165 , where @xmath166 , and @xmath167 for the u(1 ) universality class . in the renormalization group ( rg ) framework @xcite , the finite - size scaling of the rescaled correlator @xmath168 obeys the relation @xmath169 where @xmath170 is the dimensionless scaling variable , @xmath171 is the universal scaling function analytic at @xmath172 , @xmath173 is a non - universal constant , @xmath174 is the critical exponent of the leading irrelevant field @xcite , and dots represent higher - order corrections to scaling . if the irrelevant field corrections were not present , all @xmath162 curves would intersect at a unique point , @xmath155 . expanding eq . ( [ rg1 ] ) up to terms linear in @xmath21 one obtains for the crossing @xmath164 @xmath175 to employ eq . ( [ tc - fit ] ) one performs a linear fit of the sequence of @xmath164 against the right hand side of eq . ( [ tc - fit ] ) for several pairs of system sizes . the intercept of the best - fit line yields the thermodynamic limit critical temperature @xmath155 . note that if the universality class is not u(1 ) and the values of @xmath176 , @xmath177 , @xmath178 are different , or system sizes are too small to justify the scaling limit , the whole procedure fails . hence , the adopted scheme of pinpointing @xmath179 features a built - in consistency check . as has been noted in sec . [ sec : intro ] , the only relevant microscopic energy scale in the continuum unitary fermi gas is given by the fermi energy @xmath180 , where @xmath181 for a two - component fermi gas . therefore , as it was first noticed in ref . @xcite , all thermodynamic potentials feature self - similarity properties and can be written in terms of dimensionless scaling functions of the dimensionless temperature @xmath182 . all scaling functions are mutually related ; it is sufficient to know just one of them to unambiguously restore the rest . apart from the shape of scaling functions , the self - similarity at the unitary point is _ identical _ to that of a non - interacting gas fermi gas @xcite , including functional relations between different thermodynamic potentials . a derivation of scaling functions and relations between them can be found in ref . @xcite . in this section , we render the scaling analysis in a form convenient for our mc study . in terms of the dimensionless chemical potential @xmath183 , the dimensionless equation of state reads @xmath184 . the @xmath185 function can be calculated numerically . another quantity which is also available in our simulation is the dimensionless energy per particle @xmath186 . the scaling relations for other thermodynamic quantities are defined likewise . for instance , the entropy and pressure read @xmath187 , and @xmath188 . though @xmath189 and @xmath190 are not directly calculated in our simulation , and we can relate them to @xmath191 . it is also important to relate @xmath185 to @xmath191 , since this yields a consistency check for the numerical results . to establish desired relations , we start with the scaling relation for the helmholtz free energy @xmath192 which in canonical variables reads @xmath193 where @xmath194 . then , for the entropy and pressure we have ( the prime stands for the derivative ) @xmath195 the expression for energy in terms of @xmath196 is @xmath197 we thus see that @xmath198 on may also consider eq . ( [ e2 ] ) as a differential equation : @xmath199 to be integrated with respect to @xmath196 from @xmath200 down to finite @xmath201 , taking advantage of known asymptotic behaviour of @xmath196 and @xmath191 for the weakly interacting two - component gas . now , we note that from the general thermodynamic relation @xmath202 it immediately follows that @xmath203 which , in turn , leads to the following relations @xmath204 that allow one to express @xmath189 and @xmath196 functions through numerically available functions @xmath191 and @xmath185 . another useful relation is @xmath205 which allows to extract @xmath206 directly from @xmath191 and @xmath185 , and thus provides a simple check for the data consistency : the result ( [ f2 ] ) for the @xmath196 curve should be consistent with the the derivative deduced from ( [ f_prime ] ) . by integrating equation ( [ rel0 ] ) we get @xmath207 here we took into account the asymptotic ideal - gas behaviour of @xmath191 : @xmath208 and introduced the corresponding term into the integral to render the latter convergent . the free constant of integration , @xmath209 , can be restored from ( [ f2])-([ass_e ] ) combined with the asymptotic ideal - gas behaviour of @xmath210 , @xmath211 the result is @xmath212 note that if higher - order terms in the asymptotic ( @xmath213 ) behaviour of @xmath214 are also known , then ( [ f3 ] ) can be used to establish the corresponding corrections for @xmath215 and other scaling functions . for instance , it has been found in ref.@xcite that , as @xmath216 @xmath217 in accordance with ( [ f3 ] ) , this implies ( @xmath216 ) @xmath218 finally , the scaling functions @xmath219 and @xmath220 defined in ref . @xcite are related to @xmath191 and @xmath185 as follows : @xmath221 so far we have considered the uniform fermi gas , while in experimental realizations @xcite one has to deal with the parabolic trapping potential . the standard procedure , especially in systems with short `` healing length '' is to use the local density approximation ( lda ) , _ i.e. _ to replace the chemical potential with its coordinate - dependent counterpart @xmath222 . this procedure can be easily combined with mc results as follows . we introduce the dimensionless variable @xmath223 and define the scaling function for the number density as @xmath224 . this is equivalent to the parametric @xmath225 dependence of @xmath9 on @xmath226 . the scaling functions for other thermodynamic quantities are defined in a similar manner , e.g. @xmath227 for the energy , and @xmath228 for the entropy . within lda @xmath226 acquires the coordinate dependence @xmath229 which translates into the density profile @xmath230 . likewise , other thermodynamic functions are to be understood as local , coordinate - dependent quantities . consider the case of @xmath231 particles in a cigar - shaped parabolic trap , characterized by the axial and radial frequencies @xmath232 and @xmath233 . the characteristic energy in this case is @xmath234 , which would coincide with the fermi energy for a non - interacting gas in the trap . note , that we denote the fermi energy in the trap by an upright capital @xmath235 in order to avoid confusion with the uniform system fermi energy @xmath236 . by integrating over the radial coordinates one obtains the axial density profile @xmath237 ( @xmath238 is an axial coordinate ) in the form @xmath239 where @xmath240 , the aspect ratio @xmath241 , the oscillator length @xmath242 , and @xmath243 by integrating eq . ( [ n_a ] ) with respect to @xmath238 , one finally relates chemical potential to temperature : @xmath244 where @xmath245 obtaining the temperature dependence of thermodynamic functions for a non - uniform system within lda is also straightforward . for the total energy of a cloud @xmath246 we obtain @xmath247 and likewise for the entropy : @xmath248 we performed simulations outlined in previous sections for filling factors ranging from @xmath249 down to @xmath250 with up to about 300 fermions on lattices with up to @xmath251 sites . the typical rank of determinants involved in computations of acceptance ratios ( sec . [ sec : updates ] ) and estimators ( sec . [ ssec : estimators ] ) is up to @xmath252 . since we only need ratios of determinants , we use fast - update formulas @xcite to reduce the computational complexity of updates from @xmath253 down to @xmath254 . we validate our numerical procedure by comparing results against the exact diagonalization data for a @xmath255 cluster @xcite , and other simulations of the critical temperature at quarter filling @xmath256 @xcite and @xmath257 @xcite . in all cases we find agreement within statistical errors of a few percent . figure [ fig : crossing ] shows a typical example of the finite - size analysis outlined in sec . [ sec : sense ] . despite the fact that numerically accessible system sizes are quite small , ( and similar analysis for the whole range of filling factors ) supports expectations that the universality class for the sf phase transition is u(1 ) . the finite - size analysis allows us to pinpoint the phase transition temperature to within a few percent . curves . the errorbars are @xmath258 , and solid lines are the linear fits to the mc points . inset shows the finite - size scaling of the filling factor ( @xmath0 vs @xmath259 ) , which yields @xmath260 . from this plot and eq . ( [ tc - fit ] ) one obtains @xmath261 ] . [ fig : crossing ] shown in is the dependence of the critical temperature on the lattice filling factor . the critical temperature is measured in units of the fermi energy , as is natural for the unitarity limit . we define the fermi momentum for a lattice system with filling factor @xmath0 as @xmath262 and the fermi energy @xmath263 , as those of a continuum gas with the same effective mass and number density @xmath264 . it is clearly seen that the presence of the lattice suppresses the critical temperature considerably , nearly by a factor of @xmath265 , depending on the filling factor . strong dependence of @xmath179 on @xmath0 is in apparent contradiction with ref . @xcite , which claims weak or no @xmath0-dependence . this disagreement might be due to the difference in the single - particle spectra @xmath266 used : ref . @xcite employs the parabolic spectrum with spherically symmetric cutoff , while we use the tight - binding dispersion law . indeed , eqs.([gamma - a])-([gamma - b ] ) indicate that a particular choice of @xmath266 does influence lattice corrections to @xmath179 , which may even have different signs for different @xmath266 . corresponds to the half filling . the errorbars are one standard deviation . the results of ref.@xcite at quarter filling and @xmath257 are also shown for a comparison . see the text for discussion . ] it is also clear from that close to half - filling @xmath179 is essentially constant , as expected ( see , _ e.g. _ @xcite ) . the predicted @xmath267 scaling ( [ nu-13 ] ) sets in at about @xmath268 . we thus use a linear fit @xmath269 to eliminate lattice corrections in the final result . such fitting procedure results in the best - fit line given by @xmath270 . we further analyze the fit residues in order to estimate the effect of the sub - leading lattice corrections which are expected to be proportional to @xmath271 . as shown in , such corrections , if any , are smaller than the uncertainty of the @xmath272 fit . this analysis yields the final result @xmath273 for the continuum uniform gas , which is noticeably below the transition temperature in the bec limit @xmath274 . various approximate analytical treatments led in the past to @xmath179 either above @xcite , or below @xcite @xmath275 . ( circles ) . zero level is shown by the horizontal line , the blue dashed lines are linear fits to the data points for @xmath276 and @xmath277 , respectively . ] is is instructive to compare our results for @xmath179 to other numerical calculations available from the literature . the simulations of ref . @xcite yield @xmath278 , but at the value of the scattering length which has not been determined precisely . this result most probably corresponds to a deep bcs regime , where the transition temperature is exponentially suppressed . lee and schfer @xcite report an upper limit @xmath279 , based on a study of the caloric curve of a unitary fermi gas down to @xmath280 for filling factors down to @xmath256 . the caloric curve of ref . @xcite shows no signs of divergent heat capacity which would signal the phase transition . this upper limit is consistent with @xmath281 , see . the seattle group has performed simulations of the caloric curve and condensate fraction , @xmath282 , of the unitary gas , ref.@xcite . using `` visual inspection '' of the caloric curve shape the critical temperature was estimated in ref.@xcite to be @xmath283 . unfortunately , the authors did not perform the finite - size analysis and @xmath1 extrapolation . the overall shape of the caloric curve seem to be little affected by the finite volume of the system . this is hardly surprising since even in the thermodynamic limit @xmath284 and its derivative @xmath285 are continuous at the transition point . these properties also make it hard to use non - quantitative measures for reliable estimates of critical parameters from the @xmath284 curve . on the other hand , the condensate fraction which has singular properties at @xmath179 does show sizable finite - size corrections , see figure 1 of ref . @xcite . at this point we note that scaling of the condensate fraction is identical to that for @xmath151 . in we plot the data of seattle s group as @xmath286 versus temperature . the intersection of scaled curves turns out to be inconsistent with the estimate for @xmath179 derived from the caloric curve inspection . ) by the @xmath287 factor . shaded vertical strips represent results for @xmath288 of this work and ref . @xcite , respectively , solid lines are drawn to guide an eye . ] the filling factor dependence of thermodynamic quantities is similar to that of @xmath179 : figure [ fig : emu - tc ] displays the behaviour of energy and chemical potential along the critical line @xmath289 . the extrapolation towards @xmath158 yields for the continuum gas @xmath290 the numerical values for other thermodynamic functions at criticality can be easily restored using the formulas of sec.[sec : thermodynamics ] . . dots are the mc results , dashed lines are the linear fits . ] in order to elucidate the thermodynamic behaviour of the unitary gas , we performed simulations for a range of temperatures @xmath291 . shown in are the simulation results for energy and chemical potential for the continuum gas as functions of temperature . each point was obtained using data analysis similar to that depicted in . in the high - temperature region we simulated up to @xmath292 fermions on lattices with up to @xmath293 sites . in this region , the condition @xmath294 is necessary but not sufficient for extrapolation to the continuum limit , for it is crucial to keep temperature much smaller than the bandwidth : @xmath295 . as can be seen from , our results for both energy and chemical potential approach the virial expansion @xcite as @xmath296 . for @xmath297 our data are not far from the curve of ref . @xcite . though we do not have data points for @xmath298 there is still a reasonable agreement even at @xmath179 with the @xmath299 fixed - node mc values @xcite . in this region , our results are consistent with a very weak dependence of energy and chemical potential on temperature , and the numeric values of both are consistent with the experimental results @xcite . using eq . ( [ f2 ] ) and data from we deduce the dependence of free energy on temperature , see . we also use eq . ( [ f_prime ] ) to make sure that our mc data for energy and chemical potential are consistent with each other . ) . black triangles are mc results of ref . @xcite , purple star denotes the ground - state fixed - node mc result @xcite , black dotted line shows the boltzmann gas curve , and the blue dashed is the asymptotic prediction of ref . @xcite . ] as discussed in sec . [ ssec : lda ] , the thermodynamic functions for the uniform case can be used for analysis of experimental system within the local density approximation . in this section we report preliminary results of our ongoing study of trapped gas experiments . it starts with the interpolation procedure which produces continuous functional behaviour for thermodynamic functions , consistent with the discrete set of simulated points . we use a piecewise - cubic ansatz with a smooth crossover to the virial expansion , eq . ( [ h_m_f ] ) , for the free energy . temperature dependence of both energy and entropy are then deduced using numerical integration of eqs . ( [ etot ] ) and ( [ stot ] ) , respectively . as the trapped gas is cooled down , the superfluidity first sets in at the centre of the trap , where the density is the highest . equation ( [ wbarbar ] ) can be used to pinpoint this onset temperature : at @xmath300 , @xmath301 , where @xmath302 is the fermi energy of the uniform gas with the density equal to the density at the trap centre . using @xmath303 and eq . ( [ magic_numbers ] ) , one obtains @xmath304 . we quote here a conservative estimate for the uncertainty , which incorporates both the uncertainty of the critical temperature itself , and a systematic uncertainty which stems from restoring the continuous functional dependence of the chemical potential out of the finite set of the monte carlo calculated points with finite errorbars . experimentally , the temperature of the strongly interacting fermi gas is not easily accessible . on the contrary , thermometry of the non - interacting fermi gases is well established . in the adiabatic ramp experiments one starts from the non - interacting gas at some temperature [ in units of fermi energy ] @xmath305 , and slowly ramps magnetic field towards the feshbach resonance @xcite , thus adiabatically connecting the system at unitarity to a non - interacting one . assuming the entropy conservation during the magnetic field ramp , equation ( [ stot ] ) can be employed for the thermometry of the interacting gas : by matching the entropy of a non - interacting gas at the temperature @xmath306 with the entropy calculated via eq.([stot ] ) one relates the initial temperature ( before the magnetic field ramp ) to the final temperature ( after the ramp ) . we find that the onset of the superfluidity corresponds to @xmath307 ( again , we quote here the most conservative estimate for the errorbar ) . this value seems to be somewhat lower than the value suggested by the experimental results @xcite . nevertheless , given the level of the noise in figure 4 of ref . @xcite the consistency is reasonable . an alternative thermometry can be build on recent advances in the experimental technique @xcite which made it possible to directly image the _ in situ _ density profiles of the interacting system . such density profiles can be directly fit to eq . ( [ n_a ] ) , which gives the shape of the cloud depending on @xmath308 and @xmath309 . by relating the chemical potential to the temperature using eq . ( [ wbarbar ] ) , one is left with only one fitting parameter , @xmath309 ( apart from a trivial fitting parameter @xmath310 which accounts for the overall shift of the cloud image off the trap centre ) . as an illustrative example of such procedure we have analyzed the experimental density profiles measured by the rice s group @xcite , as depicted in . from this analysis we deduce the upper bound for the temperature in the experiments @xcite @xmath311 , which is consistent with the results of the measurements of the condensate fraction @xcite . since in the experiments @xcite the gas is very degenerate , one is able to put an upper limit on temperature only . ) ([wbarbar ] ) with @xmath312 ( correspondingly , @xmath313 ) , dashed blue line corresponds to @xmath314 ( @xmath315 ) , and dot - dashed green line is for @xmath316 ( @xmath317 ) . in both cases we allowed for a horizontal displacement of a whole curve in the range of @xmath318 m see the text for discussion . ] note that if the temperature is known from , e.g. the adiabatic ramp experiments , eqs . ( [ n_a])-([wbarbar ] ) must reproduce the cloud shape _ without free parameters _ ( apart from @xmath310 ) . we have developed a worm - type scheme within the systematic - error - free determinant diagrammatic monte carlo approach for lattice fermions . we applied it to the hubbard model with attractive interaction and equal number of spin - up and spin - down particles . at finite densities , the model describes ultracold atoms in optical lattice . in the limit of vanishing filling factor , @xmath1 , and fine - tuned ( to the resonance in the two - particle @xmath319-wave channel ) on - site attraction , a universal regime sets in , which is identical to the bcs - bec crossover in the continuous space . in the present work , we confined ourselves to a special value of the on - site interaction , @xmath320 , corresponding to the divergent @xmath319-scattering length . at @xmath37 and @xmath1 , the system reproduces the unitary point of the bcs - bec crossover . the unitary regime is scale - invariant : all thermodynamic potentials are expressed in terms of dimensionless scaling functions of the dimensionless ratio @xmath321 ( temperature in units of fermi energy ) . we obtained these scaling functions by extrapolating results for the hubbard model to @xmath322 . for the critical temperature of the superfluid - normal transition in the uniform case we found @xmath2 . our results form a basis for an unbiased thermometry of trapped fermionic gases in the unitary regime : in particular , we found ( within the local density approximation ) that for the parabolic confinement , the critical temperature in units of the characteristic trap energy @xmath235 is @xmath323 . for the experimentally relevant case of an isentropic conversion of a gas from the non - interacting regime to the unitary regime we find that the onset of the superfluidity corresponds to the initial temperature ( before the magnetic field ramp ) @xmath324 , which is reasonably consistent with the experimental result @xcite , to within the experimental noise . we appreciate generosity of a. bulgac , p. magierski , and j. drut , who kindly provided us with their numeric data . we are also indebted to w. li and r. hulet for providing us with their unpublished experimental data . this research was enabled by computational resources of the center for computational sciences and in part supported by the laboratory research and development program at oak ridge national laboratory . part of the simulations were performed on the `` hreidar '' cluster of eth zrich . we also acknowledge partial support by nsf grants nos . phy-0426881 and phy-0456261 . the generic diagrammatic mc rules for constructing updates are as follows @xcite . let an update @xmath325 transform a diagram @xmath98 into diagram @xmath326 . here configurations @xmath103 and @xmath104 may have different order and may or may not feature the `` worm '' two - point vertices . the update @xmath325 involves two steps . first , a modification of the configuration is proposed , with some probability density @xmath327 for new continuous / discrete variables . there are no strict requirements fixing the form of @xmath328 , it should rather be chosen on physical grounds to maximize the efficiency of the algorithm , and be simple enough to allow analytical evaluation of the normalization integral . then , the update is either accepted , with probability @xmath329 , or rejected . the complimentary update @xmath330 , which transforms @xmath326 into @xmath331 proposes the modification with the probability density @xmath332 , which , in principle , may differ from @xmath328 , and is accepted with the probability @xmath333 . for a pair of updates @xmath325 and @xmath330 to be balanced , the acceptance probabilities must obey the metropolis relations @xcite @xmath334 where @xmath335 is a solution of the detailed - balance equation @xmath336 if the probabilities of selecting updates @xmath325 and @xmath330 are not equal , they must also be included into the definition of @xmath327 and @xmath337 , respectively . the minimal ergodic set of updates , as suggested in @xcite , consists of just one pair of `` diagonal '' updates which increase / decrease the number of vertices in a configuration by one : in the forward update one adds a vertex at randomly selected point in the @xmath338 hypercube ; in the reverse update one removes a random vertex from the configuration . in the simplest version , the probability density for selecting a point is uniform , i.e. one selects a particular lattice site with the probability @xmath339 , and a temporal position with the probability density @xmath340 . a vertex to be removed is selected at random out of @xmath41 vertices present in the configuration . hence the metropolis acceptance ratio function ( [ balance - eq ] ) is given by @xmath341 here @xmath342 is the configuration with an extra vertex @xmath343 . the acceptance ratio to remove a vertex is @xmath344 the diagonal updating scheme described in the previous section is highly inefficient in the dilute regime the regime we are mostly interested in . to assess the efficiency of the `` diagonal '' scheme consider first a low - density gas in the deep bec regime . in this limit , all fermions are paired into composite bosons ( apart from exponentially rare fluctuations ) . inter - boson interactions occur only via rare collisions . in the diagrammatic language , the existence of compact composite bosons is reflected by the fact that the dominant diagrams are the multi - ladder ones ( see a ) , with the typical @xmath16-span of each ladder being of the order of @xmath345 , and the typical @xmath346-span of the order of @xmath62 . each ladder of such a diagram in fact mimics a world line of a composite boson . the analogy between ladder parts of multi - ladder diagrams and world lines of composite bosons helps elucidating the structure of typical diagrams at resonance . indeed , at the two - particle resonance composite bosons exist only virtually , but at the same time the energy cost of creating a boson is zero . thus , multi - ladder diagrams still dominate , but the @xmath16- and @xmath346-span of individual ladders becomes finite and related to the inter - particle distance , see b. resonant ladders also become more `` dilute '' since the average distance between vertices diverges , but they still contain many vertices since the resonance formation involves distances much smaller than @xmath347 . proposing diagrams which disregard the nature of the resonant interaction results in small acceptance ratios . to significantly reduce this type of slowing down , we resort to a worm - type updating scheme which allows one to , loosely speaking , directly `` draw '' the multi - ladder diagrams . these updates are the necessary ingredient of the worm - type scheme , since they switch between the diagonal ( [ conf - p ] ) and off - diagonal ( [ conf - p - corr ] ) diagrams . these updates transform diagrams as follows : @xmath348 to create a pair of two - point vertices @xmath80 and @xmath81 , one first selects @xmath16 and @xmath346 uniformly over the spatial lattice and @xmath349 interval , respectively . one then selects @xmath350 among the lattice sites in the spatial cube with the edge length @xmath351 centered around @xmath16 with equal probabilities , and selects @xmath346 uniformly on the interval @xmath352 . the values of @xmath351 and @xmath353 are arbitrary . the overall probability density is thus given by : @xmath354 the reverse update attempts to remove @xmath80 and @xmath355 provided @xmath356 ( @xmath357 ) and @xmath358 , as prescribed by the balance requirement . the solution to the detailed balance equation ( [ balance - eq ] ) then reads : @xmath359 in order to maximize the efficiency of the numerical scheme , it is advantageous to have the acceptance probabilities ( [ metropolis_1])-([metropolis_2 ] ) to be of the order unity . on the other hand , the explicit macroscopically large factors in the right - hand side of eq . ( [ r - ca ] ) render the acceptance probabilities macroscopic . the use of the extra weighting factor @xmath111 is now clear : by choosing @xmath360 one removes undesired factors from the right - hand side of eq . ( [ r - ca ] ) . the remaining freedom of choosing @xmath361 has to be employed to further fine - tune the acceptance probabilities . the required values of @xmath361 are typically of the order of unity . these updates are central for the method since they change the diagram order . the main idea of these updates is : given a pair of `` worm '' two - point vertices @xmath96 and @xmath97 , use , say , @xmath107 as a tip of a magic pen to add or remove vertices from the diagram ( note that it suffices to use either @xmath96 or @xmath97 as a dynamical variable , the choice being only a matter of taste ) : @xmath362 we have employed two versions of these updates . [ [ high - density - version ] ] high - density version + + + + + + + + + + + + + + + + + + + + in the most simple yet useful version of a forward update , one selects @xmath363 and @xmath364 uniformly in the interval @xmath365 , and the spatial cube of the edge length @xmath351 centered at @xmath16 , correspondingly . in the reverse update one selects a vertex to be removed at random out of @xmath366 choices , where @xmath366 is the number of vertices @xmath367 such that @xmath368 and @xmath369 . the solution to the detailed balance equation then reads @xmath370 [ [ low - density - version ] ] low - density version + + + + + + + + + + + + + + + + + + + the strategy just described works well close to half - filling . in the low - density regime the multi - ladder diagrams dominate and the above scheme is not optimal for a number of reasons : ( i ) it does not respect the diagram structure which should feature ( at least locally ) well - defined vertex chains ; ( ii ) the probability density according to which the value of @xmath363 is proposed should favour shifting @xmath97 towards smaller @xmath346-s . to see this , consider two particles in vacuum : the matrix elements of @xmath371 simply equal zero for @xmath372 . ] ; ( iii ) both large and small values of @xmath373 have to be accounted for . consider two fermions in vacuum again . the local probability density for a shift of length @xmath374 and @xmath375 is proportional to the free diffusion propagator : @xmath376 , which is singular as @xmath377 . however , the mean value of @xmath374 for this distribution , @xmath378 , diverges at the upper limit . these properties of resonant ladders are ignored when shifts are proposed uniformly over some @xmath379 interval . the requirements ( ii ) and ( iii ) are best met if , when adding a vertex , a new position of @xmath97 is proposed according to the probability density [ cf . ( [ leaps ] ) ] @xmath380 since the free single - particle propagator is known only numerically , we use eq . ( [ w - leap ] ) on a uniform mesh with some step @xmath132 : @xmath381 . that is , we tabulate the values @xmath382 prior to the start of computations ; we then select @xmath383 and @xmath384 according to the ( normalized ) distribution @xmath385 . the proposed new position of @xmath97 is then chosen as @xmath386 , and @xmath387 , where @xmath388 is uniform over the interval @xmath389 $ ] . we find that @xmath390 produces good results . in order to meet the requirement ( i ) , the notion of the closest in terms of a certain distance neighboring vertex is introduced : the reverse update @xmath391 _ deterministically _ deletes the closest neighbor of @xmath97 . the detailed balance than requires that the forward update should be automatically rejected whenever @xmath131 is _ not _ the closest neighbor of @xmath392 , cf . eq.([leaps ] ) . we define the distance between vertices using a simple euclidean norm @xmath393 . the acceptance ratio function ( [ balance - eq ] ) is then @xmath394 where @xmath395 is given by eq . ( [ w - leap ] ) . the scheme ( [ r - leap ] ) in the dilute regime typically yields a an order of magnitude gain in efficiency , as compared to the scheme ( [ r - le ] ) . clearly , the scheme ( [ r - leap ] ) should be supplemented by an update which allows for an easy change of the closest neighbor . the following update performs the task : @xmath396 we select @xmath350 among the nearest neighbors of the lattice site @xmath16 with equal probabilities and @xmath397 uniformly in some interval around @xmath346 . this update is self - balanced , and the acceptance ratio function @xmath335 is simply @xmath398 99 dieckmann k , stan c a , gupta s , hadzibabic z , schunck c h and ketterle w 2002 * 89 * 203201regal c a , ticknor c , bohn j l and jin d s 2003 _ nature _ * 424 * 47bourdel t , cubizolles j , khaykovich l , magalhes k m f , kokkelmans s j j m f , shlyapnikov g v and salomon c 2003 * 91 * 020402strecker k e , partridge g b and hulet r g 2003 * 91 * 080406cubizolles j , bourdel t , kokkelmans s j j m f , shlyapnikov g v and salomon c 2003 * 91 * 240401jochim s , bartenstein m , altmeyer a , hendl g , chin c , hecker denschlag j and grimm r 2003 * 91 * 240402 zwierlein m w , stan c a , schunck c h , raupach s m f , gupta s , hadzibabic z and ketterle w 2003 * 91 * 250401greiner m , regal c a and jin d s 2003 _ nature _ * 426 * 537jochim s , bartenstein m , altmeyer a , hendl g , riedl s , chin c , hecker denschlag j and grimm r 2003 _ science _ * 302 * 2101regal c a , greiner m and jin d s 2004 * 92 * 083201zwierlein m w , stan c a , schunck c h , raupach s m f , kerman a j and ketterle w 2004 * 92 * 120403kinast j , hemmer s l , gehm m e , turlapov a and thomas j e 2004 * 92 * 150402bartenstein m , altmeyer a , riedl s , jochim s , chin c , hecker denschlag j and grimm r 2004 * 92 * 203201 chin c , bartenstein m , altmeyer a , riedl s , jochim s , hecker denschlag j and grimm r 2004 _ science _ * 305 * 1128zwierlein m w , abo - shaeer j r , schirotzek a , schunck c h and ketterle w 2005 _ nature _ * 435 * 1047 partridge g b , strecker k e , kamar r i , jack m w and hulet r g 2005 . * 95 * 020404 bartenstein m , altmeyer a , riedl s , jochim s , chin c , hecker denschlag j and grimm r 2004 * 92 * 120401 partridge g b , li w , kamar r i , liao y - a and hulet r g 2006 _ science _ * 311 * 503 ohara k m , hemmer s l , gehm m e , granade s r and thomas j e 2002 _ science _ * 298 * 2179bourdel t , khaykovich l , cubizolles j , zhang j , chevy f , teichmann m , tarruell l , kokkelmans s j j m f and salomon c 2004 * 93 * 050401kinast j , turlapov f , thomas j e , chen q , stajic j and levin k 2005 _ science _ * 307 * 1296 regal c a , greiner m and jin d s 2004 * 92 * 040403 modugno g , ferlaino f , heidemann r , roati g and inguscio m 2003 a * 68 * 011601(r ) khl m , moritz h , stferle t , gnter k and esslinger t 2005 * 94 * 080403moritz h , stferle t , gnter k , khl m and esslinger t 2005 * 94 * 210401stferle t , moritz h , gnter k , khl m and esslinger t 2006 * 96 * 030401khl m , gnter k , stferle t , moritz h and esslinger t 2006 _ j. phys . _ b : at . . phys . * 39 * s47 heiselberg h and hjorth - jensen m 2000 _ phys . _ * 328 * 237 heiselberg h 2001 a * 63 * 043606 baker g a , jr 1999 , c * 60 * 054311 eagles d r 1969 * 186 * 456 legett a j 1980 , in _ modern trends in the theory of condensed matter _ ( _ lecture notes in physics _ vol 115 ) ed pekalski a and przystawa r ( berlin : springer - verlag ) nozires p and schmitt - rink s 1985 _ j. low temp . phys . _ * 59 * 195 haussmann r 1994 b * 49 * 12975 randeria m 1995 , in _ bose - einstein condensation _ , ed griffin a ( cambridge : cambridge university press ) holland m , kokkelmans s j j m f , chiofalo m l and walser r 2001 * 87 * 120406timmermans e , furuya k , milloni pw and kerman a k 2001 a * 285 * 228 ohashi y and griffin a 2002 * 89 * 103402 perali a , pieri p , pisani l and strinati g c 2004 * 92 * 220404 liu x - j and hu h 2005 a * 72 * 063613 binder k and landau d p 2000 _ a guide to monte carlo simulations in statistical physics _ ( cambridge : cambridge university press ) troyer m and wiese u - j 2005 * 94 * 170201 carlson j , chang s - y , pandharipande v r and schmidt k e 2003 * 91 * 050401 chang s - y , pandharipande v r , carlson j and schmidt k e 2004 a * 70 * 043602 astrakharchik g e boronat j , casulleras j and giorgini s 2004 * 93 * 200404 astrakharchik g e boronat j , casulleras j and giorgini s 2005 * 95 * 230405 scalapino d j and sugar r l 1981 * 46 * 519blankenbecler r , scalapino d j and sugar r l 1981 d * 24 * 2278 rombouts s m a , heide k and jachowicz n 1999 * 82 * 4155 rubtsov a n 2003 _ preprint _ cond - mat/0302228rubtsov a n and lichtenstein a i 2004 _ pisma v jetp _ * 80 * 67rubtsov a n , savkin v v and lichtenstein a i 2004 _ preprint _ cond - mat/0411344 chen j - w and kaplan d b 2004 * 92 * 257002 bulgac a , drut j e and magierski p 2006 * 96 * 090404 wingate m 2005 _ preprint _ cond - mat/0502372 lee d t and schfer t 2005 _ preprint _ nucl - th/0509018 burovski e , prokofev n , svistunov b and troyer m 2006 * 96 * 160402 fetter a l and walecka j d 1971 _ quantum theory of many - particle systems _ ( new york : mcgraw - hill ) landau l d and lifshitz e m 1980 _ statistical physics _ vol 2 ( new york : pergamon press ) burovski e , prokofev n and svistunov b 2004 b * 70 * 193101(r ) prokofev n v and svistunov b v 1998 * 81 * 2514 mishchenko a , prokofev n , sakamoto a and svistunov b 2000 b * 62 * 6317 metropolis n , rosenbluth a w , rosenbluth m n , teller a m and teller e 1953 * 21 * 1087 prokofev n v , svistunov b v and tupitsyn i s 1998 a * 238 * 253 see , e.g. fisher m e 1983 , in _ lecture notes in physics _ * 186 * ( springer - verlag , 1983 ) . binder k 1981 * 47 * 693 see , e.g. guida r and zinn - justin j 1998 * 31 * 8103 ho t -l 2004 * 92 * 090402 landau l d and lifshitz e m 1980 _ statistical physics _ vol 1 ( new york : pergamon press ) ho t - l and mueller e j 2004 * 92 * 160404 husslein t , fettes w and morgenstern i 1997 _ int . phys . _ * c8 * 397 sewer a , beck h and zotos x 2002 b * 66 * 140504 maier t a , dca / qmc , unpublished micnas r , ranninger j and robaszkiewicz s 1990 * 62 * 113 prokofev n v , svistunov b v and tupitsyn i s 1998 _ sov . phys.jetp_

we investigate the dynamical effect of the turbulence in baryonic intergalactic medium ( igm ) on the baryon fraction distribution . in the fully developed nonlinear regime , the igm will evolve into the state of turbulence , containing strong and curved shocks , vorticity and complex structures . turbulence would lead to the density and velocity fields of the igm to be different from those of underlying collisionless dark matter . consequently , the baryon fraction will deviate from its cosmic mean . we study these phenomena with simulation samples produced by the weighted essentially non - oscillatory ( weno ) hybrid cosmological hydrodynamic / n - body code , which is effective of capturing shocks and complex structures . we find that the distribution of baryon fraction is highly nonuniform on scales from hundreds kpc to a few of mpc , and varies from as low as 1% to a few times of the cosmic mean . we further show that the turbulence pressure in the igm is weakly scale - dependent and comparable to the gravitational energy density of halos with mass around h m . the baryon fraction in halos with mass equal to or smaller than h m should be substantially lower than . numerical results show that is decreasing from at halo mass scales around m to at h m and shows further decrease when halo mass is less than h m . the strong mass dependence of is similar to the observed results . although the simulated in halos are higher than the observed value by a factor of 2 , the turbulence of the igm should be an important dynamical reason leading to the remarkable missing of baryonic matter in halos with mass h m . [ firstpage ] cosmology : theory - intergalactic medium - large - scale structure of the universe - methods : numerical

1103.1058

in the concordance @xmath11cdm universe , the baryonic gas ( igm ) traces the collisionless cold dark matter in the linear regime of gravitational clustering . however , observation shows that these two components do not related to each other by a linear mapping on scales up to a few mpc . most attention on the reason of the igm - cold dark matter separation on such scales has been drawn to the thermal property of baryonic gas , mainly radiative heating and cooling , such as photo - heating , feedbacks from star formation and accretion to black holes ( e.g. , valageas & silk 1999 ; tozzi & norman 2001 ; voit et al . 2002 ; zhang & pen 2003 ; xue & wu 2003 ) . the hydrodynamical origin of the igm - dark matter separation has been noticed in the early studies of large scale structure formation ( shandarin & zeldovich 1989 ) . successive studies found that in the weak and moderate nonlinear regime of clustering , the evolution of collisionless particle can be described by the zeldovich approximation , while the dynamics of baryonic component is sketched by the adhesion approximation with dissipation of kinetic energy , or by a random force driving burgers equation ( gurbatov , saichev , & shandarin 1989 ; berera & fang 1994 ; vergassola et al . 1994 ; jones 1999 ; matarrese & mohayaee 2002 ; pando , feng , & fang 2004 ) . when the reynolds number is large , burgers fluid will be turbulent , consisting of shocks ( lssig 2000 ) . burgers shocks occur not only in high , but also in middle and even low density regions . these shocks could contribute partly to the deviation of the velocity and density fields of the igm from that of cold dark matter ( pando et al 2004 ; kim et al 2005 ) . a new progress shows that in the highly developed nonlinear regime the velocity field of the igm is no longer potential , but dominated by vorticity on scales from one and a half hundred kpc to a couple of mpc ( zhu , feng & fang 2010 ) . oblique shocks will form in inhomogeneous baryonic fluid and act as the source of vorticity of the igm . these results give supports to the scenario that in the nonlinear regime the baryonic fluid is in the state of fully developed turbulence , which can be characterized by the she - leveque ( sl ) scaling ( he et al . 2006 ) and the log - poisson hierarchy ( liu & fang 2008 ) . although tiny vorticity can also be generated in dark matter field during shell crossing , the velocity field of dark matter is still dominated by potential motion on scales around @xmath12 mpc ( pichon & bernardeau 1999 ; pueblas & scoccimarro 2008 ) . consequently , the velocity and density fields of the igm would depart from those of underlying collisionless dark matter on scales below a few mpc . some features predicted from the fully developed turbulence have been found to be consistent with observation , such as the log - poisson non - gaussianity of ly@xmath13 transmitted flux of quasar s absorption spectrum ( lu , chu , & fang 2009 ; lu et al . 2010 ) , the scaling relations among the x - ray luminosity , temperature and sz effect of clusters ( zhang et al 2006 ; yuan et al . 2009 ) and the turbulence broadening of hi and heii ly@xmath13 absorption lines ( zheng et al . 2004 ; liu et al . 2006 ) . in this paper , we extend these studies to the baryon fraction @xmath9 , which is defined as the mass ratio between the baryonic and total matter in a system . in the concordance @xmath11cdm universe , the cosmic mean of baryon fraction is @xmath14 , where @xmath15 and @xmath16 are the mean mass density parameters of the baryonic matter and total matter respectively ( dunkley et al . 2009 ; komatsu et al . the baryon fraction in gravitational bound objects is found to be lower than the cosmic mean , known as the missing baryon problem . for galaxy clusters and groups , the baryon fraction is smaller than the cosmic mean by a factor of 2 - 4 ( ettori 2003 ; giodini et al . 2009 ; dai et al . 2010 ) , while can be as low as about @xmath17 of the cosmic mean for dwarf galaxies(mcgaugh et al . generally , the baryon fraction decreases monotonically with decreasing mass of collapsed halos . this problem has also been seen in the study of galaxies abundance . the @xmath11cdm model predicts too many low mass dark matter halos in comparison with estimation from the observed luminosity function of galaxies . it implies that the star formation rate in low mass dark matter halos is substantially low . the baryon content residing in virialized objects with small mass should be much less than that given by the cosmic mean . additional physics are needed to keep the baryonic matter from overcooling ( white & frenk 1991 ) . many kinds of mechanisms have been proposed to prevent most of the baryon falling into dark matter halos . feedback from supernovae and agn and photo - heating are well investigated . feedback from massive stars and sne were once believed to be able to driving out the collapsed gaseous baryon and hence suppress the star formation(dekel & silk 1986 ; white & frenk 1991 ; kauffmann et al . theoretically , simulations and observations have shown that this mechanism is unlike to work while taking appropriate feedback efficiency , mass loss rate(e.g . mac low & ferrara 1999 ; benson et al . 2003 ) , except for low mass halos . in contrast , mechanisms in which the missing baryon are inhibited to fall into the protogalactic halos in the very beginning , i.e. , never fell into , are much preferred ( mo et al . 2005 ; anderson & bregman 2010 ) . photo - heating fulfills this category which adopt a photo - ionizing field to reheat the baryon around progalactic halos and hence keep them from collapse into potential wells(e.g . efstathiou 1992 ; gnedin 2000 ; benson 2002 ) . the efficiency of this model , however , was later found to be largely limited , only works for halos below @xmath18 m@xmath19 ( hoeft et al . 2006 ) , because the central self - shielding will delay the photo - evaporation . in our view , igm turbulence should be an important reason of the missing baryon problem . the effects of turbulence on the properties of clusters have been extensively studied by numerical simulations and theoretical works(e.g . norman & bryan 1999 ; dolag et al . 2005 ; vazza et al . 2009 , 2010 ; burns , skillman & oshea 2010 ; ruszkowski & oh 2010 ) . yet , we will focus on the dynamical and statistical effect of turbulence on the spatial distribution of baryon fraction . that is , the turbulence of igm is treated as a basic environment factor of gravitational clustering . the baryon missing in virialized objects should be a result of the formation and evolution of inhomogeneity of @xmath0 spatial distribution . one can then explain the baryon missing by considering the inhibition of gravitational collapse by turbulence pressure ( chandrasekhar 1951a , 1951b ) . this paper is organized as follows . 2 addresses the theoretic background of turbulence in the igm and its effect on the gravitational collapsing of baryonic matter . 3 gives the simulation method of producing samples . in 4 , we study the properties of the non - uniform distribution of the baryon fraction , and its relation with turbulence . 5 presents the baryon fraction in collapsed halos . we discuss our results and compare them with previous numerical studies in 6 . finally , the conclusions are given in 7 . hierarchical structure formation process will introduce shocks(ryu et al . 2003 ; pfrommer et al . 2006 ; vazza et al . the shock wakes and shear flows , arose from gas accreting into pancakes , filaments and halos , protogalactic and collapsed objects moving in complex structures and gaseous structure colliding and merging , will interplay with each other and drive instability , result in the onset of turbulence . fully developed turbulence consists of eddies on various scales and will cascade the kinetic energy of largest eddies down to smaller one ( lin 1966 ; shu 1992 ) . more specifically , the evolution of baryonic fluid in the moderate nonlinear regime can be characterized by random force driving burgers equation ( gurbatov et al . 1989 ; berera & fang , 1994 ; jones , 1999 ; matarrese & mohayaee , 2002 ) . when the effective reynolds number is large , burgers fluid will be turbulent , consisting of shocks and complex structures on various scales(polyakov , 1995 ; lssig 2000 ; boldyrev et al . 2004 ) . as dark matter is not influenced by burgers turbulence , the igm velocity field will dynamically decouple from the dark matter field on scales larger than the jeans length once burgers turbulence is developed . this will lead to the deviation of @xmath0 from @xmath20 in low and high density areas . the velocity field of baryonic fluid keeps irrotational in the moderate nonlinear regime . latter on , the velocity field of igm will no longer be potential dominated , because the vorticity can be effectively generated by oblique shocks due to baroclinic instability(zhu et al 2010 ) . when shocks propagate in inhomogeneous medium , they will generally evolve into oblique or curved shock waves ( e.g. landau & lifshitz 1987 ) , acting as the source of vorticity ( e.g. picone et al . 1984 ; emanuel 2000 ) . once triggered , the vorticity can be self - amplified by the nonlinearity of hydrodynamics ( see 2.3 ) . finally the igm evolves into a fully developed turbulent state which satisfies the she - leveque scaling and log - poisson hierarchy ( he et al . 2006 ; liu & fang 2008 ; lu et al 2009 , 2010 ) . more details about the development of turbulence in the igm along cosmit evolution can be found in ( zhu et al . 2010 ; fang & zhu 2011 ) . the jeans length of self - gravitational clustering in a statistical uniform gas with density @xmath21 and speed of sound @xmath22 with or without turbulent gas motion has been studied by chandrasekhar ( 1951a , 1951b ) . in the absence of turbulence , the jeans length is @xmath23 , i.e. gravitational clustering can occur only if the perturbation scale is larger than @xmath24 . the effect of turbulent motions on the clustering is to replace @xmath22 by an effective sound speed @xmath25 where @xmath26 measures the velocity fluctuations of turbulence . the jeans length is increased by the random velocity field of turbulence , which plays the similar role as the randomly thermal motion . the random motion of the turbulence provides an extra pressure , @xmath27 , called turbulence pressure . the turbulence pressure , joining the thermal pressure , will slow down and even halt the igm falling into a gravitational well . considering the gravitational collapsing on scale @xmath28 will not be affected by the velocity dispersion on scales that larger than @xmath28 , the velocity fluctuations with wave - numbers @xmath29 will not resist gravitational collapsing on scales larger than @xmath28 . consequently , the effect of turbulence pressure on gravitational collapsing on scale @xmath28 could be estimated by ( bonazzola et al 1987 ) @xmath30 where @xmath31 is the power spectrum of kinetic energy density @xmath32 of the turbulence . the upper limit of the integral eq.(2 ) @xmath33 corresponds to the minimal scale @xmath34 below which the turbulence is dissipated . the lower limit of the integral eq.(2 ) is the maximum of @xmath35 and @xmath36 , where @xmath37 and @xmath36 matches the upper scale of turbulence . obviously , the turbulent pressure @xmath38 is dynamical , not thermal . it can be comparable to the thermal pressure of the igm , especially in regions the temperature of igm is not very high . when we estimate the turbulence pressure with eq.(1 ) or ( 2 ) , we should separate the velocity field @xmath39 into two components : one is the random motion of turbulence and the other is the bulk velocity . the latter mainly depends on the gravitational potential , and does not contribute to the turbulence pressure . in some algorithms , the bulk velocity is identified as the mean velocity within a box , while the fluctuation with respect to the mean velocity is supposed to be the turbulent motion . as the box size is selected by hand , these algorithms might contain non - ignorable system bias . a better way to pick up the random motion of turbulence bases on the vorticity of velocity field . the vorticity field @xmath40 of the velocity field @xmath41 of the igm is defined by @xmath42 , i.e. , @xmath43 , where @xmath44 . the dynamical equation of the vorticity @xmath45 is ( zhu et al . 2010 ) @xmath46 where @xmath47 is the pressure of the igm , @xmath48 is the divergence of the velocity field and @xmath49 is the cosmic factor . tensor @xmath50 , defined as @xmath51 , is called strain rate , and @xmath52_i = s_{ij}\omega_j$ ] . a remarkable property of eq.(3 ) is the absence of the gravity term . this point is expected , as gravity is curl - free in nature . in other words , vorticity is fully given by the nonlinearity of hydrodynamics and hence is effective to measure the turbulence . an important property of the vorticity is that for a fully developed turbulence the power spectra of the vorticity field @xmath53 and the velocity field @xmath54 should satisfy @xmath55 ( batchelor 1959 , 2000 ) . this relation can be used to determine the wavenumber @xmath36 , i.e. the upper scale of fully developed turbulence . we use this criterion to ascertain @xmath36 to calculate the integral eq.(2 ) . the dynamical effect of turbulence on the gravitational clustering of the igm can be seen with the time evolution of the irrotational component of velocity field , i.e. the divergence of velocity field @xmath48 , which follows ( zhu et al 2010 ) @xmath56,\end{aligned}\ ] ] where @xmath57 is the total mass density including both cold dark and baryonic matter with @xmath58 being its mean value . as negative divergence means the increase of density , positive @xmath59 will lead to clustering , while negative one result in anti - clustering . the right hand side of eq.(4 ) can be used to compare the effects of hydrodynamical terms on clustering with that of gravity . we first identify the physical meaning of the hydrodynamical terms on the right hand sight of eq.(4 ) by considering incompressible fluid , i.e. assuming @xmath57 to be a constant @xmath60 . in this case , the term of gravity @xmath61 , and @xmath62 . the continuity equation yields @xmath63 , and eq.(4 ) gives @xmath64 this is a typical poisson equation for the scalar field @xmath47 . analogous to the field equations in electrostatics , the term @xmath65 $ ] on the right hand side of eq.(5 ) plays the role of the charge `` of a pressure field . positive charge '' produces an attraction force , while negative charge `` yields a repulsive force . back to eq.(4 ) , @xmath66 $ ] also plays the role of nonthermal pressure of turbulence . in regions with negative charge '' , i.e. @xmath67 < 0 $ ] , the turbulent fluid will prevent the gravitational clustering . the sign of the charge `` is actually determined by the levels that turbulence has developed . from the definition of vorticity and strain rate , we have @xmath68.\ ] ] for a gaussian velocity field , @xmath69 , the net effect of velocity fluctuations on the igm collapsing is null in average . however , for a non - gaussian velocity field , it can be either positive or negative , depending on the property of the velocity field . for a homogeneous and isotropic turbulence , @xmath70 ( batchelor 1959 ) , the sign of the charge '' , @xmath71 $ ] , is always negative . a fully developed turbulent flow will generally prevent the igm from gravitational collapsing . the term @xmath72 of eq.(4 ) relates to the hydrostatic pressure while does not consider fluid compressibility , which has been presented in eq.(5 ) . it is mostly negative in overdense collapsing regions , and will resist upon gravitational collapse . the term @xmath73 on the right hand side of eq.(4 ) result from the compressibility of the igm . its value would be negative when the density - pressure relation is a power law @xmath74 and @xmath75 , and then also plays the role of resisting gravitational collapsing . the physical meaning of the term @xmath76 can be shown by the ratio between @xmath77 and the gravity term @xmath78 . the ratio is roughly equal to @xmath79 , where @xmath80 is the free falling time scale , @xmath81 , @xmath82 , is the dynamical time scale on a spatial scale @xmath28 collapsing . when @xmath83 is larger than @xmath84 , the free falling time scale is larger than the dynamical time scale , and in result , the collapsing on scale @xmath28 is significantly prevented . as mentioned above , the turbulent igm contains curved shocks , vortices , and other discontinuities . to study the effect of turbulent igm with simulation , the algorithm should be qualify for capturing these complex structures . we take advantage of the development of the weighted essentially non - oscillatory ( weno ) method ( shu 1999 ) . the weno schemes have been widely used in various fields , such as high reynolds number compressible flows ( zhang et al . 2003 ) , high mach number jets ( carrillo et al . 2003 ) , magneto - hydrodynamics ( jiang & wu , 1999 ) and hypersonic boundary layer ( rehman et al 2009 ) . the shock capturing algorithm with weno scheme has past many test , including shock - boundary layer interaction ( lagha et al . 2009 ) , shocks in high - speed flows ( martin , piomelli & candler 2000 ) , shock vortex interaction ( grasso & pirzzoli 2000a , 2000b ) and shock - turbulence interaction ( pirozzoll 2002 ) . an updated review of the weno method is given in shu 2009 . in the context of cosmological hydrodynamical simulation , the wigeon code is based on eulerian description of hydrodynamics with 5th order weno finite difference scheme and particle mesh ( pm ) method for dark matter particles ( feng , shu , & zhang 2004 ) . the wigeon code can reproduce commonly accepted results such as the relation between igm temperature and density and the component of whim ( he , feng , & fang 2004 ) . in the same time , it also reveals a series of turbulent behavior of the igm ( he at al 2006 ; liu & fang 2008 ; lu et al . 2009 , 2010 ) . these features of this code fit well with our goal . the cosmological parameters are taken to be @xmath85 ( 0.274 , 0.726 , 0.705 , 0.812 , 0.0456 , 0.96 , 11.0 ) ( komatsu et al . , 2009 ) . the simulation run in a periodic cubic box of size 25 @xmath86 mpc since redshift @xmath87 with a @xmath88 grid and an equal number of dark matter particles , giving a mass resolution of @xmath89 . a uniform uv background of ionizing photons is added at @xmath90 to mimic the reionization . radiative cooling and heating are followed as theuns et al . ( 1998 ) with the primordial composition @xmath91 . star formation and its feedback are not added in our simulation , because the resolution we used makes it hard to implant these processes appropriately . snapshots are outputted at redshifts @xmath92 . to study the convergence of numerical results , we also run a simulation with @xmath93 grid and an equal number of dark matter particles . the velocity field of simulation samples has been used to show the development of vorticity and turbulence in zhu et al.(2010 ) . to locate shocks post simulation we use a algorithm also based on weno kernel , combining with conditions from ryu et al.(2003 ) ( see appendix ) . to construct the density field of dark matter on grids , we assign the mass of dark matter particles onto grids using the triangle - shaped - cloud ( tsc ) method . in order to minimize the system bias of assignment in calculating the baryon fraction on grids , we smooth over the density fields of baryonic and dark matter on grids separately using the same smooth window with radius of one grid . to identify halos of dark matter and their radius @xmath94 , we use the same process as crain et al . halos are identified by two methods : a. ) friends - of - friends ( fof ) method with a linking length parameter @xmath95 ; b. ) hop method ( eisenstein & hut 1998 ) with default parameters for group searching and @xmath96 , @xmath97 and @xmath98 for group merging . we study only halos consisting of no less than 2000 dark matter particles at @xmath99 . for each halo , we can then find the center and radius of a sphere , in which the mean mass density is equal to @xmath100 , where @xmath101 is the critical density at redshift @xmath102 . figure 1 presents an example of the spatial distributions of baryonic and dark matter density in a slice of 25x25x0.2 @xmath103 mpc@xmath104 at redshift @xmath99 . the density fields of dark matter and baryonic fluid of figure 1 display the typical sheets - filaments - knot structures on the cosmic scales . the spatial distributions of shock , temperature , vorticity and normalized baryon fraction @xmath105 on the same slice are shown in figure 2 . the shocks do not always follow the filament and sheet structures of the matter density fields . shocks can also be formed in low density areas , as demonstrated in other simulations ( e.g. ryu et al . 2003 ; kang et al . 2007 ) . the vorticity field in figure 2 is described by the dimensionless quantity @xmath106 , where @xmath107 is the cosmic time . @xmath106 accounts the number of rotation of the vorticity @xmath108 in the age of universe . the distribution of vorticity also does nt always follow the filament and sheet structures of the matter density fields , but has cloud - like structures ( zhu et al 2010 ) . although curved shocks are the sources of vorticity , the spatial distribution of vorticity field does not show the same configuration as the shocks . it is because the spatial transfer of vorticity is given by the velocity field [ eq.(3 ) ] , while the case with shock front is different . once the vorticity appears , it will depart from their source - curved shocks and spreads over the space along with the velocity field ( batchelor 2000 ) . in the plot of vorticity distribution , we add fof identified halos , which are marked as solid circles . the radius of circles are enlarged to 5 times of the real halos radius . a remarkable feature is that all the halos are located in the area of @xmath109 . it means that all the gravitational bounded halos are formed in the environment of turbulent igm . as discussed in 2.2 , the gravitational clustering should be affected by the turbulence pressure . the spatial distribution of the normalized baryon fraction @xmath110 is highly non - uniform . the value of @xmath110 spreads over three orders of magnitude from @xmath111 to @xmath112 , indicating that the separation of baryonic matter from dark matter is significant developed during the nonlinear evolution . the deviation of baryon fraction from the cosmic mean is not only occurred in or around massive halos , but also in the low and moderate density areas . the spatial distribution of baryon fraction does not display the sheets - filaments - knot structures as the density fields of dark and baryonic matter do . nevertheless , the area of high @xmath113 are surrounded by high baryon fraction area ( dark blue ) , indicating that the processes of separating baryonic matter from dark matter would be potent around high density areas . the sizes of some region in which @xmath114 , corresponding to @xmath115 , is much larger than the resolution , demonstrating that the existence of regions with @xmath116 is robust . as mentioned in 2.3 , the term @xmath117 on the right hand side of eq.(4 ) measures the gravity effect leading to clustering ( @xmath118 ) , or anti - clustering ( @xmath119 ) . the first three terms on the right hand side of eq.(4 ) describe the gas dynamical and thermal effects , namely , the vorticity and strain rate @xmath120 , the ordinary pressure @xmath72 and multiphase - related term @xmath121 . we now use simulation sample to study the general properties of these terms . figure 3 presents a comparison between @xmath76 and @xmath122 in cells randomly selected from the simulation samples at @xmath99 . we chose only the cells of @xmath123 , which drives collapsing . the variables used in fig . 3 , @xmath124 and @xmath125 , are dimensionless , where @xmath107 is the age of the universe . we can see from figure 3 , when @xmath126 , more than a half of the data points having @xmath127 and few points have @xmath128 , which corresponds to the so - called inverse density - pressure(temperature ) relation . therefore , the term @xmath77 enhanced by shocks and complex structures can effectively resist gravitational collapsing till @xmath129 . for deep gravity wells with @xmath130 , most data points show that gravity generally is larger than @xmath76 . namely , only in deep gravitational wells the motion of baryonic matter will be dominated by the gravity . figure 4 compares @xmath131 vs. @xmath77 ( left panel ) , and @xmath132 vs. @xmath133 ( right panel ) at randomly selected cells . we use the dimensionless variables by multiplying with @xmath134 . the left panel of figure 4 shows that the magnitude of @xmath135 is comparable with @xmath76 . while in the right panel , it exhibits that term @xmath136 is stronger than the term @xmath133 . on average , the term @xmath133 is smaller than @xmath77 . in a word , in terms of prevention of gravitational collapsing , the effect of vorticity and strain rate is comparable with @xmath77 and dominates over @xmath133 . both of the latter two terms are thermal pressure @xmath47 related , one can regard vorticity and strain rate as an effective thermal pressure . figure 2 shows that in most regions igm temperature are equal to or less than 10@xmath137 k. giving the comparison in the last paragraph , the effective thermal pressure of turbulence would be as strong as the thermal pressure with temperature @xmath138 k. the turbulence pressure level given by simplified estimation is consistent with the more delicate one using the energy density of turbulence in zhu et al . ( 2010 ) . as vorticity is an important dynamical factor to cause a low baryon fraction , one can expect a correlation between the baryon fraction and the vorticity . figure 5 gives the probability density function(pdf ) of @xmath139 $ ] in various ranges of the baryon fraction @xmath110 . figure 5 indeed shows that for cells have small @xmath110 , the pdf at the side of @xmath139 > 2 $ ] are higher . the cells with more serious baryon missing have larger vorticities in average . the similar pdf of @xmath140 is also shown in figure 5 . the cells with more significant baron missing generally corresponds to large @xmath141 as well . although the two panels of figure 5 look similar , the values of @xmath142 and @xmath143 actually cover different magnitude range . the former can be as large as @xmath144 , while the later is about @xmath145 . this is consistent with the right panel of figure 4 . the vorticity term @xmath146 is larger than the term @xmath147 . figure 6 plots the @xmath110 vs. mass density of dark matter in randomly selected cells in which @xmath132 is positive and @xmath132 dominant over @xmath148 . it shows clearly that the higher the density @xmath113 the lower the baryon fraction @xmath110 . since high density @xmath113 means high @xmath148 , figure 6 indicates that as long as the turbulence of the igm is dominant , the baryon fraction is lower for higher @xmath113 . we can also see from figure 6 that a significant fraction of those data point which have @xmath116 , i.e. more than @xmath149 percent of baryonic matter are missing , locate in the areas with mass density @xmath150 . the event of @xmath116 baryon missing can occur even in the regions out of collapsed clumps , indicating that the baryon missing of virialized halos should be considered as result of the evolution of baryon distribution on sizes larger than virialized halos . it has been shown that the relation @xmath55 is well satisfied by the igm velocity field in the scale range from 0.2 to 3 @xmath86 mpc at @xmath99 ( zhu et al 2010 ) , on which the igm is in the state of fully developed homogeneous and isotropic turbulence . the power spectrum of kinetic energy density of the turbulence is approximately @xmath151 . the turbulence pressure ( or the kinetic energy density ) is weakly scale - dependent @xmath152 and about the order of magnitude @xmath153 g cm@xmath3 s@xmath155 . on the other hand , the viral temperature @xmath156 of a halo at the virial radius @xmath94 is @xmath157 . the kinetic energy density of the virial motion at @xmath94 is then @xmath158 . the mass of halos with @xmath159 k is about @xmath2 @xmath86 @xmath160 . obviously , the effect of turbulence pressure on the igm collapsing would be comparable with and even larger than the gravity for halos with mass equal to and less than @xmath2 @xmath86 @xmath160 . the baryon fraction would significantly be reduced for halos with mass @xmath161 @xmath86 @xmath160 . meanwhile , the kinetic energy density of turbulence is small than that of the virial motion of halos with mass larger than times of @xmath2 @xmath86 @xmath160 . thus , the halo mass dependence of the baryon massing would show two phases : when the halo mass is larger than times of @xmath2 @xmath86 @xmath160 , the baryon fraction @xmath0 is not much affected by turbulent igm , and it will be substantially decreasing with halo mass when halo mass is lower than a few @xmath2 @xmath86 @xmath160 . with the preparation of above sections , we now analysis the baryon fraction in gravitational collapsed halos . the accuracy of the baryon fraction in selected halos relies on the calculation of the mass of baryonic matter enclosed in the viral radius . to reduce the system error , we divide each halo sphere into sub - cubic cells with size equal to @xmath162 of the sphere radius . we then use cloud - in - cell ( cic ) interpolation to determine the baryon density at each sub - cell . the halo mass @xmath163 and baryon mass are given by the sum of the mass of total matter and baryonic matter respectively in all the sub - cubic cells . this method may yield large uncertainty for halos with @xmath94 less than the size of a grid . to restrain the uncertainty , our statistical results mainly are on @xmath164 , i.e. the virial radius is about equal to or larger than two grids . figure 7 displays the baryon fraction as a function of halo mass at @xmath99 , where the halos are identified by fof ( left ) and hop ( right ) respectively . the results of the two panels are statistically the identical . at a given halo mass , there are large scatters in the baryon fraction @xmath110 , the scatters are seen to be even more significant in the fof algorithm than the hop . this is consistent with figures 6 , which indicates a large scatter in @xmath110 at a given mass density . the distribution of @xmath110 is highly non - gaussian and has a long tail on the side of small @xmath110 , conforming with statistics of turbulent fluid , of which the velocity and density fields are intermittent and their pdfs have long tails . regardless of the scatters , the distribution of @xmath110 in figure 7 can not be described by a single power law , but shows two distinct phases in halo mass - dependence . in the mass range @xmath165 @xmath86 @xmath160 , @xmath110 is weakly dependent on halo mass and takes a value of @xmath166 to @xmath167 of the cosmic mean , which is the first phase . when halo mass is less than @xmath168 @xmath160 , @xmath110 is quickly dropping off with decreasing halo mass , i.e. the second phase . at mass@xmath169 @xmath86 @xmath160 , @xmath110 is around @xmath170 of the cosmic mean . for halos with mass @xmath171 , the mean of @xmath110 is even as low as about @xmath172 . although the data points in the second phase shows a large scattering , this phase has a clear upper envelop . the strong downward in @xmath110 with decreasing halo mass is evident even taking account of the scatters . the scatter looks very small on the side of high halo mass . to test the convergence and stability of the results of fig . 7 , we 1 . ) calculate @xmath110 within radius 2@xmath94 ; 2 . ) analysis 256@xmath104 samples ( 3 ) . for the 256@xmath173 samples , we only calculate the @xmath174 within radius 2@xmath94 to avoid large system error , as @xmath94 is under the grid size at the low mass end . the results are given in fig . 8 , in which the halos are identified by the hop algorithm . it shows that @xmath110 in 2@xmath94 is about the same as that of @xmath94 within the error bars for halos larger than times of @xmath175 @xmath86 @xmath160 . for halos with mass @xmath176 @xmath86 @xmath160 , @xmath110 in @xmath94 shows a little lower than that of 2@xmath94 . this result seems to be reasonable , if noted that the baryon fraction of cluster increases with radii ( e.g. ettori & fabian 1999 , wu & xue 2000 ) . for the @xmath93 simulation , @xmath174 in radius 2@xmath94 shows about @xmath177 lower than its counterpart of the @xmath88 simulation at halo mass @xmath178 @xmath86 @xmath160 . this difference probably comes from the uncertainty caused by the size of grids . the uncertainty of @xmath110 caused by the resolution of @xmath88 simulation should be around @xmath177 at halo mass @xmath2 @xmath86 @xmath160 figure 9 . compares the baryon fraction as a function of halo mass @xmath163 at redshifts @xmath179 , and 2 . the halos in figure 9 are identified by hop method , which yields similar results as the fof method . clearly , it shows the scatter in @xmath110 becomes more significant at lower redshifts , data points have @xmath180 at @xmath181 are quite few , but significantly increases at @xmath179 . the redshift evolution of @xmath110 is parallel to the igm turbulence ( zhu et al 2010 ) . the turbulence is fully developed on scales from 0.2 @xmath86 mpc up to a couple of mpc since redshift @xmath182 . in the igm density distribution shown in figure 1 , there are about @xmath183 of volume with @xmath184 at redshift @xmath99 , while it is merely @xmath185 at redshift @xmath181 . the effect of turbulence becomes stronger to slow down the igm clustering at lower redshifts . many measurements on the baryon fraction of galaxies , groups and clusters have been done in recent years . although it is widely accepted that the baryon missing occurs in gravitational bound objects , the result has not well quantitatively settled yet . for instance , some measurements claim that the baryon fraction in most massive , relaxed galaxy clusters is close to the cosmic mean ( e.g. vikhlinin et al . 2006 ; allen et al . 2008 ) , while someone else argues that the baryon fraction of clusters with mass @xmath186 m@xmath4 is less than the cosmic mean ( giodini et al . nevertheless , observation shows that the baryon fraction of objects with mass @xmath187 m@xmath4 is systematically decreasing with decreasing masses . for galaxy groups and galaxies , the baryon fraction is probably not higher than about 0.1 of the cosmic mean ( e.g. sun et al . 2009 ; hoekstra et al . 2005 ; heymans et al . 2006 ; mandelbaum et al . 2006 ; gavazzi et al . 2007 ) . figure 10 shows the mass dependence of baryon fraction @xmath110 measured at @xmath188 ( magaugh et al . 2010 ) and @xmath94 ( dai et al ( 2010 ) . these two data show some difference at mass @xmath189 @xmath86 @xmath160 . it is probably due to the cool gas is underestimated with x - ray measurement . it may be also partially caused by the difference in redshifts of the two sample sets . the result of magaugh et al.(2010 ) can not be fitted by a single power law , but show two - phase feature . in the mass range @xmath190 @xmath86 @xmath160 , @xmath191 , almost independent of object mass . in the mass range @xmath192 @xmath86 @xmath160 , the baryon fraction decreases significantly with the halo mass . @xmath110 is only about 10% of @xmath1 at @xmath175 @xmath86 @xmath160 and further downs to 6% at halo mass of @xmath193 @xmath86 @xmath160 . in figure 10 , we also plot the mean and variance of simulation baryon fraction @xmath110 in @xmath94 for hop halos at z=0 . the simulated mass - dependence of @xmath110 shows the similar trend as observation . in the mass range of @xmath194 h@xmath3 m@xmath4 , simulation results are higher than observed data by a factor of about 2 , but within the error bars of observed data . the effect of igm turbulence is not the only reason leading to the baryon missing . yet it should be an important dynamical reason of the baryon missing , especially for halos with mass @xmath195 @xmath86 m@xmath4 . the baryon depletion has attracted many studies with cosmological hydrodynamical simulation . as revealed in this paper , the missing baryon could be caused by the igm turbulence . in order to accurately study the effect of the turbulent igm on baryon fraction distribution and hence the baryon mission , the algorithm of hydrodynamical simulation should be effective to capture curved shocks , vortices , and intermittence of turbulence in the igm . it has been shown that the smoothed particle hydrodynamics ( sph ) method may not be able to handle shocks or discontinuities as well as grid method , because the nature of sph is to smooth between particles(tasker et al . the sph method is found to have strong damping of velocity fluctuations and fluid shear instabilities with respect to grid method ( agertz et al . moreover , the artificial viscous force ( dissipation ) in sph algorithm will reduce the reynolds number , and suppress the effect of turbulence ( dolag et al . 2005 ) . these factors would be part of the reasons that some sph simulations find the baryon fraction in halos to be generally only @xmath196 lower than the cosmic mean(e.g . eke , navarro & frenk 1998 ; frenk et al . 1999 ; ettori 2006 ; crain et al . 2007 ) , where the lower in @xmath9 results from energy transfer during shock formation(navarro & white 1993 ) . the eulerian grid method , including both fixed grid and those with adaptive mesh refinement(amr ) , are shown to be effective to pick up the turbulent behavior of baryon fluid in and around galaxies and clusters ( e.g. norman & bryan 1999 ; ryu et al 2008 ; molnar et al 2009 ; vazza et al . 2009 ; burns , skillman & oshea 2010 ; ruszkowsky & oh 2010 ) . however , very few eulerian simulation has been conducted on the turbulent behavior of igm other than zhu et al.(2010 ) . oshea et al . ( 2005 ) investigated the baryon fraction in halos at z = 3 during their study on the performance comparison of enzo and sph method without looking into tur- bulence . however , their small box , merely 3 mpc , and relative poorly resolved dark matter halos , typically tens to hundreds parti- cle in a halo , makes a meaningful comparison between their results and those presented in this paper unavailable . the energy cascade would be suppressed in small box , which would significantly re- strain the development of turbulence in the igm . very recently , the turbulence of igm is studied with grid based amr code enzo in- cluding a subgrid scale model for small scale unresolved turbulence in iapichino et al . ( 2011 ) , which yields many results similar to zhu et al ( 2010 ) . however , statistical study on the baryon fraction is not carried out . the weno algorithm embeded in our simulation uses a convex combination of all the stencils in the reconstruction procedure , where each stencil is assigned with a nonlinear weight depending on the local smoothness . this algorithm provides a high order accurate way to capture the non - oscillatory property near strong discontinuities as well as complex smooth solution features . this feature makes the weno scheme has gained rapid popularity in the simulation of complex - structure of hydrodynamical field ( shu 1999 , 2009 ) . the 5-th order weno scheme used in this work makes it possible to follow high reynolds number baryonic fluid . on the other hand , like other high order non - artificial - viscosity scheme , e.g. piecewise perturbation method ( ppm ) , with fixed grid , the computation time needed is much more than the sph and amr method , which would become more obvious in higher resolution work . to estimate the effect of turbulence of igm , a key factor is the upper scale of fully developed turbulence . in some algorithm , the energy of the turbulence is calculated with the fluctuations of velocity with respect to the mean velocity in cells with selected size . an underestimated cell size generally leads to undervalued turbulence ( dolag et al . 2005 ) . in our algorithm the upper scale of turbulence is determined by the relation @xmath197 , by which the ambiguity in selecting the cell size is avoided . star formation and its feedback on the igm evolution are not considered in our simulation . the injection of hot gas and energy by supernova explosions , agn or other sources of cosmic rays is not followed . these factors can be properly added , if the star formation history and radiative transfer is well known , which in turn need very high resolution . as mentioned above , the influence of injecting hot gas and energy by supernovae is believed to yield further decrease of @xmath110 in less massive halos , while feedback from agn might works in massive central galaxies . including all these mechanisms would give better consistency with observation . meanwhile , an accurate and robust observation result of baryon fraction in objects from as massive as galaxy cluster to low mass dwarf galaxy is full of challenge . large scatter in the observed quantities and significant system bias comes from the models applied in object structure and radiative transfer urges much more efforts to deal with . the spatial resolution of our simulation is @xmath198kpc , which is not enough to simulate virialized halos with mass less than times of @xmath199 @xmath86 m@xmath4 . the lower limit of the inertial range of turbulence may be affected by the grid resolution , which is already seen in the vorticity power spectra given by samples with different resolution in zhu et al . ( 2010 ) . on the other hand , the box size of our simulation is @xmath200mpc . it may lose the effect of long wavelength perturbations and hence underestimate the nonlinear evolution of igm velocity field . nevertheless , we believe that the basic dynamical picture features revealed by the current samples would be valid when these factors are improved . since gravity is of scale free , it is generally believed that in the scenario of hierarchical clustering , the formation and evolution of halos is scale free in a large range of halo mass . however , the dynamics of the system consisting of two components , dark matter and igm , is very different from those of one component system . in the nonlinear regime , the hydrodynamical nature of the igm leads to the dynamical and statistical departure of the igm from the dark matter . the two component system is no longer to follow the scaling of gravitational clustering of pure dark matter . this deviation is a reason of the baryon missing in gravitational collapsed halos . the dynamical equation of vorticity does not contain terms of gravity . therefore , the igm turbulence characterized by the vorticity will yield , at least , two scales , which violate the scale free of the gravitational hierarchical clustering : 1 . ) the length scale on which the igm fluid has been developed to the state of fully developed turbulence ; 2 . ) the mass scale on which the turbulence pressure is comparable with gravity of halo considered . these scales play important role in the evolution of baryon fraction . with cosmological hydrodynamic simulation , we find that at @xmath99 the first scale is about 3 h@xmath3 mpc , and the second one is @xmath169 h@xmath3 m@xmath4 . with these results , we reach to the following conclusions : \i . the distribution of baryon fraction is highly nonuniform on scales from hundreds kpc to a few of mpc , and @xmath0 varies from as low as 1% to a few times of the cosmic mean . the turbulence can effectively prevent the igm from falling into potential wells of dark matter halos with mass @xmath169 h@xmath3 m@xmath4 . the @xmath0 in dark matter halos is decreasing from @xmath201 at halo mass scales around @xmath6 @xmath7 m@xmath4 to @xmath8 at @xmath2 h@xmath3 m@xmath4 due to the turbulent state of the igm . the estimated turbulence pressure at @xmath99 correspond to a random motion with r.m.s velocity of about @xmath202 km @xmath203 in the scale range from hundreds of kpc and up to @xmath204 mpc . the turbulent pressure is dynamical and non - thermal . when the turbulence dissipated , its kinetic energy becomes the thermal energy . it yields the entropy in halos ( he , feng , & fang 2005 ) . therefore , the dissipation of turbulence actually is a mechanism of heating , which gives a compensation to the cooling of gas in halos . this result is consistent with the burgers shock heating ( he et al 2004 ) . in summmary , the dynamics of turbulence can effectively affect the baryon fraction of halos . wsz acknowledges the support of the international center for relativistic center network ( icranet ) . our work takes advantage of open facilities provided by the nasa hpcc ess group at the university of washington ( http://www-hpcc.astro.washington.edu/ ) . this work is partially supported by the national science foundation of china grant nsfc 10633040 , 10725314 , 10621303 and the 973 program under contract no . 2007cb815402 . agertz o. , et al . , 2007 , mnras , 380 , 963 allen , s. w. , rapetti , d. a. , schmidt , r. w. , ebeling , h. , morris , r. g. , & fabian , a. c. , 2008 , mnras , 383 , 879 anderson , m. e. , & bregman j. n. , 2010 , apj , 714 , 320 batchelor , g.k . 1959 , _ the theory of homogeneous turbulence _ , ( canbridge university press ) batchelor , g.k . 2000 , _ an introductrion to fluid dynamics _ , ( canbridge university press ) berera , a. , & fang , l. z. , 1994 , phys . 72 , 458 benson a. j. , lacey c. g. , baugh c. m. , cole s. , frenk c. s. , 2002 , mnras , 333 , 156 benson a. j. , bower r. g. , frenk c. s. , laccy c. g. , baugh c. m. , cole s. , 2003 , apj , 599 , 38 boldyrev , linde , t. & polyakov , a. , 2004 , phs . 93 , 184503 bonazzola , s. , falgarone , heyvaerts , j. , perault , m. & puget , j. 1987 , a&a , 172 , 293 burns , j. , skillman , s. , & oshea , b. , 2010 , apj , 721 , 1105 carrillo , j. a. , gamba , i. m. , majorana , a. , & shu , c. w. 2003 , j. comput . phys . , 184 , 498 chandrasekhar , s. 1951a , proc . a210 , 18 chandrasekhar , s. 1951b , proc . a210 , 26 crain . r. a. , eke v. r. , frenk c. s. , jenkins a. , mccarthy i. g. , navarro j. f. , & pearce f. r. , 2007 , mnras , 377 , 41 croton d. j. , springel v. , white s. d. m. , de lucia g. , frenk c. s. , gao l. , jenkins a. , kauffmann g. , navarro j. f. , & yoshida n. , 2006 , mnras , 365 , 11 dai , x. y. , bregman , j. n. , kochanek , c. s. , & rasia , e. , 2009 , apj , 719 , 119 de silva , a , c . , kay , s.t . , lidde , a. r. & thomas , p. a. 2004 , mnras , 348 , 1401 dekel a. , silk j. , 1986 , apj , 303 , 39 dolag , k. , vazza , f. , brunetti g. , & tormen g. 2005 , mnras , 364 , 753 dunkley , j. et al . , 2009 , apjs , 180 , 306 eisenstein , d. j. , & hut , p. 1998 , apj , 498 , 137 eke v. r. , navarroo j. f. , frenk c.s . , 1998 , apj , 503,569 emanuel , g. 2000 , analytical fluid dynamics , 2nd ed . crc press ettori . s. 2003 , mnras , 344 , l13 ettori , s. , dolag , k. , borgani , s. , murante , g. , 2006 , mnras , 365,1021 ettori , s. & fabian , a. c. 1999 , mnras , 305 , 834 fang , l. z. & zhu , w.s . , 2011 , advances in astronomy , 2011 , artcle i d 492480 feng , l.l . , shu , c .- w . , & zhang , m.p . 2004 , apj , 612 , 1 frenk c. s. et al . , 1999 , apj , 525 , 554 gavazzi , r. , treu , t. , et al . , 2007 , apj , 667 , 176 gnedin n. y. , 2000 , apj , 542 , 535 giodini , s. , et al . 2009 , apj 703 , 982 grasso , f. , & pirzzoli , r. , 2000a , physics of fluids , 12 , 3265 grasso , f. , & pirzzoli , r. , 2000b , theor . and comp . fluid dyn.,13 , 421 gurbatov , s. n. , saichev , a. i. , & shandarin , s. f. , 1989 , mnras , 236 , 385 kauffmann g. , colberg j. m. , diaferio a. , white s. d. m. , 1999 , mnras , 303 , 188 he , p. , feng , l. l. , & fang , l. z. , 2004 , apj , 612 , 14 he , p. , feng , l.l . , & fang , l.z . 2005 , apj , 623 , 601 he , p. , liu , j. , feng , l.l . , shu , c .- w . , & fang , l.z . 2006 , phys . lett , 96 , 051302 heymans , c. , bell , e. f. et al . , 2006 , mnras , 371 , l60 hoekstra , h. , hsieh , b. c. , yee , h. k. c. , lin , h. , & gladders , m. d. 2005 , apj , 635 , 73 hoeft m. , yepes g. , gottlober s. , springel v. , 2006 , mnras , 371,401 iapichino , l. , schmidt , w. , niemeyer , j. c. , & merklein j. 2011 , ariv:1102.3352 jiang , g. & wu , c.c . 1999 , j. of comput . , 150 , 561 - 594 . jones , b. t. 1999 , mnras , 307,376 kang , h. s. , ryu , d. s. , cen , r. y. , & ostriker , j. p. , 2007 , apj , 669 , 729 keres , d. , katz , n. , dave , r. , fardal , m. , & weinberg , david h. , 2009 , mnras , 396 , 2332 kim , b. , he , p. , pando , j. , feng , l. l. & fang , l. z. 2005 , apj , 625 , 599 komatsu , e. , et al . 2009 , apjs , 180 , 330 lagha , m. , zhong , x. l. , eldredge , j. , & kim , j. , 2009 , aiaa 2009 - 1136 landau , l. , & lifshitz , e. 1987 , fluid mechanics , 2ed edition(pergamon press ) lssig , m. 2000 , phys . lett . , 84 , 2618 lin , c. c. , 1966 , theory of hydrodynamical stability ( cambridge university ) liu , j.r . , & fang , l.z . 2008 , apj , 672 , 11 liu , j.r . , jamkhedkar , p. , zheng , w. , feng , l. l. , & fang , l. z. 2006 , apj , 645 , 861 lo , s .- c . , blaisdelly , g. a. & lyrintzisz , a. s. , 2007 , aiaa 2007 - 827 lu , y. , chu , y.q . , & fang , l.z . 2009 , apj , 691 , 43 lu , y. , zhu , w. s. , chu , y.q . , feng , l. l. & fang , l.z . 2010 , mnras , 408 , 452 mac low m .- m . , ferrara a. , 1999 , apj , 513 , 142 matarrese , s. , & mohayaee , r. 2002 , mnras , 329 , 37 martin , m.p . , piomelli , u. , & candler , g.v . 2000 . theor . and comp . fluid dyn . 13 , 361 - 376 mcgaugh , s. s. , schombert , j. m. , de blok , w. j. g. , & zagursky , m. j. , 2010 , apj , 708 , l14 mandelbaum , r. , seljak , u. , cool , r. j .. blanton , m. , hirata , c. m .. & brinkmann , j. , 2006 , mnras , 372 , 758 mo h. j. , yang x. , van den bosch f. c. , katz n. , 2005 , mnras , 363 , 1155 molnar , s.m . , hearn , n. , haiman , z. , bryan , g. , evrard , a. e. & lake , g. , 2009 , apj , 696 , 1640 navarro j. f. , white s. d. m. , 1993 , mnras , 265 , 271 norman , m. , & bryan , g. , 1999 , lnp , 530 , 106 oshea , b. w. , nagamine , k. , springel , v. , hernquist , l. & norman , m.l . , 2005 , apjs , 160 , 1 pando , j. , feng , l. l. , & fang , l. z. 2004 , apjs , 154 , 475 pfrommer , c. , springel , v. , enblin , t. a. , & jubelgas , m. 2006 , mnras , 367 , 113 pichon , c. & bernardeau , f. 1999 a&a 343 , 663 picone , j.p . , oran , e.s . , boris , j.p . & young jr , t.r . 1984 , nrl memorandum report a234441 pirozzoll , s. , 2002 , journal of computational physics , 178 , 81 polyakov , a. m. 1995 , phys . , 52 , 6183 pueblas & scoccimarro 2008 phys . d , 80 , 043504 rehman , s. f. , eldredge , j. d. , zhong , x. l. , & kim , j. , 2009 , aiaa , 2009 - 941 ruszkowski , m. , & oh , s. p. 2010 , apj , 713 , 1332 ryu , d. s. , kang , h. s. , hallman , e. , & jones , t. w. , 2003 , apj , 593 , 599 ryu , d. , kang , h. , cho , j. , & das , s. 2008 , science , 320 , 909 shandarin , s. f. , & zeldovich , y. b. , 1989 , rvmp , 61 , 185 shu , f. h. , 1992 , the physics of astrophysics volume ii : gas dynamics(university science books ) shu , c .- w , 1999 , in lecture notes in computational science and engineering , 9 , 439 shu , c .- w , 2009 , siam review , 51 , 82 somerville , r. s. , hopkins , p. f. , cox , t. j. , robertson , b. e. , & hernquist , l. , 2008 , mnras , 391 , 481 sun , m. , voit , g. m. , donahue , m. , jones , c. , forman , w. , & vikhlinin , a. , 2009 , apj , 693 , 1142 tasker , e. j. et al . , 2008 , mnras , 390 , 1267 theuns , t. , leonard , a. , efstathiou , g. , pearce , f. r. , & thomas , p. a. 1998 , mnras , 301 , 478 tozzi , p. , & norman , c. 2001 , apj , 546 , 63 valageas , p. , & silk , j. 1999 , a&a , 350 , 725 vazza , f. , brunetti , g. , kritsuk , a. , wagner , r. , gheller , c. , & norman , m. 2009 , a&a , 504 , 33 vergassola , m. , dubrulle , b. , frisch , u. , & noullez , a. 1994 , a&a , 289 , 325 visbal , m. r. , & gaitonde , d. v. , 2005 , aiaa 2005 - 1265 vikhlinin , a. , kravtsov , a. , forman , w. , jones , c. , markevitch , m. , murray , s. s. , & van speybroeck l. , 2006 , apj , 640 , 691 voit , g. m. bryan , g. l. , balogh , m. l. , & bower , r. g. 2002 , apj , 576 , 601 white , s. d. m. , frenk , c. s. , 1991 , apj , 379 , 52 wu , x.p . & xue , y.j . 2000 , mnras , 311 , 825 xue , y. j. & wu , x. p. 2003 , apj , 584 , 34 yuan , q. , wan h. y. , zhang t. j. , liu j. r. , feng l. l. , & fang l. z. 2009 , newa , 14 , 152 zhang , p. j. , & pen , u. l. 2003 , apj , 588 , 704 zhang , y. t. , shi , j. , shu , c. w. , & zhou , y. , 2003 , phys . e , 68 , 046709 zhang , t. j. , et al . 2006 , apj , 642 , 625 zheng , w. , et al . 2004 , apj , 605 , 631 zhu , w. s. , feng , l. l. , & fang , l. z. , 2010 , apj , 712 , 1 in the weno scheme , the smoothness of a hydrodynamical quantities @xmath205 in any dimension is measured by @xmath206 where the definition of @xmath207 is the same as that given in the appendix of zhu et al ( 2010 ) . with @xmath208 , we can construct the one - dimensional discontinuous detector by @xmath209 where @xmath210 . we use @xmath211 and @xmath212 to detect the discontinuous surface ( visbal & gaitonde 2005 ; lo et al . for each cell , we calculate @xmath213 in three directions . the value of @xmath214 is normalized by the max of @xmath214 among all the cells . then , a threshold parameter of @xmath215 of @xmath216 is used to flag discontinuous zone with @xmath217 . finally , we pick up the shocks with mach number @xmath218 from the flagged cells by checking the conditions as that given by ryu et al .

we show that the planar chern - simons ( cs ) theory on can be described by its dimensionally reduced model . this description of cs theory can be regarded as a novel large- reduction for gauge theories on . we find that if one expands the reduced model around a particular background consisting of multiple fuzzy spheres , the reduced model becomes equivalent to cs theory on in the planar limit . in fact , we show that the free energy and the vacuum expectation value of unknot wilson loop in cs theory are reproduced by the reduced model in the large- limit . [ cols= " < " , ] + a novel large- reduction on : + demonstration in chern - simons theory goro ishiki , shinji shimasaki asato tsuchiya + _ center for quantum spacetime ( cquest ) _ + _ sogang university , seoul 121 - 742 , korea _ + _ department of physics , kyoto university _ + _ kyoto , 606 - 8502 , japan _ + _ harish - chandra research institute _ + _ chhatnag road , jhusi , allahabad 211019 , india _ + _ department of physics , shizuoka university _ + _ 836 ohya , suruga - ku , shizuoka 422 - 8529 , japan _

1001.4917

the large-@xmath1 reduction @xcite asserts that large-@xmath1 gauge theories on flat space - times are equivalent to matrix models ( reduced models ) that are obtained by dimensional reduction to lower dimensions ( for further developments , see @xcite ) . it is not only conceptually relevant because it realizes emergent space - time in matrix models , but also practically relevant because it can give a non - perturbative formulation of planar gauge theories as an alternative to lattice gauge theories . it is well - known that the equivalence does not hold naively due to the @xmath7 symmetry breaking @xcite . some remedy is needed for the equivalence to indeed hold ( for recent studies , see @xcite ) . however , as far as we know , there has been no remedy that manifestly preserves supersymmetry in gauge theories . recently , the large-@xmath1 reduction was generalized to theories on @xmath0 in @xcite ( for earlier discussions , see @xcite ) . here @xmath0 is constructed as a nontrivial @xmath8 fiber bundle over @xmath9 by expanding matrix models around a particular background corresponding to a sequence of fuzzy spheres . this novel large-@xmath1 reduction is important from both of the above two viewpoints . it would give hints to the problem of describing curved space - time @xcite in the matrix models @xcite . unlike on flat space - times , the equivalence would hold on curved space - times without any remedy since the theories are massive due to the curvature of @xmath0 . the novel large-@xmath1 reduction can , therefore , give a non - perturbative regularization of gauge theories that respects both supersymmetry and the gauge symmetry . the novel large-@xmath1 reduction has been studied mainly for two cases so far . one is @xmath10 sym on @xmath11 @xcite . the reduced model of this theory takes the form of the plane wave matrix model @xcite . the large-@xmath1 reduction provides a non - perturbative formulation of @xmath10 sym , which respects sixteen supersymmetries and the gauge symmetry so that no fine tuning of the parameters is required . thus the formulation gives a feasible way to analyze the strongly coupled regime of @xmath10 sym , for instance , by putting it on a computer through the methods proposed in @xcite , and therefore enables us to perform new nontrivial tests of the ads / cft correspondence @xcite . the other is chern - simons ( cs ) theory on @xmath0 , which has been exactly solved @xcite . in this case , the equivalence between the reduced model and the original theory can be verified explicitly , which was briefly reported in @xcite . in this paper , we present the verification of the equivalence in detail . we also discuss the large-@xmath1 reduction for cs theory on @xmath12 . our result is an explicit demonstration of the novel large-@xmath1 reduction . cs theory on @xmath0 is a topological field theory associated with the knot theory and therefore interesting in its own right . our formulation gives a new regularization method for cs theory on @xmath0 . cs theory on @xmath0 is also interpreted as open topological a strings on @xmath13 in the presence of @xmath1 d - branes wrapping @xmath0 . topological strings capture some aspects of more realistic string , so that our study is expected to give some insights into the matrix models as a non - perturbative formulation of superstring . more recently , it was shown in @xcite that the large-@xmath1 reduction holds on general group manifolds . in the case of @xmath14 , the background taken in the large-@xmath1 reduction in @xcite is different from that in @xcite . it is desirable to explicitly demonstrate this different type of the large-@xmath1 reduction , for instance , in the case of cs theory on @xmath0 . we leave this for a future study . this paper is organized as follows . in section 2 , we briefly review cs theory on @xmath0 . in section 3 , we summarize the relationships among cs theory on @xmath0 , 2d ym on @xmath9 and the matrix model which were obtained in @xcite . in section 4 , we argue how cs theory on @xmath0 is realized in the matrix model as a novel large-@xmath1 reduction . in section 5 , we show the equivalence between the theory around a particular background of the matrix model and cs theory on @xmath0 . in section 6 , by using the monte carlo simulation , we give a check of this equivalence . section 7 is devoted to conclusion and discussion . in appendices , some details are gathered . in this section , we review some known facts about cs theory on @xmath0 needed for this paper @xcite . we consider cs theory on @xmath0 with the gauge group @xmath15 , whose action is given by @xmath16 where @xmath17 must be an integer . the partition function @xmath18 defines a topological invariant of the manifold @xmath0 , which also depends on a choice of framing . given an oriented knot @xmath19 in @xmath0 , one can consider the wilson loop in an irreducible representation @xmath20 of @xmath15 @xmath21 the expectation value of the wilson loop @xmath22 defines a topological invariant of @xmath19 depending on a choice of framing as well . in this paper , we are concerned with the wilson loop for the unknot , which we denote by @xmath23 , and mainly consider the wilson loop for the unknot in the fundamental representation , which we denote by @xmath24 . @xmath0 is obtained by gluing two solid tori along their boundaries with the transformation @xmath25 where @xmath26 and @xmath27 are generators of @xmath28 , and @xmath29 and @xmath30 are arbitrary integers . @xmath27 has the following representation in the hilbert space @xmath31 , which is obtained by doing canonical quantization of cs theory on @xmath32 : @xmath33,\end{aligned}\ ] ] where @xmath34 denotes a state associated to the highest weight @xmath35 , and @xmath36 is the weyl vector of @xmath37 . the partition function is given by @xmath38 where @xmath39 stands for the trivial representation , while the expectation value of the wilson loop for the unknot is given by @xmath40 the canonical framing corresponds to @xmath41 . in the following , we first present the exact results for the partition function and the expectation value of the wilson loop in the canonical framing . the partition function ( [ partition function of cs ] ) was computed exactly : @xmath42 where the sum over @xmath43 is a sum over the elements of the weyl group @xmath44 of @xmath15 , and @xmath45 is the signature of @xmath43 . the expectation value of the wilson loop for the unknot was also computed exactly : @xmath46 one can obtain different expressions for ( [ result for the partition function ] ) and ( [ result for the wilson loop ] ) by using weyl s character formula @xmath47 and weyl s denominator formula @xmath48 where @xmath49 are positive roots , and @xmath50 is the set of weights associated to the representation @xmath20 . by applying ( [ denominator formula ] ) to ( [ result for the partition function ] ) , one obtains @xmath51 while by applying ( [ character formula ] ) and ( [ denominator formula ] ) to ( [ result for the wilson loop ] ) @xmath52 for the fundamental representation , it reduces to @xmath53 furthermore , by using ( [ denominator formula ] ) , one can show that ( [ result for the partition function ] ) is equivalent to an integral over @xmath1 variables which takes the form analogous to a partition function of a matrix model @xcite : @xmath54 where we have introduced @xmath55 which is identified with the string coupling in the context of topological strings . we call the statistical model defined by ( [ csm ] ) the chern - simons matrix model ( cs matrix model ) . correspondingly , one can find a relation @xmath56 where @xmath57 finally we consider the framing with @xmath58 , which appears in the direct evaluation of the partition function of cs theory on @xmath0 in @xcite . one can see from ( [ partition function by surgery ] ) that the partition function in this framing is related to the one in the canonical framing as @xmath59 one can also see from ( [ wilson loop by surgery ] ) that the expectation value of the wilson loop is related to the one in the canonical framing as @xmath60 in particular , because @xmath61 for the fundamental representation , one obtains @xmath62 similarly , the perturbative part of the free energy is evaluated in this framing as @xmath63 in the following sections , we will show that the reduced model of cs theory on @xmath0 reproduces the results for the original theory in the framing corresponding @xmath58 . in this section , we review part of the results in @xcite , which we use in this paper . in section 3.1 , we dimensionally reduce cs theory on @xmath0 to 2d ym on @xmath9 and a matrix model . in section 3.2 , we describe a classical equivalence between the theory around each multiple monopole background of 2d ym on @xmath9 and the theory around a certain multiple fuzzy sphere background of the matrix model . in section 3.3 , we exactly perform the integration in the matrix model . in section 3.4 , using the result in section 3.3 , we show that the equivalence in section 3.2 also holds at quantum level . in order to dimensionally reduce cs theory on @xmath0 , we expand the gauge field in terms of the right - invariant 1-form defined in ( [ right - invariant 1-form ] ) as @xmath64 we choose an @xmath65-isometric metric for @xmath0 ( [ metric of s^3 ] ) fixing the radius of @xmath0 to @xmath66 . then , we rewrite ( [ chern - simons theory on s^3 ] ) as @xmath67 where @xmath68 is the killing vector dual to @xmath69 defined in ( [ definition of killing vector ] ) , and we have used the maurer - cartan equation ( [ maurer - cartan ] ) . as we explain in appendix [ @xmath0 and @xmath9 ] , we regard @xmath0 as a @xmath8 bundle over @xmath9 . here the fiber direction is parametrized by @xmath71 . by dropping the @xmath72derivatives , we obtain a gauge theory on @xmath9 : @xmath73 where @xmath74 in ( [ chern - simons theory on s^3 ] ) must be an integer , such a restriction is not imposed on @xmath17 in ( [ bf with mass term on s^2 ] ) . ] , the radius of @xmath9 is @xmath75 , and @xmath76 are the angular momentum operators on @xmath9 given in ( [ monopole angular momentum ] ) with @xmath77 . @xmath78 will be identified with the coupling constant of 2d ym on @xmath9 below . first , we see that ( [ bf with mass term on s^2 ] ) is the bf theory with a mass term on @xmath9 . we expand @xmath79 as @xcite @xmath80 where @xmath81 and @xmath82 , @xmath83 . @xmath84 are the gauge field on @xmath9 , and @xmath85 is a scalar field on @xmath9 which comes from the component of the fiber direction of the gauge field on @xmath0 . indeed , we can rewrite ( [ bf with mass term on s^2 ] ) as @xmath86 where @xmath87 $ ] is the field strength . this is the bf theory with a mass term : the first term is the bf term and the second term is a mass term . next , we integrate @xmath85 out in ( [ bf with mass term on s^2 2 ] ) to obtain 2d ym on @xmath9 : @xmath88 furthermore , by dropping all the derivatives in ( [ bf with mass term on s^2 ] ) and rescale @xmath79 as @xmath89 , we obtain a three - matrix model : @xmath90\right ) , \label{n=1^ * matrix model}\end{aligned}\ ] ] where @xmath91 . in the sense of the dijkgraaf - vafa theory @xcite , this matrix model is regarded as a mass deformed superpotential of @xmath10 sym , which gives the so - called @xmath92 theory . the matrix model ( [ n=1^ * matrix model ] ) with the matrix size @xmath93 possesses the following classical solution , @xmath94}_i \otimes 1_{n_s } \label{matrix background}\end{aligned}\ ] ] where @xmath95}_i$ ] are the spin @xmath96 representation of the @xmath97 generators obeying @xmath98}_i , l^{[j_s]}_j]=i\epsilon_{ijk}l^{[j_s]}_k$ ] , and the relation @xmath99 is satisfied . @xmath100 runs over some integers , but its range is not specified here . it will be specified later as @xmath101 with @xmath35 an even positive integer . we put @xmath102 with @xmath103 and @xmath104 integers and take the limit in which @xmath105 where @xmath106 is the area of @xmath9 . then , it was shown classically in @xcite that the theory around ( [ matrix background ] ) is equivalent to the theory around the following classical solution of ( [ bf with mass term on s^2 ] ) with the gauge group @xmath107 , @xmath108 where @xmath109 , and @xmath110 is satisfied . @xmath111 are the angular momentum operators in the presence of a monopole with the monopole charge @xmath112 , which are given in ( [ monopole angular momentum ] ) . this theory can also be viewed as the theory around the following classical solution of ( [ bf with mass term on s^2 2 ] ) , @xmath113 where the upper sign is taken in the region @xmath114 and the lower sign in the region @xmath115 , and @xmath116 and @xmath117 represent the monopole configuration . equivalently , this theory can be viewed as the theory around the multiple monopole background with the background for the gauge field given in ( [ monopole solution ] ) of 2d ym on @xmath9 ( [ 2d ym ] ) . as reviewed in the following two subsections , the above equivalence is extended to quantum level @xcite . first , we rewrite the matrix model ( [ n=1^ * matrix model ] ) in terms of the following matrices ; @xmath118 where @xmath119 is an @xmath93 complex matrix and @xmath120 is an @xmath93 hermitian matrix . the partition function is defined by @xmath121 + ( 1-i\epsilon)zz^\dagger+(1-i\epsilon)\phi^2 \right\}.\end{aligned}\ ] ] here , in order to make the path integral well - defined and converge , we introduce @xmath122 in the exponential and @xmath123 in the action . taking the gauge in which @xmath120 is diagonal as @xmath124 and integrating @xmath119 and @xmath125 out , then we obtain @xmath126 where @xmath127 comes from the diagonalization of @xmath120 , which is the square of the vandermonde determinant . @xmath128 is the volume of @xmath129 and @xmath130 is the volume of the weyl group of @xmath131 . next let us concentrate on the factor in the integrand in ( [ calz_m in eigenvalues ] ) , @xmath132 for this factor , we use the following formula , @xmath133 where `` @xmath134 '' denotes the principal value . then we obtain @xmath135 . \label{expanded prod factor}\end{aligned}\ ] ] it is easily seen that ( [ expanded prod factor ] ) is written in the sum of terms containing @xmath136 where @xmath137 can take the value @xmath138 and we have reordered and relabeled the eigenvalues of @xmath120 as @xmath139 the size of the matrices , @xmath140 , satisfies @xmath141 . @xmath142 represents the @xmath122-th component of the @xmath143-th block . thus ( [ expanded prod factor ] ) is decomposed into terms , each of which is characterized by an @xmath140-dimensional reducible representation of the @xmath97 algebra consisting of @xmath137 blocks ( irreducible representations ) , with a @xmath8 degree of freedom in each block . we label the reducible representations by `` @xmath144 '' '' specify the number of irreducible representations of @xmath97 , @xmath137 , and the dimensions of the irreducible representations , @xmath145 , which satisfy the relation @xmath141 . ] and denote the @xmath8 part in the @xmath143-th block by @xmath146 , putting @xmath147 . then , we obtain @xmath148 where @xmath149 after some calculations , we find that ( [ decomposed form 1 ] ) results in @xmath150 thus the partition function of the matrix model is decomposed into sectors characterized by the ( reducible ) representation of the @xmath97 algebra . note that @xmath134 prevents the integration over the @xmath8 part @xmath146 from mixing the sectors . in this subsection , we show that the partition function of ym on @xmath9 is obtained from that of the matrix model @xcite . as we will see , the sector which consists of @xmath137 irreducible representations in the matrix model corresponds to @xmath151 2d ym on @xmath9 . in the @xmath137 block sector , we group the irreducible representations with the same dimension and label the groups by @xmath100 , such that @xmath152 for @xmath153 . we denote the multiplicity of the @xmath100-th group by @xmath154 . then , the relations @xmath155 and @xmath156 are satisfied . we also denote the @xmath122-th @xmath8 part in the @xmath100-th group by @xmath157 , where @xmath158 . note that in this notation @xmath159 in ( [ n_r ] ) equals @xmath160 . since the partition function of the matrix model ( [ result mm partition function ] ) is completely decomposed into sectors without overlap , we can extract the sectors consisting of @xmath137 blocks . in the @xmath137-block sector , due to the factor @xmath161 , configurations of almost equal size blocks are dominant in the limit in which @xmath162 . so , we consider configurations around the dominated configuration by setting @xmath163 where @xmath164 . substituting this into the @xmath137-block sector in ( [ result mm partition function ] ) and taking the limit ( [ limit ] ) , we obtain @xmath165 here we have absorbed overall irrelevant constants and divergences into a renormalized constant @xmath166 . we have also replaced the denominator in the integrand in ( [ result mm partition function ] ) by constant ( @xmath167 ) and taken the integral regions of @xmath157 to @xmath168 , since the exponential factor in ( [ result mm partition function ] ) oscillates rapidly around @xmath169 in the limit where @xmath170 and @xmath171 , so that the integral dominates around @xmath172 . finally , by rescaling @xmath157 by @xmath173 , and making an analytical continuation @xmath174 , we obtain @xmath175 where irrelevant constants are again absorbed into a constant @xmath176 . @xmath177 exactly agrees with the partition function of @xmath107 2d ym on @xmath9 @xcite . part agrees , and the @xmath8 part does not ( see ref.@xcite ) . however , this difference does not matter in the following arguments . ] in ym on @xmath9 , @xmath104 are identified with the monopole charges since the @xmath104 dependent term in the exponential coincides with the contribution of the classical solution ( [ monopole solution ] ) to the action of 2d ym . this fact is consistent with the classical equivalence reviewed in section 3.2 . in this section , we provide a prescription for realization of cs theory on @xmath0 in 2d ym on @xmath9 and in the matrix model . we also generalize the prescription to cs theory on @xmath12 . in section 5 , we prove the large-@xmath1 equivalence between cs theory and the matrix model based on this prescription . cs theory on @xmath0 is realized from 2d ym on @xmath9 through the large-@xmath1 reduction for the case of compact space proposed in @xcite as follows . note that the kaluza - klein ( kk ) momenta along the fiber ( @xmath178 ) direction in @xmath0 are regarded as the monopole charges on @xmath9 . hence , if one considers the particular monopole solution in 2d ym on @xmath9 such that the spectrum of the monopole charges reproduces the spectrum of the kk momenta along @xmath178 , the theory on @xmath9 expanded around such monopole solution is equivalent to cs theory on @xmath0 . more precisely , in order to realize cs theory on @xmath0 , we first put @xmath179 and @xmath180 in ( [ monopole solution 1 ] ) ( or in ( [ monopole solution ] ) ) and then make @xmath100 run from @xmath181 to @xmath182 , where @xmath35 is a positive even integer . finally we take the following limit , @xmath183 then , we see that 2d ym on @xmath9 expanded around the monopole background is equivalent to @xmath15 cs theory on @xmath0 in the planar limit . here , we can identify the difference of the monopole charges @xmath184 with the momenta @xmath185 along the @xmath178 fiber direction and @xmath35 is interpreted as the momentum cutoff . introducing @xmath1 and taking the limit @xmath186 are needed to suppress the non - planar contribution @xcite . as for the relation between the coupling constants of the two theories , we naively expect that @xmath187 . however , the last relation in ( [ limit 1 ] ) implies that we need some renormalization . in sections 5 and 6 , we show that the above prescription including the renormalization of the coupling constant is indeed valid . furthermore , by combining this equivalence with the relationship between 2d ym on @xmath9 and the matrix model reviewed in section 3.4 , we obtain the following statement : if one takes in ( [ matrix background ] ) @xmath188 and takes the limit in which @xmath189 the theory around ( [ matrix background ] ) of the matrix model ( [ n=1^ * matrix model ] ) is equivalent to @xmath15 cs theory on @xmath0 in the planar limit . in sections 5 and 6 , we confirm this statement . the particular background ( [ background ] ) and the limit ( [ s^3 limit ] ) is the same as the ones adopted in realizing @xmath10 sym on @xmath11 in pwmm @xcite . @xmath103 and @xmath35 play the role of the ultraviolet cutoffs for the angular momenta on @xmath9 and in the fiber direction , respectively . the results in section 3.4 and the above prescriptions lead us to consider the following statistical model , @xmath190 this model is obtained from ( [ 2dym mp ] ) by making an analytic continuation as @xmath191 , @xmath192 and putting @xmath179 , @xmath180 ( @xmath193 ) . it follows from the above arguments that this model reproduces the planar limit of cs theory on @xmath0 in the limit in which @xmath194 with @xmath195 in the next section , we will verify the equivalence of cs theory on @xmath0 with ( [ z of mm ] ) which is equivalent to the reduced matrix model expanded around the particular background . we can easily extend the above argument to cs theory on @xmath12 . we consider the case in which @xmath196 acts on the circle along the fiber direction in @xmath0 . if we use the notation in appendix a , the periodicity for @xmath12 is expressed as , @xmath197 so that the radius of the fiber direction is @xmath198 of that for @xmath199 and the momenta along the fiber direction are discretized as @xmath200 . then , we see that if one imposes ( [ background ] ) with the second equation replaced with @xmath201 and takes the limit ( [ s^3 limit ] ) , the theory around ( [ matrix background ] ) of the matrix model ( [ n=1^ * matrix model ] ) is equivalent to @xmath15 cs theory on @xmath12 in the planar limit . correspondingly , the statistical model for cs theory on @xmath12 is given by replacing @xmath202 in ( [ z of mm ] ) with @xmath203 . in this section , we prove our statement that the model ( [ z of mm ] ) reproduces the planar limit of cs theory on @xmath0 in the limit shown in ( [ limit 2 ] ) and ( [ limit 3 ] ) . this equivalence can be understood from a correspondence between the feynman diagrams in the two theories . the feynman diagrams in ( [ z of mm ] ) have a one - to - one correspondence to those in ( [ csm ] ) which was obtained by rewriting the partition function of cs theory on @xmath0 . furthermore , if one takes the limit given by ( [ limit 2 ] ) and ( [ limit 3 ] ) , each feynman diagram in ( [ z of mm ] ) takes the same value as the corresponding diagram in ( [ csm ] ) . we consider the free energy and the vacuum expectation value of unknot wilson loop operator in the model ( [ z of mm ] ) to see the equivalence . we show that they coincide with the results in cs theory on @xmath0 completely in the planar limit . let us first consider the feynman rule of ( [ z of mm ] ) . for this purpose , we rewrite ( [ z of mm ] ) into a manifestly @xmath15 invariant form . we will see that the result is given by a multi - matrix model with double trace interaction . this form allows us to compare the two theories explicitly . we first factorize the measure in ( [ z of mm ] ) as follows , @xmath204 the first factor on the right hand side is just a constant which we will omit in the following and the second factor gives the vandermonde determinant for each @xmath100 . we next introduce the t hooft coupling in this model as , @xmath205 and redefine the fields as @xmath206 . then , the third factor in ( [ ms in mm ] ) can be written as @xmath207 finally , defining @xmath15 covariant matrices @xmath208 such that @xmath209 is @xmath122-th eigenvalue of @xmath210 for each @xmath100 , we find that ( [ z of mm ] ) can be written as a multi - matrix model , @xmath211 where @xmath212 is given by the double trace interaction , @xmath213 ) . the right figure shows the vertex which includes @xmath214 . the number @xmath215 represents the total number of @xmath210 and @xmath216 included in the vertex.,height=94 ] the feynman rule for ( [ mm covariant ] ) is given as follows . the tree level propagator is @xmath217 in terms of the standard double line notation , this is represented by double lines with indeces @xmath100 and @xmath218 as shown in fig.[fig vertices in mm ] . the double trace interaction vertices are represented in terms of dashed lines which connect two traces . for example , fig.[fig vertices in mm ] ( right ) shows the term in ( [ v in mm ] ) which includes @xmath214 . we connect two vertices coming from @xmath219 and @xmath220 in terms of a dashed line to represent a double trace interaction vertex . the weights of the vertices can be easily read off from ( [ v in mm ] ) . the @xmath221 vertex gives a constant factor , @xmath222 ) .,height=98 ] as in the standard perturbation theory , physical observables such as the free energy are computed by summing all the connected diagrams . fig.[typical diagram in mm ] shows typical examples of connected diagrams which contribute to the free energy . more precisely , the connected diagrams in this theory are those which are connected by dashed lines or by double lines . one can show that only such diagrams indeed contribute to the calculation of observables . since we are interested in the planar limit , we consider what type of diagrams is dominant in this limit . it is easy to see that the leading order in the @xmath223 expansion is given by the diagrams that satisfy two conditions . one is that they are planar diagrams with respect to the double lines in the ordinary sense . the other is that they are divided by two parts by cutting any dashed line . we call the diagrams satisfying the latter condition ` tree ' diagrams in this paper . we can see that the diagrams in which the dashed lines form any loop ( i.e. non-`tree ' diagrams ) give subleading contribution . in fig.[typical diagram in mm ] , the left diagram contributes to the free energy in the large-@xmath1 limit while the right diagram is suppressed since the dashed lines form a loop . next , we construct a feynman rule of cs matrix model ( [ csm ] ) . we first rewrite ( [ csm ] ) into a manifestly @xmath15 invariant form as we did for ( [ z of mm ] ) in the previous subsection . the identification for the coupling constants shown in ( [ limit 3 ] ) allows us to put @xmath224 where @xmath225 is the t hooft coupling in ( [ mm covariant ] ) which is introduced in ( [ thooft in mm ] ) . if we redefine the fields as @xmath226 . we can rewrite the measure in ( [ csm ] ) as follows , @xmath227 where @xmath228 represents that we have omitted an irrelevant constant factor and @xmath229 is the zeta function . we have used the formula , @xmath230 to obtain the second line in ( [ ms ] ) . if we introduce an @xmath15 covariant matrix @xmath231 such that its @xmath122-th eigenvalue is given by @xmath232 , we obtain the following partition function for cs matrix model on @xmath0 @xcite , @xmath233 where the potential @xmath234 is given by the double trace interaction , @xmath235 we construct the feynman rule of ( [ csm covariant ] ) as follows . the tree level propagator is given by @xmath236 which is expressed by double lines as usual in the standard double line notation . the double trace interaction vertices are represented in terms of dashed lines which connect two traces as in the case of ( [ mm covariant ] ) . they are represented in the same way as in fig.[fig vertices in mm ] except that we do not need the indeces @xmath202 here . fig.[fig vertices in mm ] ( right ) without @xmath202 corresponds to the term in ( [ v in csm ] ) which includes @xmath237 . the weights of the vertices can be easily read off from ( [ v in csm ] ) . the @xmath238 vertex gives a constant factor , @xmath239 as in the case of ( [ mm covariant ] ) , connected diagrams in this theory are those which are connected by dashed lines or by double lines . in the @xmath240 limit , the leading order in the @xmath223 expansion is given by the diagrams which are planar in the ordinary sense and ` tree ' with respect to dashed lines . for instance , the right diagram in fig.[typical diagram in mm ] in the case of ( [ csm covariant ] ) ( i.e. the diagram without the indeces @xmath241 . ) is suppressed in the large-@xmath1 limit since the dashed lines form a loop . note that the feynman diagrams in ( [ csm covariant ] ) have the one - to - one correspondence to those in ( [ mm covariant ] ) . in this subsection , we show that ( [ mm covariant ] ) and ( [ csm covariant ] ) are equivalent in the limit shown in ( [ limit 2 ] ) and ( [ limit 3 ] ) . we can prove this equivalence by showing that , in this limit , each feynman diagram in ( [ mm covariant ] ) takes the same value as the corresponding diagram in ( [ csm covariant ] ) . note that we have already imposed the identification of the coupling constants shown in ( [ limit 3 ] ) through ( [ thooft in mm ] ) and ( [ thooft in csm ] ) . in order to prove the equivalence , we first show that @xmath242 holds in the limit shown in ( [ limit 2 ] ) and ( [ limit 3 ] ) , where @xmath243 and @xmath244 are the free energies of ( [ mm covariant ] ) and ( [ csm covariant ] ) , respectively . as an example of the feynman diagrams contributing to the free energy , let us consider the left diagram in fig.[typical diagram in mm ] for the both theories . since the only deference between ( [ vertex in mm ] ) and ( [ vertex in csmm ] ) is @xmath245 and @xmath246 , it is easy to see that the diagram in ( [ mm covariant ] ) divided by @xmath247 becomes equal to the corresponding diagram in ( [ csm covariant ] ) divided by @xmath248 if @xmath249 holds . in fact , this identity holds in the @xmath250 limit . we show the proof of ( [ zeta identity ] ) in appendix [ proof of identity ] . thus , we see that the two diagrams with the appropriate normalization shown in ( [ free energy agreement ] ) take the same value in the @xmath250 limit . we can show more general formula which includes ( [ zeta identity ] ) as a special case . let us consider generic ` tree ' planar diagram in ( [ mm covariant ] ) shown in fig.[tree planar diagram ] . for this diagram , we can show the following identity in the large-@xmath35 limit ) holds also for the diagrams which are ` tree ' and non - planar since ( [ zeta identity 2 ] ) depends only on the structure of the dashed lines . ] , @xmath251 where @xmath252 represents the set of indeces @xmath253 and each @xmath254 runs from @xmath181 to @xmath182 . note that the singular points such as @xmath255 are not included in the summations as in the simpler case of ( [ zeta identity ] ) . the proof of ( [ zeta identity 2 ] ) is shown in appendix [ proof of identity ] . ( [ zeta identity 2 ] ) implies that the value of the diagram in fig.[tree planar diagram ] divided by @xmath247 is equal to the same diagram in ( [ csm covariant ] ) divided by @xmath248 . we find , therefore , that ( [ free energy agreement ] ) holds to all orders in @xmath225 in the limit ( [ limit 2 ] ) . ) . each blob represents a planar diagram.,height=132 ] we can check ( [ free energy agreement ] ) also by performing an explicit perturbative calculation . the detail of the calculation is shown in appendix [ perturbative calculation ] . up to @xmath256 , the result is given by , @xmath257 while for cs matrix model , we can read off the counterpart from ( [ free energy in cs ] ) as , @xmath258 thus , we see that if one takes the limit shown in ( [ limit 2 ] ) and ( [ limit 3 ] ) , ( [ pc 1 ] ) indeed agrees with ( [ pc 2 ] ) . in terms of the formula ( [ zeta identity 2 ] ) , we can show that other physical observables also take the same values in the two theories . in the following , we show that the vev of unknot wilson loop in cs theory which is given by ( [ wl in noncanonical framing ] ) is also reproduced from the model ( [ z of mm ] ) . we introduce the wilson loop operator in @xmath92 matrix model proposed in @xcite , @xmath259 , \label{wl in mm}\end{aligned}\ ] ] where @xmath260 are coordinates on @xmath0 , @xmath261 parameterize the knot @xmath19 and @xmath262 are the right - invariant 1-forms on @xmath0 defined in ( [ right invariant 1-form ] ) . in @xcite , it is shown through taylor s t - duality that this operator is reduced to the wilson loop operator on @xmath0 in the continuum limit . in order to see the correspondence to the unknot wilson loop in cs theory , we consider ( [ wl in mm ] ) with an unknot contour . we take the simplest one , a great circle on @xmath0 , as such contour . the great circle is parameterized , for example , by @xmath263 with @xmath264 $ ] . by substituting ( [ great circle ] ) into ( [ wl in mm ] ) and setting @xmath265 , we obtain , @xmath266 as in section 3.3 , we take the gauge in which @xmath267 is diagonalized . recall that after we integrate out the delta functions shown in ( [ product of delta functions ] ) which appear in the partition function , the remaining eigenvalues of @xmath120 are represented by @xmath268 . then , in ( [ wl 0 ] ) , if we make the analytic continuation @xmath269 and the field redefinition @xmath270 , we obtain , @xmath271 therefore , we find that the unknot wilson loop operator in the fundamental representation is given by ( [ wl ] ) in ( [ z of mm ] ) . on the other hand , in cs matrix model ( [ csm ] ) , the vev of unknot wilson loop is given by ( [ wilson loop in non - canonical framing ] ) . then , the equivalence for the unknot wilson loop is expressed in terms of @xmath231 and @xmath210 as @xmath272 where @xmath273 and @xmath274 denote the expectation values in ( [ mm covariant ] ) and ( [ csm covariant ] ) , respectively . we can show ( [ equivalence for wilson loop ] ) in all orders in perturbation theory . in perturbation theory , the vevs in ( [ equivalence for wilson loop ] ) are calculated by expanding the exponentials . then , ( [ equivalence for wilson loop ] ) is satisfied if @xmath275 holds for @xmath276 . because ( [ zeta identity 2 ] ) holds for all the ` tree ' planar feynman diagrams contributing to the vevs in ( [ equivalence for wilson loop 2 ] ) , we find that ( [ equivalence for wilson loop 2 ] ) and hence ( [ equivalence for wilson loop ] ) hold in all orders in @xmath225 . thus , we have shown that the vev of unknot wilson loop in cs theory is reproduced from the reduced model . it is straightforward to check ( [ equivalence for wilson loop ] ) explicitly by calculating the vev of ( [ wl ] ) in perturbation theory . the method of calculating the vev is almost the same as in the case of the free energy . we show the detail of this calculation in appendix [ perturbative calculation ] . in the large-@xmath277 limit , the result is given by @xmath278 on the other hand , ( [ wl in noncanonical framing ] ) is expanded as , @xmath279 comparing ( [ vev of wilson loop in matrix model ] ) and ( [ unknot wilson loop in cs ] ) , we can see that ( [ equivalence for wilson loop ] ) indeed holds in the @xmath240 limit . we can show the similar equivalence for the case of cs theory on @xmath280 . although this theory possesses many nontrivial vacua , we consider the trivial vacuum only in this paper . the vacua in cs theory on @xmath12 are characterized by the holonomy along the circle on which @xmath196 acts . in @xcite , it is shown that the partition function of cs theory on @xmath12 for each vacuum sector is reduced to the partition function of a matrix model similar to ( [ csm covariant ] ) . for example , the partition function for the trivial vacuum sector is given by ( [ csm covariant ] ) with @xmath225 replaced by @xmath281 . on the other hand , as mentioned in the last part in section [ cs theory on @xmath0 from the matrix model ] , the statistical model for cs theory on @xmath12 is given by replacing @xmath202 in ( [ z of mm ] ) with @xmath203 . by following the same calculation as ( [ ms in mm ] ) and ( [ ms in mm 2 ] ) , we can see that such model is equivalent to ( [ mm covariant ] ) with @xmath225 replaced by @xmath282 . then , with the identification @xmath283 , we see that the two theories are equivalent in the limit ( [ limit 2 ] ) . thus , we find that the theory on @xmath12 is also reproduced from the matrix model . the large-@xmath1 equivalence between ( [ z of mm ] ) and cs theory on @xmath0 can be understood also from the agreement of the eigenvalue densities in the two theories . in this section , by performing a monte carlo simulation , we show that the eigenvalue density of @xmath284 at the saddle point of ( [ z of mm ] ) coincides with that of @xmath285 in ( [ csm ] ) if @xmath286 is sufficiently small compared to the cutoff @xmath182 . the eigenvalue densities for @xmath284 with @xmath100 near the cutoff have some deviation from the density of @xmath285 . however , we will see that such cutoff effect vanishes in the continuum limit and does not contribute to values of physical observables . from this property of the eigenvalue density , we can also give another check of the equivalences for free energy and unknot wilson loop operator which we have shown in the previous section . let us introduce the eigenvalue density of @xmath285 in cs theory on @xmath0 as , @xmath287 and the eigenvalue density of @xmath284 in ( [ z of mm ] ) as , @xmath288 we also define the difference between ( [ eigenvalue density in cs ] ) and ( [ eigenvalue density in mm ] ) as , @xmath289 by performing a monte carlo simulation of ( [ z of mm ] ) , we can see that @xmath290 for @xmath291 while @xmath292 if @xmath100 is near the cutoff @xmath293 . figure [ fig.1 ] shows the result of the simulation . this result shows that as one goes toward the midpoint @xmath294 , the distribution converges rapidly to the distribution of cs theory on @xmath0 . from this result , we conclude that the deviation of the eigenvalue density exists only for sufficiently large @xmath286 . we have also verified through the monte carlo simulation that @xmath295 possesses the following property , @xmath296 this fact is shown as follows . let us consider the saddle point equation of ( [ z of mm ] ) , @xmath297 because @xmath290 for @xmath291 , in order for ( [ saddle pt eq ] ) to be satisfied , @xmath295 with @xmath100 near the cutoff should behave as @xmath298 then , we can expect that @xmath299 in the limit @xmath250 so that ( [ limit of delta rho ] ) is satisfied . we confirmed ( [ logt / t ] ) through the monte carlo simulation . figure [ fig.2 ] is the result of the simulation and it shows that the quantity on the left hand side in ( [ logt / t ] ) is indeed nicely fitted by @xmath300 with @xmath301 . if we assume ( [ limit of delta rho ] ) or ( [ logt / t ] ) , we check that the equivalence of the free energy ( [ free energy agreement ] ) indeed holds . free energy of ( [ z of mm ] ) is written in terms of ( [ eigenvalue density in mm ] ) as , @xmath302 then , by substituting ( [ delta rho ] ) , we obtain @xmath303 by using the formula ( [ sinh formula ] ) , we can show that the first line on the right hand side of ( [ f and fcs ] ) is equal to @xmath304 in cs matrix model in the limit @xmath250 . we can further show that the second and the third lines vanish in @xmath250 as follows . let us introduce a sufficiently large but finite constant @xmath305 such that the eigenvalue densities for all @xmath100 fit in the interval @xmath306 $ ] . such @xmath305 exists because the attractive force in ( [ saddle pt eq ] ) coming from the gaussian potential is stronger than the repulsive force for @xmath307 so that @xmath308 has a compact support . then , the first term in the second line is bounded as @xmath309 ) , we find that this term does not contribute to @xmath310 in the @xmath250 limit . we can apply the same argument to all the terms in the second and the third lines in ( [ f and fcs ] ) . thus , we have given another check of ( [ free energy agreement ] ) . we can also show the equivalence for unknot wilson loop operator . since the vev in ( [ z of mm ] ) can be written as @xmath311 using ( [ limit of delta rho ] ) , we find that @xmath312 in the limit @xmath250 . thus , we have given another derivation of the equivalence for unknot wilson loop . in this paper , we studied a matrix model that is obtained by the dimensional reduction of cs theory on @xmath0 to zero dimension . we found that expanded around the particular background corresponding to @xmath0 , it reproduces the original theory on @xmath0 in the planar limit . we calculated the partition function and the vev of the wilson loop in the theory around the background of the matrix model and verified that they agree with those in the planar cs theory on @xmath0 . we checked these results also by performing the monte carlo simulation . we also extended this equivalence to the case of cs theory on @xmath12 . in this paper , we only consider the unknot wilson loop . it is relevant to see whether the vev of the wilson loop for a real knot is reproduced in our formulation . we should verify that the vev of the wilson loops does not change against continuous deformation of the loops , such that it indeed represents a topological invariant . we should also extend our analysis to the case of the wilson loop with general representation @xmath313 of @xmath37 . it is also interesting to see whether we can use our formulation to achieve the large-@xmath1 reduction for other theories including the cs term such as the abjm model@xcite . we would like to thank k. ohta for his collaboration in the early stage of this work and for many discussions . the work of g. i. is supported by the national research foundation of korea(krf ) grant funded by the korea government(mest ) through the center for quantum spacetime(cquest ) of sogang university with grant number 2005 - 0049409 . the work of s. s. is supported in part by the jsps research fellowship for young scientists . the work of a. t. is supported in part by grant - in - aid for scientific research ( 19540294 ) from jsps . in this appendix , we summarize some useful facts about @xmath0 and @xmath9 ( see also @xcite ) . @xmath0 is viewed as the @xmath97 group manifold . we parameterize an element of @xmath97 in terms of the euler angles as @xmath314 where @xmath315 , @xmath316 , @xmath317 . the periodicity with respect to these angle variables is expressed as @xmath318 the isometry of @xmath0 is @xmath319 , and these two @xmath97 s act on @xmath320 from left and right , respectively . we construct the right - invariant 1-forms , @xmath321 where @xmath69 are explicitly given by @xmath322 the radius of @xmath0 is @xmath66 . @xmath323 satisfy the maurer - cartan equation @xmath324 the metric is constructed from @xmath69 as @xmath325 the killing vector dual to @xmath69 is given by @xmath326 where @xmath327 and @xmath328 are inverse of @xmath329 . the killing vector is explicitly expressed as @xmath330 from the maurer - cartan equation ( [ maurer - cartan ] ) , one can show that the killing vector satisfies the @xmath97 algebra @xmath331=i\epsilon_{ijk}{\cal{l}}_k$ ] . next , let us regard @xmath0 as a @xmath8 bundle over @xmath332 . @xmath9 is parametrized by @xmath333 and @xmath334 and covered with two local patches : the patch i defined by @xmath335 and the patch ii defined by @xmath336 . in the following , the upper sign represents the patch i while the lower sign represents the patch ii . the element of @xmath97 in ( [ euler angles ] ) is decomposed as @xmath337 with @xmath338 @xmath339 represents an element of @xmath9 , while @xmath340 represents the fiber @xmath8 . the fiber direction is parametrized by @xmath341 . the zweibein of @xmath9 is given by the @xmath342 components of the left - invariant 1-form , @xmath343 , which takes the form @xmath344 this zweibein gives the standard metric of @xmath9 with the radius @xmath75 : @xmath345 making a replacement @xmath346 in ( [ killing vector ] ) leads to the angular momentum operator in the presence of a monopole with magnetic charge @xmath347 at the origin @xcite : @xmath348 where @xmath347 is quantized as @xmath349 , because @xmath71 is a periodic variable with the period @xmath350 . these operators act on the local sections on @xmath9 and satisfy the @xmath97 algebra @xmath351=i\epsilon_{ijk}l_k^{(q)}$ ] . when @xmath77 , these operators are reduced to the ordinary angular momentum operators on @xmath9 ( or @xmath352 , @xmath353 . the @xmath97 acting on @xmath320 from left survives as the isometry of @xmath9 . general ` tree ' planar diagram shown in fig.[tree planar diagram ] consists of several parts ( blobs ) which are connected by the dashed lines and each part represents a planar diagram . for each part , we define a certain quantity in order to show ( [ zeta identity 2 ] ) . for instance , if we consider the left diagram shown in fig.[typical diagram in mm ] , we define the following quantity for the part labeled by @xmath100 , @xmath354 namely , for a dashed line connecting @xmath100 and @xmath218 , we assign @xmath355 where @xmath356 is the number of fields contained in the vertex and @xmath357 is the summation of the product of them with respect to the variables other than @xmath100 . then , we can see that @xmath357 has a finite value and bounded from above by a certain constant . this is shown as follows . first , note the following identity , @xmath358 where @xmath229 and @xmath359 are the zeta function and the generalized zeta function , respectively . they are defined by , @xmath360 because @xmath361 , @xmath362 thus , we find that @xmath357 is finite even if we take the limit @xmath250 . we are able to define @xmath357 also for more general diagram shown in fig.[tree planar diagram ] . if the diagram is ` tree ' , through the same calculation just as we described above , we see that @xmath357 is bounded from above by a @xmath35-independent constant which is given by a product of the zeta functions exists also for non-`tree ' diagrams . however , this fact is not important for our purpose . ] . then , let us consider the diagram in fig.[qs ] . we can define @xmath357 for this diagram . we assume that the rectangular part in fig.[qs ] represents a ` tree ' diagram so that @xmath357 for this diagram is bounded from above . we consider the following quantity , @xmath363 since there exists a positive @xmath35-independent constant @xmath364 such that @xmath365 , the second term on the right hand side in ( [ qs1 ] ) is bounded as follows , @xmath366 where we have used ( [ zeta function ] ) to obtain the second line . because the second line in ( [ bound ] ) is @xmath367 for @xmath368 and @xmath369 for @xmath370 , this quantity goes to zero in the limit @xmath250 . therefore , we find from ( [ qs1 ] ) and ( [ bound ] ) that @xmath371 then , for instance , we can show the following equality by using ( [ formula 1 ] ) iteratively , @xmath372 this is nothing but ( [ zeta identity ] ) . since ( [ proof ] ) implies that we can replace @xmath373 on the tip of a branch in the ` tree ' diagrams with @xmath374 in the @xmath250 limit , applying the same calculation , we can show ( [ zeta identity 2 ] ) for the generic ` tree ' planar diagram shown in fig.[tree planar diagram ] . in this appendix , we explicitly evaluate the free energy of ( [ mm covariant ] ) and the vev of ( [ wl ] ) up to @xmath375 . we perform the usual perturbation theory in ( [ mm covariant ] ) . first , we will calculate the free energy . let us expand the interaction terms in terms of the power of the coupling @xmath376 and the partition function as @xmath377 where @xmath378 is the part of which the coupling dependence is @xmath379 . @xmath380 is the free part , @xmath381 and @xmath382 are given by @xmath383 we define the free energy of our matrix model as @xmath384 substituting ( [ cz_1,2,3 ] ) into ( [ free energy ] ) , we obtain @xmath385 where @xmath386 means the connected part of @xmath387 . for example , @xmath388 is calculated as follows ; @xmath389 in feynman diagram introduced in subsection [ feynman rule for our matrix model ] , one can describe this contribution as fig . [ fe_example ] . second , we perform a perturbative calculation of the vev of the unknot wilson loop ( [ wl ] ) up to @xmath375 . we expand the vev as @xmath392 where @xmath393 means the ladder part and @xmath394 means to evaluate with interaction vertices . the ladder part is easily calculated as follows @xmath395 then , up to @xmath375 the ladder part gives @xmath396 next , let us expand the third term in ( [ pert wl ] ) up to @xmath375 @xmath397 . \label{pert wl int}\end{aligned}\ ] ] for example , we can calculate the second term in ( [ pert wl int ] ) as @xmath398 the feynman diagrams corresponding to the leading part are depicted in fig . [ wl_example ] . we can calculate the other terms in the same way and the result is @xmath399 finally , gathering ( [ pert wl ladder ] ) and ( [ pert wl int 2 ] ) , we obtain perturbative expansion of the unknot wilson loop in our matrix model ( [ vev of wilson loop in matrix model ] ) . t. eguchi and h. kawai , phys . lett . * 48 * , 1063 ( 1982 ) . g. bhanot , u. m. heller and h. neuberger , phys . b * 113 * , 47 ( 1982 ) . g. parisi , phys . b * 112 * , 463 ( 1982 ) . d. j. gross and y. kitazawa , nucl . b * 206 * , 440 ( 1982 ) . s. r. das and s. r. wadia , phys . b * 117 * , 228 ( 1982 ) [ erratum - ibid . b * 121 * , 456 ( 1983 ) ] . a. gonzalez - arroyo and m. okawa , phys . d * 27 * , 2397 ( 1983 ) . r. narayanan and h. neuberger , phys . lett . * 91 * , 081601 ( 2003 ) [ arxiv : hep - lat/0303023 ] . g. ishiki , s. shimasaki , y. takayama and a. tsuchiya , jhep * 0611 * ( 2006 ) 089 [ arxiv : hep - th/0610038 ] . t. ishii , g. ishiki , s. shimasaki and a. tsuchiya , jhep * 0705 * ( 2007 ) 014 [ arxiv : hep - th/0703021 ] . t. ishii , g. ishiki , s. shimasaki and a. tsuchiya , phys . d * 77 * ( 2008 ) 126015 [ arxiv:0802.2782 [ hep - th ] ] . m. hanada , h. kawai and y. kimura , prog . * 114 * , 1295 ( 2006 ) [ arxiv : hep - th/0508211 ] . t. banks , w. fischler , s. h. shenker and l. susskind , phys . rev . d * 55 * ( 1997 ) 5112 [ arxiv : hep - th/9610043 ] . n. ishibashi , h. kawai , y. kitazawa and a. tsuchiya , nucl . phys . b * 498 * ( 1997 ) 467 [ arxiv : hep - th/9612115 ] . r. dijkgraaf , e. p. verlinde and h. l. verlinde , nucl . b * 500 * ( 1997 ) 43 [ arxiv : hep - th/9703030 ] . g. ishiki , s. w. kim , j. nishimura and a. tsuchiya , phys . lett . * 102 * , 111601 ( 2009 ) [ arxiv:0810.2884 [ hep - th ] ] . g. ishiki , s. w. kim , j. nishimura and a. tsuchiya , jhep * 0909 * , 029 ( 2009 ) [ arxiv:0907.1488 [ hep - th ] ] . y. kitazawa and k. matsumoto , arxiv:0811.0529 [ hep - th ] . d. e. berenstein , j. m. maldacena and h. s. nastase , jhep * 0204 * , 013 ( 2002 ) [ arxiv : hep - th/0202021 ] . m. hanada , j. nishimura and s. takeuchi , phys . * 99 * ( 2007 ) 161602 [ arxiv:0706.1647 [ hep - lat ] ] . k. n. anagnostopoulos , m. hanada , j. nishimura and s. takeuchi , phys . * 100 * ( 2008 ) 021601 [ arxiv:0707.4454 [ hep - th ] ] . s. catterall and t. wiseman , phys . d * 78 * , 041502 ( 2008 ) [ arxiv:0803.4273 [ hep - th ] ] . j. m. maldacena , adv . * 2 * ( 1998 ) 231 [ arxiv : hep - th/9711200 ] . e. witten , commun . phys . * 121 * , 351 ( 1989 ) . g. ishiki , s. shimasaki and a. tsuchiya , phys . d * 80 * , 086004 ( 2009 ) [ arxiv:0908.1711 [ hep - th ] ] . h. kawai , s. shimasaki and a. tsuchiya , arxiv:0912.1456 [ hep - th ] . t. ishii , g. ishiki , k. ohta , s. shimasaki and a. tsuchiya , prog . * 119 * ( 2008 ) 863 [ arxiv:0711.4235 [ hep - th ] ] . g. ishiki , k. ohta , s. shimasaki and a. tsuchiya , phys . b * 672 * , 289 ( 2009 ) [ arxiv:0811.3569 [ hep - th ] ] . m. marino , rev . phys . * 77 * , 675 ( 2005 ) [ arxiv : hep - th/0406005 ] . m. marino , arxiv : hep - th/0410165 . m. tierz , mod . a * 19 * , 1365 ( 2004 ) [ arxiv : hep - th/0212128 ] . y. dolivet and m. tierz , j. math . * 48 * , 023507 ( 2007 ) [ arxiv : hep - th/0609167 ] . y. kitazawa , nucl . b * 642 * ( 2002 ) 210 [ arxiv : hep - th/0207115 ] . r. dijkgraaf and c. vafa , nucl . b * 644 * , 3 ( 2002 ) [ arxiv : hep - th/0206255 ] ; nucl . b * 644 * , 21 ( 2002 ) [ arxiv : hep - th/0207106 ] ; arxiv : hep - th/0208048 . a. a. migdal , sov . phys . jetp * 42 * , 413 ( 1975 ) [ zh . eksp . * 69 * , 810 ( 1975 ) ] . e. witten , j. geom . * 9 * , 303 ( 1992 ) [ arxiv : hep - th/9204083 ] . j. a. minahan and a. p. polychronakos , nucl . b * 422 * , 172 ( 1994 ) [ arxiv : hep - th/9309119 ] . d. j. gross and a. matytsin , nucl . b * 429 * , 50 ( 1994 ) [ arxiv : hep - th/9404004 ] . m. blau and g. thompson , arxiv : hep - th/9310144 . m. aganagic , a. klemm , m. marino and c. vafa , jhep * 0402 * , 010 ( 2004 ) [ arxiv : hep - th/0211098 ] .

in this paper we propose and discuss a mirror construction for complete intersections in partial flag manifolds . this construction includes our previous mirror construction for complete intersection in grassmannians and the mirror construction of givental for complete flag manifolds . the key idea of our construction is a degeneration of to a certain gorenstein toric fano variety which has been investigated by gonciulea and lakshmibai . we describe a natural small crepant desingularization of and prove a generalized version of a conjecture of gonciulea and lakshmibai on the singular locus of . [ subsection ] [ theorem]example [ theorem]proposition [ theorem]proposition [ theorem]lemma [ theorem]corollary [ theorem]remark [ theorem]definition [ theorem]notations [ theorem]conjecture [ theorem]claim

math9803108

using combinatorial dualities for reflexive polyhedra and gorenstein cones together with the theory of generalized gkz - hypergeometric functions , one can extend the calculation of the number @xmath2 of rational curves of degree @xmath3 on the generic quintic threefold in @xmath4 by candelas , de la ossa , green , and parkes @xcite to the case of calabi - yau complete intersections in toric varieties @xcite . another class of examples which includes calabi - yau quintic @xmath5-folds are calabi - yau complete intersections in homogeneous fano varieties @xmath6 where @xmath7 is a semisimple lie group and @xmath8 is its parabolic subgroup . it is a priori not clear how to find an appropriate mirror family for these varieties , because @xmath6 is not a toric variety in general . in @xcite , we described a mirror construction for complete interesections in the grassmannian @xmath9 , which turned out to involve a degeneration of @xmath9 to a certain singular toric fano variety @xmath10 introduced by sturmfels in @xcite . in this paper we consider the extension of our methods to the case of complete intersections in _ arbitrary partial flag manifolds _ and give complete proofs of statements from @xcite . it turns out that the plcker embedding of any such flag manifold @xmath11 admits a flat degeneration to a gorenstein toric fano variety @xmath12 . this deformation has been studied recently by gonciulea and lakshmibai in @xcite . the `` mirror - dual '' toric variety @xmath13 associated with a reflexive polyhedron @xmath14 has a nice combinatorial description in terms of a certain graph @xmath15 that was introduced by givental for the case of the complete flag manifolds @xcite . the idea of toric degenerations has been discussed in more general framework in @xcite . using residue formula , we compute explicitly a series @xmath16 associated with the graph @xmath17 and conjecture that @xmath18 gives a solution to the quantum @xmath19-module associated with gromov - witten classes and quantum cohomology of the partial flag manifold @xmath20 . we note that there is no essential difficulty in checking the conjecture in each particular case at hand , because it involves only calculations in the small quantum cohomology ring of @xmath20 , for which explicit formulas are known @xcite . applying the `` trick with factorials '' ( see @xcite , or [ ci ] below ) to a calabi - yau complete intersection in @xmath20 , we obtain @xmath18 as a specialization of the toric gkz - hypergeometric series , from which the instanton numbers ( that is , the virtual numbers of rational curves on the calabi - yau ) can be computed via the standard procedure ( see e.g. @xcite ) . as the validity of this trick was shown recently for general homogeneous spaces @xcite , this implies that any instanton numbers computed via the usual `` mirror symmetry method '' are automatically proven to be correct in all cases for which our conjecture on @xmath18 holds . the series @xmath18 of complete flag manifolds has been investigated by schechtman @xcite . the paper is organized as follows . in section 2 we introduce main combinatorial notions used in the definition of a gorenstein toric fano variety @xmath21 associated with a given partial flag manifold @xmath22 . in section 3 we investigate singularities of @xmath21 and show that these singularities can be smoothed by a flat deformation to the partial flag manifold @xmath22 . as a consequence of our results , we prove a generalized version of a conjecture of gonciulea and lakshmibai about the singular locus of @xmath21 @xcite . in section 4 we discuss quantum differential systems following ideas of givental @xcite . finally , in section 5 we explain the mirror construction for calabi - yau complete intersections in partial flag varieties @xmath20 and the computations of the corresponding hypergeometric series @xmath18 . .2truein * acknowledgement * we would like to thank a. givental , s. katz , s .- a . strmme , and e. rdland for helpful discussions and the mittag - leffler institute for hospitality . the second and third named authors have been supported by mittag - leffler institute postdoctoral fellowships . in this section we explain how to associate to an arbitrary partial flag manifold @xmath22 certain combinatorial objects : a graph @xmath23 , a reflexive polytope @xmath24 and a gorenstein toric fano variety @xmath21 . let @xmath25 be a fixed sequence of positive integers . we set @xmath26 , @xmath27 @xmath28 , and @xmath29 . denote by @xmath22 the partial flag manifold parametrizing sequences of subspaces @xmath30 with @xmath31 . then @xmath32 to symplify notations , we shall often write @xmath20 instead of @xmath22 , if there is no confusion about the numbers @xmath33 . by a classical result of ehresmann ( @xcite ) , a natural basis for the integral cohomology of @xmath20 is given by the schubert classes . these are poincar dual to the fundamental classes of the closed schubert cells @xmath34 , parametrized by permutations @xmath35 modulo the subgroup @xmath36 in particular , the picard group of @xmath20 , which is isomorphic to @xmath37 , is generated by @xmath38 divisors @xmath39 , corresponding to the simple transpositions @xmath40 exchanging @xmath41 and @xmath42 . denote by @xmath43 the standard * ladder diagram * consisting of unit squares ( the number of unit squares in @xmath44 is equal to the dimension of @xmath20 ) corresponding to the schubert cell of maximal dimension in the flag manifold @xmath20 . we place the ladder diagram @xmath44 in the lower left corner of a @xmath45-square @xmath46 . the lower left corner of @xmath44 ( or of @xmath46 ) will be denoted by @xmath47 . we denote by @xmath48 @xmath49 the common vertex of the diagonal squares @xmath50 of size @xmath51 and @xmath52 of size @xmath53 ( the figure below illustrates the case @xmath54 ) . figure 1 let @xmath55 be the above ladder diagram . we * associate * with @xmath44 the following : * @xmath56 , the set of centers of unit squares in @xmath44 : we place a dot at the center of each unit square and call elements of @xmath57 * dots*. * @xmath58 , the set consisting of @xmath59 * stars * : an element of @xmath60 is obtained by placing a star at the @xmath61-shift of the lower left corner of each of the diagonal squares @xmath50 @xmath62 . * @xmath63 , the set of oriented horizontal and vertical segments connecting adjacent elements of @xmath64 : the vertical segments are oriented downwards , and the horizontal segments are oriented to the right . @xmath65 is the oriented graph whose set of vertices is @xmath64 , and whose set of oriented edges is @xmath66 . such a graph @xmath17 ( without the orientation ! ) is shown in figure 2 below . the edges of @xmath17 are drawn with solid lines . figure 2 we denote by @xmath67 , @xmath68 , and @xmath69 the free abelian groups ( or lattices ) generated by the sets @xmath70 and @xmath66 . we remark that the lattices @xmath71 and @xmath72 can be viewed as the groups of @xmath73-chains and @xmath74-chains of the graph @xmath17 . then the boundary map in the chain complex is @xmath75 where @xmath76 are the maps that associate to an oriented edge @xmath77 its * head * and its * tail * respectively . figure 3 a * box * @xmath78 in @xmath17 is a subset of @xmath79 edges @xmath80 which form together with their end - points a connected subgraph @xmath81 , such that the topological space associated to @xmath82 is homeomorphic to a circle . the set of boxes in @xmath17 will be denoted by @xmath83 . it is easy to see that @xmath84 we also consider the projection @xmath85 and the composed map @xmath86 since one can regard the groups @xmath72 and @xmath87 together with the homomorphism @xmath88 as the relative chain complex of the topological pair @xmath89 , we have @xmath90 a * roof*@xmath91 @xmath92 , @xmath93 is the set of @xmath94 edges of @xmath17 forming the oriented path that runs along the upper right `` boundary '' of @xmath17 between the @xmath95-th and the @xmath96-st stars in @xmath60 . the * corner * @xmath97 of a box @xmath98 is the pair of edges @xmath99 meeting at the lower left vertex of @xmath82 . so a corner @xmath97 contains one vertical edge @xmath100 and one horizontal edge @xmath101 such that @xmath102 . the roofs and corners give a decomposition of the set @xmath66 of edges of the graph @xmath17 into a disjoint union of subsets : @xmath103 this decomposition is shown in the figure below . figure 4 the * opposite corner * @xmath104 of a box @xmath105 is the pair of edges @xmath106 meeting at the upper right vertex of @xmath82 . an opposite corner @xmath104 contains one vertical edge @xmath107 and one horizontal edge @xmath108 , such that @xmath109 . by elementary arguments one obtains : the elements @xmath110 where @xmath78 runs over the set @xmath83 , form a natural @xmath111-basis of @xmath112 . moreover , the elements @xmath113 and @xmath114 form a natural @xmath111-basis of @xmath115.[basis ] we denote again by @xmath88 the @xmath116-scalar extension @xmath117 of the homomorphism @xmath118 . the polyhedron @xmath119 associated to @xmath20 is the convex hull of the set @xmath120 where the set @xmath66 is identified with the standard basis of @xmath121 . in order to describe the faces of the polyhedron @xmath122 we introduce some further combinatorial objects associated to the ladder diagram @xmath44 . \(i ) a * positive path * @xmath123 in the diagram @xmath44 is a path obtained by starting at one of the points @xmath48 @xmath124 and moving either downwards , or to the left along some @xmath125 edges of @xmath44 , until the lower left corner @xmath47 is reached ( see fig . 5 ) . we denote by @xmath126 the set of positive paths , and by @xmath127 the set of positive paths connecting @xmath48 and @xmath47 , so that @xmath128 note that the number of elements in @xmath127 is @xmath129 \(ii ) a * meander * is a collection of positive paths @xmath130 @xmath131 , with the property that the union @xmath132 is a _ tree _ with endpoints @xmath133 . the set of all meanders is denoted by @xmath134 . figure 5 [ meande ] there is a natural bijection between the codimension @xmath74 faces of @xmath122 and the set @xmath134 of meanders . _ proof : _ since every face @xmath135 of @xmath122 is given by its supporting hyperplane , it follows from the exact sequence @xmath136 that this hyperplane can be described by a linear function @xmath137 which vanishes on @xmath138 and satisfies the conditions @xmath139 and @xmath140 let us show that every meander @xmath141 defines such a linear function @xmath142 . we define the value of @xmath142 on @xmath77 by the formula : @xmath143 it follows that @xmath144 if the meander @xmath145 does nt intersect @xmath100 , and @xmath146 is negative if @xmath145 intersects @xmath100 . now we show that the linear function @xmath142 satisfies the requirement @xmath147 . by [ basis ] , it suffices to prove that @xmath148 for all @xmath93 , and @xmath149 for all @xmath150 . we remark first that every roof @xmath92 , @xmath93 contains exactly one edge @xmath151 intersecting the positive path @xmath152 , for which @xmath153 on the other hand , @xmath154 , for each @xmath155 , @xmath156 . it follows that @xmath157 now let @xmath158 be an arbitrary box . since the positive paths of the meander @xmath145 form a tree , only the following @xmath5 cases can occur : case 1 . the meander @xmath145 does nt intersect edges in @xmath78 . then @xmath159 for all @xmath79 edges of @xmath78 , hence ( [ b - relation ] ) holds . the meander @xmath145 intersects exactly two edges in @xmath78 . then @xmath145 intersects exactly one edge @xmath160 belonging to @xmath97 and exactly one edge @xmath161 belonging to @xmath104 . by the formula ( [ supp ] ) for @xmath142 , we have @xmath162 . so again the relation ( [ b - relation ] ) holds . the meander @xmath145 intersects exactly three edges in @xmath78 . then @xmath145 intersects both edges @xmath163 belonging to @xmath104 and exactly one edge @xmath164 belonging to @xmath97 . by ( [ supp ] ) @xmath165 again the relation ( [ b - relation ] ) holds . therefore , by proposition [ basis ] , @xmath147 . let @xmath166 be the face of @xmath122 defined by the supporting affine hyperplane @xmath167 . we claim that @xmath166 has codimension @xmath74 . since @xmath166 is the convex hull of the lattice points in @xmath122 corresponding to the edges @xmath168 on which @xmath142 takes the value @xmath74 , it is sufficient to show that any linear function @xmath169 satisfying @xmath170 and @xmath171 for all @xmath172 with @xmath173 must coincide with @xmath142 . indeed , by [ basis ] , the value of such a linear function @xmath169 is uniquely determined on each edge @xmath100 of each roof @xmath92 @xmath174 : @xmath175 next we remark that if for some box @xmath158 we have shown that @xmath176 holds @xmath177 , then , by [ basis ] and ( [ b - relation ] ) , we obtain @xmath178 and therefore @xmath179 since only one edge @xmath180 can be intersected by @xmath145 ( see cases 1 - 3 ) . since we have established the equality @xmath181 for all @xmath182 , the above arguments imply the equality @xmath181 for all @xmath77 . now we prove that any codimension-@xmath74 face @xmath135 of @xmath122 can be obtained from some meander @xmath183 . for this purpose , it suffices to show that if a supporting linear function @xmath184 defines a face @xmath185 , then there exists a meander @xmath183 with @xmath186 . the latter is equivalent to the condition @xmath187 for all edges @xmath77 such that @xmath188 . first we remark that the linear function @xmath184 can not attain the value @xmath74 on all edges of the roof @xmath189 , because @xmath184 vanishes on the element @xmath190 ( see [ basis ] ) . now start a positive path @xmath191 at @xmath192 whose first nonempty intersection with edges of the opposite corner @xmath193 of some box @xmath194 occurs on an edge @xmath195 with @xmath196 . since @xmath197 the value of @xmath184 on at least one of the two edges of @xmath97 has to be strictly less then @xmath74 . we prolong our path through that edge and enter a next box , where the same reasoning applies . continuing this , we complete a positive path @xmath191 from @xmath192 to @xmath47 crossing only edges where @xmath184 is strictly less than @xmath74 . now we repeat this construction for each of the @xmath48 in subsequent order , starting at @xmath198 ... etc . if in the process of constructing a positive path @xmath199 we collide with some already constructed positive path @xmath200 @xmath201 , we just follow from this point the path @xmath200 . in the end , we produce a meander with the required property . we conclude that @xmath166 @xmath202 are all the codimension-@xmath74 faces of @xmath122 . @xmath203 is a reflexive polyhedron . [ reflex ] _ proof . _ the statement follows immediately from theorem [ meande ] and from the integrality of the supporting linear function @xmath142 ( see definition 4.1.5 in @xcite ) . the complete rational polyhedral fan @xmath204 is the fan defined as the collection of cones over all faces of @xmath122 . the toric variety @xmath205 associated to the fan @xmath206 will be denoted by @xmath207 using one of the equivalent characterizations of reflexive polyhedra ( see theorem 4.1.9 in @xcite ) , we obtain from [ reflex ] : @xmath208 is a gorenstein toric fano variety . define @xmath209 to be the toric variety @xmath210 associated to the fan @xmath211 , obtained by refining the fan @xmath206 to a simplicial one , whose one - dimensional cones are the same as the ones of @xmath206 ( i.e. , they are generated by the lattice vectors @xmath212 ) and whose combinatorial structure is given by the following @xmath213 primitive collections : @xmath214 in other words , the cones of maximal dimension of the fan @xmath211 are defined by taking all edges @xmath77 except one from each roof and from each corner . the variety @xmath215 is a small toric desingularisation of @xmath8 . [ s - desing ] _ proof : _ we have to show that each cone of @xmath211 is contained in a cone of @xmath216 , and each cone of @xmath211 is generated by a part of a basis . it suffices to prove the above properties for cones of @xmath211 of maximal dimension . choose an edge @xmath217 in each roof @xmath92 @xmath218 and an edge @xmath219 in each corner @xmath97 @xmath158 . this choice determines a @xmath220-dimensional cone @xmath221 in @xmath211 . for each @xmath222 there exists a unique positive path from @xmath48 to @xmath47 with the following two properties : * @xmath199 crosses the edge @xmath217 ; * if @xmath199 enters a box @xmath78 , then it crosses the edge @xmath219 . it is easy to see that the union @xmath223 of these paths is a meander . indeed , if a union of positive paths as above is not a tree , then there must exist a box @xmath150 with both edges of the corner @xmath97 intersecting the union of positive paths . this contradicts the second of the above conditions . therefore the set of edges @xmath224 defines uniquely a meander @xmath183 , and the cone @xmath221 is contained in the cone over the face @xmath225 . on the other hand , the elements @xmath226 and @xmath227 together with the set @xmath228 form a @xmath111-basis of @xmath72 . by [ basis ] , the set of generators of @xmath221 ( i.e. , the @xmath88-image of @xmath229 ) is a @xmath111-basis of @xmath87 . the desingularization morphism @xmath230 induced by the refinement @xmath211 of @xmath206 is small ( i.e. , contracts no divisor ) , because the sets of @xmath74-dimensional cones in @xmath211 and @xmath206 are the same . there is another way to describe @xmath215 , namely as an iterated toric fibration over @xmath231 : one starts with the product of projective spaces @xmath232 corresponding to the roofs . then one chooses a corner @xmath97 of a box @xmath158 whose opposite corner @xmath104 belongs to a roof . this choice allows to define a toric bundle over @xmath231 with the fibre @xmath233 . then one adds a new corner @xmath234 of a box @xmath235 whose opposite corner @xmath236 is contained in the union of roofs and @xmath97 @xmath237 etc . at each stage of this process one gets a toric fibre bundle over @xmath231 , with fibre the space constructed in the previous step . using this description of @xmath215 , one obtains an alternative proof of the fact that the anti - canonical divisor on @xmath8 is cartier and ample , i.e. , that the polyhedron @xmath122 is reflexive . let @xmath158 be an arbitrary box . define @xmath238 to be the closure of the torus orbit in @xmath8 corresponding to the @xmath5-dimensional cone @xmath239 generated by the @xmath88-image of the @xmath79-element set @xmath78 . the singular locus of @xmath8 consists of codimension @xmath5 strata @xmath240 , @xmath158 . these are conifold strata , i.e. transverse to a generic point of @xmath240 the variety @xmath8 has an ordinary double point . [ sing ] _ proof : _ since the desingularization morphism @xmath241 is small , @xmath8 is smooth in codimension @xmath242 . moreover , the singular locus of @xmath8 is precisely the union of toric strata in @xmath8 over which the morphism @xmath243 is not bijective . according to the main result of @xcite , the exceptional locus @xmath244 ( i.e. , @xmath245 ) is the union of toric strata covered by rational curves contracted by @xmath243 . on the other hand , since @xmath215 is an iterated toric bundle , the mori cone @xmath246 is a simplicial cone generated by the classes of the primitive relations @xmath247 and @xmath248 ( see sections 2 and 4 in @xcite ) . since the morphism @xmath243 is defined by the semi - ample anticanonical class of @xmath215 , it contracts exactly the extremal rays in @xmath246 defined by the primitive relations corresponding to the boxes @xmath158 . the rational curves representing each such class cover the codimension-@xmath242 strata @xmath249 , corresponding to the @xmath242-dimensional cones in @xmath211 spanned by the @xmath88-images of the edges forming the opposite corner @xmath104 . these strata are contracted , with @xmath231-fibres , to the codimension @xmath5 strata @xmath240 in @xmath8 corresponding to the @xmath5-dimensional cones @xmath250 over the quadrilateral faces @xmath251 of @xmath122 whose vertices are @xmath88-images of the edges in @xmath78 ( @xmath158 ) . it follows that @xmath252 is exactly the singular locus of @xmath8 . let @xmath253 be a partial flag manifold . the semiample line bundles @xmath254 associated to the schubert divisors @xmath255 define the plcker embedding of @xmath20 into a product of projective spaces : @xmath256 where @xmath129 . we will always consider @xmath20 as a smooth projective variety together with this embedding . we describe now an embedding of @xmath8 in the same product of projective spaces . for each @xmath168 , let @xmath257 be the toric weil divisor on @xmath8 determined by the @xmath74-dimensional cone of @xmath206 spanned by the vector @xmath258 . for every edge @xmath259 which is part of a roof , denote by @xmath260 the subset of @xmath66 consisting of the edge @xmath100 , together with all edges @xmath261 which are either directly below @xmath100 in the graph @xmath17 , if @xmath100 is horizontal , or directly to the left of @xmath100 , if @xmath100 is vertical . fix @xmath262 . for @xmath155 consider the weil divisor @xmath263 . [ cartier ] for each @xmath155 , the weil divisor @xmath263 is cartier . moreover , if @xmath264 is another edge in the same roof , then the associated divisor @xmath265 is linearly equivalent to @xmath263 . _ proof : _ to each edge @xmath266 , and each positive path @xmath267 , joining @xmath48 with @xmath47 , we associate a linear function @xmath268:l(e){\longrightarrow}{{\bf z}}$ ] defined by @xmath269(g)=\left\ { \begin{array}{rl } 0 , & \mbox { \rm if $ \pi\cap g=\emptyset$ and $ g\notin u(e)$,}\\ 0 , & \mbox { \rm if $ \pi\cap g\neq\emptyset$ and $ g\in u(e)$,}\\ -1,&\mbox { \rm if $ \pi\cap g=\emptyset$ and $ g\in u(e)$,}\\ 1,&\mbox { \rm if $ \pi\cap g\neq\emptyset$ and $ g\notin u(e)$,}\end{array } \right.\ ] ] it is an elementary exercise to check that @xmath268 $ ] vanishes on the elements @xmath270 and @xmath271 it follows from proposition [ basis ] that @xmath268 $ ] descends to a linear function on @xmath87 . to show that @xmath263 is cartier , it suffices to construct for each maximal dimensional cone @xmath221 in @xmath206 an integral linear function @xmath272 which vanishes on @xmath273 , and satisfies * @xmath274 , for all @xmath275 such that @xmath276 and @xmath277 . * @xmath278 , for all @xmath275 such that @xmath276 and @xmath279 . by theorem [ meande ] , every maximal cone @xmath221 is determined by a meander @xmath280 , and @xmath276 if and only if the meander does not intersect the edge @xmath108 ( cf . the proof of [ meande ] ) . it follows that @xmath281\ ] ] satisfies the above conditions , where @xmath199 is the positive path in @xmath145 which joins @xmath48 with @xmath47 . hence @xmath263 is cartier . note that the functional @xmath282 defined above does not depend on the positive paths @xmath283 in @xmath145 that do not intersect the roof @xmath92 . to prove the second part of the lemma , define an integral linear function @xmath284 , by @xmath285 as above , one can easily check that @xmath286 vanishes on @xmath273 , hence it descends to a linear function on @xmath87 . the descended linear function defines a rational function on @xmath8 , whose divisor is @xmath287 . this finishes the proof of the lemma . for each @xmath288 , the line bundle associated to the roof @xmath92 is @xmath289 for some edge @xmath155.[linebundles ] it follows from lemma [ cartier ] that @xmath290 does not depend on the choice of the edge @xmath155 . we note that for each maximal dimensional cone @xmath221 the linear function @xmath282 defined in the proof of [ cartier ] satisfies @xmath291 for all @xmath292 such that @xmath293 . this implies that the line bundle @xmath294 is generated by global sections ( cf . @xcite , p.68 ) . we will now identify the space of global sections . the cartier divisor @xmath263 determines a rational convex polyhedron @xmath295 $ ] in the dual vector space @xmath296 , given by @xmath297=\left\{\lambda\in l(d)^*\otimes{{\bf r}}\ ; : \ ; \begin{array}{l}\lambda(\delta(g))\geq -1,\ ; \forall g\in u(e),\\ \lambda(\delta(g))\geq 0,\ ; \forall g\in e\setminus u(e ) \end{array}\right\}.\ ] ] the space of global sections of the line bundle @xmath294 has a natural basis , indexed by the lattice points in @xmath295 $ ] . by its very definition , for each positive path @xmath267 , the linear function @xmath268 $ ] introduced in the proof of [ cartier ] gives such a lattice point . [ global - sect ] for each @xmath288 , the space of global sections of @xmath290 has a natural basis parametrized by the set @xmath127 of positive paths connecting @xmath48 and @xmath47 . _ proof : _ choose an edge @xmath155 . we have to show that the only lattice points in @xmath295 $ ] are the ones given by @xmath268 $ ] , @xmath267 . let @xmath298 be any linear function vanishing on @xmath273 , and such that the descended linear function is in @xmath295 $ ] . since on the one hand @xmath184 vanishes on every @xmath299 and on the other hand @xmath184 can be negative only on edges in @xmath260 , there are exactly two possibilities : @xmath300 @xmath301 for all @xmath302 ; @xmath303 @xmath304 , there exists an edge @xmath107 in the @xmath305 roof @xmath92 with @xmath306 , and @xmath301 for all @xmath307 . if @xmath300 holds , then we start a positive path @xmath123 at @xmath48 that intersects the roof @xmath92 at the edge @xmath100 . let @xmath78 be the box containing @xmath100 in its opposite corner and let @xmath101 be the other edge in @xmath260 contained in this box . if @xmath308 , we prolong the path through the edge @xmath101 , and enter a next box @xmath309 , where we have the same situation as before ( i.e. , there is an other edge @xmath310 , and if @xmath311 , then we prolong the path through @xmath312 etc . ) . so we may assume that @xmath313 . the edge @xmath101 is part of the corner @xmath97 of @xmath78 . let @xmath314 be the other edge in @xmath97 . since @xmath184 vanishes on all elements @xmath315 @xmath316 must be strictly positive ( hence at least @xmath74 ) . we prolong the path @xmath123 through the edge @xmath314 , and enter a next box @xmath317 , for which @xmath314 is part of the opposite corner . now @xmath184 is nonnegative on all four edges of @xmath317 , and @xmath318 . it follows that there must be an edge @xmath319 in the corner @xmath320 , with @xmath321 . we prolong the path through this edge , and enter a next box , where the same reasoning applies . continuing this , we complete eventually a positive path @xmath123 . consider the linear function @xmath322 $ ] on @xmath72 . by construction , and the definition of @xmath268 $ ] , the functional @xmath323 is nonnegative on all edges @xmath275 . on the other hand , @xmath323 vanishes on the generators of @xmath273 described in [ basis ] , since both @xmath184 and @xmath268 $ ] do . we claim that @xmath323 is identically zero on @xmath324 . indeed , since @xmath323 is nonnegative and @xmath325 for @xmath326 , it follows that @xmath323 takes the value zero on each edge in the union of all roofs . similarly , if @xmath323 vanishes on each of the edges of the opposite corner @xmath104 of some box , then it must vanish on each of the edges of the corner @xmath327 as well . from these two facts , one obtains inductively that @xmath323 takes the value zero on every @xmath275 . hence @xmath328 $ ] . assume now that @xmath303 holds . in this case , we start a positive path @xmath123 at @xmath48 that intersects the roof @xmath92 at the edge @xmath107 . a reasoning entirely similar to that in case @xmath300 shows that the path @xmath123 can be completed such that the functional @xmath329 $ ] is nonnegative on every edge . hence we obtain again @xmath328 $ ] . the @xmath330-dimensional cone @xmath331 associated to the flag manifold @xmath20 is the convex polyhedral cone in the space @xmath332 spanned by the vectors @xmath333 with @xmath77 . we denote by @xmath334 the dual cone in the dual space @xmath335 . [ lambdapi ] let @xmath336 be any positive path @xmath336 . we associate to @xmath123 a linear function @xmath337 by setting @xmath338 if the path @xmath123 crosses the edge @xmath100 , and @xmath339 if it does nt . if the path @xmath123 enters a box @xmath150 , then it does so by crossing an edge which is part of the opposite corner @xmath104 , and it has to leave @xmath78 by crossing an edge which is part of the corner @xmath104 . it follows that the corresponding functional @xmath340 is zero on @xmath341 , hence it descends to a functional on @xmath342 , still denoted by @xmath340 . by definition , @xmath340 is a lattice point in the dual cone @xmath343 . [ cone ] the semigroup of lattice points in @xmath334 is minimally generated by the set of all @xmath340 , where @xmath123 runs over the set @xmath126 of positive paths . _ proof : _ let @xmath344 with @xmath345 and @xmath346 for all @xmath77 . we define the _ weight _ of @xmath184 to be @xmath347 it is clear that @xmath348 , and that @xmath349 iff @xmath350 . note also that @xmath351 for all @xmath336 . the statement of theorem will be proved if we show that @xmath352 for all non - zero integral linear functions @xmath344 with @xmath345 and @xmath346 for all @xmath77 , and , moreover , any such @xmath184 is a non - negative integral linear combination of @xmath340 ( @xmath336 ) . by [ basis ] , the requirement @xmath345 is equivalent to @xmath353 for all @xmath150 , or @xmath354 as in the proof of proposition [ global - sect ] , the above condition implies that if @xmath355 , then there exists a roof @xmath92 containing an edge @xmath100 on which @xmath184 is nonzero ( hence @xmath356 ) . we start to construct a positive path @xmath199 from @xmath48 by choosing its edges in such a way that @xmath100 is the first edge of the graph @xmath17 intersected by @xmath199 . let @xmath158 be a box containing @xmath100 in its opposite corner ( i.e. , @xmath357 ) . since @xmath353 , there must be an edge @xmath358 such that @xmath359 ( hence , @xmath360 ) . we prolong the path @xmath199 through @xmath101 and enter a next box @xmath309 , for which @xmath361 . again there must exist an edge @xmath362 such that @xmath363 . @xmath364 etc . continuing this process , we eventually obtain a positive path @xmath199 , which only crosses edges @xmath100 of @xmath17 having the property @xmath356 . this shows that @xmath365 is again an integral non - negative linear functional on @xmath366 . on the other hand , @xmath367 since @xmath368 , this shows that @xmath352 . by induction on @xmath369 , we can assume that @xmath169 is already a non - negative integral linear combination of @xmath340 , hence so is @xmath370 . we define a partial ordering on the set @xmath126 of positive paths by declaring @xmath371 if the path @xmath123 runs above the path @xmath372 . figure 6 it is easy to see that the set @xmath126 of positive paths together with the above partial ordering is a distributive lattice . the maximum @xmath373 for any two paths @xmath123 and @xmath372 is the path bounding the union of the regions under @xmath123 and @xmath372 ; similarly , @xmath374 bounds the intersection of these regions . consider the partition of the set of independent variables @xmath375 into @xmath38 disjoint subsets @xmath376 and define @xmath377 to be the subvariety of @xmath378 given by the @xmath38-homogeneous quadratic equations @xmath379 for all pairs of noncomparable elements @xmath380 . the variety @xmath381 has been investigated by n. gonciulea and v. lakshmibai in the papers @xcite , where the following result has been proved : \(i ) @xmath382 is a @xmath220-dimensional , irreducible , normal , toric variety . \(ii ) there exists a flat deformation @xmath383)\ ] ] such that @xmath384 , and @xmath385 for all @xmath386 . [ deform ] the next theorem describes an isomorphism @xmath387 let @xmath388 be the toric variety associated with a partial flag manifold @xmath389 . the line bundles @xmath290 @xmath218 define an embedding @xmath390 whose image coincides with the toric variety @xmath391 . [ identif ] _ proof : _ we have @xmath392/{{\cal i}})$ ] , with @xmath393 the ideal generated by the quadratic polynomials in ( [ eq ] ) , and @xmath394 is taken with respect to the @xmath395-grading given by @xmath396 if we identify @xmath126 with the set @xmath397 , then @xmath393 is the _ toric ideal _ ( see the definition in @xcite , p. 31 ) associated to this set ( this is a standard fact about the ideals associated to distributive lattices ; see for example thm . 4.3 in @xcite for a proof ) . let @xmath398 be the affine toric variety @xmath399/{{\cal i}})$ ] . by theorem [ cone ] and prop . 13.5 in @xcite , @xmath398 coincides with the affine toric variety defined by the cone @xmath400 , i.e. @xmath401/{{\cal i}}$ ] can be identified with the ring @xmath402 $ ] determined by the semigroup @xmath403 of lattice points in the dual cone @xmath334 . pick an edge @xmath404 for each @xmath262 , and identify the line bundle @xmath290 with @xmath405 . for each @xmath262 , let @xmath406\subset l(d)^*\otimes{{\bf r}}$ ] be the supporting polyhedron for the global sections of the line bundle @xmath290 ( cf . proposition [ global - sect ] ) ; recall that the lattice points in @xmath406 $ ] are given by the linear functions @xmath407 $ ] ( @xmath267 ) defined in the proof of lemma [ cartier ] . define now for each @xmath95 a linear function @xmath408:l(e){\longrightarrow}{{\bf z}}$ ] by @xmath409(f)=\left\ { \begin{array}{rl } 1,&\mbox { \rm if $ f\in u(e)$,}\\ 0 , & \mbox { \rm otherwise.}\end{array } \right.\ ] ] it is clear that @xmath408 $ ] descends to a functional on @xmath410 , and that for every path @xmath267 the functional @xmath340 in definition [ lambdapi ] coincides with @xmath407+v[e_i]$ ] . for each @xmath262 , let @xmath411 be the cone over the translated polyhedron @xmath408+\delta[e_i]$ ] . then the minkowski sum @xmath412 of these cones coincides with the cone @xmath334 , since both @xmath221 and @xmath334 are generated by the vectors @xmath413 . it follows that @xmath414)$ ] , where @xmath394 is taken with respect to the natural @xmath395-grading induced by the decomposition of @xmath334 into the minkowski sum of the @xmath415 s . for each @xmath262 , choose an ordering @xmath416 of the set @xmath127 . let @xmath417 denote the section determined by @xmath418 ( @xmath419 , @xmath420 ) . the line bundles @xmath421 define a morphism @xmath422 @xmath423,\ldots , [ s_{\pi_{l,1}}(x):\ldots : s_{\pi_{l , n_l}}(x)]\right ) .\ ] ] by the above arguments , @xmath424 is the isomorphism @xmath425){\longrightarrow}proj({{\bf c}}[z_{\pi}\ ; ; \ ; \pi\in\pi]/{{\cal i}}),\ ] ] and the theorem is proved . from [ deform ] and [ identif ] we obtain there exists a flat deformation @xmath383)\ ] ] such that @xmath426 and @xmath385 for all @xmath386 . a description of the singular locus of @xmath8 was conjectured by n. gonciulea and v. lakshmibai in the case @xmath20 a grassmannian ( see @xcite ) . our theorem [ sing ] proves this conjecture and its generalization for arbitrary partial flag manifolds @xmath20 . in order to explain our mirror construction , we give a short overview of the quantum cohomology @xmath19-module . the reader is refered to @xcite for details . let @xmath427 be a smooth projective variety . denote @xmath428 and @xmath429 two homogeneous bases of @xmath430 , dual with respect to the poincar pairing , i.e. such that @xmath431 we will consider only the even degree part of @xmath430 and will assume that @xmath432 and @xmath433 are torsion - free . we denote by @xmath74 the fundamental class of @xmath427 . to simplify the exposition , suppose there is a basis @xmath434 of @xmath432 consisting of nef divisors . let @xmath435 be the mori cone of @xmath427 . introduce formal parameters @xmath436 , @xmath437 and let @xmath438 $ ] be the ring of formal power series . the small quantum cohomology ring of @xmath427 will be denoted by @xmath439 . this is the free @xmath440$]-module @xmath441 $ ] , together with a new multiplication given by @xmath442 with @xmath443 the _ @xmath5-point , genus @xmath73 , gromov - witten invariants _ of @xmath427 . for the case of a partial flag manifold , the small quantum cohomology ring is well - understood . a presentation of this ring is known ( @xcite ) , as well as explicit formulas for quantum multiplication ( @xcite ) . the operators of quantum multiplication with the generators @xmath444 give the _ quantum differential system _ , a consistent first order partial differential system ( see e.g. @xcite ) : @xmath445 @xmath446 where @xmath447 is an @xmath430-valued function in formal variables @xmath448 , and @xmath449 , @xmath437 . here @xmath450 is an additional parameter . remarkably , a complete set of solutions to this system can be written down explicitly in terms of the so - called _ gravitational descendants _ @xcite : @xmath451 } \frac{e_1^*(e^{pt/\hbar } t_a)}{\hbar -c}\cup e^*_2(t_b ) \right ) .\\ \end{array}\ ] ] here @xmath452 is kontsevich s space of stable maps , with evaluation morphisms @xmath453 at the two marked points , @xmath454 $ ] is the _ virtual fundamental class _ ( @xcite ) , and @xmath455 is the first chern class of the line bundle over @xmath452 given by the cotangent line at the first marked point . finally , @xmath456 and @xmath457 are shorthand notations for @xmath458 and @xmath459 respectively . the _ quantum @xmath19-module _ of @xmath427 is the @xmath19-module generated by the functions @xmath460 for all solutions @xmath447 to the above differential system . a general conjecture about the structure of quantum @xmath19-modules is givental s version of the mirror conjecture @xcite : there exists a family @xmath461 of ( possibly noncompact ) complex manifolds @xmath462 , having the same dimension as @xmath427 , together with holomorphic functions @xmath463 , and holomorphic volume forms @xmath464 such that the @xmath19-module generated by integrals @xmath465 where @xmath466 are suitable morse - theoretic middle dimension cycles of the function @xmath467 , is equivalent to the quantum @xmath19-module of @xmath427 . [ mirror ] now assume that @xmath427 is fano . let @xmath381 be the zero locus of a generic section of a decomposable rank @xmath468 vector bundle @xmath469 such that each @xmath470 is generated by global sections . in such a situation one can also define a quantum ring @xmath471 over the coefficient ring @xmath472 $ ] , which encodes some of the enumerative geometry of rational curves on the complete intersection @xmath381 . this leads to a quantum differential system for @xmath473 ( see @xcite ) . we define degrees of @xmath436 s by requiring that @xmath474 furthermore , we suppose that all degrees of @xmath436 are nonnegative ( this is equivalent to the condition that @xmath475 is nef ) . one can write down a similar complete set of solutions to the quantum differential system for @xmath473 @xcite : @xmath476 } \frac{e_1^*(e^{pt/\hbar } t_a)}{\hbar -c}\cup e^*_2(t_b ) \cup e_{\beta } \right ) , \\ \end{array}\ ] ] where @xmath477 is the euler class of the vector bundle on @xmath452 whose fibre over a point @xmath478 is the subspace of @xmath479 consisting of sections vanishing at @xmath480 , and the rest of notations are as above . consider the cohomology valued functions @xmath481 and @xmath482 these functions are given explicitly by the expressions @xmath483 and @xmath484 where now @xmath485 is the euler class of the vector bundle on @xmath452 whose fibre over a point @xmath478 is the subspace of @xmath479 consisting of sections vanishing at @xmath486 . if we view @xmath381 as an abstract variety , the general theory in [ qdmod ] gives a @xmath487-valued function @xmath488 . the functions @xmath488 and @xmath489 are closely related . for example , if @xmath490 , where @xmath491 is the inclusion , then @xmath492 . now consider a new cohomology valued function @xmath493 in general it is very hard to compute @xmath488 or @xmath489 explicitly . however , note that @xmath494 can be computed directly from the function @xmath495 associated to the ambient manifold , which in many cases turns out to be more tractable . it is therefore extremely useful to have a result relating @xmath489 and @xmath494 . extending ideas of givental , b. kim @xcite has recently proved the following theorem , which applies to the cases considered in this paper : [ qlef ] if @xmath427 is a homogeneous space and @xmath496 is the zero locus of a generic section of a nonnegative decomposable vector bundle @xmath497 , then @xmath489 and @xmath494 coincide up to a weighted homogeneous triangular change of variables : @xmath498 where @xmath499 are weighted homogeneous formal power series supported in @xmath500 , with @xmath501 , and @xmath502 , @xmath503 . in particular , this implies that the coefficient @xmath504 of the cohomology class @xmath505 in @xmath495 , and the coefficient @xmath506 of @xmath507 in @xmath494 ( specialized to @xmath508 , @xmath509 ) are related in a very simple way . namely , if @xmath510 @xmath511 then @xmath512 we will refer to theorem [ qlef ] as the _ quantum hyperplane section theorem_. the relation @xmath513 above was called the `` trick with factorials '' in @xcite . in this section we give a partially conjectural mirror construction for partial flag manifolds , and use it to obtain an explicit hypergeometric series as the power series expansion of the integral representation . the case of calabi - yau complete intersections is then discussed in some detail . let @xmath514 be a partial flag manifold . in the notations of section [ toric ] , we introduce @xmath38 independent variables @xmath436 , @xmath288 ( each @xmath436 corresponds to the roof @xmath92 ) , @xmath515 independent variables @xmath516 , @xmath158 , and @xmath517 independent variables @xmath518 , @xmath519 . consider the following set of algebraically independent polynomial equations : + @xmath520 roof equations : for @xmath521 @xmath522 @xmath520 box equations : for @xmath523 @xmath524 where @xmath525 . this set of equations was discussed by givental @xcite , and was used to give an integral representation for the solutions to the quantum cohomology differential equations for the special case of complete flag manifolds . the results in that paper were the starting point for our investigations . we describe below givental s result and our ( conjectural ) generalization to a general partial flag manifold . let @xmath526 be the complex affine space with the coordinates @xmath518 @xmath527 . for fixed parameters values of @xmath528 we obtain an affine variety : @xmath529 if all components of @xmath530 are non - zero , @xmath531 is isomorphic to the torus @xmath532 . let @xmath541 denote the set of edges in the diagram @xmath44 that intersect @xmath17 . we orient the vertical edges in @xmath541 upwards and the horizontal edges to the right . let @xmath542 . for @xmath543 , the _ tail _ @xmath544 of @xmath545 , is defined to be the box @xmath546 where @xmath545 starts . similarly , the _ head _ @xmath547 of @xmath545 is the box @xmath548 where @xmath545 ends . if @xmath545 crosses the roof @xmath92 , so that its head `` is outside the graph @xmath17 , we put @xmath549 , and if the tail '' of @xmath545 is outside @xmath17 , we put @xmath550 . in the sense of duality of planar graphs , the graph with vertices @xmath427 , edges @xmath541 , and incidence given by @xmath551 is _ dual _ to the graph @xmath17 with all stars collapsed to one point . for each cone @xmath552 of maximal dimension we define a cycle @xmath553 in @xmath554 by @xmath555 note that the @xmath556 s with @xmath557 are determined uniquely by the @xmath518 s with @xmath558 and the roof and box equations @xmath559 , @xmath560 . the cycle @xmath466 is a real torus , of dimension equal to @xmath561 . since it is defined over the entire family of @xmath554 s , it is invariant under monodromy . the integral over this special cycle will be denoted by @xmath562 . _ proof : _ by leray s theorem , the integral is equal to @xmath567 where @xmath568 is the tube map . for @xmath569 , @xmath570 , the cycle @xmath571 is homologous to the cycle @xmath572 in the complement of the hypersurfaces @xmath573 . we now expand all the terms in the integrand : @xmath574 @xmath575 @xmath576 where @xmath577 makes up the corner and @xmath578 the opposite corner of the box @xmath579 . the integral picks up precisely the constant coefficient of the following power series in the @xmath518 s , with parameters the @xmath580 s and @xmath581 s : @xmath582 now there are three types of edges . type i : @xmath583 for some @xmath419 . then @xmath100 also is edge of the opposite corner of a unique box @xmath78 . only the terms with @xmath584 will give a contribution . type ii : @xmath585 , for two boxes @xmath78 and @xmath309 . we can then assume that @xmath100 is part of the corner of @xmath78 , and the opposite corner of @xmath309 . only the terms with @xmath586 will give a contribution . type iii : @xmath100 is contained in a unique @xmath158 . in this case @xmath100 is part of the corner of @xmath78 . only the terms with @xmath587 will give a contribution . hence we see that the integral is given by the series : @xmath588 where for each edge the number @xmath589 is determined by the @xmath590 and @xmath591 by the above equations . we can rewrite this coefficient nicely in terms of binomial coefficients as follows . each edge @xmath592 of the diagram @xmath44 intersects precisely one edge @xmath77 of type @xmath593 or type @xmath594 . the corresponding coefficient @xmath589 is then given by @xmath595 trivially , @xmath596 the heads of arrows @xmath592 which are _ not _ tails are the heads of arrows intersecting the edges of type i. the tails of arrows @xmath592 which are not heads are in bijection to the edges of type iii . hence , when we pull out a factor @xmath597 from the denominator of the left hand side of the above equality , the other terms in numerator and denominator can precisely be combined into the product @xmath598 this proves the result . note that @xmath599 is the generalized hypergeometric series for the smooth toric variety @xmath215 defined in section 3 . the parameters @xmath600 correspond to the generators of @xmath601 coming from the singular variety @xmath8 ( the pull - backs of the line bundles @xmath602 ) , while @xmath603 correspond to the additional generators of @xmath601 . [ s0 ] the series @xmath614 is the coefficient of the cohomology class 1 in the @xmath615-valued function @xmath616 describing the quantum @xmath19-module of @xmath20 , that is , @xmath617 where @xmath618 stands for @xmath619 , with @xmath620 the schubert basis of @xmath621 , and @xmath622 is the cohomolgy class of a point . @xmath623 besides the case of complete flag manifolds ( cf . remark [ complete ] ) , there is another case for which the above conjecture agrees with previously known results . consider the partial flag manifold @xmath624 of flags @xmath625 . the pl " ucker embedding , identifies @xmath20 with a @xmath626 hypersurface in @xmath627 . the hypergeometric series for @xmath627 is @xmath628 ( cf . @xcite ) , and by the quantum hyperplane section theorem ( @xcite , @xcite ) we obtain that the hypergeometric series for @xmath20 is @xmath629 on the other hand , the recipe of theorem [ hyperg ] gives the formula @xmath630 for the hypergeometric series of @xmath20 . the identity @xmath631 implies that the series @xmath632 and @xmath633 coincide . @xmath634 the quantum pieri formula @xcite gives explicitly the quantum product of a special schubert class with a general one , and in particular the quantum product of a schubert divisor with any other schubert class . using this , one can write down in reasonably low dimensional cases the quantum differential system for @xmath20 and reduce this first order system to higher order differential equations satisfied by the components . in particular , one can write down the differential operators annihilating the component @xmath460 of any solution @xmath447 , and check by direct computation that the hypergeometric series @xmath635 of theorem [ hyperg ] is annihilated by these operators . in @xcite this is done for the grassmannians containing complete intersection calabi - yau @xmath5-folds . for the complete flag manifolds , the operators are known to be the operators for the quantum toda lattice ( see @xcite ) recall that @xmath637 is generated by the line bundles @xmath638 , @xmath639 , which also generate the ( closed ) khler cone . hence any line bundle @xmath640 on @xmath20 which is globally generated is of the form @xmath641 , with @xmath642 nonnegative . the common zero locus of @xmath468 general sections of line bundles @xmath643 will be denoted by @xmath644 . assuming conjecture [ s0 ] , it follows from the quantum hyperplane section theorem that the hypergeometric series @xmath506 has the expression @xmath645 where @xmath646 are the coefficients of @xmath18 in theorem [ hyperg ] . from now on the complete intersection @xmath647 is assumed to be a calabi - yau manifold . the construction of mirrors described in @xcite for the case when @xmath20 is a grassmannian can be extended to the case of a general @xmath20 as follows : @xmath381 can be regarded as the intersection of @xmath648 with @xmath468 general hypersurfaces @xmath649 in @xmath650 , with @xmath651 of multidegree @xmath652 . let @xmath398 be the calabi - yau complete intersection of the same hypersurfaces with the toric degeneration @xmath8 of @xmath20 . for each edge @xmath259 , which is part of a roof , define polynomials @xmath653 where @xmath654 are generically chosen complex numbers ( recall that we have defined @xmath260 as the set consisting of @xmath100 , together with all edges in the graph @xmath17 which are either directly below @xmath100 , if @xmath100 is horizontal , or directly to the left of @xmath100 , if @xmath100 is vertical ) . partition each of the roofs @xmath92 , @xmath639 into @xmath468 disjoint subsets @xmath655 such that @xmath656 . it follows from definition [ linebundles ] and theorem [ identif ] that the toric weil divisor @xmath657 is cartier , and @xmath658 introduce additional independent variables @xmath662 , one for each generator of the lattice @xmath87 . for every edge @xmath168 , set @xmath663 where , as before , @xmath664 ( resp . @xmath665 ) is the head ( resp . tail ) of @xmath100 . the torus @xmath568 can be identified with @xmath666)$ ] , with the embedding @xmath667 induced by the ring homomorphism @xmath668\longrightarrow{{\bf c}}[x_d , x_d^{-1}\ ; ; \ ; d\in d],\;\;\ ; y_e\mapsto x^{\delta(e)}.\ ] ] with this identification , we obtain laurent polynomials @xmath669 for @xmath670 , let @xmath671 be the newton polyhedron of the laurent polynomial @xmath672 the polyhedra @xmath673 define a nef - partition of the anticanonical class of @xmath674 ( see definitions in @xcite ) , and according to @xcite and @xcite , the mirror family @xmath675 of the calabi - yau complete intersection @xmath676 consists of calabi - yau compactifications of the general complete intersections in @xmath568 defined by the equations @xmath677 [ cy ] let @xmath678 be a calabi - yau compactification of a general complete intersection in @xmath568 defined by the equations @xmath679 , with the additional requirement that the coefficients satisfy the relation @xmath680 whenever @xmath681 make up a box @xmath150 , with @xmath682 forming the corner @xmath97 of @xmath78 . then a minimal desingularization of @xmath678 is a mirror of a generic complete intersection calabi - yau @xmath683 . the main period of the mirror @xmath675 of @xmath398 is given by @xmath684 where the extra factors @xmath685 come from the nef - partition of the anti - canonical class of @xmath674 described above . specifically,@xmath686 by direct expansion of the integral defining @xmath687 ( as in theorem [ hyperg ] ) , followed by the specialization @xmath688 , @xmath150 , one gets exactly the hypergeometric series @xmath506 . first , as discussed in @xcite , our construction can be interpreted via _ conifold transitions_. indeed , by theorem [ sing ] , if @xmath381 is generic , then its degeneration @xmath689 is a singular calabi - yau @xmath5-fold , whose singular locus consists of finitely many nodes . the resolution of singularities @xmath690 induces a small resolution @xmath691 . in other words the ( nonsingular ) calabi - yau s @xmath381 and @xmath692 are related by a conifold transition , and conjecture [ cy ] essentially states that their mirrors are related in a similar fashion . second , it is well understood ( see e.g. @xcite ) that the knowledge of the hypergeometric series @xmath506 for a calabi - yau @xmath5-fold gives the virtual numbers of rational curves on @xmath381 via a formal calculation . in @xcite we have used the hypergeometric series @xmath693 to compute these numbers for complete intersections in grassmannians . a ( general ) complete intersection calabi - yau 3-fold in @xmath20 is the common zero locus of @xmath697 general sections @xmath698 , where @xmath699 are line bundles with @xmath700 . hence , if @xmath20 contains a complete intersection calabi - yau @xmath5-fold , then necessarily @xmath701 the flag manifold @xmath702 sits as a @xmath626 hypersurface in @xmath627 . hence these cases ( as well as the case when @xmath20 is projective space ) can be viewed as particular instances of complete intersection calabi - yau s in toric varieties . we list below all the partial flag manifolds ( not excluded by proposition [ bound ] ) for which the inequality ( [ constraint ] ) is satisfied . the anticanonical class of @xmath20 , denoted by @xmath715 , is expressed in terms of the natural schubert basis of the picard group . the last column of the table below contains the possible splittings of the anticanonical class into @xmath716 nonnegative divisors . in general , there is a natural duality isomorphism @xmath717 this is taken into account by listing only one of the two isomorphic flag manifolds . it may also be that the flag manifold is self - dual , i.e. ( [ iso ] ) is an automorphism , and two families of complete intersection calabi - yau @xmath5-folds corresponding to different splittings of the anticanonical class are interchanged by the duality automorphism . whenever this happens ( e.g. , when @xmath20 parametrizes complete flags ) , only one of the two splittings of @xmath715 is listed . v. v. batyrev , d. van straten , _ generalized hypergeometric functions and rational curves on calabi - yau complete intersections in toric varieties _ , commun . math . * 168 * ( 1995 ) , 493 - 533 . ( alg - geom/9307010 ) v. v. batyrev , i. ciocan - fontanine , b. kim , and d. van straten,_conifold transitions and mirror symmetry for calabi - yau complete intersections in grassmannians _ , to appear in nucl . b ( alg - geom/9710022 ) i. ciocan - fontanine , _ on quantum cohomology rings of partial flag varieties _ , institut mittag - leffler preprint , 1997 ( available electronically at http://www.math.okstate.edu/@xmath719ciocan/research.html ) . m. reid , _ decomposition of toric morphisms _ , in _ arithmetic and geometry _ , papers dedicated to i.r . shafarevich on the occasion of his 60th birthday ( m. artin and j. tate , eds . ) vol ii , geometry , progress in math . 36 , birkhuser , boston , basel , stuttgart , ( 1983 ) , 395 - 418 .

multiparticle symmetrization effects are contributions to the spectra of bose - symmetrized states which are not the product of pairwise correlations . usually they are neglected in particle interferometric calculations which aim at determining the geometry of the boson emitting source from the measured momentum distributions . based on a method introduced by zajc and pratt , we give a calculation of all multiparticle symmetrization effects to the one- and two - particle momentum spectra for a gaussian phase space distribution of emission points . our starting point is an ensemble of-particle bose - symmetrized wavefunctions with specified phase space localization . in scenarios typical for relativistic heavy ion collisions , multiparticle effects steepen the slope of the one - particle spectrum for realistic particle phase space densities by up to 20 mev , and they broaden the relative momentum dependence of the two - particle correlations . we discuss these modifications and their consequences in quantitative detail . also , we explain how multiparticle effects modify the normalization of the two - particle correlator . the resulting normalization conserves event probabilities , which is not the case for the commonly used pair approximation . finally , we propose a new method of calculating bose - einstein weights from the output of event generators , taking multiparticle correlations into account . -1 cm

nucl-th9801009

most hadrons are emitted in the final stage of a relativistic heavy ion collision . they do not probe directly the hot and dense intermediate stages where quarks and gluons are expected to be the relevant physical degrees of freedom for equilibration processes . the geometrical size and dynamical state of the hadronic phase space emission region , however , depends sensitively on the entire evolution of the collision . this motivates current attempts to reconstruct its spatial and dynamical state from the experimental hadron spectra and to use it as a starting point for a dynamical back extrapolation into the hot and dense intermediate stages @xcite . two - particle correlations of identical particles which are sensitive to the space - time characteristics of the collision @xcite , play a crucial role in this approach . the reconstruction program based on their analysis has very good prospects : due to the increasing event multiplicities and larger statistics of the cern sps lead beam program ( and the yet better quality data expected from relativistic heavy ion collider rhic at bnl ) , particle interferometric measurements start showing statistical errors on the few percent level . also , systematic errors are increasingly better understood . theoretical calculations should aim for a similar accuracy and control necessary approximations quantitatively . one uncontrolled approximation used so far in almost all particle interferometric calculations is to neglect for the particle momentum spectra of bose - einstein symmetrized @xmath0-particle states all multiparticle correlations which can not be written in terms of simpler pairwise ones . this reduces the number of @xmath1 terms contributing to the two - particle correlator @xmath2 to a manageable sum over all particle pairs . beyond this approximation , two approaches have been used in the literature . first , zajc has employed monte carlo techniques @xcite to generate events with realistic multiparticle correlations . this amounts to a shifting prescription @xmath3 @xmath4 @xmath5 which modifies the momentum distribution of simulated events according to unit weights ( which themselves depend on the space - time structure of the source ) . second , zajc has found @xcite the first steps towards a calculational scheme brought into final form by pratt @xcite : for model distributions , the @xmath0-particle spectra are given by a simple algorithm involving only two types of terms : @xmath6 and @xmath7 . in practice , this reduces the sums over all @xmath1 permutations , typical for the calculation of momentum spectra , to sums over all partitions of @xmath0 . for both these approaches , there are first numerical calculations @xcite and related analytical attempts @xcite to control multiparticle effects to the one- and two - particle spectra , but a detailed study of their momentum dependence is missing , even for simple models this work aims at filling this gap , making quantitative statements about the extent to which the slope of the one - particle spectrum and the relative momentum dependence of the two - particle correlations are modified due to multiparticle symmetrization effects . our investigation takes the set @xmath8 of phase space emission points @xmath9 as _ initial condition_. for notational simplicity , we restrict the discussion to one particle species , like - charge pions say . to the set @xmath8 , we associate a symmetrized @xmath0-particle wave function @xcite @xmath10 where the sum runs over all permutations @xmath11 of the @xmath0 indices , @xmath12 is a shorthand for the @xmath0 3-dimensional coordinates @xmath13 , and the functions @xmath14 denote single particle wavepackets centered around @xmath15 at initial time @xmath16 and propagated according to the free time - evolution . final state interactions , which imply a structure of the @xmath0-particle state different from ( [ 1.1 ] ) , will not be considered in the present work . the wave function @xmath17 _ defines _ the boson emitting source . what we are interested in is the calculation of the one- and two - particle momentum spectra resulting from ( [ 1.1 ] ) , the information they contain about the initial distribution of the ` emission points ' @xmath18 , the extent to which these results modify the predictions of the pair approximation , and finally the algorithm which implements the numerical calculation of multiparticle spectra from the initial distribution @xmath8 . our work is organized as follows : section [ sec2 ] shortly sets up and illustrates the general formalism via which particle momentum spectra are calculated from an @xmath0-particle state . in section [ sec3 ] , we discuss the properties of gaussian wavepackets which we choose for the single - particle states @xmath19 in ( [ 1.1 ] ) . section [ sec4 ] contains the main results . it gives a complete qualitative and quantitative analysis of multiparticle contributions to the one- and two - particle spectra for a boson emitting source of gaussian phase space distribution . in section [ sec5 ] , we discuss the implications of this model study for an algorithm which calculates the hanbury brown twiss ( hbt ) radius parameters and momentum spectra for an arbitrary distribution of emission points . results and related conceptual questions are summarized and discussed in the conclusions . we want to determine for the @xmath0-particle symmetrized wave function @xmath17 the detection probability for measuring @xmath0 identical bosons at time @xmath20 at the positions @xmath13 and momenta @xmath21 . we calculate first the @xmath0-particle wigner phase space density for @xmath17 @xcite [ 2.1 ] @xmath22 here , @xmath23 denotes the momentum operator acting on @xmath17 . the one - particle pseudo - wigner functions @xmath24 provide the basic building blocks for the calculation of the @xmath0-particle momentum spectrum which is obtained by integrating ( [ 2.1a ] ) over all spatial coordinates , [ 2.2 ] @xmath25 where @xmath26 is a normalization constant , @xmath27 , ensuring that the probability of detecting @xmath0 particles is one . for a free time evolution of @xmath17 , the integration over the spatial components of ( [ 2.2a ] ) leads to a time - independent expression @xmath28 , since interactions between the particles are necessary to change the momentum distribution in time . in contrast , integrating @xmath29 over all momenta leads to the detection probability of the @xmath0 bosons at positions @xmath13 , which is a time - dependent quantity since free evolving bosons change their positions in time . the calculation of the @xmath0-particle momentum spectrum according to ( [ 2.2a ] ) involves a sum over @xmath30 terms . due to the factorization of @xmath31 , this reduces to a sum over @xmath1 terms , @xmath32 in what follows , we are especially interested in the one- and two - particle momentum spectra @xmath33 , @xmath34 associated with the @xmath0-particle state @xmath17 . these are obtained by integrating @xmath28 over all but one , respectively two momenta , [ 2.4 ] @xmath35 all particle momentum spectra are given in terms of the building blocks @xmath36 ( which determine @xmath37 ) and @xmath38 . once the analytical form of the single particle wavefunctions @xmath14 is specified , these are readily calculated . in what follows , capital letters denote measurable position and momentum coordinates , small letters @xmath39 , @xmath40 , @xmath16 denote the centers of wavepackets which are not directly measurable . the only exception to this is the measurable relative momentum @xmath41 of the two - particle correlator @xmath42 which we denote by a small letter . dynamical correlations between particles in the source are reflected in correlations in the set of emission points @xmath9 . if there are no correlations , then the initial distribution of the centers of single particle wavepackets is given by a one - particle probability distribution @xmath43 . the @xmath44-particle spectra for a set of events with multiplicity @xmath0 are obtained by averaging over this distribution [ 2.5 ] @xmath45 a particular model distribution with correlations which assumes that the emission probability of a boson is increased if there is another emission in its vicinity @xcite , is e.g. obtained by replacing in ( [ 2.5a ] ) @xmath46 here @xmath47 is an averaged normalization defined below . the technical advantage of adopting ( [ 2.5ext ] ) is that the @xmath48-dependent normalization factor @xmath26 in the spectrum @xmath49 of ( [ 2.5a ] ) is canceled . this allows to write without approximation all spectra in terms of the building blocks @xcite [ 2.6 ] @xmath50 the resulting zajc - pratt ( zp ) algorithm for the calculation of one- and two - particle spectra reads @xcite @xmath51 these spectra are normalized to unity . in appendix [ appb ] , we give their derivation in some combinatorical detail . due to the zp - algorithm , the calculation of the @xmath0-particle spectra is reduced from sums over all permutations to sums over all partitions @xmath52 of a set of @xmath0 points into @xmath53 subsets of @xmath44 points , @xmath54 . the number of partitions of @xmath0 grows asymptotically like @xmath55 . for explicit calculations with event multiplicities in the hundreds , it is hence important to get control over the @xmath56-dependence of the pratt terms @xmath6 and @xmath7 . this is our strategy in section [ sec4 ] . to illustrate the above formalism , we consider a normalized @xmath0-particle density matrix @xmath57 for multiparticle states @xmath58 , created by repeated operation of the single - particle creation operator @xmath59 , [ 2.10 ] @xmath60 this density matrix specifies in particular the one- and two - particle spectra @xmath61 $ ] and @xmath62 @xmath63 $ ] . for the gaussian model distribution @xcite [ 2.11 ] @xmath64\ , , \label{2.11a } \\ \rho({\bf x } ) & = & ( \pi r^2)^{\footnotesize{-3/2 } } \exp\left[- { \bf x}^2/ r^2 \right ] \ , , \label{2.11b } \end{aligned}\ ] ] the effects of multiparticle correlations on the hbt - radius parameters have been considered already by zajc . his discussion however is restricted to an explicit calculation of three - particle symmetrization effects and to qualitative estimates of higher order contributions . here , we demonstrate that the zp - algorithm allows for a complete quantitative analysis of the zajc model . the first step is to identify the building blocks of the particle spectra ( [ 2.4 ] ) , [ 2.12 ] @xmath65\ , , \label{2.12a } \\ f_{ij } & = & \exp\left [ { - { p_0 ^ 2\over 2}({\bf x}_i - { \bf x}_j)^2}\right]\ , . \label{2.12b } \end{aligned}\ ] ] the calculation of the terms @xmath6 reduces then to @xmath56-dimensional gaussian integrations . one finds @xmath66 and [ 2.13 ] @xmath67 where @xmath68 denotes chebysheff polynomials of the first kind @xcite . the momentum - dependent terms read [ 2.14 ] @xmath69 the main message of these involved but explicit expressions is contained in the m - dependence of the terms @xmath70 and @xmath71 . these specify the @xmath72- and @xmath73-dependence of the building blocks @xmath7 and hence of all spectra . here , they are functions of the phase space volume @xmath74 only , and their behaviour can be understood by simple arguments : the factors @xmath71 generically increase with increasing @xmath56 , see figure [ fig1 ] . especially , in a limiting case , one finds @xmath75 the reason is that bose - einstein symmetrization effects enhance the low momentum region of the one - particle spectrum , leading to steeper slopes . this one - particle spectrum @xmath76 is a linear superposition of gaussian terms @xmath77 , which due to @xmath71 show increasingly steeper slopes . the @xmath73-dependence of @xmath7 governs the @xmath73-dependence of the two - particle spectrum . the terms @xmath70 depicted in figure [ fig1 ] decrease with increasing @xmath56 and have the limiting values @xmath78 zajc has concluded on the basis of this behaviour that @xcite `` the two - particle correlation function becomes a superposition of terms with successively broader distribution in @xmath79 , leading to an increasingly smaller value for the inferred radius . '' the reason is that bose - einstein symmetrization effects enhance the probability of finding bosons closer together in configuration space and hence result in broader @xmath73-distributions of the two - particle spectra . the above arguments explain the effect of multiparticle correlations qualitatively . for a quantitative understanding , the weights of higher order terms contributing to the one- and two - particle spectra are important . these weights are governed by the terms @xmath6 which ( up to a correction factor @xmath80 of order unity ) essentially decrease like @xmath81th powers of the inverse phase space volume . for event multiplicities in the hundreds , a quantitative analysis can then be done numerically , using the analytical expressions ( [ 2.13 ] ) and ( [ 2.14 ] ) . we defer such a study to a slightly more general model in section [ sec4 ] where certain analytical simplifications allow for a more transparent discussion . a short comparison of the qualitative and quantitative properties of the models studied here and in section [ sec4 ] is given in the text following eq . ( [ 4.4 ] ) . in the example of section [ sec2b ] , we have started from single particle creation operators @xmath82 whose momentum support @xmath83 does not carry a label @xmath84 . as a consequence , the @xmath0-particle states considered in section [ sec2b ] are build up from @xmath0 single particle wavefunctions with _ identical _ phase space localization . we now adopt a more general setting in which a set of @xmath0 phase space points @xmath85 is associated with @xmath0 gaussian wavepackets @xmath14 , centered at initial time @xmath16 around the points @xmath15 with spatial width @xmath86 , @xcite @xmath87 the fourier transform of @xmath14 is proportional to @xmath88 @xmath89 $ ] and can be compared to the momentum support of @xmath82 in ( [ 2.10a ] ) . the corresponding @xmath0-particle symmetrized states reduce to those considered in section [ sec2b ] for @xmath90 and @xmath91 , when the momentum support function becomes independent of the particle label @xmath84 . the one boson state ( [ 3.1 ] ) is optimally localized around @xmath15 in the sense that it saturates the heisenberg uncertainty relation @xmath92 , with @xmath93 at initial time @xmath94 . different choices of single particle wavefunctions can lead to different results for the calculated spectra . we have some control over the extent to which different choices matter by finally checking the @xmath86-dependence of our results . the model parameter can range between @xmath95 $ ] . in refs . @xcite , it was argued that a realistic value for @xmath86 is for pions of the order of the compton wavelength . the free time evolution of @xmath14 is specified by the one - particle hamiltonian @xmath96 which acts as a multiplication operator in momentum space , @xmath97 the resulting building blocks for the @xmath0-particle spectra are [ 3.3 ] @xmath98 to streamline our notation , we have neglected in @xmath99 a factor @xmath100 $ ] . the product of these factors cancels in the calculation of the wigner function ( [ 2.1b ] ) and a fortiori in all the functions derived from it . the function @xmath101 denotes the probability that a boson in the state @xmath14 is detected with momentum @xmath102 . this measured momentum @xmath102 has a gaussian distribution around the central momentum @xmath39 of the wavepacket . the functions @xmath38 in ( [ 2.4c ] ) characterize the overlap between the wavepackets @xmath14 and @xmath103 and play an important role in the zp - algorithm . they take a particularly simple form if all particles are emitted in a flash , @xmath104 \nonumber \\ & & \times \exp \left [ -{i \over 2 } ( { \bf p}_i+{\bf p}_j)\cdot ( { \bf r}_i-{\bf r}_j)\right]\ , . \label{3.4 } \end{aligned}\ ] ] all terms contributing to @xmath105 or @xmath106 contain the same number of factors @xmath38 and hence , the normalization of @xmath38 does not matter in what follows . for notational convenience , we change it by fixing @xmath107 . the functions @xmath38 measure the distance @xmath108 between the phase space points @xmath84 and @xmath109 . this distance measure depends on the wavepacket width @xmath86 but leaves the phase space volume independent of @xmath86 , @xmath110\ , , \nonumber \\ z_j & = & { 1\over \sigma}r_j + i\sigma p_j\ , . \label{3.5 } \end{aligned}\ ] ] we now determine quantitatively multiparticle correlation effects for a source of @xmath0 identical bosons whose wavepackets of spatial width @xmath86 are emitted instantaneously according to a gaussian @xmath0-particle phase space distribution ( [ 2.5ext ] ) with @xmath111\ , . \label{4.1}\ ] ] our main aim is to study for this model the extent to which multiparticle correlations modify the slope of the one - particle spectrum and the width of the two - particle correlator . to this end , we calculate first the building blocks of the zp - algorithm . having explicit expressions for @xmath70 , @xmath71 and @xmath6 in terms of simple polynomials will simplify our discussion considerably . the corresponding first order pratt terms are [ 4.2 ] @xmath112 all higher order pratt terms can be calculated explicitly as averages over relative and average pair distributions . details are given in appendix [ appa ] . the momentum - independent terms read [ 4.3 ] @xmath113 here , @xmath114 denotes the greatest integer not larger than @xmath115 ( ` floor ' ) , and @xmath116 the least integer not smaller than @xmath115 ( ` ceiling ' ) . the notational shorthands @xmath117 and @xmath118 range between 0 and 1 depending on the phase space localization of the wavepacket centers , i.e. , they map the whole parameter space @xmath119 , @xmath120 of the model ( [ 4.1 ] ) onto a finite region . the momentum - dependent terms are given by [ 4.4 ] @xmath121 the comparison of the present model calculation with the zajc model ( [ 2.11 ] ) is not straightforward . as mentioned in the sequel of eq . ( [ 3.1 ] ) , the wavepackets used in both models can be compared by setting the wavepacket centers @xmath91 . however , the integral over @xmath122 in ( [ 2.10b ] ) performs an average over the positions @xmath123 while in the model ( [ 4.1 ] ) we do not average over the position @xmath124 but over the centers of wavepackets @xmath39 , @xmath40 . due to these different starting points , the zajc model is not a simple limiting case of ( [ 4.1 ] ) . nevertheless , main features of the zajc model ( [ 2.11 ] ) can be reproduced qualitatively in the present model . the leading contribution of the momentum - independent pratt terms @xmath6 shows again a power law behaviour . also , the @xmath56-dependence of the terms @xmath70 and @xmath71 of the zajc model is recovered in certain limiting cases , [ 4.5 ] @xmath125 in general , however , the @xmath56-dependence of the terms @xmath70 and @xmath71 is much weaker in the present model . especially , there is no limit in which both @xmath126 and @xmath127 . these differences between both models may provide a first idea about the extent to which the choice of the model distribution affects our conclusions . the normalization @xmath47 is not a direct physical observable , but it determines the weights with which multiparticle correlations contribute to the particle spectra . to see this , we consider the one - particle spectrum , @xmath128 the @xmath56-th order contributions @xmath129 are normalized to one , and the weights @xmath130 add up to unity , [ 4.8 ] @xmath131 the lowest order contribution @xmath132 , by which the one - particle spectrum is typically approximated , contributes a fraction @xmath133 only , the value @xmath134 characterizes the importance of higher order contributions . for a quantitative analysis we now determine the dependence of the normalization @xmath47 and the weights @xmath130 on the event multiplicity @xmath0 and the phase space density of the emission region . we consider the terms @xmath6 , the building blocks of @xmath47 . for the present model , these are given in ( [ 4.3a ] ) . the factor @xmath135 in this equation ranges between @xmath136 and @xmath137 , and can be written as [ 4.9 ] @xmath138 the correction factor @xmath139 appears only linearly in the expressions for @xmath6 , rather than as an @xmath56-th power , and it is of order @xmath140 ( it is exactly @xmath141 for a choice of parameters @xmath142 and @xmath143 such that @xmath144 ) . this allows for the approximation [ 4.10 ] @xmath145 with which the normalization @xmath47 takes the simple form @xmath146 here , the combinatorical factors @xmath147 denote the number of permutations of @xmath0 elements which contain exactly @xmath148 cycles . they are commonly refered to as stirling numbers of the first kind . we have used their generating function in terms of @xmath149-functions @xcite . we can now determine the weights of multiparticle contributions to the one- and two - particle spectra . for the one - particle spectrum , we find by inserting ( [ 4.11 ] ) into the zp - algorithm [ 4.12 ] @xmath150 the approximation in the last line is valid for large multiplicities , when @xmath151 . similarly , the weights @xmath152 for the different contributions to the two - particle momentum spectrum can be calculated . using the power law behaviour @xmath153 , we find [ 4.13 ] @xmath154 where @xmath155 again , the approximation in the last line is valid for large multiplicities , when @xmath151 . to sum up : multiparticle correlations account for a fraction @xmath156 of the one - particle spectrum . for the two - particle spectrum , they are somewhat more important : the pure pairwise correlations receive only a weight @xmath157 . for sufficiently large event multiplicities , the weights @xmath130 and @xmath152 of multiparticle contributions are not separate functions of @xmath158 and @xmath0 , but functions of the product @xmath159 only . the physics entering @xmath159 can be most easily illustrated in the large phase space volume limit , when @xmath160 we hence call @xmath159 a `` phase space density of emission points '' . this notion should not be taken too literally : the product of the volumes of three - dimensional spheres in position and momentum space is @xmath161 , and hence , @xmath159 is for large sources approximately a factor 10 larger than the particle number per unit phase space cell . also , for realistic source sizes , the value of @xmath158 deviates significantly from the approximation ( [ 4.15 ] ) , and a calculation of @xmath159 starting from ( [ 4.10b ] ) is preferable . one can ask whether in the large @xmath0 limit , the normalization @xmath47 becomes a function of @xmath159 only . to clarify this , we recall that a product @xmath162 with @xmath163 has a @xmath164-limit if and only if @xmath165 converges . for the normalization ( [ 4.11 ] ) , we find @xmath166 @xmath167 , i.e. , for fixed phase space density , @xmath168 there is no physical reason why @xmath47 should remain finite . it is not an observable . what matters in a quantitative study of multiparticle correlation effects is the phase space density of emission points and not the particle multiplicity , what matters are the weights @xmath130 and @xmath152 , and not the normalization @xmath47 . we finally estimate realistic values for the phase space density @xmath159 in heavy ion collisions . for a choice of model parameters @xmath169 fm , @xmath170 fm , @xmath171 mev say , we find @xmath172 . with multiplicities of like - sign pions in the hundreds , this leads to @xmath173 . the present static , spherically symmetric model is however unrealistic in so far that it does not take the strong longitudinal expansion into account which significantly increases the volume out of which particles are emitted . from these heuristic considerations , we expect realistic phase space densities to lie in the range @xmath174 depending on the precise value of @xmath159 in this range , the importance of multiparticle contributions to the one- and two - particle spectrum varies significantly . for @xmath175 , higher order contributions start dominating , while they account for @xmath176 10 % of the signal if @xmath177 . the two @xmath56-dependent terms @xmath71 and @xmath70 in ( [ 4.4 ] ) control the dependence of @xmath178 on the relative pair momentum @xmath73 and the average pair momentum @xmath72 . they determine the momentum dependence of all particle spectra . in figure [ fig2 ] , these factors are shown as a function of @xmath118 for a source with @xmath179 fm and @xmath180 fm . the first message of this plot is that even for very high multiparticle contributions ( e.g. @xmath181 ) , the momentum dependence of all building blocks @xmath7 can be calculated exactly . secondly , the factors @xmath70 and @xmath182 show the interesting property that irrespective of the value chosen for @xmath143 ( and hence for @xmath118 ) , they rapidly converge to an @xmath56-independent quantity . in contrast to the limiting case ( [ 4.5a ] ) , the @xmath56-dependence of @xmath70 is much weaker for realistic model parameters @xmath117 , @xmath118 , @xmath183 analogously , for realistic model parameters @xmath117 , @xmath118 , @xmath184 while for the limit ( [ 4.5b ] ) of the parameter space , a strong @xmath56-dependence remains . we have checked that the conclusions ( [ 4.18 ] ) , ( [ 4.19 ] ) hold for a large range of the model parameter space , including the part realistic for heavy ion collisions . going to smaller source sizes @xmath142 ( and hence to smaller values for @xmath117 ) , the factor @xmath185 is found to deviate significantly from unity . also , values for @xmath186 vary significantly . the rapid convergence to the limiting behaviour ( [ 4.18 ] ) , ( [ 4.19 ] ) however is observed for all values @xmath187 . for choices of the model parameters realistic for heavy ion collisions , the factors @xmath182 and @xmath70 are hence well approximated by an @xmath56-independent constant for sufficiently large @xmath56 . we can write the higher order pratt terms as [ 4.20 ] @xmath188 in subsection [ sec4d ] , we exploit this simple @xmath56-dependence of the terms @xmath71 , @xmath70 . for the discussion of the one - particle spectrum , we introduce the temperature @xmath189 via @xmath190 the model @xmath191 in ( [ 4.1 ] ) describes then a phase space distribution of emission points with boltzmann temperature @xmath189 . our aim is to determine how the slope and shape of the observed one - particle spectrum @xmath192 changes with the slope @xmath189 of this distribution @xmath191 and to what extent it is affected by multiparticle correlations . @xmath192 is a superposition of gaussians of different widths [ 4.22 ] @xmath193 where @xmath194 , and @xmath195 characterizes the slope of the @xmath56-th order contribution . according to ( [ 4.22 ] ) , the one - particle spectrum can not be characterized by a single slope parameter . for a qualitative understanding , we consider first the largest slope parameter @xmath196 and the smallest slope parameter @xmath197 . here , the superscripts _ pair _ and _ mult _ stand for ` pairwise ' and ` multiparticle ' correlations . if all multiparticle contributions vanish , then the momentum distribution of @xmath192 coincides with that of @xmath198 . the slope @xmath199 of @xmath198 is shown in figure [ fig3]a as a function of the model temperature @xmath189 for the pion mass @xmath200 mev and different wavepacket localizations @xmath86 . @xmath199 is always larger than the model temperature @xmath189 . for a spatial wavepacket width @xmath180 fm e.g. , the term @xmath201 takes a value of @xmath202 mev . even for small model temperatures @xmath189 as input , the quantum contribution @xmath201 accounts for a slope parameter @xmath199 comparable to the hagedorn temperature . for this apparently leading effect , the notion ` quantum temperature ' was coined @xcite . figure [ fig3]b shows that multiparticle contributions can change this picture qualitatively , if they are dominant . the slope parameter @xmath203 still depends significantly on the choice of the wavepacket localization . but for model temperatures in the range @xmath204 mev @xmath205 @xmath206 mev , there is always a value for @xmath86 , @xmath207 fm @xmath208 @xmath136 fm , such that the observed temperature @xmath203 coincides with the model temperature @xmath189 . for sufficiently large @xmath86 , the multiparticle effect can even overcompensate the broadening due to the quantum mechanical localization , @xmath209 . this illustrates that multiparticle symmetrization effects tend to populate the low momentum region of the one - particle spectrum , thereby increasing the slope of the spectrum . both effects , this narrowing and the broadening due to the quantum mechanical localization , are governed by the same scale @xmath86 and hence they cancel at least to some extent . the lowest order term @xmath210 with slope parameter @xmath199 contributes a fraction @xmath211 to the one - particle spectrum only . hence , the one - particle spectrum @xmath212 is not monoexponential , but can be be characterized by local slope parameters which specify the tangent onto @xmath212 at @xmath194 . going to very large values of @xmath213 , the local slope always coincides with that of @xmath199 , since the exponential of slowest decrease dominates in ( [ 4.22a ] ) . for sufficiently small @xmath213 ( below @xmath214 gev say ) , however , and for realistic phase space densities , neither the pair nor the multiparticle contributions can be neglected . the local slope lies between @xmath199 and @xmath203 . for a more quantitative statement , we have plotted in figure [ fig4 ] the one - particle spectrum for different phase space densities @xmath159 as a function of @xmath213 . increasing @xmath159 and hence the contribution of multiparticle correlations , the local slope of the spectrum becomes steeper . also , the superposition of gaussians of different width leads to a slight curvature of the spectrum . fitting the spectra in the range 0 - 1 gev naively with a monoexponential thermal distribution , a variation of @xmath159 between @xmath215 and @xmath216 results in a change of the slope parameter @xmath217 this has to be taken properly into account in a quantitative analysis of hadron spectra . we finally compare the spectrum ( [ 4.22a ] ) to that of a bose - einstein distribution as obtained from an arbitrary initial phase space distribution @xmath191 , keeping the bosons in a box till they have equilibrated , [ 4.24 ] @xmath218 for the spectrum @xmath212 in ( [ 4.22 ] ) being of bose - einstein form , the weights @xmath130 and the pratt terms @xmath7 would have to show a particular @xmath56-dependence , [ 4.25 ] @xmath219\ , . \label{4.25b } \end{aligned}\ ] ] in the present model , ( [ 4.25a ] ) is satisfied by setting @xmath220 @xmath221 , see ( [ 4.12b ] ) . however , the @xmath56-dependence of the terms @xmath222 is in general weaker than what is required to match ( [ 4.25b ] ) . for the zajc model , the @xmath56-dependence of @xmath7 is compatible with the bose - einstein distribution in the infinite volume limit . however , the pratt terms @xmath6 differ significantly from a power law which indicates a deviation from ( [ 4.25a ] ) . the reason for these differences is , that our model calculations of @xmath212 take a particular distribution @xmath191 as initial condition and include bose - einstein symmetrization , but they do not contain an equilibration mechanism : the particles are emitted and propagate freely . also , the @xmath0-particle state @xmath17 is not an equilibrium state . hence , the spectrum @xmath212 depends in contrast to ( [ 4.24 ] ) on @xmath191 . in general , it is not a bose - einstein distribution . in this section , we first discuss the normalization of the two - particle correlator @xmath223 and how it is calculated in the so - called pair approximation . then we turn to the study of multiparticle effects . there has been some debate recently about the appropriate normalization of the two - particle correlation function used in the hbt analysis of multiparticle production . for a compilation of the different normalizations used , and their problems , see @xcite . in the present work , we use the normalization @xcite @xmath224 where @xmath225 is the total pion cross section . this originates from normalizing both the single- and the double - differential cross sections to unity . setting these cross sections proportional to the one- and two - particle spectra @xmath226 , @xmath227 , the two - particle correlator reads @xmath228 the simplest way to make further progress is to assume that the approximation ( [ 4.20 ] ) for @xmath7 is valid for all @xmath229 . then , the two - particle spectrum can be written as @xmath230 and the correlator reads @xmath231 the normalization @xmath232 obtained in this approximation remains unchanged if the full @xmath56-dependence of the pratt terms @xmath7 is included , though the momentum dependence of ( [ 4.30 ] ) is then much more involved . the structure of the normalization @xmath232 is important for what follows : the term @xmath233 contains exactly @xmath234 terms , while @xmath47 contains @xmath1 terms . by integrating @xmath235 over @xmath236 and using the power law behaviour @xmath153 , we find [ 4.31 ] @xmath237 the normalization of the correlator is smaller than unity , the offset depends on the phase space volume occupied by the source . we shortly comment on the importance of this result : the two - particle correlator can be viewed as the factor relating two - particle differential cross sections @xmath238 of the real world ( where bose - einstein symmetrization exists ) to an idealized world in which bose - einstein correlations are absent , @xmath239 if @xmath240 everywhere , this implies that bose - einstein symmetrization effects increase the total cross section . for heavy ion collisions , there is no direct test of whether such an enhancement exists . in @xmath241-collisions we know that bose - einstein correlations do not affect total cross sections appreciably , since e.g. perturbative qcd predicts the production characteristics of @xmath242 to a per mille level without invoking them . in our calculation , we start from an @xmath0-particle state and we require that after final state bose - einstein symmetrization , @xmath0 particles are detected . hence , we explicitly assume that final state bose - einstein correlations do not affect the total cross sections . according to the above calculation , this automatically implies an offset of the normalization of @xmath223 below unity . the offset @xmath158 in ( [ 4.31b ] ) measures the system size from which particles are emitted . this is intrinsically consistent : for smaller system sizes ( @xmath158 large ) , the correlator shows an enhancement in a broader @xmath73 momentum region , and the normalization @xmath243 is smaller . this ensures that @xmath244 . we finally note that irrespective of the offset @xmath245 of the normalization , the correlator in ( [ 4.30 ] ) changes by a factor @xmath214 between the limits @xmath246 and @xmath247 . this is a consequence of the approximation ( [ 4.20 ] ) which we have justified for the present model in subsection [ sec4b ] . in general , the @xmath72- and @xmath73- dependence of the pratt terms @xmath7 leads to a more complicated dependence of the correlator @xmath223 according to ( [ 4.13 ] ) . depending on the model , this may affect the intercept of @xmath223 at @xmath246 . indeed , a decrease of the intercept with increasing event multiplicities was reported in recent model studies @xcite . in contrast to the present study , these models however do not work with fixed event multiplicities and show a significantly different physics , including pion lasing effects @xcite . we only conclude from the present study that multiparticle symmetrization effect do not lead automatically to a strong decrease of the intercept parameter . we now explain why the offset @xmath243 of the normalization of @xmath223 is not obtained in conventional calculations where multiparticle symmetrization effects are neglected . for the wavepackets introduced in section [ sec3 ] , the wavepacket overlap functions @xmath38 in ( [ 3.4 ] ) shows a gaussian decrease with the phase space distance @xmath248 . in general , the @xmath249 is an uncontrolled approximation to this gaussian behaviour . it becomes exact in certain limiting cases @xmath250 in the pair approximation , all higher order pratt terms vanish @xmath251 and one regains the results of the conventional calculations where multiparticle effects are neglected : the two - particle correlator is normalized to unity , @xmath252 and ( [ 2.5ext ] ) does not change the phase space distribution since @xmath253 . the term @xmath254 in ( [ 4.13 ] ) then insures that @xmath255 . according to ( [ 4.32 ] ) , this contradicts the statement @xmath256 . the origin of this difficulty can be traced back to the integral @xmath257 which should vanish according to ( [ 4.35 ] ) . the pair approximation does not treat the integral @xmath258 and the integrand of ( [ 4.37 ] ) on an equal footing . it sets @xmath259 but uses @xmath260 for the calculation of the bose - einstein enhancement . it is this inconsistency which leads to a correlator @xmath240 everywhere . once the momentum - dependent higher order pratt terms @xmath7 and their weights @xmath130 , @xmath152 are known , the determination of the two - particle correlator ( [ 4.28 ] ) is a matter of straightforward numerical calculation . from @xmath223 , the hbt radius parameters are obtained by fitting the gaussian ansatz @xmath261\ , . \label{4.38}\ ] ] we restrict our discussion to this one - dimensional parametrization of @xmath223 since our model ( [ 4.1 ] ) is spherically symmetric . the general , analytical form of the two - particle correlator ( [ 4.28 ] ) is obtained from ( [ 4.7 ] ) and ( [ 4.13 ] ) and it is quite involved . for a transparent discussion , we hence turn first to limiting cases . in the pair approximation , when all multiparticle effects are neglected , we obtain from ( [ 4.2b ] ) the well - known result @xcite @xmath262 the other , equally unrealistic limiting case is that multiparticle contributions dominate completely . the resulting two - particle correlator is given in ( [ 4.30 ] ) and the corresponding hbt - radius parameter reads @xmath263 the difference between these two expressions is significant . for a discussion of the main qualitative and quantitative effects , we now focus on the parameter region @xmath264 relevant for relativistic heavy ion collisions . in this regime , the @xmath265-dependent parts of ( [ 4.39 ] ) and ( [ 4.40 ] ) can be neglected , and according to the study of section [ sec4b ] , @xmath266 . we then find the simple relation @xmath267 this is the factor @xmath214 , obtained first by zajc @xcite in his comparison of the first and second order pratt terms , see ( [ 2.16 ] ) . zajc has given an estimate that in his model , multiparticle effects change the hbt radius parameter by as much as a factor 0.67 for a particle density of @xmath136 per unit phase space cell @xcite . this would affect estimates of the source volume or the energy density by a factor 3 . in zajc s model , however , higher order multiparticle effects ( @xmath268 ) change the @xmath73-dependence of @xmath2 much more dramatically than in the present study , where @xmath70 has a very weak @xmath56-dependence , negligible for @xmath264 . in our model , the correlator is a weighted superposition of gaussians in @xmath73 whose widths differ by as much as a factor @xmath269 . in figure [ fig5 ] , we have plotted the resulting two - particle correlator for relatively low phase space densities where the lowest order pratt term @xmath210 is still the leading contribution . the result of the pair approximation describes the main behaviour , but deviations due to multiparticle contributions affect the hbt - radius parameter on a 10 % -level for moderate phase space densities . this 10% effect translates into underestimating the volume by 30 % and overestimating the corresponding energy density by a similar amount . in models which show a stronger @xmath56-dependence of the momentum - dependent part of @xmath7 , the effect may be significantly stronger . in general , the degree to which multiparticle effects modify hbt radius parameters depends significantly on the phase space density of emission points . in this section , we shortly discuss how the calculation of multiparticle correlation effects can be extended to more realistic source distributions where the analytical techniques used in section [ sec4 ] are not applicable . pratt has shown already how to calculate the terms @xmath6 and @xmath178 for _ continuous _ source functions @xcite : up to order @xmath270 , straightforward monte carlo methods can be used , and an improved metropolis method allows to push numerical calculations up to @xmath271 . this seems to be sufficient for all practical purposes , since according to our model study , the weights @xmath130 , @xmath152 for @xmath272 are negligible for realistic phase space densities @xmath159 . here , we propose yet a different technique which is applicable to events characterized by a set of @xmath0 _ discrete _ phase space points @xmath9 . we interprete these points as the centers of gaussian single - particle wavepackets , i.e. , @xmath273 an arbitrary distribution of @xmath0 discrete phase space points is the most general ` source model ' in the present framework . we first discuss algorithms which allow for the calculation of the two - particle correlator from a set @xmath85 . then we comment on applications of these algorithms to event generators . we start with the simplest case , the pair approximation , which reduces the sum over @xmath1 terms in the two - particle spectrum to a sum over all @xmath274 particle pairs @xmath275 . each pair is weighted with the pair probability @xmath276 , calculated from the corresponding two - particle symmetrized state , @xcite @xmath277 to compare this expression to the zp - algorithm , we introduce the quantities [ 5.3 ] @xmath278 here , @xmath279 is the first order pratt term for the discrete phase space distribution @xmath85 and @xmath280 is a finite multiplicity correction , correcting for the double counting of identical pairs @xmath281 in @xmath282 in the large multiplicity limit , when @xmath280 can be neglected , this expression is equivalent to the pair approximation of the zp - algorithm , see e.g. ( [ 4.13a ] ) . the one - particle spectrum @xmath283 and the 2-particle correlation obtained in this way have been discussed already extensively in @xcite . they can be rewritten in terms of the single - particle probabilities @xmath284 of ( [ 3.3b ] ) , [ 5.5 ] @xmath285 the number of numerical operations needed to calculate this correlator grows linearly with the multiplicity @xmath0 rather than quadratic as in ( [ 5.2 ] ) and this makes it particularly suitable for a numerical algorithm . at given observed momenta @xmath72 , @xmath73 , one calculates for each event the one particle probabilities @xmath286 according to ( [ 3.3b ] ) and performes the sums in ( [ 5.5 ] ) . the @xmath286 are continuous in the measured momentum , and hence the obtained correlator is a continuous function of @xmath72 and @xmath73 , i.e. , no binning is required . at high multiplicities , when the factors @xmath280 are negligible , the correlator ( [ 5.5 ] ) can be rewritten as a fourier transform over an emission function @xmath287 , thereby regaining the well - known starting point of most model studies , ( see refs . @xcite for further details ) [ 5.6 ] @xmath288 the above discussion shows , that in the pair approximation , a numerical algorithm for the calculation of one- and two - particle spectra can be based on a discrete version of the first order pratt terms @xmath210 . here , we propose to extend this approach to take multiparticle correlations into account by calculating [ 5.7 ] @xmath289 the sums in these expressions run over all sets of @xmath56 out of @xmath0 integers and over all @xmath290 permutations of each set . this implies that for event multiplicities in the hundreds , only terms up to order @xmath291 can be calculated in a reasonable amount of cpu time . a tentative strategy for calculating multiparticle correlation effects on the basis of ( [ 5.7 ] ) then proceeds as follows : 1 . fit the calculated terms @xmath6 to a power law @xmath153 and use the parameter @xmath158 thus determined for a calculation of the weights @xmath130 , @xmath152 in ( [ 4.7 ] ) and ( [ 4.13a ] ) . fit gaussians in @xmath72 and @xmath73 to the momentum - dependent pratt terms @xmath178 . this allows to extract the terms @xmath71 and @xmath70 which govern the momentum dependence of multiparticle effects . calculate the one- and two - particle spectra by including up to the numerically determined order , @xmath292 say , all contributions exactly , approximating the momentum dependence of higher order terms by setting @xmath293 , @xmath294 for @xmath295 . this scheme draws on the experience gained from the study of the gaussian toy model in section [ sec4 ] . we expect that it allows to estimate multiparticle contributions for model distributions @xmath85 . many event generators for the simulation of heavy ion collisions have been developed in recent years @xcite , and irrespective of the large variety of physical inputs present in their codes , the typical output of their event simulation contains for each event a set @xmath8 of @xmath0 phase space points @xmath9 which one associates to the final state particles produced . however , a choice of interpretation is involved in comparing an event generator output @xmath18 to experimental data . usually , the measured one - particle spectra are compared directly to the binned momenta @xmath39 , i.e. one implicitly interprets the simulated phase space points @xmath296 as defining momentum eigenstates , which a fortiori carry no space - time information . for the calculation of two - particle correlations , this interpretation is not suitable , it leads to a sharp @xmath297-like correlator @xcite . also , a classical interpretation of @xmath296 which takes both @xmath40 and @xmath39 as `` sharp '' information , is problematic : ignoring the correct quantum mechanical localization leads to quantitatively and qualitatively unreliable results @xcite . on the other hand , including a quantum mechanical localization width @xmath86 , one changes both the one- and two - particle spectrum by a prescription which has no dynamical foundation . the origin of these difficulties in finding a consistent interpretation of the event generator output @xmath18 is well - known , see e.g. @xcite : quantum mechanical processes require a description in terms of amplitudes , while event simulations are formulated in a probabilistic setting . bose - einstein correlations occur by symmetrizing the production amplitudes of identical particles and are hence not encoded for in event generators . the mere fact that the simulated output is a discrete phase space distribution of emission points without identical multiparticle correlations ( rather than a set of detected momenta which includes all multiparticle correlations and hence all space - time information available from the measurement ) , is the very consequence of using a probabilistic language . in practice , this forces one to take recourse to an algorithm which associates _ a posteriori _ bose - einstein weights with the simulated phase space distribution @xmath8 , rather than obtaining these correlations from the dynamical propagation of properly ( anti)-symmetrized n - particle states . this a posteriori modification of an incomplete quantum - dynamical evolution creates the interpretational problems of the output @xmath18 . phenomenologically motivated numerical simulations of heavy ion collisions have remarkable success despite these fundamental problems . it is of some interest to confront them with experimental two - particle correlations , since this provides a test of their spatio - temporal ( rather than only their momentum - dependent ) properties . we expect that the algorithms discussed in section [ sec5a ] are useful in such comparisons . the zajc - pratt algorithm provides a simple technique for the calculation of multiparticle symmetrization effects in one- and two - particle spectra . based on this algorithm , we have studied to what extent multiparticle correlations steepen the slope of the one - particle and broaden the width of the two - particle spectrum . the scale of both effects depends sensitively on the particle phase space density @xmath159 in the emission region . multiparticle correlations contribute a fraction @xmath298 to the one - particle spectrum and a fraction @xmath299 to the two - particle spectrum . also , the @xmath56-dependence of higher order pratt terms , i.e. , the extent to which the momentum dependence of multiparticle contributions @xmath7 changes with increasing order @xmath56 , plays an important role . in a gaussian source model with instantaneous emission and moderate particle phase space densities , the slope parameter @xmath300 of the one - particle spectrum changes by up to 20 mev , and the change in the hbt radius parameters is of the order of 10 percent , if multiparticle effects are neglected . for the reconstruction program advocated in @xcite , our results indicate that neglecting multiparticle effects , one underestimates the source temperature significantly and one overestimates the energy density by up to 30 % . here , the caveat is however that our calculation does not include multiparticle final state coulomb interactions . bose - einstein and coulomb effects typically arise on the same scale and compensate each other at least to some extent . this may significantly reduce the effect of multiparticle correlations in the measured @xmath301 and @xmath302 spectra on which most of the phenomenological analysis is based currently . it will hence be very interesting to compare the slopes of the one - particle spectra of charged and neutral pions @xcite . on the basis of the above heuristic ideas , one may expect the slope of the @xmath303-spectrum to be somewhat steeper , since multiparticle coulomb interactions can not compensate for the multiparticle bose - einstein symmetrization effects . our analysis of multiparticle effects has also shed new light on a recent discussion about the normalization of the two - particle correlator . defining the correlator via ( [ 4.26 ] ) , it is normalized for large relative momenta @xmath73 to @xmath243 , and not to unity . the importance of this result was already discussed in section [ sec4d1 ] : the normalization ( [ 4.31b ] ) corrects the result of the commonly used pair approximation and leads to bose - einstein unit weights which conserve event probabilities . we close by embedding the present work into a wider perspective : many aspects of the dynamical evolution of relativistic heavy ion collisions are mesoscopic , and this makes it very difficult to decide whether certain physical observables have a known conventional interpretation or are indicative of the new in medium properties which the current experimental programs at cern and rhic aim for . the physics of final state hadrons however , though being mesoscopic and hence difficult to treat , is known in principle . it allows hence for a detailed quantitative description on the basis of present day knowledge . in a second step , such a description will pave the way for quantitative statements about the geometry and dynamics of the final stage of the collision process , thereby setting up an experimentally motivated starting point for a dynamical back extrapolation @xcite . we hope that the present work , by quantifying one mesoscopic multiparticle effect , will prove useful in refining this reconstruction program and in separating known physics from the in medium properties we are looking for . the author thanks miklos gyulassy and bill zajc for helpful discussions and the hospitality extended to him at columbia university . they and their groups ( and especially yang pang and stephen vance ) provided a stimulating atmosphere throughout this work . also , it is a pleasure to acknowledge conversations in the `` regensburg cappuccino group '' with rainer fries and martin maul , concerning the determinants of band matrices , and with claus slotta , concerning the physics underlying eq . ( [ 2.16 ] ) . discussions with klaus kinder - geiger and tjorbjrn sjstrand have sharpened the ideas presented in section [ sec5 ] . last but not least , the author thanks ulrich heinz for a critical reading of the manuscript and a discussion focussing on eq . ( [ 4.24 ] ) . this work was supported by a dfg habilitandenstipendium and by the us department of energy under contract no . de - fg-02 - 86er40281 . so far , there is no self - contained derivation of the zajc - pratt algorithm in the literature . for the convenience of the reader , we give here the main arguments . _ notation : _ we denote the term @xmath304 by an arrow from @xmath305 to @xmath306 . @xmath56 of these terms form a closed @xmath56-cycle @xmath307 . the only other structure appearing in all spectra are open @xmath56-cycles @xmath308 , see figure [ fig6 ] . here , @xmath309 is represented by a cross from which the arrow starts , @xmath310 by a cross at which the arrow ends . the pratt terms @xmath6 , @xmath7 are obtained by averaging the emission points in figure [ fig6 ] over a model - distribution @xmath311 . dots without indices attached indicate that this average was carried out , see figure [ fig7 ] . _ normalization : _ the normalization @xmath312 sums the products @xmath313 over all partitions @xmath52 . each partition of @xmath52 has @xmath314 possible realizations . these have to be multiplied by the @xmath315 different ways to combine each set of @xmath44 points to a closed cycle . this leads to the prefactor @xmath316 in ( [ 2.7 ] ) . there exists a simple diagrammatic prescription to construct from the terms of @xmath317 the terms of @xmath318 . one adds to each diagram of @xmath317 an @xmath319th dot by * 1 . ) placing the dot once disconnected from all other cycles . changing closed @xmath56-cycles @xmath6 into @xmath56 closed @xmath320-cycles @xmath321 . this prescription can be checked e.g. in figure [ fig7 ] . it automatically insures , that @xmath317 is represented by @xmath1 diagrams , the maximal value for @xmath317 is @xmath1 . _ one - particle spectrum _ : each term of @xmath322 has the structure @xmath323 : it contains exactly one open @xmath56-cycle @xmath324 , the @xmath325 other dots being contained in closed cycles . to construct from it the terms of @xmath326 , one has to use the replacement @xmath327 diagrammatically , this is realized by employing the two diagrammatic rules given above and supplementing them with * 3 . ) change open @xmath56-cycles @xmath7 into @xmath56 open @xmath320-cycles @xmath328 . the prescription can be checked e.g. in figure [ fig8 ] . using the recursion relation ( [ b1 ] ) , one can prove by complete induction that the averaged one - particle spectrum @xmath76 takes the form ( [ 2.8 ] ) . _ two - particle spectrum : _ in @xmath329 , each term contains two open cycles of lengths @xmath84 and @xmath109 , the @xmath330 remaining dots being contained in closed cycles . on the endpoints of the two open cycles @xmath331 and @xmath332 , the momenta @xmath333 and @xmath236 can be attached in different combinations : for @xmath334 , there are two possiblities , for @xmath335 , there are four . this combinatorics is properly encoded in the definition of the product @xmath336 @xmath337 @xmath338 , see eq . ( [ 4.13b ] ) . an @xmath339 recursion relation for the two - particle spectrum is obtained by using the rule to change open @xmath56-cycles into @xmath56 open @xmath320-cycles , and properly treating the combinatorics for @xmath334 , @xmath340 the expression ( [ 2.9 ] ) for the two - particle spectrum then follows from ( [ b2 ] ) by complete induction . here , we give details of the calculation of higher order pratt terms @xmath6 and @xmath7 for the gaussian source model ( [ 4.1 ] ) in section [ sec4 ] . due to the structure of the functions @xmath38 , it is advantageous to change to relative and average momentum variables . for an average over @xmath56 phase space points , we introduce the integration variables for the gaussian emission probability ( [ 4.1 ] ) , the probability @xmath343 of having two particles with wavepackets centered around phase space points @xmath344 , @xmath345 , is given by the product @xmath346 @xmath347 and factorizes into a probability distribution for the relative and average particle pair coordinates , we then define the higher order pratt terms as averages over pair distribution probabilities @xmath349 this is a gaussian integral which can be calculated analytically . its exponent is diagonal in all integration variables @xmath350 and in @xmath351 , and doing the corresponding integrals leads to 9 u. heinz in proc . of the xxvth int . workshop on gross properties of nuclei and nuclear exitations , hirschegg , austria , jan 13 - 18,1997 , nucl - th/9701054 ; u.a . wiedemann in proc . of the xxxiind rencontres de moriond , les arc , france , march 22 - 29 , 1997 , nucl - th/9705042 ; u.a . wiedemann , talk at the 13th int . conf . on ultra - relativistic nucleus - nucleus collisions ( quark matter97 ) , tsukuba , japan , dec 1 - 5 1997 , nucl - th/9801017 , to be published in nucl.phys . * a*. m. gyulassy , s.k . kauffmann , and l.w . wilson , phys . c*20 * , 2267 ( 1979 ) . zajc , phys . rev . d*35 * , 3396 ( 1987 ) ; w.a . zajc et al . , phys . c*29 * , 2173 ( 1984 ) . s. pratt , phys . b*301 * , 159 ( 1993 ) ; phys . c*50 * , 469 ( 1994 ) ; s. pratt and v. zelevinsky , phys . lett . * 72 * , 816 ( 1994 ) . w.n . zhang et al . , phys . c*47 * , 795 ( 1993 ) . a. bialas and a. krzywicki , phys b354 * , 134 ( 1995 ) . n. amelin and r. lednicky , heavy ion physics * 4 * , 241 ( 1996 ) . k. fialkowski and r. wit , hep - ph/9703227 and hep - ph/9709205 , acta phys . polon . * b28 * , 2039 ( 1997 ) . j. wosiek , phys . lett . * b*399 , 130 ( 1997 ) . t. csrg and j. zimnyi , phys . lett . * 80 * , 916 ( 1998 ) ; j. zimnyi and t. csrg , hep - ph/9705432 . u.a . wiedemann , p. foka , h. kalechofsky , m. martin , c. slotta and q.h . zhang , phys . rev . c*56 * , r614 ( 1997 ) . m. hillery , r.f . oconell , m. scully and e.p . wigner , phys . rep . * 106 * , 121 ( 1984 ) . s. padula , m. gyulassy and s. gavin , nucl . * b329 * , 357 ( 1990 ) . h. merlitz and d. pelte , z. phys . * a357 * , 175 ( 1997 ) . q.h . zhang , u.a . wiedemann , c. slotta and u. heinz , phys . lett . * b*407 , 33 ( 1997 ) . h. sorge , h. stcker and w. greiner , ann . phys . * 192 * , 266 ( 1989 ) . k. werner , phys . rep . * 232 * , 87 ( 1993 ) . y. pang , t.j . schlagel and s.h . kahana , phys . 68 * , 2743 ( 1992 ) ; t.j . schlagel , y. pang and s.h . kahana , phys . lett . * 69 * , 3290 ( 1992 ) . k. geiger , comp . phys . com . * 104 * , 70 ( 1997 ) . x.n . wang and m. gyulassy , phys . rev . * d44 * , 3501 ( 1991 ) ; comp . phys . com . * 83 * , 307 ( 1994 ) . l. lnnblad and t. sjstrand , phys . lett . * b351 * , 293 ( 1995 ) ; l. lnnblad and t. sjstrand , eur . j. c2 , 165 ( 1998 ) . j. aichelin , nucl . * a617 * , 510 ( 1997 ) . m. martin , h. kalechofsky , p. foka and u.a . wiedemann , nucl - th/9612023 , eur . j. c in press . d. mikowiec and s. voloshin , nucl - ex/9704006 . i.s . gradshteyn and i.m . ryzhik , _ table of integrals , series and products _ ( academic press , ney york , 1980 ) . s. wolfram , the mathematica book , 3rd edition , 1996 , cambridge university press . t. peitzmann for the wa98 collaboration , talk at the 13th int . conf . on ultra - relativistic nucleus - nucleus collisions ( quark matter97 ) , tsukuba , japan , dec 1 - 5 1997 , to be published in nucl.phys . * a*. the results presented in section iv are not the results for the factorizable _ one - particle _ distribution ( 4.1 ) , in contrast to our statement . all results of section iv were calculated from the probability distribution ( b3b ) for relative and average _ pair _ coordinates according to ( b4 ) . the use of this pair distribution ( b3b ) is equivalent to the one - particle distribution ( 4.1 ) only for the lowest order pratt terms ( 4.2 ) on which the conventional two - particle symmetrized calculation of the hanbury - brown / twiss effect is based . for the higher order pratt terms , however , our expressions ( 4.3 ) , ( 4.4 ) differ from the results for the model ( 4.1 ) , given in ref . [ 1 ] . especially , all observables of model ( 4.1 ) can be shown [ 2 ] to depend only on the two effective source parameters @xmath359 and @xmath360 . the pratt terms ( 4.3 ) and ( 4.4 ) on which our discussion is based , however , depend on all three parameters @xmath142 , @xmath143 and @xmath86 . this shows that the calculation of higher order pratt terms via pair distributions ( b4 ) destroys the one - particle factorizability . we withdraw the statement that the pratt terms ( 4.3 ) , ( 4.4 ) are solutions of the model distribution ( 4.1 ) . they are solutions of ( b4 ) . [ 2 ] u. heinz , hep - ph/9806512 , conference proceedings , cris98 ( catania , june 8 - 12 , 1998 ) , `` cris98 : measuring the size of things in the universe : hbt interferometry and heavy ion physics '' , ( s. costa et al . , eds . ) , world scientific , singapore , 1998 .

we extend the standard model by adding two gauge - singlet-symmetric scalar fields that interact with visible matter only through the higgs particle . one is a stable dark matter wimp , and the other one undergoes a spontaneous breaking of the symmetry that opens new channels for the dark matter annihilation , hence lowering the mass of the wimp . we study the effects of the observed dark matter relic abundance on the annihilation cross section and find that in most regions of the parameters space , light dark matter is viable . we also compare the elastic scattering cross - section of our dark matter candidate off nucleus with existing ( cdmsii and xenon100 ) and projected ( supercdms and xenon1 t ) experimental exclusion bounds . we find that most of the allowed mass range for light dark matter will be probed by the projected sensitivity of xenon1 t experiment .

1101.0365

cosmology tells us that about 25% of the total mass density in the universe is dark matter that can not be accounted for by conventional baryons @xcite . alongside observation , intense theoretical efforts are made in order to elucidate the nature and properties of this unknown form of matter . in this context , electrically neutral and colorless weakly interacting massive particles ( wimps ) form an attractive scenario . their broad properties are : masses in the range of one to a few hundred gev , coupling constants in the milli - weak scale and lifetimes longer than the age of the universe . recent data from the direct - detection experiments dama / libra @xcite and cogent cogent , and the recent analysis of the data from the fermi gamma ray space telescope @xcite , if interpreted as signal for dark matter , require light wimps in the range of 5 to 10 gev @xcite . also , galactic substructure requires still lighter dark matter masses @xcite . in this regard , it is useful to note in passing that the xenon100 collaboration has provided serious constraints on the region of interest to dama / libra and cogent @xcite , assuming a constant extrapolation of the liquid xenon scintillation response for nuclear recoils below 5 kev , a claim disputed in @xcite . most recently , the cdms collaboration has released the analysis of their low - energy threshold data @xcite which seems to exclude the parameter space for dark matter interpretation of dama / libra and cogent results , assuming a standard halo dark matter model with an escape velocity @xmath1 km / s and neglecting the effect of ion channeling @xcite . however , with a highly anisotropic velocity distribution , it may be possible to reconcile the cogent and dama / libra results with the current exclusion limits from cdms and xenon @xcite ( see also comments on p.6 in @xcite about the possibility of shifting the exclusion contour in @xcite above the cogent signal region ) . in addition , cresst , another direct detection experiment at gran sasso , which uses cawo@xmath2 as target material , reported in talks at the idm 2010 and wonder 2010 workshops an excess of events in their oxygen band instead of tungsten band . if this signal is not due to neutron background a possible interpretation could be the elastic scattering of a light wimp depositing a detectable recoil energy on the the lightest nuclei ( oxygen ) in the detector @xcite . while this result has to await confirmation from the cresst collaboration , it is clear that it is important as well as interesting to study light dark matter . the most popular candidate for dark matter is the neutralino , a neutral @xmath3-odd supersymmetric particle . indeed , they are only produced or destroyed in pairs , thus rendering the lightest susy particle ( lsp ) stable @xcite . in the minimal version of the supersymmetric extension of the standard model , the neutralino is a linear combination of the fermionic partners of the neutral electroweak gauge bosons ( gauginos ) and the neutral higgs bosons ( higgsinos ) . they can annihilate through a t - channel sfermion exchange into standard model fermions , or via a t - channel chargino - mediated process into @xmath4 , or through an s - channel pseudoscalar higgs exchange into fermion pairs and they can undergo elastic scattering with nuclei mainly through scalar higgs exchange @xcite . if these neutralinos were light , they would then be overproduced in the early universe and , in the minimal model , would not have an elastic - scattering cross section large enough to account for the dama / libra and cogent results due to constraints from other experiments such as the lep , tevatron , and rare decays @xcite . therefore , with no clear clue yet as to what the internal structure of these wimps is , if any , a pedestrian approach can be attractive . in this logic , the simplest of models is to extend the standard model by adding a real scalar field , the dark matter , a standard - model gauge singlet that interacts with visible particles via the higgs field only . to ensure stability , it is endowed with a discrete @xmath0 symmetry that does not spontaneously break . such a model can be seen as a low - energy remnant of some higher - energy physics waiting to be understood . in this cosmological setting , such an extension has first been proposed in @xcite and further studied in @xcite where the unbroken @xmath0 symmetry is extended to a global u(1 ) symmetry . a more extensive exploration of the model and its implications was done in burgess - pospelov - veldhuis , specific implications on higgs detection and lhc physics discussed in @xcite and one - loop vacuum stability looked into and perturbativity bounds obtained in @xcite . the work of @xcite considers also this minimal extension and uses constraints from the experiments xenon10 @xcite and cdmsii @xcite to exclude dark matter masses smaller than 50 , 70 and 75gev for higgs masses equal to 120 , 200 and 350 gev respectively . in order to allow for light dark matter , it is therefore necessary to go beyond the minimal one - real - scalar extension of the standard model . the natural next step is to add another real scalar field , endowed with a @xmath0 symmetry too , but one which is spontaneously broken so that new channels for dark matter annihilation are opened , increasing this way the annihilation cross - section , hence allowing smaller masses . this auxiliary field must also be a standard model gauge singlet . after this brief introductory motivation , we present the model in the next section . we perform the spontaneous breaking of the electroweak and the additional @xmath0 symmetries in the usual way . we clarify the physical modes as well as the physical parameters . there is mixing between the physical new scalar field and the higgs , and this is one of the quantities parametrizing the subsequent physics . in section three , we impose the constraint from the known dark matter relic density on the dark - matter annihilation cross section and study its effects . of course , as we will see , the parameter space is quite large , and so , it is not realistic to hope to cover all of it in one single work of acceptable size . representative values have to be selected and the behavior of the model as well as its capabilities are described . our main focus in this study is the mass range 0.1gev 100gev and we find that the model is rich enough to bear dark matter in most of it , including the very light sector . in section four , we determine the total cross section @xmath5 for non relativistic elastic scattering of dark matter off a nucleon target and compare it to the current direct - detection experimental bounds and projected sensitivity . for this , we choose the results of cdmsii and xenon100 , and the projections of supercdms @xcite and xenon1 t @xcite . here too we can not cover all of the parameter space nor are we going to give a detailed account of the behavior of @xmath5 as a function of the dark matter mass , but general patterns are mentioned . the last section is devoted to some concluding remarks . note that as a rule , we have avoided in this first study narrowing the choice of parameters using particle phenomenology . of course , such phenomenological constraints have to be addressed ultimately and this is left to a forthcoming investigation , contenting ourselves in the present work with a limited set of remarks mentioned in this last section . finally , we have gathered in an appendix the partial results regarding the calculation of the dark matter annihilation cross section . we extend the standard model by adding two real , spinless and @xmath6-symmetric fields : the dark matter field @xmath7 for which the @xmath0 symmetry is unbroken and an auxiliary field @xmath8 for which it is spontaneously broken . both fields are standard model gauge singlets and hence can interact with ` visible ' particles only via the higgs doublet @xmath9 . this latter is taken in the unitary gauge such that @xmath10 , where @xmath11 is a real scalar . we assume all processes calculable in perturbation theory . the potential function that includes @xmath7 , @xmath11 and @xmath8 writes as follows:@xmath12where @xmath13 , @xmath14and @xmath15 and all the coupling constants are real positive numbers . in the standard model scenario , electroweak spontaneous symmetry breaking occurs for the higgs field , which then oscillates around the vacuum expectation value @xmath16 particle - data - group . the field @xmath8 will oscillate around the vacuum expectation value @xmath17 . both @xmath18 and @xmath19 are related to the parameters of the theory by the two relations:@xmath20it is assumed that the self - coupling constants are sufficiently larger than the mutual ones . writing @xmath21 and @xmath22 , the potential function becomes , up to an irrelevant zero - field energy:@xmath23where the mass - squared ( quadratic ) terms are gathered in @xmath24 , the cubic interactions in @xmath25 and the quartic ones in @xmath26 . the quadratic terms are given by:@xmath27where the mass - squared coefficients are related to the original parameters of the theory by the following relations:@xmath28replacing the vacuum expectation values @xmath18 and @xmath19 by their respective expressions ( [ vacua ] ) will not add clarity . in this field basis , the mass - squared matrix is not diagonal : there is mixing between the fields @xmath29 and @xmath30 . denoting the physical mass - squared field eigenmodes by @xmath31 and @xmath32 , we rewrite:@xmath33where the physical fields are related to the mixed ones by a @xmath34 rotation:@xmath35here @xmath36 is the mixing angle , related to the original mass - squared parameters by the relation:@xmath37and the physical masses in ( [ u_quad diagonal ] ) by the two relations:@xmath38 ; \notag \\ m_{1}^{2 } & = & \frac{1}{2}\left [ m_{h}^{2}+m_{1}^{2}-\varepsilon \left ( m_{h}^{2}-m_{1}^{2}\right ) \sqrt{\left ( m_{h}^{2}-m_{1}^{2}\right ) ^{2}+4m_{1h}^{4}}\right ] , \label{physical masses}\end{aligned}\]]where @xmath39 is the sign function . written now directly in terms of the physical fields , the cubic interaction terms are expressed as follows:@xmath40where the cubic physical coupling constants are related to the original parameters via the following relations:@xmath41 -\lambda _ { 1}v\sin ^{3}\theta ; \notag \\ \lambda _ { 2}^{\left ( 3\right ) } & = & \lambda _ { 1}v\cos ^{3}\theta -\frac{1}{2}\sin 2\theta \left [ \left ( 2\lambda _ { 1}-\eta _ { 1}\right ) v_{1}\cos \theta + \left ( 2\lambda _ { 1}-\lambda \right ) v\sin \theta \right ] + \lambda _ { 1}v_{1}\sin ^{3}\theta . \notag\end{aligned}\]]also , in terms of the physical fields , the quartic interactions are given by:@xmath42where the physical quartic coupling constants are written in terms of the original parameters of the theory as follows:@xmath43 \sin 2\theta ; \notag \\ \lambda _ { 2}^{\left ( 4\right ) } & = & \lambda _ { 1}\cos ^{2}2\theta -\frac{1}{4}\left ( 2\lambda _ { 1}-\eta _ { 1}-\lambda \right ) \sin ^{2}2\theta ; \notag \\ \lambda _ { 3}^{\left ( 4\right ) } & = & \frac{1}{2}\left [ \left ( \eta _ { 1}-3\lambda _ { 1}\right ) \cos ^{2}\theta -\left ( \lambda -3\lambda _ { 1}\right ) \sin ^{2}\theta \right ] \sin 2\theta . \label{physical quartic coeff}\end{aligned}\ ] ] finally , after spontaneous breaking of the electroweak and @xmath0 symmetries , the part of the standard model lagrangian that is relevant to dark matter annihilation writes , in terms of the physical fields @xmath31 and @xmath32 , as follows:@xmath44the quantities @xmath45 , @xmath46 and @xmath47 are the masses of the fermion @xmath48 , the @xmath49 and the @xmath50 gauge bosons respectively , and the above coupling constants are given by the following relations:@xmath51 the original theory ( [ u ] ) has nine parameters : three mass parameters @xmath52 , three self - coupling constants @xmath53 and three mutual coupling constants @xmath54 . perturbativity is assumed , hence all these original coupling constants are small . the dark - matter self - coupling constant @xmath55 does not enter in the calculations of the lowest - order processes of this work @xcite , so effectively , we are left with eight parameters . the spontaneous breaking of the electroweak and @xmath0 symmetries for the higgs and @xmath8 fields respectively introduces the two vacuum expectation values @xmath18 and @xmath19 given to lowest order in ( [ vacua ] ) . the value of @xmath18 is fixed experimentally to be @xmath56 and for the present work , we fix the value of @xmath19 at the order of the electroweak scale , say @xmath57 . hence we are left with six parameters . four of these are chosen to be the three physical masses @xmath58 ( dark matter ) , @xmath59 ( @xmath32 field ) and @xmath60 ( higgs ) , plus the mixing angle @xmath36 between @xmath32 and @xmath31 . we will fix the higgs mass to @xmath61 and give , in this section , the mixing angle @xmath36 the two values @xmath62 ( small ) and @xmath63 ( larger ) . the two last parameters we choose are the two physical mutual coupling constants @xmath64 ( dark matter higgs ) and @xmath65 ( dark matter @xmath32 particle ) , see ( u_quar ) . in the framework of the thermal dynamics of the universe within the standard cosmological model @xcite , the wimp relic density is related to its annihilation rate by the familiar relations:@xmath66the notation is as follows : the quantity @xmath67 is the hubble constant in units of 100km/(s@xmath68mpc ) , @xmath69 the planck mass , @xmath58 the dark matter mass , @xmath70 the ratio of the dark matter mass to the freeze - out temperature @xmath71 and @xmath72 the number of relativistic degrees of freedom with a mass less than @xmath71 . the quantity @xmath73 is the thermally averaged annihilation cross - section of a pair of two dark matter particles multiplied by their relative speed in the center - of - mass reference frame . solving ( [ relic - density - relation ] ) with the current value for the dark matter relic density @xmath74 @xcite gives : @xmath75for a range of dark matter masses between roughly @xmath76 to @xmath57 and @xmath77 between 19.2 and 21.6 , with about 0.4 thickness weinberg . the value in ( [ v12sigma_annihi ] ) for the dark matter annihilation cross - section translates into a relation between the parameters of a given theory entering the calculated expression of @xmath73 , hence imposing a constraint on these parameters which will limit the intervals of possible dark matter masses . this constraint can be exploited to examine aspects of the theory like perturbativity . for example , in our model , we can obtain via ( v12sigma_annihi ) the mutual coupling constant @xmath78 for given values of @xmath64 , study its behavior as a function of @xmath58 and tell which dark - matter mass regions are consistent with perturbativity . note that once the two mutual coupling constants @xmath64 and @xmath78 are perturbative , all the other physical coupling constants will be . in the study of this section , we choose the values @xmath79 ( very weak ) , @xmath80 ( weak ) and 1 ( large ) . we also let the two masses @xmath58 and @xmath59 stretch from 0.1gev to 120gev , occasionally @xmath58 to 200gev . finally , note that we do not incorporate the uncertainty in ( [ v12sigma_annihi ] ) when imposing the relic - density constraint , something that is sufficient in view of the descriptive nature of this work . the dark matter annihilation cross sections ( times the relative speed ) through all possible channels are given in the appendix . the quantity @xmath73 is the sum of all these contributions . imposing @xmath81 dictates the behavior of @xmath82 , which is displayed as a function of the dark matter mass @xmath58 . of course , as the parameters are numerous , the behavior is bound to be rich and diverse . we can not describe every bit of it . also , one has to note from the outset that for a given set of values for the parameters , the solution to the relic - density constraint is not unique : besides positive real solutions ( when they exist ) , we may find negative real or even complex solutions . it is beyond the scope of the present work to investigate the nature and behavior of all the solutions . we are only interested in finding the smallest positive real solution @xmath83 when it exists , looking at its behavior and finding out when it is small enough to be perturbative . let us describe briefly and only partly how the mutual @xmath7 @xmath32 coupling constant @xmath83 behaves as a function of the @xmath7 mass @xmath58 . we start by a small mixing angle , say @xmath84 , and a very weak mutual @xmath7 higgs coupling constant , say @xmath85 . let us also fix the @xmath32 mass first at the small value @xmath86 . the corresponding behavior of @xmath83 versus @xmath58 is shown in fig . [ eta014_theta-10_lambda04 - 001_m1 - 10 ] . the range of @xmath58 shown is from @xmath87 to @xmath88 , cut in four intervals to allow for ` local ' features to be displayed . we see that the relic - density constraint on @xmath7 annihilation has no positive real solution for @xmath89.@xmath90 , and so , with these very small masses , @xmath7 can not be a dark matter candidate . in other words , for @xmath86 , the particle @xmath7 can not annihilate into the lightest fermions only ; inclusion of the @xmath91-quark is necessary . note that right about @xmath92 , the @xmath91 threshold , the mutual coupling constant @xmath83 starts at about @xmath93 , a value , while perturbative , that is roughly eighty - two - fold larger than the mutual @xmath7 higgs coupling constant @xmath94 . then @xmath83 decreases , steeply first , more slowly as we cross the @xmath95 mass towards the @xmath96 mass . just before @xmath97 , the coupling @xmath83 hops onto another solution branch that is just emerging from negative territory , gets back to the first one at precisely @xmath97 as this latter carries now smaller values , and then jumps up again onto the second branch as the first crosses the @xmath58-axis down . it goes up this branch with a moderate slope until @xmath58 becomes equal to @xmath59 , a value at which the @xmath32 annihilation channel opens . right beyond @xmath59 , there is a sudden fall to a value @xmath98 that is about half the value of @xmath94 , and @xmath83 stays flat till @xmath99 where it starts increasing , sharply after 60gev . in the mass interval @xmath100 , there is a desert with no positive real solutions to the relic - density constraint , hence no viable dark matter candidate . beyond @xmath101 , the mutual coupling constant @xmath83 keeps increasing monotonously , with a small notch at the @xmath49 mass and a less noticeable one at the @xmath50 mass . note that for this value of @xmath102(@xmath76 ) , all values reached by @xmath83 in the mass range considered , however large or small with respect to @xmath94 , are perturbativily acceptable . increasing @xmath59 to moderate values does not change the above qualitative features . as an illustration , fig . [ eta014_theta-10_lambda04 - 001_m1 - 30 ] shows the behavior of @xmath83 as a function of @xmath58 for @xmath103 , keeping the mixing angle @xmath84 , still small , and the mutual @xmath7 higgs coupling constant @xmath85 , still very weak . the first thing to note is that not all values of @xmath83 are perturbative . indeed , @xmath83 does not start until @xmath104 , but with the very large value 89.8 . it decreases very sharply right after , to 2.04 at about 1.6gev . it continues to decrease with a pronouncedchange in the slope at the @xmath96 threshold . effects at the masses@xmath105 and @xmath59 similar to those of figure eta014_theta-10_lambda04 - 001_m1 - 10 do occur here too . there is a desert that lies in this case in the mass interval 66.5gev 76.5gev . at the upper bound , the coupling @xmath83 takes the value 2.15 and decreases very slowly till @xmath106 . right after this mass , it plunges down to catch up with a solution branch that is just emerging from negative values . this solution branch increases steadily with two small notches at the @xmath49 and @xmath50 masses . a similar global behavior occurs at other moderate @xmath59 masses , with varying local features . because of the very - small-@xmath58 deserts described and visible on fig . [ eta014_theta-10_lambda04 - 001_m1 - 10 ] , one may ask whether the model ever allows for very light dark matter . to look into this , we fix @xmath58 at small values and let @xmath59 vary . take first @xmath107 and see fig . [ eta014_theta-10_lambda04 - 001_m0 - 02 ] . the allowed @xmath7 annihilation channels are the very light fermions @xmath108 and @xmath109 , plus @xmath32 when @xmath110 . note that we still have @xmath111 and @xmath112 . qualitatively , we notice that in fact , there are no solutions for @xmath113 ( @xmath114 here ) , a mass at which @xmath82 takes the very small value @xmath115 . it goes up a solution branch and leaves it at @xmath116 to descend on a second branch that enters negative territory at @xmath117 , forcing @xmath118 to return onto the first branch . there is an accelerated increase till @xmath119 , a value at which @xmath120 . and then a desert , no positive real solutions , no viable dark matter . increasing @xmath58 until about @xmath121 does not change these overall features : some ` movement ' for very small values of @xmath59 and then an accelerated increase till reaching a desert with a lower bound that changes with @xmath58 . for example , the desert starts at @xmath122 for @xmath123 and @xmath124 for @xmath125 . note that in all these cases where @xmath126 , all values of @xmath83 are perturbative . therefore , the model can very well accommodate very light dark matter with a restricted range of @xmath32 masses . however , the situation changes after the inclusion of the @xmath95 annihilation channel . indeed , as fig . eta014_theta-10_lambda04 - 001_m0 - 14 - 15 shows , for @xmath127 , though the overall shape of the behavior of @xmath83 as a function of @xmath59 is qualitatively the same , the desert threshold is pushed significantly higher , to @xmath128 . but more significantly , @xmath83 starts to be larger than one already at @xmath129 , therefore loosing perturbativity . for @xmath130 , the desert is effectively erased as we have a sudden jump to highly non - perturbative values of @xmath83 right after @xmath131 . such a behavior stays with larger values of @xmath58 . but for @xmath132 ( case @xmath133 ) , the values of @xmath83 are smaller than one and physical use of the model is possible if needed . what are the effects of the relic - density constraint when we vary the parameter @xmath94 ? let us keep the higgs @xmath32 mixing angle small ( @xmath84 ) and increase @xmath94 , first to @xmath80 and later to 1 . for @xmath134 , figure eta014_theta-10_lambda04 - 02_m1 - 20 shows the behavior of @xmath83 as a function of the dark matter mass @xmath58 when @xmath135 . we see that @xmath83 starts at @xmath136 with a value of about @xmath137 . it decreases with a sharp change of slope at the @xmath96 threshold , then makes a sudden dive at about 5 gev , a change of branch at @xmath97 down till about @xmath138 where it jumps up back onto the previous branch just before going to cross into negative territory . it drops sharply at @xmath139 and then increases slowly until @xmath140 . beyond , there is nothing , a desert . this is of course different from the situation of very small @xmath94 like in fig . [ eta014_theta-10_lambda04 - 001_m1 - 10 ] and fig . eta014_theta-10_lambda04 - 001_m1 - 30 above : here we see some kind of natural dark - matter mass ` confinement ' to small - moderate viable for @xmath141 are all perturbative . ] values . still for @xmath134 with @xmath142 this time , fig . [ eta014_theta-10_lambda04 - 02_m1 - 60 ] shows @xmath83 starting very high ( @xmath143 at @xmath144 , decreasing quickly with a first sudden drop at 2.7 gev and a second one to zero at 26.2 gev . a solution branch is then picked up left briefly at @xmath97 until 49 gev and then nothing . what is peculiar here is that , in contrast with previous situations , the desert starts at a mass @xmath145 , i.e. , before the opening of the @xmath32 annihilation channel . in other words , the dark matter is annihilating into the light fermions only and the model is perturbatively viable in the range 20gev 49gev . the case @xmath146 with @xmath147 is displayed in fig . [ eta014_theta-10_lambda04 - 1_m1 - 20 ] . there are no solutions below @xmath144 at which @xmath148 . from this value , @xmath83 slips down very quickly to pick up less abruptly when crossing the @xmath95 threshold . there is a significant change in the slope at the crossing of the @xmath96 mass . note the absence of a solution at @xmath97 , which is a new feature , present for other values of @xmath59 not displayed here . beyond @xmath97 , there is a slight change in the downward slope , a change of solution branch , and that goes until @xmath149 where @xmath83 jumps to catch up with the previous branch . it goes down this branch until about @xmath150 where the desert starts . we have studied the behavior of @xmath83 as a function of @xmath58 for other values of @xmath59 between @xmath151 and @xmath57 while keeping @xmath84 and @xmath146 . for @xmath152 , the behavior is qualitatively quite similar to that shown in fig . [ eta014_theta-10_lambda04 - 1_m1 - 20 ] , but beyond this mass , there is a highly non - perturbative branch @xmath83 jumps onto at small and moderate values of @xmath58 . this highly non - perturbative region stretches in size as @xmath59 increases . fig . eta014_theta-10_lambda04 - 1_m1 - 793 displays this new feature . note that on this figure , not all of the range of @xmath83 is shown in order to allow the small - coupling regions to be displayed ; the high values of @xmath118 are in the two thousands . note also that it is the same highly non - perturbative solution branch @xmath83 jumps onto for other large values of @xmath59 . last in this descriptive study is to see the effects of larger values of the @xmath32 higgs mixing angle @xmath36 . we give it here the value @xmath153 and tune back the mutual @xmath7 higgs coupling constant @xmath94 to the very small value @xmath154 . figure eta014_theta-40_lambda04 - 001_m1 - 20 shows the behavior of @xmath83 as a function of @xmath58 for @xmath147 . one recognizes features similar to those of the case @xmath84 , though coming in different relative sizes . the very - small-@xmath58 desert ends at about @xmath155 . there are by - now familiar features at the @xmath91 and @xmath96 masses , @xmath97 and @xmath59 . two relatively small forbidden intervals ( deserts ) appear for relatively large values of the dark matter mass : @xmath156 and @xmath157 . the @xmath49 mass region is forbidden but there is action as we cross the @xmath50 mass . other values of @xmath59 , not displayed because of space , behave similarly with an additional effect , namely , a sudden drop in slope at @xmath158 coming from the ignition of @xmath7 annihilation into @xmath32 and higgs . we have also worked out the cases @xmath134 and 1 for @xmath153 . the case @xmath134 is displayed in fig . [ eta014_theta-40_lambda04 - 02_m1 - 20 - 60 ] and presents differences with the corresponding small - mixing situation @xmath84 . indeed , for @xmath147 , the first feature we notice is a smoother behavior ; compare with fig . [ eta014_theta-10_lambda04 - 02_m1 - 20 ] . here , @xmath118 starts at @xmath159 with the small value @xmath160 and goes up , faster at the @xmath91 mass and with a small effect at the @xmath96 mass . it increases very slowly until @xmath97 and decreases very slowly until @xmath139 , and then there is a sudden change of branch followed immediately by a desert not displayed on fig . [ eta014_theta-40_lambda04 - 02_m1 - 20 - 60 ] . ] . so here too the model naturally confines the mass of a viable dark matter to small - moderate values , a dark matter particle annihilating into light fermions only . what is also noticeable is that there is stability of @xmath118 around the value of @xmath94 in the interval @xmath161 @xmath162 here@xmath163 . the case @xmath164 presents also overall similarities as well as noticeable differences with the corresponding case @xmath84 , see fig . [ eta014_theta-10_lambda04 - 02_m1 - 60 ] . the first difference is that all values of @xmath83 are perturbative . this latter starts at @xmath165 with the value @xmath166 , goes down and jumps to catch up with another solution branch emerging from negative territory when crossing the @xmath95 mass . it increases , kicking up when crossing the @xmath96-quark mass . it changes slope down at @xmath97 and goes to zero at about @xmath167 . it jumps up onto another branch that goes down to zero also at about @xmath168 , just below @xmath59 , and then there is a desert , except for the small interval @xmath169 . the case @xmath146 is shown in fig . eta014_theta-40_lambda04 - 1_m1 - 20 - 60 . global similarities with the previous case are apparent . all values of @xmath83 are perturbative and the mass range is naturally confined to the interval @xmath170 for @xmath147 , and @xmath171 for @xmath164 . we note action at the usual masses and , in particular , we see there are no solutions at @xmath172 like in the case @xmath84 . we note here too the quasi - constancy of @xmath83 for most of the available range . finally , we note that we have worked out larger mixing angles , notably @xmath173 . in general , these cases do not display any new features worth discussing : the overall behavior is mostly similar to what we have seen , with expected relative variations in size . experiments like cdms ii @xcite , xenon 10/100 xenon10 - 1,xenon100 , dama / libra @xcite and cogent @xcite search directly for a dark matter signal . such a signal would typically come from the elastic scattering of a dark matter wimp off a non - relativistic nucleon target . however , throughout the years , such experiments have not yet detected an unambiguous signal , but rather yielded increasingly stringent exclusion bounds on the dark matter nucleon elastic - scattering total cross section @xmath5 in terms of the dark matter mass @xmath58 . in order for a theoretical dark - matter model to be viable , it has to satisfy these bounds . it is therefore natural to inquire whether the model we present in this work has any capacity of describing dark matter . hence , we have to calculate @xmath5 as a function of @xmath58 for different values of the parameters ( @xmath174 and project its behavior against the experimental bounds . we will limit ourselves to the region 0.1gev 100gev as we are interested in light dark matter . as experimental bounds , we will use the results from cdmsii and xenon100 , as well as the future projections of supercdms supercdms and xenon1 t @xcite . the results of cogent , dama / libra and cresst will be discussed elsewhere @xcite . as the figures below show @xcite , in the region of our interest , xenon100 is only slightly tighter than cdmsii , supercdms significantly lower and xenon1 t the most stringent by far . but it is important to note that all these results loose reasonable predictability in the very light sector , say below 5gev . the scattering of @xmath175 off a sm fermion @xmath48 occurs via the t - channel exchange of the sm higgs and @xmath176 . in the non - relativistic limit , the effective lagrangian describing this interaction reads @xmath177 where @xmath178.\end{aligned}\ ] ] in this case the total cross section for this process is given by : @xmath179 ^{2}. \label{s_0 f cross - section}\]]at the nucleon level , the effective interaction between a nucleon @xmath180 or @xmath181 and @xmath175 has the form @xmath182 where the effective nucleon-@xmath175 coupling constants is given by:@xmath183 . \label{nucleon h and s_0 coupling}\]]in this relation , @xmath184 is the nucleon mass and @xmath185 the baryon mass in the chiral limit @xcite . the total cross section for non - relativistic @xmath7 @xmath186 elastic scattering is therefore:@xmath187 ^{2}. \label{s_0 n cross - section}\ ] ] the rest of this section is devoted to a brief discussion of the behavior of @xmath5 as a function of @xmath58 . we will of course impose the relic - density constraint on the dark matter annihilation cross section ( v12sigma_annihi ) . but in addition , we will require that the coupling constants are perturbative , and so impose the additional requirement @xmath188 . also , here too , the choices of the sets of values of the parameters ( @xmath189 ) can by no means be exhaustive but only indicative . furthermore , though a detailed description of the behavior of @xmath5 could be interesting in its own right , we will refrain from doing so in this work as there is no need for it , and content ourselves with mentioning overall features and trends . generally , as @xmath58 increases , the detection cross section @xmath5 starts from high values , slopes down to minima that depend on the parameters and then picks up moderately . there are features and action at the usual mass thresholds , with varying sizes and shapes . excluded regions are there , those coming from the relic - density constraint and new ones originating from the additional perturbativity requirement . close to the upper boundary of the mass interval considered in this study , there is no universal behavior to mention as in some cases @xmath5 will increase monotonously and , in some others , it will decrease or ` not be there ' at all . let us finally remark that the logplots below may not show these general features clearly as these latter are generally distorted . let us start with the small higgs @xmath32 mixing angle @xmath84 and the very weak mutual @xmath7 higgs coupling @xmath85 . [ exp_det_theta-10_lambda04 - 001_m1 - 20 ] shows the behavior of @xmath5 versus @xmath58 in the case @xmath147 . we see that for the two mass intervals @xmath190 and @xmath191 , plus an almost singled - out peaks at @xmath172 , the elastic scattering cross section is below the projected sensitivity of supercdms . however , xenon1 t will probe all the these masses , except @xmath192 gev and @xmath193 gev . also , as we see in fig . [ exp_det_theta-10_lambda04 - 001_m1 - 20 ] , most of the mass range for very light dark matter is excluded for these values of the parameters . is this systematic ? in general , smaller values of @xmath59 drive the predictability ranges to the lighter sector of the dark matter masses . figure exp_det_theta-05_lambda04 - 001_m1 - 05 illustrates this pattern . we have taken @xmath194 , just above the lighter quarks threshold . in the small - mass region , we see that supercdms is passed in the range @xmath195 . again , all this mass ranges will be probed by xenon1 t experiment , except a sharp peak at @xmath196 , but for such a very light mass , the experimental results are not without ambiguity . reversely , increasing @xmath59 shuts down possibilities for very light dark matter and thins the intervals as it drives the predicted masses to larger values . for instance , in figure exp_det_theta-10_lambda04 - 001_m1 - 40 where @xmath197 , in addition to the peak at @xmath97 that crosses supercdms but not xenon1 t , we see acceptable masses in the ranges 40gev 65gev and 78gev up . here too the intervals narrow as we descend , surviving xenon1 t as spiked peaks at 62 gev and around 95 gev . a larger mutual coupling constant @xmath94 has the general effect of squeezing the acceptable intervals of @xmath58 by pushing the values of @xmath5 up . as an illustration , see figure exp_det_theta-10_lambda04 - 02_m1 - 60 where we have taken @xmath198 and a larger value of @xmath164 . in this example , already xenon100 excludes all the masses below 100 gev except a relatively narrow peak at @xmath97 . increasing the mixing angle @xmath36 has also the general effect of increasing the value of @xmath5 . figure exp_det_theta-40_lambda04 - 001_m1 - 20 shows this trend for @xmath199 ; compare with fig . [ exp_det_theta-10_lambda04 - 001_m1 - 20 ] . the only allowed masses by the current bounds of cdmsii and xenon100 are between @xmath200 gev and @xmath201 gev , the narrow interval around @xmath97 , and another very sharp one , at about 94gev . the projected sensitivity of xenon1 t will probe all mass ranges except those at @xmath202 and 94gev . finally , it happens that there are regions of the parameters for which the model has no predictability . see figure exp_det_theta-10_lambda04 - 04_m1 - 60 for illustration . we have combined the effects of increasing the values of the two parameters @xmath203 and @xmath204 . as we see , we barely get something at @xmath97 that can not even cross xenon100 down to supercdms . in this work , we presented a plausible scenario for light cold dark matter , ( for masses lighter than @xmath205 gev ) . this latter consists in enlarging the standard model with two gauge - singlet @xmath206-symmetric scalar fields . one is the dark matter field @xmath175 , stable , while the other undergoes spontaneous symmetry breaking , resulting in the physical field @xmath176 . this opens additional channels through which @xmath175 can annihilate , hence a reducing its number density . the model is parametrized by three quantities : the physical mutual coupling constant @xmath94 between @xmath7 and the higgs , the mixing angle @xmath36 between @xmath32 and the higgs and the mass @xmath59 of the particle @xmath32 . we first imposed on @xmath7 annihilation cross section the constraint from the observed dark - matter relic density and studied its effects through the behavior of the physical mutual coupling constant @xmath83 between @xmath7 and @xmath32 as a function of the dark matter mass @xmath58 . apart from forbidden regions ( deserts ) or others where perturbativity is lost , we find that for most values of the three parameters , there are viable solutions in the small - moderate masses of the dark matter sector . deserts are found for most of the ranges of the parameters whereas perturbativity is lost mainly for larger values of @xmath59 . through the behavior of @xmath83 , we could see the mass thresholds which mostly affect the annihilation of dark matter , and these are at the @xmath91 , @xmath95 and @xmath96 masses , as well as @xmath97 and @xmath59 . the current experimental bounds from cdmsii and xenon100 put a strong constraint on the @xmath175 masses in the range between @xmath207 to @xmath200 gev . for small values of @xmath204 , very light dark matter is viable , with a mass as small as one gev . this is of course useful for understanding the results of the experiments dama / libra , cogent , cresst @xcite as well as the recent data of the fermi gamma ray space telescope @xcite . the projected sensitivity of future wimp direct searches such as xenon1 t will probe all the @xmath175 masses between @xmath208 gev and @xmath205 gev . the next step to take is to test the model against the phenomenological constraints . indeed , one important feature of the model is that it mixes the @xmath32 field with the higgs . this must have implications on the higgs detection through the measurable channels . current experimental bounds from lep ii data can be used to constrain our mixing angle @xmath36 , and possibly other parameters . in addition , a very light @xmath175 and/or @xmath176 will contribute to the invisible decay of @xmath209 and @xmath210 mesons and can lead to a significant branching fraction . these constraints can be injected back into the model and restrain further its domain of validity . these issues are under current investigation @xcite . also , in this work , the @xmath32 vacuum expectation value @xmath19 was taken equal to 100gev , but a priori , nothing prevents us from considering other scales . however , taking @xmath211 much larger than the electro - weak scale requires @xmath212 to be very tiny , which will result in the suppression of the crucial annihilation channel @xmath213 . also , we have fixed the higgs mass to @xmath61 , which is consistent with the current acceptable experimental bounds @xcite . yet , it can be useful to ask here too what the effect of changing this mass would be . finally , in this study , besides the dark matter field @xmath7 , only one extra field has been considered . naturally , one can generalize the investigation to include @xmath186 such fields and discuss the cosmology and particle phenomenology in terms of @xmath186 . it just happens that the model is rich enough to open new possibilities in the quest of dark matter worth pursuing . the cross - sections related to the annihilation @xmath7 into the scalar particles are as follows . for the @xmath214 channel , we have : @xmath215 ( m_{h}^{2}-2m_{0}^{2})}+\frac{\left ( \lambda ^{(3)}\right ) ^{2}\left ( \lambda _ { 0}^{(3)}\right ) ^{2}}{(4m_{0}^{2}-m_{h}^{2})^{2 } } \notag \\ & & + \left . \frac{\left ( \lambda _ { 1}^{(3)}\right ) ^{2}\left ( \eta _ { 01}^{(3)}\right ) ^{2}}{(4m_{0}^{2}-m_{1}^{2})^{2}+\epsilon _ { 1}^{2}}+\frac{2\lambda _ { 0}^{(3)}\lambda _ { 1}^{(3)}\lambda ^{(3)}\eta _ { 01}^{(3)}(4m_{0}^{2}-m_{1}^{2})}{\left [ ( 4m_{0}^{2}-m_{1}^{2})^{2}+\epsilon _ { 1}^{2}\right ] ( 4m_{0}^{2}-m_{h}^{2})}\right ] .\end{aligned}\]]the @xmath216 function is the step function . for the @xmath217 channel , we have the result:@xmath218 ( m_{1}^{2}-2m_{0}^{2})}+\frac{\left ( \eta _ { 01}^{(3)}\right ) ^{2}\left ( \eta _ { 1}^{(3)}\right ) ^{2}}{(4m_{0}^{2}-m_{1}^{2})^{2 } } \notag \\ & & + \left . \frac{\left ( \lambda _ { 0}^{(3)}\right ) ^{2}\left ( \lambda _ { 2}^{(3)}\right ) ^{2}}{(4m_{0}^{2}-m_{h}^{2})^{2}+\epsilon _ { h}^{2}}+\frac{2\eta _ { 01}^{(3)}\eta _ { 1}^{(3)}\lambda _ { 0}^{(3)}\lambda _ { 2}^{(3)}(4m_{0}^{2}-m_{h}^{2})}{\left [ ( 4m_{0}^{2}-m_{h}^{2})^{2}+\epsilon _ { h}^{2}\right ] ( 4m_{0}^{2}-m_{1}^{2})}\right ] .\end{aligned}\]]for the @xmath219 channel , we have:@xmath220 \left [ 4m_{0}^{2}-(m_{h}+m_{1})^{2}\right ] } } { 128\pi m_{0}^{4}}\theta ( 2m_{0}-m_{h}-m_{1 } ) \notag \\ & & \left [ \left ( \lambda _ { 01}^{(4)}\right ) ^{2}+\frac{8\lambda _ { 01}^{(4)}\eta _ { 01}^{(3)}\lambda _ { 0}^{(3)}}{m_{h}^{2}+m_{1}^{2}-4m_{0}^{2}}+\frac{2\lambda _ { 01}^{(4)}\lambda _ { 0}^{(3)}\lambda _ { 1}^{(3)}}{4m_{0}^{2}-m_{h}^{2}}+\frac{2\lambda _ { 01}^{(4)}\eta _ { 01}^{(3)}\lambda _ { 2}^{(3)}}{4m_{0}^{2}-m_{1}^{2}}\right . \notag \\ & & + \frac{16\left ( \eta _ { 01}^{(3)}\right ) ^{2}\left ( \lambda _ { 0}^{(3)}\right ) ^{2}}{\left ( m_{h}^{2}+m_{1}^{2}-4m_{0}^{2}\right ) ^{2}}+\frac{8\left ( \lambda _ { 0}^{(3)}\right ) ^{2}\eta _ { 01}^{(3)}\lambda _ { 1}^{(3)}}{\left ( m_{h}^{2}+m_{1}^{2}-4m_{0}^{2}\right ) ( 4m_{0}^{2}-m_{h}^{2 } ) } \notag \\ & & + \frac{8\left ( \eta _ { 01}^{(3)}\right ) ^{2}\lambda _ { 0}^{(3)}\lambda _ { 2}^{(3)}}{\left ( m_{h}^{2}+m_{1}^{2}-4m_{0}^{2}\right ) ( 4m_{0}^{2}-m_{1}^{2})}+\frac{\left ( \lambda _ { 0}^{(3)}\right ) ^{2}\left ( \lambda _ { 1}^{(3)}\right ) ^{2}}{(4m_{0}^{2}-m_{h}^{2})^{2 } } \notag \\ & & + \left . \frac{2\eta _ { 01}^{(3)}\lambda _ { 0}^{(3)}\lambda _ { 1}^{(3)}\lambda _ { 2}^{(3)}}{(4m_{0}^{2}-m_{h}^{2})(4m_{0}^{2}-m_{1}^{2})}+\frac{\left ( \lambda _ { 2}^{(3)}\right ) ^{2}\left ( \eta _ { 01}^{(3)}\right ) ^{2}}{(4m_{0}^{2}-m_{1}^{2})^{2}}\right ] .\end{aligned}\ ] ] the annihilation cross - section into fermions is:@xmath221 \left [ \left ( 4m_{0}^{2}-m_{1}^{2}\right ) ^{2}+\epsilon _ { 1}^{2}\right ] } \right ] . \label{annihilation s_0 fermions}\end{aligned}\]]the annihilation cross - section into @xmath49 s is given by:@xmath222 \notag \\ & & \times \left [ \frac{\left ( \lambda _ { 0}^{\left ( 3\right ) } \lambda _ { hw}^{(3)}\right ) ^{2}}{\left ( 4m_{0}^{2}-m_{h}^{2}\right ) ^{2}+\epsilon _ { h}^{2}}+\frac{\left ( \eta _ { 01}^{\left ( 3\right ) } \lambda _ { 1w}^{(3)}\right ) ^{2}}{\left ( 4m_{0}^{2}-m_{1}^{2}\right ) ^{2}+\epsilon _ { 1}^{2}}\right . \notag \\ & & + \left . \frac{2\lambda _ { 0}^{\left ( 3\right ) } \eta _ { 01}^{\left ( 3\right ) } \lambda _ { hw}^{(3)}\lambda _ { 1w}^{(3)}\left ( 4m_{0}^{2}-m_{h}^{2}\right ) \left ( 4m_{0}^{2}-m_{1}^{2}\right ) } { \left [ \left ( 4m_{0}^{2}-m_{h}^{2}\right ) ^{2}+\epsilon _ { h}^{2}\right ] \left [ \left ( 4m_{0}^{2}-m_{1}^{2}\right ) ^{2}+\epsilon _ { 1}^{2}\right ] } \right ] . \label{annihilation s_0 w}\end{aligned}\]]last , the annihilation cross - section into @xmath50 s is:@xmath223 \notag \\ & & \times \left [ \frac{\left ( \lambda _ { 0}^{\left ( 3\right ) } \lambda _ { hz}^{(3)}\right ) ^{2}}{\left ( 4m_{0}^{2}-m_{h}^{2}\right ) ^{2}+\epsilon _ { h}^{2}}+\frac{\left ( \eta _ { 01}^{\left ( 3\right ) } \lambda _ { 1z}^{(3)}\right ) ^{2}}{\left ( 4m_{0}^{2}-m_{1}^{2}\right ) ^{2}+\epsilon _ { 1}^{2}}\right . \notag \\ & & + \left . \frac{2\lambda _ { 0}^{\left ( 3\right ) } \eta _ { 01}^{\left ( 3\right ) } \lambda _ { hz}^{(3)}\lambda _ { 1z}^{(3)}\left ( 4m_{0}^{2}-m_{h}^{2}\right ) \left ( 4m_{0}^{2}-m_{1}^{2}\right ) } { \left [ \left ( 4m_{0}^{2}-m_{h}^{2}\right ) ^{2}+\epsilon _ { h}^{2}\right ] \left [ \left ( 4m_{0}^{2}-m_{1}^{2}\right ) ^{2}+\epsilon _ { 1}^{2}\right ] } \right ] . \label{annihilation s_0 z}\end{aligned}\ ] ] the quantities @xmath224 and @xmath225 are regulators at the respective resonances . the decay rates @xmath226 and @xmath227 are calculable in perturbation theory . we have for @xmath31:@xmath228 \theta \left ( m_{h}-2m_{w}\right ) ; \notag \\ \epsilon _ { h\rightarrow zz } & = & \frac{\left ( \lambda _ { hz}^{\left ( 3\right ) } \right ) ^{2}}{4\pi } \left ( 1-\frac{4m_{z}^{2}}{m_{h}^{2}}\right ) ^{\frac{1}{2}}\left [ 1+\frac{\left ( m_{h}^{2}-2m_{z}^{2}\right ) ^{2}}{8m_{z}^{4}}\right ] \theta \left ( m_{h}-2m_{z}\right ) ; \notag \\ \epsilon _ { h\rightarrow s_{0}s_{0 } } & = & \frac{\left ( \lambda _ { 0}^{\left ( 3\right ) } \right ) ^{2}}{32\pi } \left ( 1-\frac{4m_{0}^{2}}{m_{h}^{2}}\right ) ^{\frac{1}{2}}\theta \left ( m_{h}-2m_{0}\right ) ; \notag \\ \epsilon _ { h\rightarrow s_{1}s_{1 } } & = & \frac{\left ( \lambda _ { 2}^{\left ( 3\right ) } \right ) ^{2}}{32\pi } \left ( 1-\frac{4m_{1}^{2}}{m_{h}^{2}}\right ) ^{\frac{1}{2}}\theta \left ( m_{h}-2m_{1}\right ) . \label{decays h}\end{aligned}\ ] ] for @xmath32 , we have similar expressions:@xmath229 \theta \left ( m_{1}-2m_{w}\right ) ; \notag \\ \epsilon _ { s_{1}\rightarrow zz } & = & \frac{\left ( \lambda _ { 1z}^{\left ( 3\right ) } \right ) ^{2}}{4\pi } \left ( 1-\frac{4m_{z}^{2}}{m_{1}^{2}}\right ) ^{\frac{1}{2}}\left [ 1+\frac{\left ( m_{1}^{2}-2m_{z}^{2}\right ) ^{2}}{8m_{z}^{4}}\right ] \theta \left ( m_{1}-2m_{z}\right ) ; \notag \\ \epsilon _ { s_{1}\rightarrow s_{0}s_{0 } } & = & \frac{\left ( \eta _ { 01}^{\left ( 3\right ) } \right ) ^{2}}{32\pi } \left ( 1-\frac{4m_{0}^{2}}{m_{1}^{2}}\right ) ^{\frac{1}{2}}\theta \left ( m_{1}-2m_{0}\right ) ; \notag \\ \epsilon _ { s_{1}\rightarrow hh } & = & \frac{\left ( \lambda _ { 1}^{\left ( 3\right ) } \right ) ^{2}}{32\pi } \left ( 1-\frac{4m_{h}^{2}}{m_{1}^{2}}\right ) ^{\frac{1}{2}}\theta \left ( m_{1}-2m_{h}\right ) . \label{decays s_1}\end{aligned}\ ] ] where @xmath230 is equal to @xmath231 for leptons and @xmath232 for quarks . 99 d. n. spergel _ et al_. [ wmap collaboration ] , astrophys . j. suppl . 170 ( 2007 ) 377 ( ` [ arxiv : astro - ph/0603449 ] . ` ) ; a. c. pope _ et al . _ [ the sdss collaboration ] , astrophys . j. * 607 * ( 2004 ) 655 ( ` arxiv : astro - ph/0401249 ` ) . s. andreas , t. hambye and m. h. g. tytgat , jcap * 0810 * , 034 ( 2008 ) ( ` arxiv:0808.0255 [ hep - ph ] ` ) ; s. chang , j. liu , a. pierce , n. weiner , i. yavin , jcap * 1008 * ( 2010 ) 018 ( ` arxiv:1004.0697 [ hep - ph ] ` ) ; r. essig , j. kaplan , p. schuster and n. toro , arxiv:1004.0691 [ hep - ph ] ; s. andreas , c. arina , t. hambye , fu - sin ling , m. h. g. tytgat , ` arxiv:1003.2595 [ hep - ph ] ` ; d. hooper , j. i. collar , j. hall , d. mckinsey and c. kelso , phys . * d88 * ( 2010 ) 123509 ( ` arxiv:1007.1005 [ hep - ph ] . ` ) ; r. foot , phys . b * 692 * , 65 ( 2010 ) ( ` arxiv:1004.1424 [ hep - ph ] ` ) ; a. l. fitzpatrick , d. hooper and k. m. zurek , phys . rev . * d81 * ( 2010 ) 115005 ( ` arxiv:1003.0014[hep - ph ] ` ) . . e. m. drobyshevski , ` arxiv:0706.3095 [ astro - ph ] ` ; r. bernabei et al , eur . phys . j. * c53 * , 205 ( 2008 ) ; n. bozorgnia , g. b. gelmini and p. gondolo , jcap * 1011 * ( 2010 ) 019 ( ` arxiv:1006.3110 [ astro-ph.co ] ` ) ; n. bozorgnia , g. b. gelmini and p. gondolo , jcap * 1011 * ( 2010 ) 028 ( ` arxiv:1008.3676 [ astro-ph.co ] ` ) . a. bottino , n. fornengo and s. scopel , phys . * d67 * ( 2003 ) 063519 ( ` arxiv : hep - ph/0212379 ` ) ; a. bottino , f. donato , n. fornengo and s. scopel , phys . rev . * d78 * ( 2008 ) 083520 ( ` arxiv:0806.4099 [ hep - ph ] ` ) ; v. niro , a. bottino , n. fornengo and s. scopel , phys . * d80 * ( 2009 ) 095019 ( ` arxiv:0909.2348 [ hep - ph ] ` ) ; a. bottino , f. donato , n. fornengo and s. scopel , phys . rev . * d72 * ( 2005 ) 083521 ( ` arxiv : hep - ph/0508270 ` ) . z. ahmed _ et al_. [ cdms collaboration ] , phys . * 102 * ( 2009 ) 011301 ( ` arxiv:0802.3530 [ astro - ph ] ` ) ; z. ahmed _ et al_. [ cdms collaboration ] , science * 327 * ( 2010 ) 1619 ( ` arxiv:0912.3592 [ astro-ph.co ] ` ) . the effect of @xmath55 in the one - real - scalar extension of the standard model is discussed in d.n . spergel and p. j. steinhardt , phys . * 84 * ( 2000 ) 3760 ( ` astro - ph/9909386 ` ) . see also @xcite .

a positive definite completion problem pertains to determining whether the unspecified positions of a partial ( or incomplete ) matrix can be completed in a desired subclass of positive definite matrices . in this paper we study an important and new class of positive definite completion problems where the desired subclasses are the spaces of covariance and inverse - covariance matrices of probabilistic models corresponding to directed acyclic graph models ( also known as bayesian networks ) . we provide fast procedures that determine whether a partial matrix can be completed in either of these spaces and thereafter proceed to construct the completed matrices . we prove an analog of the positive definite completion result for undirected graphs in the context of directed acyclic graphs , and thus proceed to characterize the class of dags which can always be completed . we also proceed to give closed form expressions for the inverse and the determinant of a completed matrix as a function of only the elements of the corresponding partial matrix . _ key words : _ directed acyclic graph , partial matrices , positive definite matrices , positive definite completion , cholesky decomposition , perfect graph , decomposable graph . _ ams 2000 subject classifications : _ 15b48 , 15b57 , 15b99 , 05c50 , 05c20 , 05c17 .

1201.0310

a pattern of a given @xmath0 matrix is defined to be a subset of @xmath1 , i.e. , a set of positions in the matrix in which the entries are present . a ( symmetric ) partial matrix specified by a pattern is an @xmath0 symmetric matrix in which the entries corresponding to the positions listed in the pattern are specified , but the rest of the entries are unspecified and thus free to be chosen . for example @xmath2 is a @xmath3 partial matrix specified by the pattern @xmath4 a matrix completion problem asks whether for a given pattern the unspecified entries of each partial matrix can be chosen in such a way that the resulting conventional matrix is of a desired type . recent literature in general , and in particular matrix theory , computer science , statistics and signal processing have studied a variety of matrix completion problems , such as positive definite completion @xcite , low - rank completion @xcite and singular value completion @xcite . + the positive definite completion problem , one of the most well studied matrix completion problems , asks which partial matrices have positive definite completions , with or without additional features . it is clear that if a partial matrix has a positive definite completion , then it must be partial positive definite , i.e. , each fully specified principal submatrix is positive definite . + the work of grone et al . @xcite is one of the important contributions in the area of positive definite completion . the authors prove that every partial positive matrix corresponding to a given pattern has a positive definite completion if and only if the specified pattern , considered as a set of edges , forms a chordal ( or equivalently decomposable ) graph . a chordal graph is an undirected graph that has no induced cycle of length greater than or equal to 4 . although a positive definite completion is not necessarily unique , they furthermore prove that a positive definite completion matrix @xmath5 is unique if one requires that @xmath6 for each unspecified position @xmath7 . interestingly , such a positive definite completion for @xmath5 arose in an earlier paper @xcite by dempster within the context of maximum likelihood estimation for gaussian graphical models . in light of advances in the area of graphical models in recent years , the connection between positive definite completion problems and probabilistic models corresponding to undirected graphs has been thoroughly exploited ( see for example in @xcite @xcite , @xcite , @xcite , @xcite ) . + in this paper we study a new class of positive definite completion problems that corresponds to probabilistic models over directed acyclic graphs , abbreviated dag models henceforth . dag models , better known as bayesian networks , are arguably one of the most widely used classes of graphical models . the need for studying this new class of problems naturally arises when studying spaces of covariance and inverse covariance matrices corresponding to dag models @xcite . these spaces are essential features of a dag model . in the dag setting , we consider specific positive definite completions of partial matrices that are specified by a pattern determined by the edges of a directed acyclic graph @xmath8 . here the partial matrices are desired to be completed in the space of covariance or inverse - covariance matrices corresponding to dag models . + a great advantage of positive definite completion problems for dags , as we shall determine in this paper , is that we are able to present fast decision procedures to not only determine whether a given partial matrix can be completed in either of the aforementioned spaces above , but we are also able to fully construct the completed matrices . this highlights the tractable nature of completion problems for dags as compared with those of undirected graphs , where no such completion procedures are known to exist , unless the graph is decomposable . even in case of a decomposable undirected graph @xmath9 , we shall see that the same positive definite completion can be achieved under a directed version @xmath8 of @xmath9 . in this sense our result aims to generalize the existing results for completing partial positive definite matrices corresponding to undirected graphs . + the organization of the paper is as follows . in [ sec : pre ] we briefly review the basics of graphical models and in particular gaussian graphical models for both undirected and directed graphs . in [ sec : pdcp ] first we formally define two types of positive definite completion problems for dags and then , among some other results , we present two fast procedures of polynomial complexity that determine whether a partial matrix can be completed in the desired space , and specify a way to uniquely construct the completed matrix . the uniqueness of this completion has tremendous benefits for bayesian analysis of dag models ( see @xcite for more details ) . in [ sec : gen ] we prove an analog of grone et al.s @xcite theorem but in the context of dags , and also demonstrate subtle differences between the completion problems for dags vs. undirected graphs . in [ sec : det_inv ] we provide expressions for directly computing the inverse and the determinant of the positive definite completion of a partial matrix , without actually carrying out the completion . a graph @xmath9 is a pair of objects @xmath10 , where @xmath11 and @xmath12 are two disjoint finite sets representing , respectively , the vertices and the edges of @xmath9 . each edge @xmath13 is either an ordered pair @xmath14 or an unordered pair @xmath15 , for some @xmath16 . an edge @xmath17 is called directed where @xmath18 is said to be a parent of @xmath19 , and @xmath19 is said to be a child of @xmath18 . we write this as @xmath20 . the set of parents of @xmath18 is denoted by @xmath21 , and the set of children of @xmath18 is denoted by @xmath22 . the family of @xmath18 is @xmath23 . an edge @xmath24 is called undirected where @xmath18 is said to be a neighbor of @xmath19 , or @xmath19 a neighbor of @xmath18 . we write this @xmath25 . the set of all neighbors of @xmath18 is denoted by @xmath26 . we say @xmath18 and @xmath19 are adjacent if there exists either a directed or an undirected edge between them . the boundary of @xmath18 , denoted by @xmath27 , is the union of parents and neighbors of @xmath18 . a loop in @xmath9 is an ordered pair @xmath28 , or an unordered pair @xmath29 in @xmath12 . for ease of notation , in this paper we always shall assume that the edge set @xmath12 contains all the loops , although we shall draw the respective graphs without the loops . + an undirected graph is a graph with all of its edges undirected , whereas a directed graph , digraph " , is a graph with all of its edges directed . in this section , we shall use the symbol @xmath9 to denote a general graph , and make clear within the context in which it is used , whether @xmath9 is directed or undirected . + we say that the graph @xmath30 is a subgraph of @xmath31 , denoted by @xmath32 , if @xmath33 and @xmath34 . in addition , if @xmath32 and @xmath35 , we say that @xmath36 is an induced subgraph of @xmath37 . we shall consider only induced subgraphs in what follows . for a subset @xmath38 , the induced subgraph @xmath39 is said to be the graph induced by @xmath40 . a graph @xmath37 is called complete if every pair of vertices are adjacent . a clique of @xmath37 is an induced complete subgraph of @xmath37 that is not a subset of any other induced complete subgraphs of @xmath37 . more simply , a subset @xmath38 is called a clique if the induced subgraph @xmath41 is a clique of @xmath37 . the set of the cliques of @xmath9 is denoted by @xmath42 . + a path in @xmath9 of length @xmath43 from a vertex @xmath18 to a vertex @xmath19 is a finite sequence of distinct vertices @xmath44 in @xmath11 such that @xmath45 or @xmath46 are in @xmath12 for each @xmath47 . we say that the path is directed if at least one of the edges is directed . we say @xmath48 leads to @xmath49 , denoted by @xmath50 , if there is a directed path from @xmath48 to @xmath49 . a graph @xmath31 is called connected if for any pair of distinct vertices @xmath51 there exists a path between them . an @xmath52-cycle in @xmath37 is a path of length @xmath52 with the additional requirement that the end points are identical . a directed @xmath52-cycle is defined accordingly . a graph is acyclic if it does not have any cycles . an acyclic directed graph , denoted by dag , is a directed graph with no cycles of length greater than 1 . + the undirected version of a graph @xmath31 , denoted by @xmath53 , is the undirected graph obtained by replacing all the directed edges of @xmath37 by undirected ones . an immorality in a directed graph @xmath37 is an induced subgraph of the form @xmath54 . moralizing an immorality entails adding an undirected edge between the pair of parents that have the same children . then the moral graph of @xmath37 , denoted by @xmath55 , is the undirected graph obtained by first moralizing each immorality of @xmath37 and then making the undirected version of the resulting graph . naturally there are dags which have no immoralities and this leads to the following definition . a dag @xmath37 is said to be perfect " if it has no immoralities ; i.e. , the parents of all vertices are adjacent , or equivalently if the set of parents of each vertex induces a complete subgraph of @xmath37 . given a directed acyclic graph ( dag ) , the set of ancestors of a vertex @xmath48 , denoted by @xmath56 , is the set of those vertices @xmath57 such that @xmath58 . similarly , the set of descendants of a vertex @xmath48 , denoted by @xmath59 , is the set of those vertices @xmath49 such that @xmath50 . the set of non - descendants of @xmath48 is @xmath60 . an undirected graph @xmath37 is said to be decomposable if no induced subgraph contains a cycle of length greater than or equal to four . a constructive definition in terms of the cliques and the separators of the graph @xmath9 can also be specified . the reader is referred to lauritzen @xcite for all the common notions of decomposable graphs that we will use here . decomposable ( undirected ) graphs and ( directed ) perfect graphs have a deep connection . in particular , it can be shown @xcite that if @xmath9 is decomposable , then there exists a directed version of @xmath9 , i.e. , a digraph @xmath8 such that @xmath61 , where @xmath8 is a perfect dag . a graphical model over a graph @xmath9 is a family of probability distributions on a common probability space such that each distribution satisfies the set of conditional independences described by @xmath9 . two important classes of graphical models are markov random fields ( or undirected graphical models ) and bayesian networks ( or directed graphical models ) . henceforth in this paper , we shall assume that @xmath62 is an undirected graph and @xmath63 is a directed acyclic graph ( dag ) , both with the same vertex set @xmath64 . a random vector @xmath65 belongs to a markov random field over @xmath9 if it satisfies the pairwise markov property has a positive density w.r.t . a borel measure these two properties are equivalent . @xmath9 , i.e. , @xmath66 for a set @xmath38 let @xmath67 . the random vector @xmath68 belongs to a bayesian network over a dag @xmath8 if it satisfies the directed local markov property w.r.t . @xmath8 , i.e. , @xmath69 we now give examples of such models . * let us consider the four cycle @xmath70 in figure [ fig : c4 ] . then a random vector @xmath71 satisfies the global markov property w.r.t . @xmath70 implies @xmath72 * let us consider the dag @xmath8 given in figure [ fig : ddc4 ] . note that @xmath8 is a directed version of @xmath70 . now a random vector @xmath73 satisfies the directed markov property w.r.t . @xmath8 implies @xmath74 important subclasses of the models defined above arise when @xmath68 is multivariate gaussian ; namely the gaussian markov random field over @xmath9 and the directed gaussian random field over @xmath8 , denoted by @xmath75 and @xmath76 , respectively . more precisely , @xmath75 denotes the family of multivariate normal distributions @xmath77 , @xmath78 , @xmath79 ( abbreviated @xmath80 henceforth ) , that obey the local markov property w.r.t . the family of distributions @xmath75 is often referred to as the gaussian graphical or gaussian undirected model over @xmath9 . likewise , @xmath76 denotes the family of multivariate normal distributions @xmath77 that obey the directed local markov property w.r.t . @xmath8 . we shall refer to @xmath76 as the gaussian directed acyclic graph or dag model over @xmath8 . it turns out that in both of these gaussian graphical models , the required markov property is reflected in @xmath5 in terms of certain algebraic equations of the entries of @xmath5 depending on the structure of the underlying graph . the description of the equations in @xmath75 is quite simple . we have @xmath81 if and only if @xmath82 in order to express the equations satisfied by @xmath5 when @xmath83 , first we establish a few ( and to some degree standard ) notations for dag models @xcite . + first note that the relation @xmath20 defines a partial order on the vertex set of @xmath8 . since any partial order can be extended to a total order , we can therefore assume without loss of generality that the vertices are numbered in such a way that @xmath84 implies that @xmath85 , @xmath86 . by this convention a random vector @xmath87 obeys the directed local markov property w.r.t . @xmath8 ( or more precisely an equivalent version of it called the ordered markov property - see @xcite for details ) if and only if @xmath88 where @xmath89 is called the set of predecessors of @xmath90 . + now for @xmath91 let @xmath92 denote the @xmath93 submatrix of @xmath5 with rows indexed by @xmath40 and columns indexed by @xmath94 . we often write @xmath95 for the principal submatrix @xmath96 , @xmath97 for @xmath98 and similarly @xmath99 for @xmath100 . by a result in @xcite @xmath101 if and only if @xmath80 and @xmath102 note that if @xmath103 , then equation implies that @xmath104 . in particular , this implies that @xmath105 . to include this situation in equation we shall use the convention @xmath106 and @xmath107 . equation also illustrates how the directed local markov property is reflected in the entries of @xmath5 in terms of algebraic equations . we now proceed to define the schur complement of a symmetric positive definite matrix . consider a @xmath108 symmetric matrix @xmath109 partitioned as follows : @xmath110 where @xmath111 is a partition of @xmath112 . note that the matrix @xmath109 is positive definite if and only if @xmath113 is positive definite and @xmath114 is positive definite . the matrix @xmath115 is called the _ schur complement _ of @xmath113 in @xmath109 . in this section we propose two polynomial time procedures involving rational functions for completing partial positive definite matrices to positive definite matrices that correspond to gaussian bayesian networks @xmath76 . in addition , these completion problems , as we will explain later , are also generalizations of the classical positive definite problem in @xcite . to formalize the completion problem that we discuss in this paper , we introduce some definitions and notation . let @xmath63 be a dag . a @xmath8-partial matrix is a symmetric function @xmath116 where @xmath12 is the edge set of the undirected version of the dag @xmath117 . the set of all @xmath8-partial matrices , denoted by @xmath118 , is a real linear space of dimension @xmath119 . recall that @xmath120 denotes the set of cliques of @xmath121 . now for each clique @xmath122 , the restriction of @xmath123 to @xmath124 , denoted by @xmath125 , is a @xmath126 matrix @xmath127 . a partial positive definite matrix over @xmath8 is a @xmath8-partial matrix @xmath123 such that @xmath128 for each @xmath129 . the set of all partial positive definite matrices over @xmath8 is denoted by @xmath130 . a partial matrix over an undirected graph @xmath9 , or a partial positive definite matrix over an undirected graph @xmath9 , can be similarly defined . next we define two sets ; the set of covariance matrices and the set of inverse - covariance matrices corresponding to a gaussian bayesian network @xmath76 . more precisely , these spaces are , respectively : @xmath131 similarly , for an undirected graph @xmath62 we define @xmath132 note that by equation @xmath133 if and only if @xmath6 whenever @xmath134 . a characterizing feature of a gaussian bayesian network @xmath76 is that the structure of the underlying dag ( i.e. , the graph itself ) , in terms of the missing arrows , can be fully recovered from the lower triangular matrix in the cholesky decomposition of @xmath135 . the following remark formalizes this fact . [ rem : chol_decom_pd ] let @xmath136 denote the linear space of all lower triangular matrices with unit diagonal entries such that @xmath137 then @xmath138 if and only if there exists a lower triangular matrix @xmath139 and a diagonal matrix @xmath140 , with strictly positive diagonal entries , such that in the modified cholesky decomposition @xmath141 @xcite . in addition , the modified cholesky decomposition is unique . we now discuss the relationships between the spaces @xmath142 ( or @xmath143 ) and its undirected counterpart @xmath144 ( or @xmath145 ) . in particular , the modified cholesky decomposition property of @xmath146 in remark [ rem : chol_decom_pd ] above implies the following . [ lem : wermuth ] suppose @xmath8 is an arbitrary dag . then @xmath147 , where the undirected graph @xmath148 is the moral graph of @xmath8 . suppose @xmath149 . let @xmath150 be the modified cholesky decomposition of @xmath146 . it is required to prove that @xmath151 is also in @xmath152 . in light of equation [ eq : graphical_model_prop ] it suffices to show that if @xmath153 and @xmath90 are non - adjacent in @xmath148 , then @xmath154 . on the contrary , suppose that @xmath155 . therefore @xmath156 since @xmath157 , this implies that there exists @xmath158 such that @xmath159 and @xmath160 . consequently , @xmath161 and @xmath162 . hence @xmath153 and @xmath90 are parents of @xmath158 which in turn implies that @xmath153 and @xmath90 are adjacent in @xmath148 , yielding a contradiction . [ prop : equiv ] if @xmath8 is a perfect dag , and @xmath163 is the undirected version of @xmath8 , then @xmath164 . suppose @xmath8 is a perfect dag . thus @xmath165 and by lemma [ lem : wermuth ] @xmath166 . now to establish the other inclusion , assume that @xmath133 . let @xmath167 be the modified cholesky decomposition of @xmath168 . by remark [ rem : chol_decom_pd ] it suffices to show that if @xmath169 and @xmath85 , then @xmath170 . note however that @xmath169 implies that @xmath171 . therefore @xmath172 assume to the contrary that @xmath173 , then there exists an index @xmath158 such that @xmath159 and @xmath160 . this in turn implies that there exists an immorality @xmath174 since by assumption @xmath169 . we have thus reached a contradiction to the fact that @xmath8 is perfect . therefore @xmath175 and consequently @xmath176 or @xmath149 . note that the statement of proposition [ prop : equiv ] can be rephrased as follows : if @xmath8 is a perfect dag , then a normal distribution obeys the directed local markov property w.r.t . @xmath8 if and only if it obeys the pairwise markov property w.r.t . @xmath177 , the undirected version of @xmath8 . it can be easily shown that if @xmath8 is a perfect dag , the above statement holds in more generality than just for normal distributions @xcite . convention : hereafter in this paper , and unless otherwise stated , we assume that @xmath178 is the undirected version of the dag @xmath179 , i.e. , @xmath180 . let @xmath181 be a subset of the space of @xmath182 symmetric matrices , denoted by @xmath183 . we say that a @xmath8-partial matrix @xmath123 can be completed in @xmath181 if there exists a matrix @xmath184 such that @xmath185 for each @xmath186 . we refer to @xmath187 as a completion of @xmath123 in @xmath181 , or simply a completion of @xmath123 if @xmath181 is @xmath183 . similar definitions can also given in the context of undirected graphs , i.e , the completion of @xmath9-partial matrices . + [ cor : per_dec ] let @xmath8 be a perfect dag and let @xmath163 denote the undirected version of @xmath8 . then @xmath188 can be completed in @xmath142 if and only if it can be completed in @xmath189 . the proof is immediate from proposition [ prop : equiv ] . an important question in the probabilistic analysis of directed markov random fields / dags is whether a @xmath8-partial matrix can be completed in @xmath143 , i.e. , whether a given @xmath8-partial matrix corresponds to an inverse covariance matrix of a dag model . similar questions can be asked about completions in @xmath142 . these are inherently algebraic questions . note that remark [ rem : chol_decom_pd ] implies that one can potentially recover the full matrix @xmath151 merely from the entries that correspond to the edge set of @xmath8 . we formalize this statement in the proposition below . [ prop : completion_in_pd ] let @xmath123 be a @xmath8-partial matrix in @xmath118 . if @xmath190 , then * almost everywhere ( w.r.t . lebesgue measure on @xmath118 ) , there exist a unique lower triangular matrix @xmath175 and a unique diagonal matrix @xmath191 such that @xmath192 is a completion of @xmath123 . * the matrix @xmath193 is the unique positive definite completion of @xmath123 in @xmath143 if and only if the diagonal entries of @xmath140 are all strictly positive . @xmath194 first we shall show that , almost everywhere w.r.t . lebesgue measure on @xmath118 , @xmath123 can be uniquely completed to a matrix @xmath193 in @xmath183 , not necessarily positive definite , such that @xmath192 , for some @xmath175 and a diagonal matrix @xmath195 . we shall use the cholesky factorization algorithm @xcite to construct @xmath140 and @xmath196 , column by column , in the following steps . * set @xmath170 for each @xmath169 . * set @xmath197 for each @xmath198 and set @xmath199 . * if @xmath200 , then set @xmath201 and proceed to step @xmath202 , otherwise @xmath196 and @xmath140 are constructed such that they satisfy the condition in part @xmath203 * set @xmath204 and proceed to the next step . * for each @xmath205 if @xmath206 , then set @xmath207 , and return to step @xmath208 . if @xmath209 , then no completion of @xmath123 exists that satisfies the condition in part @xmath194 . consequently , @xmath123 can not also be completed in @xmath143 . note that for each @xmath210 , the expression for @xmath211 given by @xmath212 in step @xmath202 , considered as a function of @xmath213 , is a rational function , say @xmath214 . in particular , almost everywhere w.r.t . lebesgue measure on @xmath118 , @xmath215 for all @xmath216 . therefore , almost everywhere , the process above yields matrices @xmath196 and @xmath140 that satisfy the condition in part @xmath194 . + @xmath217 part @xmath217 now follows from part @xmath194 and remark [ rem : chol_decom_pd ] . [ rem : no_guarantee_pd ] it is clear from proposition [ prop : completion_in_pd ] above that a positive definite completion of @xmath123 in @xmath143 is not always guaranteed . moreover , the ability to complete in @xmath143 is not known beforehand , and is determined as a byproduct of having gone through the completion process itself . having said this , if at any stage a @xmath211 becomes negative or zero , it is evident from proposition [ prop : completion_in_pd ] that a completion in @xmath143 is no longer possible and the completion process can be terminated . we now demonstrate the completion outlined above on a partial matrix @xmath188 . .,width=113 ] consider the dag @xmath8 given by figure [ fig : dag_pg ] and the @xmath8-partial matrix @xmath123 given by @xmath218 now by applying the completion process in proposition [ prop : completion_in_pd ] to @xmath123 we obtain @xmath219 which yields the completed matrix @xmath193 given as follows : @xmath220 however , as the diagonal elements of @xmath140 are not strictly positive , @xmath123 can not be completed in @xmath143 . an equally important question is whether a @xmath8-partial matrix can be completed in @xmath142 , the space of covariance matrices corresponding to the dag model @xmath76 . recall that from equation we have @xmath221 by recursively applying equation we show below that it is possible to determine whether a @xmath8-partial matrix can be completed in @xmath142 . the procedure is described in the following proposition . [ prop : completion_in_pdd ] let @xmath222 , then * there exists a completion process of polynomial complexity that can determine whether @xmath123 can be completed in @xmath142 ; * if a completion exists , this completion is unique and can be determined constructively using the following process : + @xmath223 set @xmath224 for each @xmath225 and set @xmath226 . + @xmath227 if @xmath228 , then set @xmath229 and proceed to the next step , otherwise @xmath5 is successfully completed . + @xmath230 if @xmath231 , then proceed , the submatrix @xmath232 is fully determined by step _ ( 2)_. ] to the next step , otherwise the completion in @xmath142 does not exist . + @xmath233 if @xmath234 is empty , then return to step @xmath227 , otherwise proceed to the next step . + @xmath235 if @xmath236 is non - empty , then set @xmath237 , @xmath238 and return to step @xmath227 . if @xmath236 is empty , then set @xmath239 and return to step @xmath227 . + note that we can shorten step @xmath235 by making the following convention . for any two distinct sets @xmath240 if @xmath241 , then set @xmath242 and @xmath243 . this convention will automatically take into account the case when @xmath236 is empty . note that the completion process above starts with the the highest label vertex in @xmath8 ( this vertex is called a source " node and does not have any parents ) and the algorithm proceeds in a descending manner . in the process above the positive definiteness condition of the completed matrix will be guaranteed by requiring that each principal submatrix , starting from the highest label and moving down , is positive definite at every step , i.e. , positive definiteness is maintained layer by layer . now suppose that down to an integer @xmath244 , the process described above has succeeded in uniquely constructing a positive definite matrix that corresponds to the principal submatrix @xmath245 of @xmath5 . therefore , the process returns to step @xmath227 with @xmath246 and then proceeds to step @xmath230 . by this step note that @xmath232 is fully determined because , ( a ) the submatrix @xmath247 is specified since it is a submatrix of @xmath245 , where the latter is already determined by the end of the previous step , and ( b ) the row @xmath248 is specified since it corresponds to directed edges in the dag ( i.e. , @xmath222 ) . hence @xmath232 is fully determined by the beginning of this step . if @xmath232 is not positive definite , then the completion in @xmath249 can not exist as all the principal submatrices of a positive definite matrix also have to be positive definite , i.e. , @xmath231 is a necessary condition to continue with the completion process . we now proceed to show that the condition @xmath231 is also sufficient for the completion process . in particular , the condition @xmath231 and equation as in step @xmath235 uniquely determine the unspecified entries of the @xmath90-th column and row of the new submatrix @xmath250 . to see this write @xmath251 thus we have @xmath252 note that in the second last step we have used the expression @xmath253 from step ( 5 ) . the positive definiteness in the last step follows respectively from the facts that , ( a ) @xmath254 as this is equivalent to assuming @xmath231 since @xmath255 , and , ( b ) @xmath256 since it corresponds to a schur complement of a principal submatrix of the positive definite matrix @xmath257 . note that in proposition [ prop : completion_in_pdd ] if we require @xmath232 in step @xmath230 to be only invertible , instead of positive definite , then by a similar argument as in the proof of [ prop : completion_in_pd ] , we can show that , almost everywhere w.r.t lebesgue measure on @xmath118 , the process in proposition [ prop : completion_in_pdd ] yields a matrix , not necessarily positive definite , @xmath258 that satisfies equation . we now proceed to illustrate the completion process in proposition [ prop : completion_in_pdd ] using two examples . the first example illustrates the completion process symbolically and the second example applies to a @xmath8-partial matrix with numerical entries . [ ex : completion_on_pdd ] @xmath194 consider the dag @xmath8 given in figure [ fig : dc4 ] . a partial matrix corresponding to @xmath8 can be written symbolically as @xmath259 where incomplete entries in @xmath123 are denoted by @xmath260 . we now proceed in layers using the steps in proposition [ prop : completion_in_pdd ] as @xmath90 decreases from @xmath261 to @xmath262 . + + in step @xmath223 of the procedure described in proposition [ prop : completion_in_pdd ] we have @xmath263 + in step @xmath227 let @xmath264 . in step @xmath230 either @xmath265 , otherwise the completion in @xmath142 does not exist . assuming the former , we proceed to step @xmath235 . since @xmath266 , the layer down to @xmath267 is thus completed . + + we now return to step @xmath227 with @xmath268 . in step @xmath230 we check whether @xmath269 . assuming @xmath270 , then in step @xmath235 , as @xmath271 , we set @xmath272 and the layer down to @xmath273 is thus completed . + + now the process is returned to step @xmath227 with @xmath274 . in step @xmath230 we first check whether @xmath275 assuming @xmath276 , then in step @xmath235 , as @xmath277 we set @xmath278 now all the unspecified entries are determined and the completed matrix @xmath279 is said to be the completion of @xmath123 in @xmath142 . + @xmath217 consider the dag @xmath8 given in figure [ fig : dg5 ] and let @xmath280 now by applying the procedure in proposition [ prop : completion_in_pdd ] , we start with @xmath281 . first note that @xmath282 . in step @xmath235 we set @xmath283 , since @xmath284 . in the next layer we have @xmath267 and it can be easily verified that @xmath285 now @xmath286 is empty so by step @xmath233 we return to step @xmath227 since @xmath235 in the completion process is redundant . now in step @xmath230 for @xmath273 it is obvious that @xmath287 therefore we proceed to step @xmath235 and calculate @xmath288 . moving to @xmath199 , it is easily verified that @xmath289 finally , we proceed to step ( 4 ) , where we have @xmath290 and we set @xmath291 the process now terminates and the unique completion of @xmath123 in @xmath142 is given by @xmath292 in order to double check that @xmath5 is indeed in @xmath142 we first compute @xmath168 : @xmath293 now the lower triangular matrix @xmath294 in the standard cholesky decomposition of @xmath151 is given by @xmath295 which clearly shows that the lower triangular matrix @xmath196 in the corresponding modified cholesky decomposition of @xmath151 is in @xmath136 . proposition [ prop : completion_in_pdd ] establishes conditions under which a @xmath8-partial matrix can be completed in @xmath142 . it therefore establishes conditions for the existence and uniqueness of the completion . a natural question to ask is if there are simple conditions which guarantee this completion . we now deduce from proposition [ prop : completion_in_pdd ] that if the digraph @xmath8 is a perfect dag then @xmath296 can always be completed in @xmath142 . [ cor : completion_per ] let @xmath8 be a perfect dag and @xmath297 . then a necessary and sufficient condition for completing @xmath123 in @xmath142 is that @xmath296 . moreover , if @xmath296 , then @xmath123 can be simply completed to @xmath149 as follows : + @xmath194 set @xmath224 for each @xmath298 , + @xmath217 set @xmath299 and @xmath300 for each @xmath301 . ( @xmath302 ) if @xmath303 , then @xmath123 can not be completed to a positive definite matrix . this follows easily from the fact that the principal minors of a positive definite matrix are all strictly positive . in particular , if a completion for @xmath123 in @xmath142 exists , then @xmath128 for each @xmath304 , which implies that @xmath296 . ( @xmath305 ) now assume that @xmath296 . as @xmath8 is perfect , by definition there are no immoralities present in @xmath8 , ( i.e. , all parents of each node are adjacent ) . hence @xmath306 is a complete subset of @xmath11 for each @xmath90 . it is therefore contained in a clique of @xmath8 , and hence @xmath307 since @xmath296 . thus step @xmath230 of the completion process in proposition [ prop : completion_in_pdd ] is always satisfied and can thus be omitted from the procedure . we can therefore conclude that @xmath123 can alway be completed in @xmath142 . now since @xmath247 are all already available step @xmath235 can be performed for @xmath308 . the proof of proposition [ prop : completion_in_pdd ] demonstrates that after performing these steps , we obtain the completion of @xmath123 in @xmath142 . an alternative but similar procedure to that in proposition [ prop : completion_in_pdd ] for completing a @xmath8-partial matrix @xmath123 in @xmath142 is to construct a finite sequence of dags , @xmath309 such that @xmath310 at the end of this sequence is perfect , and then a partial matrix , @xmath311 over @xmath310 . the first dag in this sequence is @xmath312 . if @xmath313 is not perfect , then for each immorality of the form @xmath174 in @xmath313 with @xmath85 we add a directed edge @xmath84 to the edge set of @xmath313 . let @xmath314 be the dag with added edges . it is clear that @xmath313 is an induced subgraph of @xmath314 . we continue this process until we obtain a perfect dag . therefore , we have a finite sequence @xmath309 of dags such that @xmath310 at the end of this sequence is perfect . now starting from the largest vertex @xmath90 , we use equation to compute the entries @xmath315 that correspond to added edges @xmath84 in @xmath310 . unless for some @xmath90 the requirement @xmath255 is not met in equation , this process succeeds in filling in those unspecified entries of @xmath5 that correspond to @xmath310 , i.e. , we obtain a partial matrix @xmath311 over @xmath310 . since @xmath310 is a perfect dag , by corollary [ cor : completion_per ] , the partial matrix @xmath311 can be completed in @xmath316 if and only if it belongs to @xmath317 . furthermore , @xmath311 can be completed by following the simple non - recursive completion procedure described in corollary [ cor : completion_per ] . it is clear that completion of @xmath311 in @xmath316 is also the completion of @xmath123 in @xmath142 . we illustrate this alternative procedure by an example . let @xmath8 be the dag in figure [ fig : alt_figa ] . starting from @xmath318 , the only immorality in this dag is @xmath319 . by adding the directed edge @xmath320 we obtain @xmath321 in figure [ fig : alt_figb ] . next we obtain the perfect dag @xmath322 in figure [ fig : alt_figc ] , by adding the directed edge @xmath323 corresponding to the immorality @xmath324 in @xmath321 . now consider the completion of the @xmath8-partial matrix @xmath325 from equation [ eq : dl_normal_repeat ] we compute @xmath326 thus we obtain the following partial matrix over the perfect dag @xmath322 ( or over the decomposable graph in figure [ fig : alt_figd ] ) : @xmath327 a pertinent question in the positive definite completion problem for dags is the class of dags for which the completion of a partial matrix in @xmath142 is certain to exist . corollary [ cor : completion_per ] asserts that if @xmath8 is perfect and the @xmath8-partial matrix @xmath296 then a completion in @xmath328 always exists . is the class of perfect graphs maximal in the sense that only for this class of graphs is completion guaranteed for all @xmath296 ? it is evident that a necessary condition for the existence of the completion in @xmath328 , or any positive definite completion for that matter , is that @xmath123 is a partial positive definite matrix over @xmath8 , i.e. , @xmath296 . [ thm : existence_completion ] every partial positive definite matrix over @xmath8 can be completed in @xmath142 if and only if @xmath8 is a perfect dag . we proceed using a proof by contradiction embedded in an induction argument . assume the statement of the theorem is true for any dag s.t . @xmath329 . we shall prove the theorem for @xmath330 . the case @xmath331 is trivially true , hence let @xmath332 . let @xmath333}$ ] denote the induced dag on @xmath334 . + @xmath335 suppose that every partial positive definite matrix over @xmath121 can be completed in @xmath142 . let @xmath336}$ ] be an arbitrary element in @xmath337}}$ ] . now let us define the @xmath8- partial matrix @xmath123 such that for each @xmath338 @xmath339 } & \text{otherwise . } \end{cases}\ ] ] it is clear that @xmath123 is a partial positive definite matrix in @xmath130 . by assumption , @xmath123 can be completed to a positive definite matrix @xmath5 in @xmath142 . now note that the principal submatrix @xmath340 is positive definite and satisfies equation w.r.t . @xmath333}$ ] . in particular , @xmath340 is the completion of @xmath336}$ ] in @xmath341}}$ ] . by the induction hypothesis this implies that @xmath333}$ ] is a perfect dag . therefore , to show that @xmath8 is perfect , it suffices to show that @xmath342 is a complete subset of @xmath11 . on the contrary , if @xmath342 is not complete , then there are non - adjacent vertices @xmath343 such that @xmath344 . assume w.l.o.g that @xmath345 . in particular , this implies that @xmath346 is a predecessor of @xmath347 , i.e. , @xmath348 . fix an arbitrary number @xmath349 in the open interval @xmath350 . consider the @xmath8-partial matrix @xmath123 that is defined for each @xmath351 as @xmath352 one can easily check that @xmath353 . let @xmath5 denote the completion of @xmath123 in @xmath142 . since @xmath354 from equation we have @xmath355 in particular , @xmath356 . therefore @xmath357 however , this matrix is positive definite if and only if @xmath358 , yielding a contradiction . + @xmath359let @xmath8 be a perfect dag and @xmath360 . let @xmath336}$ ] be the restriction of @xmath123 to @xmath334 and @xmath361}$ ] the completion of @xmath336}$ ] in @xmath341}}$ ] . now define @xmath5 as follows : @xmath362},\ : \sigma_{\fa(1 ) , 1}=\gamma_{\fa(1 ) , 1 } \:\ ; \text{and } \:\ ; \sigma_{\pr(1 ) , 1}=\sigma_{\pr(1 ) , \pa(1)}(\sigma_{\pa(1)})^{-1}\sigma_{\pa(1 ) , 1}.\ ] ] then using equation when @xmath199 shows that @xmath80 . notice that the @xmath359 part of theorem [ thm : existence_completion ] was also proved in corollary [ cor : completion_per ] . in the context of undirected graphs , grone et al . @xcite prove that every partial positive definite matrix can be completed to a positive definite matrix if and only if the underlying graph is decomposable . theorem [ thm : existence_completion ] above is the corresponding result in the dag context with the caveat that the result in @xcite for undirected graphs does not imply the result below . in particular , @xcite implies that @xmath363 has to be decomposable if a positive definite completion is to be guaranteed for any arbitrary partial positive definite matrix . the requirement that the graph @xmath363 be decomposable does not however mean that @xmath8 is perfect . [ cor : completion_ug ] suppose @xmath9 is a decomposable graph . then every @xmath364 can be completed to a unique @xmath5 in @xmath189 consequently , every partial positive definite matrix over a decomposable graph has a positive definite completion . let @xmath123 be in @xmath365 and let @xmath8 be a perfect dag version of @xmath9 . then @xmath222 . the result is immediate by applying corollary [ cor : completion_per ] followed by corollary [ cor : per_dec ] . there is an interesting contrast between completing a given partial positive definite matrix @xmath366 in @xmath189 vs. completing it in @xmath142 . in particular , theorem 3 . in @xcite asserts that @xmath364 can be completed in @xmath189 if _ any _ positive completion exists . a completion in @xmath142 is therefore sufficient to guarantee a completion in @xmath189 . the other way around is not true . in particular , @xmath123 may not be completed in @xmath328 even when it can be completed in @xmath189 . this is because completion in @xmath142 is more restrictive than completion in @xmath189 . we illustrate this distinction in the following example . , width=105 ] [ ex : no_completion ] consider the following partial positive definite matrix over the dag in figure [ fig : nonpdag4 ] . @xmath367 although @xmath8 is not a perfect dag we have @xmath9 , the undirected version of @xmath8 , is decomposable and therefore by corollary [ cor : completion_ug ] it can be completed to a positive definite matrix in @xmath189 . however , completion of @xmath123 in @xmath142 requires that @xmath368 and one can check that the completed matrix @xmath369 is not positive definite . consequently , @xmath123 can not be completed in @xmath142 . now suppose that @xmath9 is an undirected graph and @xmath370 . when @xmath9 is decomposable it has a dag version @xmath8 that is perfect . corollary [ cor : completion_per ] and the fact that @xmath371 ( see proposition [ prop : equiv ] ) imply that @xmath123 can be explicitly completed in @xmath372 . however , when @xmath9 is not decomposable no algorithm for completing @xmath123 in @xmath189 is known , unless a positive definite completion in @xmath373 is given or guaranteed . given @xmath9 , in practice , it is useful to know whether @xmath123 can be completed in _ some _ dag version of this undirected graph @xmath374 . this is because if @xmath123 can be completed w.r.t a dag version of @xmath9 , then the completion process given in proposition [ prop : completion_in_pdd ] can be exploited to obtain a positive definite completion , and thus ensuring the existence of a completion in @xmath189 . in the following example we show that even when @xmath123 can be completed in @xmath189 the completion may still not exist for any dag version of @xmath9 . from example [ cycle_ug_dag ] , width=132 ] [ cycle_ug_dag ] consider the partial matrix @xmath375 over the four cycle @xmath70 as given in figure [ fig : ac4 ] . it is clear that when @xmath376 , @xmath123 is a partial positive definite matrix over @xmath70 . it is shown in @xcite that @xmath123 can be completed to a positive definite matrix @xmath5 if and only if @xmath377 now consider the list of the dag versions of @xmath70 as given in table [ table : dc4 ] . a simple enumeration will demonstrate that the list in table [ table : dc4 ] is exhaustive and contains all dag version of @xmath70 . .dag versions of @xmath70 from example [ cycle_ug_dag ] . [ cols="^,^,^,^ " , ] [ table : dc4 ] table [ table : dc4 ] gives for each dag version of @xmath70 the edges labeled with the corresponding entries of @xmath123 . the dashed edges and their labels correspond to the missing edges and entries computed by using equation . for example , the dag version in table [ table : dc4 ] ( 1 ) corresponds to the partial matrix @xmath378 here the missing entry @xmath379 is computed as @xmath380 , which is the label on the dashed edge @xmath381 in table [ table : dc4 ] ( 1 ) . note that once the dashed edges are included all the dags in table [ table : dc4 ] are perfect . therefore , if the partial matrix over the corresponding perfect dag in table [ table : dc4 ] is partial positive definite , then by corollary [ cor : completion_per ] the completion in @xmath142 ( where @xmath8 is the original dag version of @xmath70 ) is guaranteed . the above reasoning allows us derive the required system of inequalities that sensure that @xmath123 can be completed in @xmath142 . for example , the partial matrix corresponding to the perfect dag in table [ table : dc4 ] ( 1 ) is @xmath382 a simple calculation will show that the lower right @xmath383 submatrix of @xmath40 is positive definite . hence @xmath40 is partial positive definite if and only if @xmath384 . this is equivalent to requiring that @xmath385 similarly , we can show that the partial matrix corresponding to each of the perfect dags given in table [ table : dc4 ] ( 2 ) , ( 3 ) or ( 4 ) respectively , is partial positive definite if and only if @xmath386 it is easy to check from table [ table : dc4 ] that the inequalities obtained under the perfect dags given in table [ table : dc4 ] ( 5 ) , ( 6 ) , ( 7 ) and ( 8) are the same inequalities already listed above . the partial matrix corresponding to each of the perfect dags given in table [ table : dc4 ] ( 9 ) or ( 10 ) respectively is , partial positive definite if and only if @xmath387 now consider the following choices for @xmath388 : @xmath389 , @xmath390 , @xmath391 , and @xmath392 . then we have @xmath393 , but @xmath394 hence when the above values for @xmath388 are substituted in @xmath123 we obtain a partial positive definite matrix that can not be completed in @xmath142 for any dag version @xmath8 of @xmath70 , although it can be completed in @xmath395 . in this section we give closed form expressions for the inverse and the determinant of a completed matrix @xmath142 as a function of only the elements of the corresponding partial matrix . first we need the following notion for undirected graphs . + * for three disjoint subsets @xmath396 and @xmath397 of @xmath11 we say that @xmath397 separates @xmath40 from @xmath94 in @xmath9 if every path from a vertex in @xmath40 to a vertex in @xmath94 intersects a vertex in @xmath397 . * let @xmath123 be a @xmath9-partial matrix . the zero - fill - in of @xmath123 in @xmath9 , denoted by @xmath398^{v}$ ] , is a @xmath399 matrix @xmath187 such that @xmath400 [ lem : inv_det ] let @xmath63 be an arbitrary dag and let @xmath178 denote the undirected version of @xmath8 . let @xmath149 and let @xmath401 be a partition of @xmath11 such that @xmath397 separates @xmath40 from @xmath94 in @xmath148 . then we have + @xmath402^{v}+\left [ \left(\sigma_{b\cup s}\right)^{-1}\right]^{v } -\left [ \left(\sigma _ { s}\right)^{-1}\right]^{v}$ ] and + @xmath403 [ prop : inv_det ] let @xmath123 be a partial positive definite matrix in @xmath130 that can be completed to a positive definite matrix @xmath5 in @xmath142 . then we have : + @xmath404^{v}-\left[\left(\sigma_{\pa(i)}\right)^{-1}\right]^{v}\right)$]and + @xmath405 . suppose that by mathematical induction the assertion of the proposition holds for any dag with number of vertices less than @xmath406 . we proceed to prove the proposition for @xmath8 with @xmath406 vertices . the case @xmath331 holds trivially . thus assume that @xmath407 . let @xmath408}$ ] denote @xmath334 . the triple @xmath409 is a partition of @xmath11 , and @xmath410 separates @xmath411 from @xmath412 in @xmath148 . thus by lemma [ lem : inv_det ] , @xmath413^{v } + \left [ \left(\sigma_{\fa(1)}\right)^{-1 } \right]^{v}-\left[\left(\sigma_{\pa(1 ) } \right)^{-1 } \right]^{v}.\ ] ] invoking the same notation used in the proof of theorem [ thm : existence_completion ] , note that the partial matrix @xmath336}$ ] is positive definite over @xmath333}$ ] with positive completion @xmath340 in @xmath341}}$ ] . by the induction hypothesis applied to @xmath414 we have @xmath415^{v_{[1]}}-\left[\left(\psi_{\pa(i)}\right)^{-1}\right]^{v_{[1]}}\right).\ ] ] since @xmath333}$ ] is an ancestral subgraph of @xmath8 , it is clear that for each @xmath416 we have @xmath417 by replacing these in equation and zero - fill - in in @xmath177 we obtain @xmath418^{v}=\sum_{i=2}^{p}\left(\left[\left(\sigma_{\fa(i)}\right)^{-1}\right]^{v}-\left[\left(\sigma_{\pa(i)}\right)^{-1}\right]^{v}\right).\ ] ] finally , substituting equation into equation yields the formula in part @xmath419 . part @xmath420 follows similarly by using part @xmath227 in lemma [ lem : inv_det ] . [ ex : inv_det ] let @xmath8 be the dag given in figure [ fig : inv_ex ] . + @xmath194 suppose that the @xmath8-partial matrix @xmath421 can be completed to a positive definite matrix @xmath5 in @xmath142 . by applying part @xmath419 of proposition [ prop : inv_det ] we have @xmath422^{v}+\left[(\sigma_{\{2,5\}})^{-1}\right]^{v}+ \left[(\sigma_{\{3,4,5\}})^{-1}\right]^{v}+\left[\sigma_{44}^{-1}\right]^{v}\\ & + \left[\sigma_{55}^{-1}\right]^{v}-\left[(\sigma_{\{2,4\}})^{-1}\right]^{v}-\left[\sigma_{55}^{-1}\right]^{v } -\left[(\sigma_{\{4,5\}})^{-1}\right]^{v}.\end{aligned}\ ] ] note that all the entries of the matrices involved in this expression are given in @xmath123 , except for @xmath423 and @xmath379 . by using equation it is easy to check that @xmath424 . hence , @xmath425^{v}+ \left [ \left ( \begin{matrix } 1&\sigma_{25}\\ \sigma_{52}&1 \end{matrix } \right)^{-1 } \right]^{v}+ \left [ \left ( \begin{matrix } 1 & \sigma_{34}&\sigma_{35}\\ \sigma_{43}&1&0\\ \sigma_{53}&0&1 \end{matrix } \right)^{-1 } \right]^{v}\\ & + \left ( \begin{matrix } 0&0&0&0&0\\ 0&0&0&0&0\\ 0&0&0&0&0\\ 0&0&0&1&0\\ 0&0&0&0&0 \end{matrix } \right)- \left ( \begin{matrix } 0&0&0&0&0\\ 0&1&0&0&0\\ 0&0&0&0&0\\ 0&0&0&1&0\\ 0&0&0&0&0 \end{matrix } \right)- \left ( \begin{matrix } 0&0&0&0&0\\ 0&0&0&0&0\\ 0&0&0&0&0\\ 0&0&0&1&0\\ 0&0&0&0&1 \end{matrix } \right)\\ & = \frac{1}{1-\sigma_{12}^{2}-\sigma_{14}^{2}}\left ( \begin{matrix } 1&-\sigma_{12 } & 0 & -\sigma_{14 } & 0 \\ -\sigma_{12}&1-\sigma_{14}^{2 } & 0 & \sigma_{12}\sigma_{14 } & 0 \\ 0 & 0 & 0 & 0&0 \\ -\sigma_{14}&\sigma_{12}\sigma_{14 } & 0&1-\sigma_{12}^{2 } & 0 \\ 0&0 & 0 & 0 & 0 \end{matrix } \right ) + \frac{1}{1-\sigma_{25}^{2 } } \left ( \begin{matrix } 0&0&0&0&0\\ 0&1&0&0&-\sigma_{25}\\ 0&0&0&0&0\\ 0&0&0&0&0\\ 0&-\sigma_{25}&0&0&1 \end{matrix } \right)\\ & + \frac{1}{1-\sigma_{34}^{2}-\sigma_{35}^{2 } } \left ( \begin{matrix } 0&0 & 0&0 & 0 \\ 0 & 0&0 & 0 & 0 \\ 0 & 0 & 1&-\sigma_{34 } & -\sigma_{35 } \\ 0 & 0 & -\sigma_{34 } & 1-\sigma_{35 } ^{2 } & \sigma_{34 } \sigma_{35 } \\ 0 & 0 & -\sigma_{35 } & \sigma_{34 } \sigma_{35 } & 1-\sigma_{34 } ^{2 } \end{matrix } \right ) + \left ( \begin{matrix } 0 & 0 & 0 & 0&0 \\ 0&-1 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & -1 & 0 \\ 0 & 0 & 0&0 & -1 \end{matrix } \right).\end{aligned}\ ] ] by combining these terms into one matrix we have @xmath168 is equal to : @xmath426 similarly , using part @xmath420 of proposition [ prop : inv_det ] we have @xmath427^{-1}.\ ] ] @xmath217 let us apply the computation of the inverse matrix in part @xmath194 to the following specific @xmath8-partial matrix @xmath428 using the completion process in proposition [ prop : completion_in_pdd ] one can check that the completion of @xmath123 in @xmath142 is given as @xmath429 from this we obtain @xmath430 however , without completing @xmath5 and with less computation , we can compute @xmath168 using the results in proposition [ prop : inv_det ] . for this , we apply part @xmath194 above . first let @xmath431 and @xmath432 . thus @xmath433 if @xmath434 denotes the completion of @xmath435 in @xmath142 , then by using part @xmath194 we obtain @xmath436 from this we can obtain @xmath437 , which as expected turns out to be same matrix as the one given in equation . example [ ex : inv_det ] demonstrates that the inverse and the determinant of the completion of a partial matrix @xmath123 in @xmath142 can be computed using the expressions given in proposition [ prop : inv_det ] , and generally , avoid recourse to the whole completion process . this is especially so when @xmath8 is a perfect dag . since in this case @xmath438 , and consequently @xmath439 , are already blocks of the partial matrix @xmath123 , the computations can be carried out without recourse to the completion process in proposition [ prop : completion_in_pdd ] . this fact about perfect dags can also be deduced from their relationship to undirected decomposable graphs ( see lemma 5.5 in @xcite ) . _ acknowledgments : _ ben - david was supported in part by a seed grant from the cardiovascular institute , stanford university school of medicine and by the national science foundation under grant no . dms - cmg-1025465 . rajaratnam was supported in part by the national science foundation under grant nos . dms-0906392 , dms - cmg-1025465 , ags-1003823 , dms-1106642 and grants nsa h98230 - 11 - 1 - 0194 , darpa - yfa n66001 - 11 - 1 - 4131 , and suwievp10-sufsc10-smscvisg0906 .

legendre - gauss - lobatto ( lgl ) grids play a pivotal role in nodal spectral methods for the numerical solution of partial differential equations . they not only provide efficient high - order quadrature rules , but give also rise to norm equivalences that could eventually lead to efficient preconditioning techniques in high - order methods . unfortunately , a serious obstruction to fully exploiting the potential of such concepts is the fact that lgl grids of different degree are not nested . this affects , on the one hand , the choice and analysis of suitable _ auxiliary spaces _ , when applying the _ auxiliary space method _ as a principal preconditioning paradigm , and , on the other hand , the efficient solution of the auxiliary problems . as a central remedy , we consider certain nested hierarchies of dyadic grids of locally comparable mesh size , that are in a certain sense properly associated with the lgl grids . their actual suitability requires a subtle analysis of such grids which , in turn , relies on a number of refined properties of lgl grids . the central objective of this paper is to derive just these properties . this requires first revisiting properties of close relatives to lgl grids which are subsequently used to develop a refined analysis of lgl grids . these results allow us then to derive the relevant properties of the associated dyadic grids .

1311.0028

studying the distribution of zeros of orthogonal polynomials is a classical theme that has been addressed in an enormous number of research papers . the sole reason for revisiting this topic here is the crucial role of _ legendre - gauss - lobatto ( lgl ) grids _ , formed by the zeros of corresponding orthogonal polynomials , for the development of efficient preconditioners for high order finite element and even spectral discretizations of pdes , see e.g. @xcite . in connection with elliptic pdes , which often possess very smooth solutions and thus render high order methods at least potentially extremely efficient , a key constituent is the fact that interpolation at lgl grids give rise to fully robust ( with respect to the polynomial degrees ) isomorphisms between high order polynomial spaces and low order finite element spaces on lgl grids . however , unfortunately , one quickly faces some serious obstructions to fully exploiting this remarkable potential of lgl grids when simultaneously trying to exploit the flexibility of _ discontinuous galerkin _ ( dg ) schemes , namely locally refined grids and locally varying polynomial degrees . indeed , as explained in @xcite , the essential source of the problems encountered then is the fact that lgl grids are _ not nested_. this affects the choice and analysis of suitable _ auxiliary spaces _ , when using the auxiliary space method , see e.g. @xcite as preconditioning strategy , as well as the efficient solution of the corresponding auxiliary problems . as a crucial remedy , certain hierarchies of nested _ dyadic _ grids have been introduced in @xcite which are _ associated _ in a certain sense with lgl grids . the term `` associated '' encapsulates a number of properties of such dyadic grids , some of which have been used and claimed in @xcite but will be proved here which is the central objective of this paper . the layout of the paper is as follows . after collecting , for the convenience of the reader , in section [ sect : ultraspherical ] some classical facts and tools used in the sequel , we formulate in section [ sect : main ] the main results of the paper . the first one , theorem [ thm : a ] , is concerned with the quasi - uniformity of lgl grids as well as with a certain notion of equivalence between lgl grids of _ different _ order . this is important for dealing with dg - discretizations involving locally varying polynomial degrees , see @xcite . the second one , theorem [ thm : b ] , concerns certain hierarchies of nested dyadic grids that are associated in a very strong sense with lgl grids . both theorems play a crucial role for the design and analysis of preconditioners for dg systems . section [ sect : proofa ] is devoted to the proof of theorem [ thm : a ] . this requires deriving a number of _ refined properties _ of lgl grids which to our knowledge can not be found in the literature . in particular , we need to revisit in section [ sect : cgl ] some close relatives namely _ chebyshev - gauss - lobatto ( cgl ) nodes _ since they have explicit formulae that help deriving sharp estimates . the central subject of section [ sect : dyadic ] is the generation of dyadic grids associated in a certain way with a given other grid as well as the analysis of the properties of these dyadic grids . in particular , the results obtained in this section lead to a specific hierarchy for which the properties claimed in theorem [ thm : b ] are verified . throughout the paper we shall employ the following notational convention . by @xmath0 we mean that the quantity @xmath1 can be bounded by a constant multiple of @xmath2 uniformly in the parameters @xmath1 and @xmath2 may depend on . likewise @xmath3 is equivalent to @xmath4 and @xmath5 means @xmath6 and @xmath7 . a central notion in this work concerns _ grids _ @xmath8 induced by zeros of certain orthogonal polynomials , especially , the first derivatives of legendre polynomials . it will be important though that those are special cases of _ ultraspherical or gegenbauer polynomials _ whose definition is recalled for the convenience of the reader . let the parameters @xmath9 and @xmath10 be fixed . the ultraspherical or gegenbauer polynomial @xmath11 of degree @xmath12 is defined as orthogonal polynomial on the interval @xmath13 $ ] with respect to the weighting function @xmath14 , i.e. @xmath15 with @xmath16 and normalization @xmath17 the following useful properties of ultraspherical polynomials can be found in ( * ? ? ? * chapter 4.7 ) . for any @xmath10 the ultraspherical polynomials fulfill the symmetry property @xmath18 and the differentiation rule @xmath19 holds . the understanding of these polynomials hinges to a great extent on a fact that will be used several times , namely that the ultraspherical polynomial @xmath11 is a solution of the linear homogeneous ordinary differential equation ( ode ) of second order @xmath20 see , e.g. , ( * ? ? ? * ( 4.2.1 ) ) . by an elementary transformation , based on the product ansatz for the solution @xmath21 , the first order term can be eliminated , see ( * ? ? ? * section 1.8 ) , so that the transformed ode @xmath22 has the solutions @xmath23 , see ( * ? ? ? * ( 4.7.10 ) ) . for the convenience of the reader we recall next some standard tools that are used to estimate the zeros of classical orthogonal polynomials and their spacings . [ theo : sturmcomparison ] let @xmath24\subset { \ensuremath{\mathbb{r}}}$ ] an interval and @xmath25\rightarrow { \ensuremath{\mathbb{r}}}$ ] and @xmath26\rightarrow { \ensuremath{\mathbb{r}}}$ ] two continuous functions . let @xmath27 and @xmath28 , respectively , be nontrivial solutions of the homogeneous linear ordinary differential equations @xmath29 respectively . 1 . [ it : sct1 ] let @xmath30 and @xmath31 . if @xmath32 for @xmath33 , then @xmath34 for @xmath35 , where @xmath36 is the smallest zero of @xmath28 in the interval @xmath37 $ ] . [ it : sct2 ] let @xmath30 and @xmath32 for @xmath33 . then the smallest zero of @xmath28 in @xmath37 $ ] occurs left of the smallest zero of @xmath27 in @xmath37 $ ] . [ it : sct3 ] under the hypothesis of ( ii ) , the @xmath38-th zero of @xmath28 in @xmath37 $ ] occurs before the @xmath38-th zero of @xmath27 to the right of @xmath1 . one application of the sturm comparison theorem is the sturm convexity theorem , where two intervals between zeros of the same function are compared . [ theo : sturmconvexity ] let @xmath27 be a nontrivial solution of @xmath39 where @xmath25 \rightarrow { \ensuremath{\mathbb{r}}}$ ] is continuous and non - increasing . then the sequence of zeros of @xmath40 is convex , i.e. , for consecutive zeros @xmath41 , @xmath42 , @xmath43 $ ] of @xmath40 we get @xmath44 we also record the following consequence of a theorem by markoff , see ( * ? ? ? * theorem 6.21.1 , pp . 121 ff ) , that @xmath45 in other words , taking the symmetry into account , we observe that the zeros of the ultraspherical polynomial @xmath11 move towards the center of the interval @xmath13 $ ] with increasing parameter @xmath46 . we recall next for later purposes the main relations between ultraspherical , legendre and chebyshev polynomials from ( * ? ? ? * chapter iv ) . here we are mainly interested in the _ legendre polynomials _ @xmath47 of degree @xmath48 , which are the ultraspherical polynomials @xmath49 , and their first derivatives . later we shall use the chebyshev polynomials of first kind @xmath50 and of second kind @xmath51 as a tool . our primary interest concerns the special case of _ legendre polynomials_. for @xmath52 the lgl nodes @xmath53 of _ order _ @xmath12 are the @xmath54 zeros of the polynomial @xmath55 , where @xmath56 is the first derivative of the legendre polynomial of degree @xmath12 . the lgl nodes are sorted in increasing order , i.e. , we have @xmath57 for @xmath58 . their collection @xmath59 forms the lgl grid of order @xmath12 . given the lgl grid @xmath60 , we define for @xmath61 the corresponding lgl intervals as @xmath62 \subset [ -1,1]$ ] , which are of length @xmath63 . we also define the corresponding lgl angles @xmath64 $ ] as @xmath65 . + our notation deviates from the standard literature on orthogonal polynomials because the lgl nodes are usually enumerated in decreasing order , see for example ( * ? ? ? * chapter vi ) . note that @xmath66 and @xmath67 , i.e. @xmath68 and @xmath69 . figure [ fig : enumeration ] exemplifies the enumeration scheme used for the lgl points and intervals for odd and even values of @xmath12 . by the differentiation rule for ultraspherical polynomials , we have in the special case of legendre polynomials @xmath70 for the special case of @xmath71 , the transformed ode @xmath72 has the solution @xmath73 , which has its zeros exactly at the lgl nodes of order @xmath54 . by the symmetry relation , the lgl nodes are symmetric with respect to the origin , i.e. , @xmath74 for @xmath52 , in particular @xmath75 if @xmath12 is even . for the lgl angles , symmetry yields @xmath76 for @xmath52 , in particular @xmath77 if @xmath12 is even . consequently , the lgl grids on @xmath13 $ ] are _ symmetric _ around zero and an analogous property holds , of course , for their affine images in general intervals @xmath24 $ ] with respect to their midpoint @xmath78 . in what follows all grids under consideration are assumed without further mentioning to be symmetric in this sense . we close this section with recalling a well - known fact about the _ monotonicity _ of lgl grids since this will be used frequently . [ theo : monotonicitylength ] the lgl interval lengths are strictly increasing from the boundary towards the center of @xmath13 $ ] , i.e. , for @xmath79 and @xmath80 we have @xmath81 . by symmetry it suffices to consider the left half of the intervall @xmath13 $ ] . the lgl points of order @xmath54 are the zeros of the polynomial @xmath73 , which is a solution of the transformed ode . since @xmath82 is monotonically decreasing on @xmath83 $ ] , by the sturm convexity theorem [ theo : sturmconvexity ] , we have @xmath84 which is the assertion . note that for @xmath85 there is only one interval and for @xmath86 the two intervals are of same length by symmetry . in this section we present the main results of this paper . as pointed out in the introduction , they are relevant for the development and analysis of robust preconditioners for high order dg - discretizations . these results can be roughly grouped into two parts , namely ( a ) results on refined properties of lgl grids themselves and ( b ) results on the relationship between lgl grids and certain families of _ associated _ dyadic grids . for ready use of the subsequent results in the dg context we formulate the results in this section for a generic interval @xmath24 $ ] . as mentioned before lgl nodes and corresponding grids on such an interval are always understood as affine images of the respective quantities on @xmath13 $ ] using for simplicity the same notation . to be precise in what follows one should distinguish between a _ grid _ and a _ partition _ or mesh induced by a grid . in general , a grid @xmath8 in @xmath24 $ ] is always viewed as an ordered set @xmath87 of strictly increasing points @xmath88 - the nodes , with @xmath89 and @xmath90 , which induces a corresponding partition @xmath91 formed by the closed intervals @xmath92 $ ] , @xmath93 . the length @xmath94 of @xmath95 is denoted by @xmath96 . conversely , a partition @xmath97 of @xmath24 $ ] into consecutive intervals induces an ordered grid @xmath98 formed by the endpoints of the intervals . [ def : quasi - uniform ] a family of grids @xmath99 is called _ locally quasi - uniform _ if there exists a constant @xmath100 such that @xmath101 the next notion concerns a certain comparability of two grids . [ def : loc - unif - equiv ] given two constants @xmath102 , the grid @xmath8 is said to be _ locally _ @xmath103-_uniformly equivalent _ to the grid @xmath104 if the following condition holds : @xmath105 the main result concerning lgl grids reads as follows . [ thm : a ] ( i ) the family of lgl grids @xmath106 , is locally quasi - uniform and the constant @xmath107 in satisfies @xmath108 ( ii ) assume that @xmath109 with @xmath110 for some fixed constant @xmath111 . then , the lgl grid @xmath112 is locally @xmath103-uniformly equivalent to the grid @xmath60 , with @xmath113 and @xmath114 depending on @xmath36 but not on @xmath115 and @xmath12 . the fact ( i ) that lgl grids are quasi - uniform seems to be folklore . since we could not find a suitable reference we restate this fact here as a convenient reference for @xcite and include later a proof with a concrete bound for the constant @xmath100 . claim ( ii ) is essential for establishing optimality of preconditioners for high order dg discretization with varying polynomial degrees , see @xcite . numerical experiments show that the quotient of the length of two consecutive lgl intervals is maximized for @xmath116 and this term grows monotonically in @xmath12 . the value of the ( smallest ) constant @xmath117 is approximately @xmath118 . a further important issue concerns the comparison of an lgl grid with a `` stretched '' version of itself . to explain this we introduce the _ stretching operator _ @xmath119}:[a,(a+b)/2 ] \rightarrow [ a , b ] , \ , x \mapsto 2x - a$ ] . there is an apparent relation concerning the behavior of lgl grids under stretching which is needed later for establishing subsequent results on associated dyadic grids . the relevant property , illustrated by figure [ fig : comparisonstretched ] , reads as follows . [ def : s ] a family of symmetric grids @xmath99 on @xmath24 $ ] has property @xmath120for some @xmath121 , if for any @xmath122 with @xmath123 the following holds : for any @xmath124 with @xmath125 $ ] and any @xmath126 such that @xmath127 , one has @xmath128 . we have verified numerically that property @xmath120holds for lgl grids up to order @xmath129 . this supports the following conjecture . [ conj : lglcomparisonstretched ] the family of lgl grids @xmath106 has property @xmath120for all @xmath121 . with its stretched version . ] lgl grids are unfortunately not nested , i.e. , an interval in @xmath130 can not be written as the union of intervals in @xmath131 which is a severe impediment on the use of lgl grids in the context of preconditioning . hierarchies of nested grids are conveniently obtained by successive dyadic splits of intervals . more precisely , a single split , i.e. , replacing an interval @xmath132 in a given grid by the two intervals @xmath133 obtained by subdividing @xmath132 at its midpoint , gives rise to a _ refinement _ of the current grid . hence successive refinements of some initial interval give rise to a sequence of nested grids . we shall see next that that one can associate in a very strong sense with any lgl grid @xmath60 a dyadic grid @xmath134 that inherits relevant features from @xmath60 while in addition the family @xmath135 is _ nested_. moreover , when using the associated dyadic grids in the context of the auxiliary space method ( see @xcite ) for preconditioning the systems arising form dg - discretizations , one may obtain meshes with hanging nodes . it then turns out that it is important that the dyadic grids are `` closed under stretching '' in the following sense . [ closedness under stretching ] [ def : closedness ] consider the stretching operator @xmath136}$ ] introduced above . then we call a dyadic grid @xmath137 on @xmath24 $ ] _ closed under stretching _ if @xmath136}({\ensuremath{{\cal d}}}\cap [ a,(a+b)/2 ] ) \subset { \ensuremath{{\cal d}}}$ ] . [ thm : b ] given any lgl grid @xmath60 on any interval @xmath24 $ ] , there exists a dyadic grid @xmath134 on @xmath24 $ ] with the following properties : * @xmath134 is symmetric and monotonic in the sense of theorem [ theo : monotonicitylength ] , i.e. , the intervals @xmath138 increase in length from left to right in the left half of @xmath24 $ ] . * the family of grids @xmath135 is locally quasi - uniform . * the family of grids @xmath135 is nested . * for each @xmath10 the grid @xmath134 is locally @xmath103-uniformly equivalent to @xmath60 , where the constants @xmath113 and @xmath114 do not depend on @xmath12 . in particular , one has @xmath139 , uniformly in @xmath12 . * whenever the family of lgl grids @xmath106 has property @xmath120for @xmath140 , then the dyadic grids @xmath134 are closed under stretching for @xmath141 . [ rem : stretch ] as mentioned before , @xmath120has been confirmed numerically to hold for @xmath142 , i.e. , the dyadic grids @xmath134 are closed under stretching for @xmath143 which covers all cases of practical interest in the context of preconditioning . conjecture [ conj : lglcomparisonstretched ] actually suggests that all dyadic grids @xmath134 referred to in theorem [ thm : b ] are closed under stretching . for the application of the above results in the context of preconditioning @xcite quantitative refinement of theorem [ thm : b ] ( ii ) is important . we call a dyadic grid @xmath137 _ graded _ if any two adjacent intervals @xmath144 in @xmath145 satisfy @xmath146 , i.e. , they differ in generation by at most one . [ rem : graded ] the dyadic grids @xmath134 referred to in theorem [ thm : b ] are graded for @xmath147 . moreover , we conjecture that gradedness holds actually for _ all _ @xmath148 , see the discussion following conjectures [ conj : quotients ] and [ conj : quotientsbessel ] in section [ sect : quasiuniformitylgl ] and proposition [ prop : graded ] . the remainder of the paper is essentially devoted to proving theorems [ thm : a ] and [ thm : b ] . this requires on the one hand refining our understanding of lgl grids , which , in particular , leads to some results that are perhaps interesting in their own right , and , on the other hand , developing and analyzing a concrete algorithm for creating @xmath134 for any given @xmath12 . since all the properties claimed in theorems [ thm : a ] and [ thm : b ] are invariant under affine transformations it suffices to consider from now on only the reference interval @xmath24=[-1,1]$ ] . in this section we proceed establishing further quantitative facts about lgl grids needed for the proof of theorem [ thm : a ] . we address first interrelations of lgl nodes within a single grid @xmath60 of arbitrary order @xmath12 . we begin recalling the following estimates on lgl angles which can be found , for instance , in @xcite and @xcite @xmath149 \quad \text{for } \quad 1 \le k \le { \ensuremath{\left\lfloor \frac{n-1}{2 } \right\rfloor}}\\ \quad \text{with } \quad { \underline{\eta}}^{n}_{k } : = \pi \frac{k}{n } \quad \text{and } \quad { \overline{\eta}}^{n}_{k}:=\pi \frac{2k+1}{2n+1 } . \end{aligned}\ ] ] the lower bound follows from and the fact that the ultraspherical polynomial @xmath150 can , up to a non - zero normalization factor , be identified as the chebyshev polynomial of second kind @xmath151 , which has the zeros @xmath152 for @xmath153 , see e.g. ( * ? ? ? * ( 2.3.15 ) on p. 77 ) . the upper bound is given in ( * ? ? ? * lemma 1 ) . note that the lower bound obtained from is sharper than the lower estimate in ( * ? ? ? * lemma 1 ) . next we consider the spacings between two subsequent lgl nodes . to this end , in the following , we will repeatedly make use of the following elementary estimates : for any @xmath154 $ ] one has @xmath155 moreover , we shall use the prostapheresis trigonometric identity @xmath156 note that for @xmath157 $ ] and @xmath158 the arguments of the sine function @xmath159 and @xmath160 are both within @xmath161 $ ] . recall that @xmath162 as the lgl nodes include the boundary points . for @xmath85 the lgl nodes are @xmath163 and @xmath164 , for @xmath165 we have @xmath166 , @xmath167 and @xmath168 . for higher order lgl nodes , unfortunately there is no simple analytical expression . in order to study lgl intervals for larger @xmath12 , we estimate the length of the lgl intervals from below and above using . for the proof we consider separately the cases where the position of a lgl node is the endpoint or the center of the interval @xmath13 $ ] . [ lem : lgleilengths ] for @xmath169 and @xmath170 , we can bound the length of the interval @xmath171 by @xmath172 for @xmath173 the length of the boundary interval @xmath174 can be estimated by @xmath175 moreover , if @xmath12 is odd and @xmath173 , we get @xmath176 whereas if @xmath12 is even and @xmath177 , we have @xmath178 from for @xmath169 and @xmath170 we can derive the upper estimate @xmath179 analogously , we have in the same case as lower estimate @xmath180 in the special case of the boundary interval @xmath181 $ ] for @xmath173 , we have @xmath182 therefore we obtain @xmath183 and @xmath184 if @xmath12 is odd ( @xmath185 ) and @xmath79 , there is a central lgl interval @xmath186 of size @xmath187 , see figure [ fig : enumeration](a ) . we estimate its length by @xmath188 and @xmath189 if @xmath12 is even ( @xmath190 ) and @xmath177 , the center of the interval is a lgl node , i.e. , we have @xmath77 , @xmath75 , see figure [ fig : enumeration](b ) . the adjacent interval @xmath191 has the length @xmath192 . this can be estimated by @xmath193 and @xmath194 which closes the proof . the estimates for lgl interval lengths in property [ lem : lgleilengths ] show that @xmath195 for the boundary interval and @xmath196 for the intervals at the origin . in this subsection we present the proof of theorem [ thm : a ] ( i ) , i.e. , we show that two consecutive lgl intervals differ in length at maximum by a constant factor independent of @xmath12 . [ lem : quasiuniformity ] the lgl grids @xmath60 satisfy @xmath197 where the constant @xmath198 is bounded by the right hand side in . by symmetry , it suffices to consider all subintervals that have a nonzero intersection with the left half interval @xmath199 . for @xmath86 the two lgl intervals @xmath200 $ ] and @xmath201 $ ] are of the same size . for @xmath173 we apply property [ lem : lgleilengths ] for the quotient of two consecutive interval lengths . for @xmath202 we have @xmath203 and @xmath204 for the quotient including the boundary interval @xmath174 , we have @xmath205 and @xmath206 if @xmath12 is even ( @xmath190 ) and @xmath207 , we can estimate the quotient from above by @xmath208 because the very but last term is monotonically decreasing for @xmath207 . from below we can estimate the quotient by @xmath209 on the other hand , if @xmath12 is odd ( @xmath185 ) and @xmath173 , we can estimate the quotient from above by @xmath210 because the very but last term is monotonically decreasing for @xmath211 . for the estimate from below we have @xmath212 overall , we have the desired estimate for @xmath100 given by , as claimed . for later purposes it is important to have as tight a bound for @xmath100 as possible and the one derived above will not quite be sufficient . the remainder of this section is devoted to a discussion of possible quantitative improvements and related conjectures . [ def : quotients ] we say that the family of lgl grids @xmath106 has property @xmath213for some @xmath121 , if the quotients @xmath214 satisfy 1 . @xmath215 for @xmath216 , i.e. , the quotients increase monotonically in @xmath12 for each fixed @xmath217,[quotientsboundaryincwithn ] 2 . @xmath218 for @xmath219 , i.e. , the quotients decrease monotonically in @xmath220 for any fixed @xmath221 . [ quotientsdecwithk ] in the present context property ( i ) is only relevant for @xmath222 . the following observations are immediate and recorded for later use . [ rem : smallestconstant ] assume that property @xmath213holds for some @xmath140 . then , the value of the ( smallest ) constant @xmath223 in for the family of lgl grids @xmath224 is @xmath225 . moreover , when omitting the outermost intervals , the constant @xmath226 satisfying @xmath227 is bounded by @xmath228 . specifically , one can verify numerically that @xmath213holds for @xmath142 , in which case one has @xmath229 numerical evidence supports the following [ conj : quotients ] the lgl grids @xmath60 have property @xmath213 for all @xmath121 . in order to determine a constant @xmath223 that holds for all @xmath148 , one can exploit the well known fact that the asymptotic behavior of the lgl nodes can be expressed by means of the zeros of the bessel function @xmath230 . [ theo : lglasymptotics ] the asymptotic behavior of the lgl angles @xmath231 is given by the formula @xmath232 where @xmath233 are the nonnegative zeros of the bessel function @xmath234 . the first nonnegative zeros of the bessel function @xmath234 are given in table [ tab : besselj1zeros ] . .the first nonnegative zeros of the bessel function @xmath230 of first kind . @xcite [ cols="^,^,^,^,^,^",options="header " , ] in this subsection we analyze the behavior of lgl interval lengths with increasing order . the situation in the following theorem is depicted in figure [ fig : displacement ] . the result is an essential ingredient of the proof of theorem [ thm : b ] ( iv ) . [ theo : displgl2 ] let @xmath48 with @xmath235 and @xmath236 . if @xmath237 for @xmath238 with @xmath239 and @xmath240 , then for the corresponding lgl intervals @xmath241 $ ] and @xmath242 $ ] , we have the inequality @xmath243 . by symmetry it is sufficient to consider only the left half of @xmath13 $ ] . if @xmath244 , the assertion follows directly from theorem [ theo : monotonicitylength ] . otherwise we observe that the function @xmath245 see , is non - increasing on @xmath83 $ ] . furthermore @xmath246 for @xmath247 and @xmath248 . let @xmath40 be a solution of the ode @xmath249 , having the zeros @xmath250 for @xmath251 . then @xmath252 is a solution of the ode @xmath253 with zeros @xmath254 for @xmath251 . now we choose @xmath255 such that @xmath256 and @xmath257 coincide . then , by the properties noted above and because of @xmath237 , we have @xmath258 by the sturm comparison theorem , the first zero @xmath259 of @xmath260 to the right of @xmath261 occurs before the first zero of @xmath262 to the right of @xmath263 , i.e. , @xmath264 which completes the proof . in order to prove theorem [ thm : a ] ( ii ) we make a small digression establishing first corresponding properties for another class of grids associated with ultraspherical polynomials , namely _ chebyshev - gauss - lobatto ( cgl ) grids_. this is a much easier task since the corresponding nodes can be expressed by explicit formulae . recall that the chebyshev polynomials of first kind @xmath265 of degree @xmath48 are also special instances of ultraspherical polynomials . in fact , for @xmath266 one has @xmath267 with appropriate nonzero normalization constants @xmath268 . [ def : cgl ] for @xmath48 we define the cgl nodes @xmath269 by @xmath270 their collection @xmath271 forms the cgl grid of order @xmath12 . we also define the corresponding cgl intervals @xmath272 $ ] for @xmath58 . we recall that the points @xmath273 with indices @xmath274 are the local extrema of the chebyshev polynomial of the first kind @xmath275 , namely the zeros of the chebyshev polynomial of the second kind @xmath151 . we can immediately calculate the lengths of the intervals bounded by two consecutive cgl points using the prostapheresis trigonometric identity . note that for @xmath157 $ ] and @xmath158 the arguments of the sine function @xmath159 and @xmath160 are both within @xmath161 $ ] . the following properties whose analogs for lgl grids have already been established before are simpler in this case but will be needed below . [ prop : cglilengths ] the length of the @xmath38-th cgl interval @xmath276 is @xmath277 we derive next two types of facts about cgl nodes and corresponding intervals , namely first _ monotonicity _ statements for any given polynomial degree and second the evolution of interval lengths for increasing degrees . let us consider the quotient of the lengths of two consecutive cgl intervals @xmath278 for @xmath153 . [ property : cgl - uniform ] the lengths of the cgl intervals increase monotonically in the left half of the interval @xmath13 $ ] , i.e. , @xmath279 furthermore , the cgl nodes form a quasi - uniform decomposition of the interval @xmath13 $ ] , i.e. , we have @xmath280 with @xmath281 for @xmath153 . by symmetry , it is sufficient to consider the left half of @xmath13 $ ] , i.e. , the cgl intervals that have nonempty intersection with @xmath199 . if @xmath12 is even , @xmath282 is a zero of the chebyshev polynomial @xmath283 . therefore , in this case @xmath284 $ ] if and only if @xmath285 . otherwise @xmath12 is odd and there is additionally a central interval @xmath286 $ ] that is symmetric with respect to the origin , i.e. , the cgl interval @xmath287 has nonempty intersection with @xmath199 if and only if @xmath288 . combining both cases , we conclude that the cgl interval @xmath287 has nonempty intersection with @xmath199 if and only if @xmath289 . in this case , the arguments of the sine functions in the numerator and the denominator of the last term of are in @xmath161 $ ] and since the sine function is monotonically increasing on @xmath161 $ ] , we have @xmath290 , which is the first part of the assertion . to estimate @xmath291 from above , we bound the sine function on @xmath161 $ ] using ( [ eq : estimatetrig ] c ) , and obtain @xmath292 which is the second part of our claims . as the degree @xmath12 increases , the cgl intervals decrease in length and move towards the end points of @xmath13 $ ] . to quantify both effects , we formulate the following proposition on the monotonicity of the interval lengths . as before , by symmetry , we can restrict ourselves to the left half of the interval @xmath13 $ ] . the following proposition is not strictly needed for our present purposes . but since cgl grids are also used in numerical analysis their association with dyadic grids may also be of interest so that we pause including the following result for completeness . [ lem : dispcgl ] let @xmath48 , @xmath293 and @xmath289 be given . if for @xmath294 we have @xmath295 , then the corresponding cgl intervals satisfy @xmath296 . using the assumption on the location of the cgl nodes and the monotonicity properties of the cosine function on @xmath161 $ ] , we note that @xmath297 is equivalent to @xmath298 which , in turn , is equivalent to @xmath299 . since the sine function increases monotonically on @xmath161 $ ] , we can conclude for the lengths of the corresponding cgl intervals , upon using several times , @xmath300 which is the desired estimate . we are now prepared to establish locally uniform equivalence of grids of comparable order , associated with a family of ultraspherical polynomials , namely for the chebyshev and legendre cases . here the chebyshev case will serve as a major tool for deriving later an analog for the legendre case . [ prop : lgl - uniequiv - cheb ] assume that @xmath301 , with @xmath110 , for some fixed constant @xmath111 . then , the cgl grid @xmath302 in the interval @xmath13 $ ] is locally @xmath103-uniformly equivalent to the grid @xmath302 , with @xmath113 and @xmath114 depending on @xmath36 but not on @xmath115 and @xmath12 . recall from definition [ def : cgl ] that the nodes of the cgl grid of order @xmath12 are defined as @xmath303 for @xmath52 ; @xmath304 $ ] is the @xmath38-th interval of this grid , with @xmath305 , whose length is given by property [ prop : cglilengths ] . suppose that @xmath306 , i.e. , there exists @xmath307 such that @xmath308 \;,\end{aligned}\ ] ] and @xmath309 \;.\end{aligned}\ ] ] the invertibility of the cosine function in @xmath310 $ ] yields @xmath311 i.e. , @xmath312 \;.\end{aligned}\ ] ] then , @xmath313 and @xmath314 , i.e , @xmath315 now , using property [ prop : cglilengths ] and @xmath316 for small @xmath307 , as well as @xmath317 , we have @xmath318 therefore it is enough to prove the uniform equivalence of the two sines . to this end , we first observe that , by , we have @xmath319 whence , noting that @xmath320 , we get @xmath321 exchanging the roles of @xmath38 and @xmath322 , we obtain @xmath323 whence @xmath324 putting together and , we see that we have two angles @xmath325 and @xmath326 , that we can assume to be in @xmath327 $ ] , which satisfy @xmath328 for some constants @xmath329 and @xmath330 . this immediately implies that there exist constants @xmath331 and @xmath332 with the same properties such that @xmath333 indeed , fix any @xmath334 and let @xmath335 such that @xmath336 for all @xmath337 ; then , let @xmath338 to be determined in a moment , and let @xmath339 be such that @xmath340 for all @xmath341 . now , if @xmath342 , then also @xmath343 as well as @xmath344 , whence @xmath345 conversely , if @xmath346 , then @xmath347 and @xmath348 . choosing @xmath349 , we have both @xmath350 and @xmath351 , whence @xmath352 this concludes the proof of the theorem . in order to establish the analogous property for the lgl grids , we need the following auxiliary result . [ lem : cgl ] for any @xmath305 , let @xmath353 and @xmath354 be , respectively , the @xmath38-th interval of the lgl grid and the cgl grid of the same order @xmath12 . then @xmath355 uniformly in @xmath38 and @xmath12 . for the lgl nodes , by we have for @xmath356 that @xmath357 thus @xmath358 , whence @xmath359 , i.e. , @xmath360 so that @xmath361 . since we already know that contiguous elements of the same grid have uniformly equivalent lengths , we obtain @xmath362 uniformly in @xmath38 and @xmath12 . in order to obtain the reverse bound , we use the following lower estimate from property [ lem : lgleilengths ] : @xmath363 and we already know from property [ prop : cglilengths ] that @xmath364 now , it is immediate to observe that for the given interval of variation of @xmath38 we have @xmath365 on the other hand , we write @xmath366 and we observe that @xmath367 whence @xmath368 since we have seen in the proof of the previous theorem that uniformly equivalent arguments imply uniformly equivalent sines , we conclude that @xmath369 and the proof of the lemma is complete for @xmath356 . for @xmath370 , we recall that @xmath371 and @xmath372 as seen above . this yields @xmath373 , whence the result for @xmath370 . finally , we observe that intervals are symmetrically placed around the origin . the following corollary finishes the proof of theorem [ thm : a ] ( ii ) . [ prop : lgl - uniequiv - leg ] assume that @xmath110 for some fixed constant @xmath111 . then , the lgl grid @xmath112 in the interval @xmath13 $ ] is locally @xmath103-uniformly equivalent to the grid @xmath374 , with @xmath113 and @xmath114 depending on @xmath36 but not on @xmath115 and @xmath12 . suppose that @xmath375 for some @xmath38 and @xmath322 ; then , recalling , the set @xmath376 has a non - empty intersection with the set @xmath377 , which implies that @xmath378 for some @xmath379 and @xmath380 . using theorem [ prop : lgl - uniequiv - cheb ] and lemma [ lem : cgl ] , as well as the fact that contiguous intervals of any cgl and lgl grid have uniformly comparable lengths ( see property [ property : cgl - uniform ] , theorem [ thm : a ] ( i ) ) , we obtain @xmath381 which is the claim . the main objective of this section is to construct for the family @xmath382 , of lgl grids an associated family @xmath383 , of dyadic grids that are locally @xmath103-uniformly equivalent with constants @xmath384 , independent of @xmath12 , which in addition offer the additional advantage of being nested . this construction is the essential basis for the proof of theorem [ thm : b ] . therefore , we formulate the construction for a general interval @xmath24 $ ] . although we shall be concerned in this section with more general classes of ordered grids @xmath385 $ ] , @xmath386 , we retain some notation used earlier only for lgl grids such as the notation @xmath387 for corresponding intervals . we always assume that @xmath8 is symmetric around the midpoint of @xmath24 $ ] and monotonic in the spirit of theorem [ theo : monotonicitylength ] . recall that we denote by @xmath388 the partition of subintervals of @xmath24 $ ] induced by @xmath8 ; conversely , @xmath389 denotes the grid defined by a partition @xmath97 of @xmath24 $ ] comprised of the endpoints of its intervals . the following notions will serve as important tools for the envisaged construction . we exploit the symmetry of @xmath8 and consider for any subinterval @xmath390 $ ] the largest and shortest overlapping subcell in @xmath91 @xmath391 note that due to the monotonicity property of @xmath8 and the restriction of @xmath132 to one half of the base interval , @xmath392 are always uniquely defined . for any interval @xmath393 in some dyadic partition @xmath97 with @xmath394 $ ] we shall denote by @xmath395 its parent interval which has @xmath393 as a subinterval created by splitting @xmath396 at its midpoint . the following inequalities will be useful in the sequel : @xmath397 it will be convenient to associate with a dyadic partition @xmath97 of @xmath24 $ ] the corresponding _ rooted binary tree _ @xmath398 whose nodes are those subintervals generated during the refinement process yielding @xmath97 . thus , @xmath399 has the interval @xmath24 $ ] as its _ root _ and the parent - child relation is given by the inclusion of intervals . those nodes in the tree that have no children are called _ leaves_. note that the binary tree @xmath399 associated with a dyadic grid is a _ full _ binary tree , i.e. , every node has either no or two child nodes in the tree . furthermore , note that the binary tree @xmath399 is full if and only if the set of leaves of @xmath399 is a partition of @xmath24 $ ] . given a real number @xmath400 , a grid @xmath8 on the interval @xmath24 $ ] and an initial dyadic grid @xmath401 , the following algorithm [ algo : dyadic ] creates a certain refined dyadic grid @xmath402 through suitable successive refinements of partitions . * for any @xmath407 , @xmath408 , one has @xmath409 * assume that @xmath410 satisfies @xmath411 then , for any @xmath412 , @xmath413 , one has @xmath414 hence , for any @xmath415\}}$ ] , the parent interval @xmath396 of length @xmath416 , whose halving produced @xmath393 , satisfies @xmath417 * whenever for each @xmath418 from the initial partition @xmath419\}}$ ] the parent @xmath396 of @xmath418 satisfies condition , then this is inherited by @xmath406 , i.e. , for each @xmath420\}$ ] the parent @xmath396 of @xmath393 also satisfies . it is easy to see that algorithm [ algo : dyadic ] terminates after finitely many steps . in fact , the maximal dyadic level of @xmath137 is @xmath421 where @xmath422 is the finest resolution in the original grid @xmath8 . indeed , for any dyadic cell @xmath393 of maximal level @xmath423 we have @xmath424 hence , such a cell can not be halved in the algorithm . we shall see next that @xmath425 and the original grid @xmath8 are still locally of comparable size whenever the initial dyadic grid @xmath401 satisfies a condition like , but with a potentially smaller constant @xmath426 , which will be important later . [ lem : dyadic_ver2 ] assume that holds for a given grid @xmath8 and assume that the initial dyadic partition @xmath410 satisfies the following condition : there exists some @xmath427 such that for all @xmath428\}$ ] , the parent @xmath429 satisfies @xmath430 then , the grid @xmath431 is locally @xmath103-uniformly equivalent to @xmath8 with @xmath432 , @xmath433 , i.e. , one has @xmath434 the lower bound follows directly from in remark [ rem : relations ] ( i ) . we shall show next that there exists an @xmath435 such that for any @xmath436 @xmath437 to see this , consider any @xmath438\}$ ] and recall from in remark [ rem : relations ] ( ii ) that its parent @xmath396 satisfies the inequality @xmath439 suppose first that @xmath396 intersects at most two intervals from @xmath91 ; these two intervals therefore have to be @xmath440 . using , and , we obtain @xmath441 i.e. , holds with @xmath442 . suppose next that one has @xmath443 . note that , if @xmath444 is not contained in @xmath396 , it must have a neighbor @xmath445 fully contained in @xmath396 . therefore , by and , @xmath446 which means that in this case is valid with @xmath447 . this concludes the proof of . the value of the parameter @xmath403 influences the deviation of the cardinality of @xmath431 from that of @xmath8 : the larger is @xmath403 , the smaller is the number of refinements of the initial grid @xmath401 induced by @xmath8 . when @xmath448 , a choice of @xmath403 around @xmath449 produces comparable cardinalities for @xmath431 and @xmath8 . this is clearly documented in figure [ fig : gridsizesdyadic ] for the lgl grids @xmath60 of increasing order @xmath12 . assume that @xmath461 are as in definition [ def : monotonicity ] ; suppose by contradiction that @xmath462 which means @xmath463 . since @xmath464 is full @xmath393 is not contained in @xmath465 , the parent of @xmath466 , and hence @xmath467 . since @xmath393 is a left neighbor of @xmath466 so that @xmath468 we infer from @xmath469 which is a contradiction . if a symmetric grid @xmath8 and an initial dyadic grid @xmath401 are locally quasi - uniform ( see ) , one readily infers from proposition [ lem : dyadic_ver2 ] that , due to the local uniform @xmath103-equivalence of @xmath8 and the dyadic grid @xmath137 , generated by the algorithm @xmath453 , the grid @xmath137 is also locally quasi - uniform . numerical experiments indicate that this can be further quantified in terms the following notion of _ gradedness_. [ lem : graded ] assume that @xmath8 is a symmetric , monotonic , and locally quasi - uniform grid . if @xmath401 is graded and if the constant @xmath100 from satisfies @xmath472 then the output @xmath137 of @xmath453 is graded . consider any two neighboring dyadic intervals @xmath473 , @xmath474 $ ] , where as before @xmath393 is the left neighbor of @xmath466 . by proposition [ lem : monotonicitydyadic ] , we know that @xmath475 and since the lengths of the dyadic intervals can only differ by powers of two , it remains to show that @xmath476 . suppose now to the contrary that @xmath477 when @xmath478 there is nothing to prove since either both @xmath479 , in which case contradicts the hypothesis on @xmath480 , or @xmath481 which means that the original right neighbor of @xmath393 in @xmath480 has been refined which also contradicts . now let @xmath482 denote the left neighbor of @xmath483 . from and we infer that @xmath484 thus , denoting by @xmath396 the parent of @xmath393 , we conclude that @xmath485 . hence , since @xmath486 , and the previous estimate assert that @xmath487 which is a contradiction for @xmath488 . this finishes the proof . in order to apply this to lgl grids we note first that the above argument is local in the following sense . again , considering by symmetry only the left half of the base interval , suppose that holds only for intervals of @xmath91 that are equal to or on the right of some @xmath489 . then the above argument implies gradedness of @xmath137 for all @xmath490 for which @xmath491 agrees with or is on the right of @xmath492 . recall that property @xmath213has been verified numerically to hold at least for @xmath497 . thus , we can invoke remark [ rem : smallestconstant ] and note that for @xmath226 given by and implies that for @xmath498 that condition is satisfied for all quotients not involving the outermost lgl intervals . the same arguments apply , on account of proposition [ prop : muldoon ] , for all @xmath148 provided that conjecture [ conj : quotients ] , holds . hence , it suffices to verify gradedness for the intervals adjacent to the left end point of the base interval . let @xmath499 share the left end point . then either @xmath418 has a right sibling of equal size in @xmath97 or equals half the base interval . in both cases gradedness holds trivially . the next observation is that the right neighbor @xmath500 of the parent @xmath396 of two siblings next to the left end point of the base interval must satisfy @xmath501 since otherwise these intervals can not belong to a partition that stems from successive dyadic splittings . the next possibility for breaking gradedness would be the transition to @xmath502 . but , again to be part of the leaf set of a dyadic tree one must have @xmath503 which again implies gradedness of @xmath504 . to obtain a jump of two levels at the left boundary of @xmath502 one must have that the two children @xmath505 also belong to @xmath97 , i.e. @xmath506 , @xmath507 . but this means ( since as in the proof of lemma [ lem : graded ] we can assume that @xmath508 ) that @xmath509 . therefore , since @xmath510 and @xmath511 , we see that @xmath512 does not contain the left end point of @xmath24 $ ] and hence is not an extreme interval . since under the assumption that conjecture [ conj : quotients ] holds for all interior intervals of @xmath97 lemma [ lem : graded ] applies , this finishes the proof . let us recall the stretching operator @xmath514}:[a,(a+b)/2]\rightarrow[a , b ] , x \mapsto 2x - a$ ] as defined in section [ sect : b ] . under the assumption that conjecture [ conj : lglcomparisonstretched ] is true , we can show that for @xmath515 and an lgl input grid @xmath60 , the dyadic grid generated by algorithm [ algo : dyadic ] is closed under stretching , in the sense of definition [ def : closedness ] . [ prop : stretched ] let @xmath515 , @xmath48 and @xmath516 , where @xmath401 is closed under stretching . then the validity of conjecture [ conj : lglcomparisonstretched ] implies that @xmath137 is closed under stretching . as usual , we set @xmath517 and @xmath410 . by affine invariance it suffices to consider @xmath24=[-1,1]$ ] , see remark [ rem : dyadicaffineinvariance ] . first note that since @xmath518)= [ -1,0]$ ] and @xmath519)= [ 0,1]$ ] , it suffices , on account of symmetry and monotonicity , to show that @xmath520 for any @xmath521 , @xmath522 $ ] . for @xmath523 there is nothing to show by our assumption on @xmath401 . suppose now that @xmath524 is a node that is split during the execution of algorithm @xmath525 , i.e. , due to the condition in line [ inalgo : split ] of algorithm [ algo : dyadic ] , we know that @xmath526 , where @xmath527 . since @xmath510 we have , in particular , @xmath528 . the assertion follows as soon as we have shown that the stretched version @xmath529 must also be split , i.e. , we have to show that @xmath530 to show this , let us observe first that it suffices to verify for @xmath522 $ ] . in fact , any @xmath531 $ ] gets mapped by @xmath532 into @xmath201 $ ] . it then follows from the monotonicity and symmetry of @xmath137 that the midpoint of such a @xmath393 must be contained in @xmath533 $ ] . so it remains to consider @xmath522 $ ] . first , there is nothing to show when the left end point of @xmath393 is @xmath534 , since then @xmath529 is the parent of @xmath393 . we may therefore assume that @xmath393 does not contain @xmath534 . by the above comment @xmath535 . clearly , @xmath536 contains the left end point of @xmath529 and therefore has to intersect @xmath537 . under the assumption that conjecture [ conj : lglcomparisonstretched ] is valid , it follows that @xmath538 which finishes the proof . [ [ construction - of - ensuremathcal - d_n - and - the - proof - of - theoremthmb ] ] construction of @xmath134 and the proof of theorem [ thm : b ] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ a first natural attempt to construct a dyadic grid associated with a given lgl grid @xmath60 would be to take @xmath539 for some @xmath540 $ ] . in fact , the initial dyadic grid @xmath541 trivially satisfies all the assumptions on @xmath401 used in the derivation of the various properties above . unfortunately , although this seems to occur very rarely , the grids @xmath542 are _ not _ always nested , as shown by numerical evidence . for instance , for @xmath543 , the first pair @xmath544 of polynomial degrees where such dyadic grids are not nested , occurs for @xmath545 and @xmath546 . for corresponding extensive numerical studies and further examples of non - nestedness , we refer the reader to @xcite . by construction , the grids @xmath134 are nested . moreover , one inductively concludes from the results of the preceding sections that the @xmath134 are symmetric and monotonic . they are also closed under stretching for any range of degrees @xmath12 for which property @xmath548holds . moreover , they are also graded ( beyond any numerically confirmed range of @xmath12 ) if conjecture [ conj : quotients ] is valid . a little care must be taken to confirm the desired locally @xmath103-uniform equivalence of @xmath134 with @xmath60 . again the lower inequality is ensured by , see remark [ rem : relations ] ( i ) . as for the upper inequality , a certain obstruction lies in the fact that is not necessarily inherited for the specific value @xmath403 . in fact , the following situation may occur which again is a consequence of the fact that intervals in lgl grids not only decrease in size but also move outwards with increasing degree . let for a given ( dyadic ) interval @xmath393 , @xmath549 be the @xmath322-th interval in @xmath550 . then it could happen that @xmath393 no longer intersects the @xmath322-th interval in @xmath551 , i.e. @xmath552 is the @xmath553-st interval in @xmath551 and may therefore have larger size than @xmath554 . as a consequence @xmath555 is not necessarily true . nevertheless , the following can be shown . [ lem : dyadic-2 ] for all @xmath556 , the dyadic grids @xmath557 produced by algorithm [ algo : dyadicnested ] are locally @xmath103-uniformly equivalent to @xmath60 with constants @xmath384 specified below : @xmath558 furthermore , we have @xmath559 due to remark [ rem : dyadicaffineinvariance ] we can assume without loss of generality that @xmath24=[-1,1]$ ] and , by symmetry , consider only intervals in the left half @xmath200 $ ] . to be able to apply proposition [ lem : dyadic_ver2 ] , we shall exploit the fact that with increasing @xmath12 outward moving intervals decrease in size . more precisely , by theorem [ theo : displgl2 ] , one has for any @xmath560 @xmath561\in { \ensuremath{{\cal p}}}_n,\,\ \ \delta^{n+m}_j= [ \xi^{n+m}_{j-1},\xi^{n+m}_{j}]\in { \ensuremath{{\cal p}}}_{n+m},\\ \xi^{n+m}_{j-1 } \leq \xi^n_{i-1 } \quad \implies~ { \ensuremath{|\delta^{n+m}_j| } } \leq { \ensuremath{|\delta^n_i|}}\;. \end{aligned}\ ] ] now , on account of proposition [ lem : dyadic_ver2 ] , the assertion follows as soon as we have shown that holds with @xmath562 . to that end , suppose @xmath563 is not subdivided in @xmath564 . without loss of generality we may assume that @xmath565\}$ ] . thus , @xmath393 must have been created by splitting @xmath566 for some @xmath560 . by the condition in line [ inalgo : split ] of algorithm [ algo : dyadic ] , @xmath567 satisfies @xmath568 . now let @xmath569 be the right neighbor of @xmath570 , so that by , @xmath571 . one readily concludes from the monotonicity of the intervals in the lgl grids , see theorem [ theo : monotonicitylength ] , that the left end point of @xmath572 must be smaller than the left end point of @xmath457 . therefore , we infer from that @xmath573 which is . this finishes the proof of . finally , follows immediately from this result . in summary , all the claims stated in theorem [ thm : b ] have been verified to be satisfied by the dyadic grids @xmath134 , generated by algorithm [ algo : dyadicnested ] , which completes the proof of theorem [ thm : b ] . ( solid lines ) as a function of the size of @xmath374 for different values of @xmath403 . the sizes of the dyadic grids as in figure [ fig : gridsizesdyadic ] are also given for a comparison ( marks ) . the black graph indicates the line of equal cardinalities of lgl and dyadic meshes.,width=287 ] in figure [ fig : gridsizesdyadicnested ] the sizes of the nested dyadic grids @xmath134 are plotted against the sizes of the lgl grids @xmath374 for different values of @xmath403 . we observe that usually only very few points , in comparison with running just @xmath574 , are added to the grid to ensure nestedness . this work was supported in part by the dfg project `` optimal preconditioners of spectral discontinuous galerkin methods for elliptic boundary value problems '' ( da 117/23 - 1 ) , the excellence initiative of the german federal and state governments ( rwth aachen seed funds , distinguished professorship projects , graduate school aices ) , and nsf grant dms 1222390 . brix , k. : robust preconditioners for @xmath577-discontinuous galerkin discretizations for elliptic problems . doctoral thesis , institut fr geometrie und praktische mathematik , rwth aachen ( in preparation ) . brix , k. , campos pinto , m. , canuto , c. , dahmen , w. : multilevel preconditioning of discontinuous - galerkin spectral element methods , part i : geometrically conforming meshes . ( submitted ) . arxiv : http://arxiv.org / abs/1301.6768[1301.6768 [ math.na ] ] . brix , k. , canuto , c. , dahmen , w. : robust preconditioners for dg - discretizations with arbitrary polynomial degrees . in : erhel , proceedings of the 21st international conference on domain decomposition methods , rennes , france , june 25th29th , 2012 . springer verlag , heidelberg ( to appear ) . arxiv : http://arxiv.org / abs/1212.6385[1212.6385 [ math.na ] ] . canuto , c. : stabilization of spectral methods by finite element bubble functions . methods appl . . eng . * 116 * , 1326 ( 1994 ) . doi : http://dx.doi.org/10.1016/s0045-7825(94)80004-9[10.1016/s0045-7825(94)80004-9 ] jordaan , k. and tookos , f. : convexity of the zeros of some orthogonal polynomials and related functions . j. comput . appl . math . * 233*(3 ) , 762767 ( 2009 ) . doi : http://dx.doi.org/10.1016/j.cam.2009.02.045[10.1016/j.cam.2009.02.045 ] reiser , s. : nullstellen der besselfunktionen erster art @xmath578 , @xmath230 und @xmath579 i m intervall @xmath580 $ ] auf @xmath581 stellen gerundet . http://stefanreiser.de/besjzero.txt ( 2010 ) . accessed 01 october 2013 xu , j. : the auxiliary space method and optimal multigrid preconditioning techniques for unstructured grids . computing * 56 * , 215235 ( 1996 ) . doi : http://dx.doi.org/10.1007/bf02238513[10.1007/bf02238513 ]

the coma cluster was observed with an x - ray imaging spectrometer ( xis ) onboard suzaku in six pointings , including the central x - ray peak region , the 14 west offset region , 30 and 34 north - west offset regions , and 44 and 60 south - west offset regions . owing to its lower background level , suzaku has better sensitivity to fe k lines than other satellites . using precise fe line measurements , we studied the temperature structure , possible bulk motions , and iron abundance distributions in the intracluster medium ( icm ) . the observed spectra were well - represented by a single - temperature model , and two- or three- temperature model did not improve substantially . the temperature , derived from k line ratios of h - like and he - like fe , agree with those derived from the single - temperature model . because the line ratio is a steep function of temperature , the consistency supports the accuracy of temperature measurements conducted with suzaku . within the 34 region , the redshift derived from the central energy of the he - like fe line is consistent with that from optical observations , within a calibration error of 18 ev or 818 km s in the line of sight . this value is smaller than the sound velocity of icm , which is 1500 km s . the central energy of fe lines at the 44 offset region around the ngc 4839 subcluster is also consistent with those within the 34 region . these results on the temperature and velocity structure suggest that the core of the cluster is in a relaxed state , and non - thermal electrons relevant to the radio halo are accelerated by intracluster turbulence rather than large - scale shocks . fe abundance is almost constant at 0.4 solar within the 34 region , and decreases with radius . this value is slightly lower than those of other clusters , which means the gas have been mixed well during a past merger associated with the growth of the cluster .

1109.0154

clusters of galaxies are thought to grow into larger systems through complex interactions between smaller systems . signatures of merging events include temperature and density inhomogeneities and bulk motions in the intracluster medium ( icm ) . numerical simulations predict simplified bulk motions with a substantial fraction of virial velocity ( @xmath41000 km s@xmath3 ) , lasting several gyr after each merging event ( e.g. , @xcite ; @xcite ) . measurements of the icm temperature and velocity structure , therefore , provide crucial information for understanding the evolution of clusters . velocity measurements of bulk motion and turbulence are essential for mass estimation and cosmological studies . if icm has a significant bulk velocity compared to its thermal velocity , the associated non - thermal pressure would threaten the assumption of hydrostatic equilibrium in deriving the total gravitational mass in a cluster . for example , factors of 23 discrepancies between the x - ray and lensing mass estimates in some objects ( e.g. , @xcite ; @xcite ) could be partly due to this effect . @xcite discovered a significant bulk motion in abell 2256 , which is a well known merging cluster . @xcite performed a fourier analysis of xmm - newton data of the coma cluster , which revealed the presence of a scale - invariant pressure fluctuation ranging between 29 and 64 kpc , and @xmath410% of the total pressure in turbulent form . however , in relaxed clusters , significant bulk motions and turbulence have not been detected . on the basis of asca and chandra data , @xcite claimed a large velocity gradient of @xmath42400 km s@xmath3 in the centaurus cluster , which is a relaxed cluster with a cool core . however , @xcite found a negative result of @xmath5 km s@xmath3 on the basis of suzaku observations of the centaurus cluster . @xcite also searched for possible bulk motions in the awm 7 cluster with suzaku , and observed no significant bulk motions . the upper limit of gas velocity was @xmath6 2000 km s@xmath3 . @xcite do not detect the variation of the redshift of the icm in the ophiuchus cluster , and the upper limit of the velocity difference is 3000 km s@xmath3 . @xcite studied the icm flow in abell 2319 with suzaku , and detected no velocity differences within the observed region . using spectra with a reflection grating spectrometer ( rgs ) onboard xmm - newton , @xcite provided upper limits on turbulence velocities in cluster core regions : at least 15 sources had less than 20% of the thermal energy density in turbulence . the coma cluster ( @xmath7 ) , also known as abell 1656 , is one of the most studied clusters @xcite and has been observed at all wavelengths from radio to hard x - ray bands . however , the physical state , particularly the dynamical state , of the coma cluster has not been completely understood . measurements of the velocity dispersion and distribution of galaxies belonging to the coma cluster constrain the dynamical history of the coma cluster . @xcite found that two central galaxies , ngc 4889 and ngc 4874 have a significant difference in velocity , which is evidence of a recent merger . @xcite argued that ngc 4874 was the original dominant galaxy of the main cluster , and that ngc 4889 belonged to a subgroup , that recently merged with the main cluster . a substructure associated with the galaxy , ngc 4839 , located 40@xmath0 south - west of the cluster core , was found by @xcite and @xcite . the existence of this subcluster was confirmed by @xcite . the thermal x - ray emission of the coma cluster has been observed with several x - ray satellites . @xcite constructed an x - ray map of the coma cluster using rosat all - sky - survey data and determined the x - ray surface brightness profile of the cluster out to a radius of roughly 100@xmath0 . the temperature maps of clusters provide knowledge about the history of past subcluster mergers . on the basis of asca observations , @xcite and @xcite found that icm is not isothermal . @xcite used xmm - newton to study the temperature structure in the central region of the coma cluster . the projected temperature distribution around ngc 4889 and ngc 4874 is remarkably homogeneous , which suggests that the core is mostly in a relaxed state . except at the center , the temperature decreases slightly with radius . a cool filament of an x - ray emission in the direction of the galaxy ngc 4911 , which is located in the south - east of the cluster center , was detected with chandra @xcite and xmm @xcite . rosat observed a substructure around the ngc 4839 subgroup @xcite , and a hot region in the direction of ngc 4839 was observed by @xcite . by xmm - newton observations , @xcite found compelling evidence for the subgroup around ngc 4839 was on its first infall into the coma cluster , but it had not passed its core yet . this interpretation is different from that of @xcite , who suggested that the ngc 4839 group was moving out of the cluster after having already passed through the coma cluster . non - thermal electrons , observed via diffused radio synchrotron emission , have been detected in more than 50 clusters , and all of them are undergoing mergers ( e.g. , @xcite ; @xcite ) . the coma cluster also has a cluster - wide synchrotron radio halo , emitted by relativistic electrons due to merging @xcite . recent non - thermal detections have been claimed by @xcite with rxte and by @xcite with bepposax , although the latter detection is controversial @xcite . long observations ( @xmath41 ms ) by integral have imaged extended diffuse hard x - ray emission from the coma cluster , although it was found to be completely consistent with thermal emission @xcite . using the temperature map of icm obtained with the pn detector onboard xmm - newton , @xcite analyzed data obtained from the suzaku hard x - ray detector ( hxd ) @xcite and derived the strongest upper limit of non - thermal emission . this paper reports results from six suzaku observations of the coma cluster out to 60@xmath0 , @xmath81.5 mpc , conducted using the x - ray imaging spectrometer ( xis ) @xcite onboard suzaku @xcite . owing to its lower background level , suzaku has better sensitivity to the iron k@xmath1 lines than previous satellites . in addition , the accurate calibration of xis allows us to measure the velocity of icm precisely . we used the hubble constant , @xmath9 km s@xmath3 mpc@xmath3 . the distance to the coma cluster is @xmath10 mpc , and @xmath11 corresponds to 28.9 kpc . we used the abundance ratio by @xcite , in which the solar fe abundance relative to h is 2.95@xmath1210@xmath13 . errors are quoted at a 90% confidence level for a single parameter . the remainder of this paper is organized as follows : we present the observations in section 2 and describe the data analysis in section 3 . we describe the results of spectral fittings and temperature structure in subsection 4.1 , bulk motions in subsection 4.5 , and the fe abundance in subsection 4.6 . in section 5 , we discuss the results , and in section 6 , we summarize the findings . cccccccc field name & target name & sequence & observation & ra & dec & exposure & distance from the + & & number & date & ( deg ) & ( deg ) & ( ks ) & x - ray peak + center & coma radio halo & 801097010 & 2006 - 05 - 31 & 194.9267 & 27.9061 & 150 & + offset & coma cluster offset & 801044010 & 2006 - 05 - 30 & 194.6939 & 27.9466 & 79 & + offset & coma 11 & 802082010 & 2007 - 06 - 19 & 194.6305 & 28.3939 & 53 & + offset & coma bkg2 & 802084010 & 2007 - 06 - 21 & 194.3428 & 28.1403 & 30 & + offset & coma 45 & 802047010 & 2007 - 12 - 02 & 194.2558 & 27.5714 & 30 & + offset & coma 60 & 802048010 & 2007 - 12 - 04 & 194.0251 & 27.4252 & 35 & + offset & coma bkg & 802083010 & 2007 - 06 - 21 & 198.7472 & 31.6480 & 30 & + ( 80 mm , 60mm)figure1_a.eps ( 80 mm , 60mm)figure1_b.eps suzaku carried out six observations in the coma cluster . the details of the observations are summarized in table [ tab : ob ] . two central observations were carried out in may 2006 , during the suzaku phase - i period . the first observation , `` coma radio halo '' , has a pointing that is @xmath14 offset from the x - ray peak of the coma cluster , which has coordinates ( ra , dec ) = ( 194.9367 , 27.9472 ) in degrees . the second observation , `` coma cluster offset '' , was located @xmath15 west offset from the first observation . the other observations , in the @xmath16 and @xmath17 north west offset regions and @xmath18 and @xmath19 south west offset regions , were carried out in 2007 , during the suzaku phase - ii period . to eliminate background emissions , archival suzaku data with a 5 degree offset from the coma cluster , observed in june 2007 was also used , and is shown in table [ tab : ob ] . xis was operated in its normal mode during observations . the xis instrument consists of four sets of x - ray ccd ( xis0 , 1 , 2 , and 3 ) . xis1 is a back - illuminated ( bi ) sensor , while xis0 , 2 , and 3 are front - illuminated ( fi ) sensors . xis2 was not used during the four offset observations beyond 30@xmath0 from the x - ray peak . figure [ fig : image ] shows a 0.54.0 kev image of xis0 . we performed data reduction using heasoft version 6.6.3 . xis event lists created by the rev 2.0 pipeline processing were filtered using the following additional criteria : the geomagnetic cut - off rigidity ( cor2 ) @xmath20 6 gv , and elevation angle from the earth limb @xmath21 . the data formats of 5x5 and 3x3 editing modes were added . the exposure times after the data selection are shown in table [ tab : ob ] . to study the temperature structure and search for possible bulk motions of icm in the coma cluster on the scale of a few arcminutes , we divided the 18@xmath22 square xis field of view ( fov ) of the center and 14@xmath0 offset regions into @xmath23 and @xmath24 cells , respectively , as shown in figure [ fig : image ] . each region in the center and @xmath15 offset regions subtends @xmath12 ( 117@xmath25 kpc ) , and 9@xmath26 ( 234@xmath27 kpc ) , respectively . regions around calibration sources were excluded . luminous point sources in which the flux limit corresponds to @xmath4 10@xmath28 counts s@xmath3 in the 2.010.0 kev energy range were also excluded as circular regions with a radius of 1@xmath0 . at most 10 point sources were excluded for each of the observed regions . we extracted spectra from the whole xis fov for regions more than @xmath29 from the x - ray peak . the non x - ray background ( nxb ) was subtracted from each spectrum using a database of night earth observations with the same detector area and cor distribution @xcite . we included the degradation of energy resolution due to radiation damage in the redistribution matrix file ( rmf ) generated by the xisrmfgen ftools task . we also created an ancillary response file ( arf ) using the xissimarfgen ftools task @xcite . a decrease in the low - energy transmission of the xis optical blocking filter ( obf ) was also included in arf . to generate arf files for the center and 14@xmath0 offset regions , we used a @xmath30-model profile for the simulated surface brightness profile in @xcite . for @xmath16 , @xmath17 , @xmath18 , @xmath19 , and 5 degree offset regions , we generated arfs assuming a uniformly extended emission from an encircled region with a 20@xmath0 radius , because the gradient of surface brightness within each fov is small . we used the xspec_v11.3.2ag package for spectral analysis . we have investigated the temperature structure , bulk motions , and iron abundance distribution of the coma cluster . we fitted the spectra with a thermal plasma model ( apec : @xcite ) to obtain the temperature , redshift and fe abundance of icm . a detailed spectral analysis is presented in subsection [ subsec : single ] . we used three gaussian models to obtain the temperature using the normalization ratio of k@xmath1 lines of h - like and he - like fe , and redshift using the shift of line centroid energy , which is presented in subsection [ subsec : ratios ] . ( 80 mm , 60mm)figure2.eps cccccccc & & & + @xmath31 & normalization & temperature & normalization & temperature & normalization & + & & ( kev ) & @xmath32 & ( kev ) & @xmath33 & @xmath2/d.o.f & @xmath2/d.o.f + @xmath34 & @xmath35 & 0.07 ( fixed ) & @xmath36 & @xmath37 & @xmath38 & 188/141 & 223/154 + first , we fitted the spectra in the 5@xmath39 offset region to determine the local x - ray background . as shown in @xcite , the background emission of suzaku xis can be fitted with a three component model : two thermal plasma models ( apec : @xcite ) for the local hot bubble ( lhb ) and the milky way halo ( mwh ) , and a power - law model for the extragalactic cosmic x - ray background ( cxb ) . mwh and cxb components were convolved with an absorption in the galaxy . we , therefore , fitted the spectra simultaneously in the 0.58.0 kev energy range for xis fis and the 0.358.0 kev energy range for xis bi using the following model formula : @xmath40-@xmath41 . we assumed a zero redshift and a solar abundance for lhb and mwh components . the temperature of lhb was fixed at 0.07 kev @xcite , while the temperature of mwh was variable . the column density of the galactic neutral hydrogen was fixed at 1.0@xmath42 @xmath43 @xcite . to avoid systematic uncertainties in the background , we ignored the 1.401.55 kev energy range , where a relatively large uncertainty exists because of an instrument al line @xcite , and energies above 8 kev . we also excluded the narrow energy band between 1.82 and 1.84 kev in the fits because of incomplete responses around the si edge . results of the spectral fit are shown in figure [ fig : bkgfit ] and the resultant parameters of the fit are shown in table [ tab : bkgtable ] . the derived photon index and normalization of the cxb component agree with @xcite and @xcite . the derived temperature and normalizations of the two thermal models are consistent with those obtained from xmm and suzaku observations @xcite . we generated the simulated spectra of the x - ray background including lhb , mwh , and cxb emissions derived from the resultant parameters of the fits in the 5@xmath39 offset region . we used the simulated spectra as the background to fit the spectra for icm of the coma cluster . figure [ fig : bgd ] shows the comparison between the xis0 spectra of the 60@xmath0 offset region of the coma cluster and backgrounds . above 7 kev , nxb dominates the observed spectra of the 60@xmath0 offset region . we also show the background - subtracted spectrum of the 60@xmath0 offset region of the coma cluster in figure [ fig : bgd ] ( blue ) . the x - ray background spectra are well reproduced . ( 80 mm , 60mm)figure3.eps we fitted the xis spectra in each region of the coma cluster using a single temperature model ( apec ) with the galactic absorption , @xmath44 . each spectral bin contained 50 or more counts . we fitted the spectra in two energy bands : 1.08.0 and 5.08.0 kev . to avoid systematic uncertainties in the background , we ignored energies higher than 8 kev . we also excluded the narrow energy band between 1.82 and 1.84 kev in the fits because of incomplete responses around the si edge . temperature , redshift and normalization of the single temperature model were free parameters , and @xmath44 was fixed to the galactic value , @xmath45 , in the direction of the coma cluster . abundances of he , c , n , and al were fixed to be a solar . we divided the other metals into three groups : o , ne and mg ; si , s , ar and ca ; and fe and ni , and allowed them to vary . the resultant parameters are summarized in table [ tab:1 - 8vapecresults ] , and a sample of the spectra is shown in figure [ fig : spec ] . since our data have very high statistics , especially in the central region , there remain residual structures around the si edge structure . these residuals are probably caused by systematic uncertainties in the response matrix because the bi and fi detectors gave different residuals . therefore , to avoid systematic uncertainties in all regions , we excluded the energy range of 1.72.3 kev and fitted the spectra again . the derived temperatures hardly changed , although the reduced @xmath2 decreased by 20% . the resultant @xmath2 are summarized in table [ tab:1 - 8vapecresults ] . to constrain the temperature structure , we fitted the spectra of the center and offset regions with a two - temperature ( apec ) model in the energy range of 1.08.0 kev , ignoring the energy range of 1.72.3 kev . the @xmath46 are improved slightly by less than a few percent . the derived temperatures of one component are similar to those derived from the single - temperature model fits , and those of the other component are either above 10 kev or below 2 kev . to obtain more detailed constraints for the lower and higher temperature components , we fitted the spectra with a three - temperature ( apec ) model in the energy range of 1.08.0 kev , excluding the 1.72.3 kev energy range to avoid uncertainties in the response matrix . in this model , the lowest , middle and highest temperatures are restricted to below 6 kev , within 6.010.0 kev , and above 10.0 kev , respectively . the resultant @xmath46 values are shown in table [ tab:1 - 8vapecresults ] . to derive the fe abundance and velocity of the bulk - motion from fe - k lines , we fitted the spectra of each region with the single - temperature apec model in an energy range of 5.08.0 kev . the results are shown in table [ tab:1 - 8vapecresults ] . to precisely determine the bulk velocity of icm , the accuracy of photon energy measurements is crucial . some sensors have large uncertainties in the determination of the line centroid energy . thus , we averaged the spectrum over fov in ( i ) all detectors , and ( ii ) all detectors except for different mn k@xmath1 line centroid energies to check the spectral shape and measured redshift in the center and ngc 4839 regions ( 44@xmath0 offset region ) . cccccccccc & & & gaussians + region & @xmath47 & @xmath46/d.o.f . & @xmath46/d.o.f . & @xmath46/d.o.f . & @xmath46/d.o.f . & fe & redshift & @xmath46/d.o.f . & @xmath47 from line + & [ kev ] & ( 1 t ) & ( 1 t ) & ( 2 t ) & ( 3 t ) & [ solar ] & [ @xmath48 & & ratio [ kev ] + center & & & & & & & & & + @xmath49 & @xmath50 & @xmath51 & @xmath52 & @xmath53 & @xmath54 & @xmath55 & @xmath56 & @xmath57 & @xmath58 + @xmath59 & @xmath60 & @xmath61 & @xmath62 & @xmath63 & @xmath64 & @xmath65 & @xmath66 & @xmath67 & @xmath68 + @xmath69 & @xmath70 & @xmath71 & @xmath72 & @xmath73 & @xmath74 & @xmath75 & @xmath76 & @xmath77 & @xmath78 + @xmath79 & @xmath80 & @xmath81 & @xmath82 & @xmath83 & @xmath84 & @xmath85 & @xmath86 & @xmath87 & @xmath88 + @xmath89 & @xmath90 & @xmath91 & @xmath92 & @xmath93 & @xmath94 & @xmath95 & @xmath96 & @xmath97 & @xmath98 + @xmath99 & @xmath100 & @xmath101 & @xmath102 & @xmath103 & @xmath104 & @xmath105 & @xmath106 & @xmath107 & @xmath108 + @xmath109 & @xmath110 & @xmath111 & @xmath112 & @xmath113 & @xmath114 & @xmath115 & @xmath116 & @xmath117 & @xmath118 + @xmath119 & @xmath120 & @xmath121 & @xmath122 & @xmath123 & @xmath124 & @xmath125 & @xmath126 & @xmath127 & @xmath128 + @xmath129 & @xmath130 & @xmath131 & @xmath112 & @xmath132 & @xmath133 & @xmath134 & @xmath135 & @xmath136 & @xmath137 + @xmath138 & @xmath139 & @xmath140 & @xmath141 & @xmath142 & @xmath143 & @xmath144 & @xmath145 & @xmath146 & @xmath147 + @xmath148 & @xmath149 & @xmath150 & @xmath151 & @xmath123 & @xmath152 & @xmath153 & @xmath154 & @xmath155 & @xmath156 + @xmath157 & @xmath158 & @xmath159 & @xmath160 & @xmath161 & @xmath162 & @xmath163 & @xmath164 & @xmath165 & @xmath166 + @xmath167 & @xmath168 & @xmath169 & @xmath170 & @xmath171 & @xmath172 & @xmath173 & @xmath174 & @xmath175 & @xmath176 + @xmath177 & @xmath178 & @xmath179 & @xmath180 & @xmath181 & @xmath182 & @xmath183 & @xmath184 & @xmath185 & @xmath186 + offset & & & & & & & & & + @xmath187 & @xmath188 & @xmath189 & @xmath190 & @xmath191 & @xmath192 & @xmath193 & @xmath194 & @xmath97 & @xmath195 + @xmath196 & @xmath197 & @xmath198 & @xmath199 & @xmath200 & @xmath201 & @xmath202 & @xmath203 & @xmath204 & @xmath205 + @xmath206 & @xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath215 + @xmath216 & @xmath217 & @xmath218 & @xmath219 & @xmath220 & @xmath221 & @xmath222 & @xmath223 & @xmath224 & @xmath225 + offset & @xmath226 & @xmath227 & @xmath228 & @xmath229 & @xmath230 & @xmath231 & @xmath232 & @xmath233 & @xmath234 + offset & @xmath235 & @xmath236 & @xmath237 & @xmath238 & @xmath239 & @xmath240 & @xmath241 & @xmath242 & @xmath243 + offset & @xmath244 & @xmath245 & @xmath246 & @xmath247 & @xmath248 & @xmath249 & @xmath250 & @xmath251 & @xmath252 + offset & @xmath253 & @xmath254 & @xmath255 & @xmath256 & @xmath257 & @xmath258 & @xmath259 & @xmath260 & @xmath261 + ( 80 mm , 60mm)figure4a.eps ( 80 mm , 60mm)figure4b.eps ( 80 mm , 60mm)figure4c.eps ( 80 mm , 60mm)figure4d.eps ( 80 mm , 60mm)figure4e.eps ( 80 mm , 60mm)figure4f.eps to derive the icm temperature from the normalization ratio of k@xmath1 lines of h - like and he - like fe , and the possible bulk motions of icm from the central energy of the lines , we fitted the spectra in the energy range of 5.08.0 kev with a sum of bremsstrahlung and three gaussian components , which correspond to k@xmath1 lines of he - like and h - like fe , and a mixture of the k@xmath1 line of he - like ni and the k@xmath30 line of he - like fe . the temperature and normalization of the bremsstrahlung component , and the central energy and normalizations of the gaussian components were free parameters , while the line width of gaussian models was fixed at 0 . the best - fit spectra are shown in figure [ fig : bremss_spec ] . ( 80 mm , 60mm)figure5a.eps ( 80 mm , 60mm)figure5b.eps ( 80 mm , 60mm)figure5c.eps ( 80 mm , 60mm)figure5d.eps ( 80 mm , 60mm)figure5e.eps ( 80 mm , 60mm)figure5f.eps ( 80mm,60mm)figure6a.eps ( 80mm,60mm)figure6b.eps figure [ fig : kt_apec ] shows a radial profile and map of the derived temperatures from the spectral fits with the single temperature ( apec ) model in the energy range of 1.08.0 kev . the radial temperature profile of the coma cluster is relatively flat within @xmath4500 kpc , which corresponds to the 14@xmath0 offset region , and decreases with the radius to the outer region . while the regions observed with suzaku are limited , the resultant temperature profile is similar to that derived from xmm - newton observations , which cover nearly the entire coma cluster out to @xmath41200 kpc @xcite . excluding the south - east quadrant ( cell numbers : 2,3,6 , and 7 ) , and the 14@xmath0 offset region , temperatures of cells in the center region have a flat distribution within a small range between @xmath48 kev and @xmath49 kev . within 300 kpc from the x - ray peak in the center region , the weighted mean temperature with statistical errors of the 16 cells is [email protected] kev . the south - east quadrant ( cell numbers : 2,3,6 , and 7 ) in the center region has a slightly cooler temperature : the weighted average with statistical errors of the four cells is [email protected] kev . this value is close to that of 5.07.0 kev in the cool temperature region observed by xmm - newton , which would be a counterpart of the south - east quadrant region @xcite . on the other hand , the north - west quadrant in the center region ( cell numbers : 8,13 , and 14 ) and the nearby area of the 14@xmath0 offset region ( cell numbers : 17,18 ) have slightly higher temperatures of @xmath49 kev . the 34@xmath0 offset region located to the north - west of the 14@xmath0 offset region , also has a relatively high temperature of @xmath489 kev . this direction corresponds to the hot spot , @xmath263 kev in @xcite with asca , and @xmath2640.4 kev in @xcite with xmm - newton . recent integral observations have shown a surface brightness excess relative to xmm observations , which is well - represented by the extended hot thermal emission with @xmath265 kev @xcite . the temperature in the 30@xmath0 offset region is around @xmath467 kev , and is clearly cooler than temperatures in the inner region . this value is consistent with the previous results in @xcite , and @xcite . the 44@xmath0 and 60@xmath0 offset regions include the ngc 4839 subcluster , and temperatures decreases with radius to [email protected] kev and [email protected] kev , respectively . these values are cooler than those in @xcite with asca , but are consistent with those in @xcite and @xcite with xmm - newton . temperatures of the core and tail of ngc 4839 with xmm - newton , which is located in the 44@xmath0 offset region with suzaku , are @xmath266 kev and @xmath267 kev , respectively @xcite . the temperature of the main body of the subcluster is @xmath268 kev @xcite . suzaku provided more significant temperatures than previous measurements in these regions . in summary , the temperature distribution observed with suzaku is fairly consistent with the previous results observed with asca , xmm and integral . in addition , we note that statistical and systematic errors with suzaku are much smaller than those in previous results . ( 80mm,60mm)figure7a.eps ( 80mm,60mm)figure7b.eps as shown in figure [ fig : spec ] , the spectra in the energy range of 1.08.0 kev are well - represented by the single - temperature model . except for discrepancies between fi and bi detectors around the si - edge , there are no systematic residuals . figure [ fig:3tchi ] shows a comparison of the resultant @xmath2 fitted with single and three temperature ( apec ) models in the energy range of 1.08.0 kev , excluding the 1.72.3 kev energy range . a more statistically rigorous approach would be to apply the f - test to compare single - temperature and three temperature models . the f - test indicates that the three - temperature model is significantly better than the single - temperature model in regions 1 , 5 , 6 , 7 , 9 , 10 , and 11 of the center region and 17 of the 14@xmath0 offset region , and better in regions 2 , and 14 of the center region , and 16 , and 19 of the 14@xmath0 offset region , 30@xmath0 offset region , 44@xmath0 offset region , and 60@xmath0 offset region . the largest f statistic value is 11.4 in cell number 7 of the center region , and the least f statistic value is 0.238 in cell number 8 of the center region . however , the single - temperature model fit still represents the spectra fairly well . ( 80mm,60mm)figure8a.eps ( 80mm,60mm)figure8b.eps the spectra accumulated over each region may contain multi temperature components , although the single temperature ( apec ) model represented those spectra well . because the ratio of he - like and h - like fe k@xmath1 lines strongly depends on the plasma temperature , comparisons of temperatures derived from the line ratio and spectral fits with the thermal model are useful for understanding the temperature structure of icm . figure [ fig : kt_ratio ] shows the ratios of the resultant normalizations of the fe k@xmath1 lines , plotted versus the temperature derived from the spectral fits with the single temperature ( apec ) model in the energy range of 1.08.0 kev . to convert line ratios to plasma temperature , we generated mock spectra assuming the single temperature apec or mekal @xcite plasma code models , convolved with xis energy resolution . we then fitted the mock spectra with a sum of a bremsstrahlung and three gaussian models . the theoretical line ratios from apec and mekal plasma codes are also plotted in figure [ fig : kt_ratio ] . the derived line ratios agreed with these theoretical line ratios of the apec plasma code assuming the single temperature model , while the mekal model gave a line ratio value several % lower at a given temperature . assuming a single temperature plasma , we converted the observed line ratios to plasma temperatures using the theoretical relation for the apec model . results are shown in table [ tab:1 - 8vapecresults ] and figure [ fig : kt_ratio ] . as shown in figure [ fig : kt_ratio ] , line temperatures agreed with those derived from the spectral fits very well . ( 80 mm , 60mm)figure9a.eps ( 80 mm , 60mm)figure9b.eps @xcite claimed a significant temperature difference of icm at the edge of the radio halo on the basis of xmm - newton observations : the temperature at the inner edge of the radio halo is [email protected] kev , while the temperature at the outer edge of the radio halo is [email protected] kev . to examine this temperature difference , we extracted spectra from counterpart regions in the 34@xmath0 offset region , which are represented by two red boxes in figure [ fig : fino ] , and fitted the spectra with the single temperature ( apec ) model . as shown in figure [ fig : fino ] , there are no significant temperature differences for each box in the 34@xmath0 offset region . we then divided fov of the 34@xmath0 offset region into 2@xmath269 cells as shown in figure [ fig : fino ] , and fitted the spectra with the single temperature ( apec ) model . the derived temperatures are also consistent with each other . to constrain the line - of - sight gas motion using the he - like fe - k@xmath1 line , precise calibration of the xis energy scale is crucial . because the six data sets used in the present analysis have three different observation epochs ( may 2006 , june 2009 and december 2009 ) , it is necessary to check the energy scale for each observation . in addition , the energy gain may vary from place to place on the same ccd chip owing to the charge transfer inefficiency . we thus estimate the xis calibration uncertainty in the following two ways : ( i ) measurements of the calibration source energy and ( ii ) comparisons of redshift values derived from four or three sets of xis . \(i ) the fiducial absolute energy scale can be examined by measuring the centroid energy of the mn k@xmath1 line from built - in calibration sources , which illuminate two corners of each xis chip . the spectra accumulated from the calibration source regions were fitted to the power - law model for continuum and two gaussian functions for the mn k@xmath1 line at 5.894 kev and mn k@xmath30 line at 6.490 kev . because the mn k@xmath30 line has an intrinsically complex line shape and is contaminated by the cluster iron emission , only the mn k@xmath1 line is used for the calibration . figures [ fig : cal ] and [ fig : cal_result ] show the result of spectral fitting and the mn k@xmath1 line centroid energy , respectively . as shown in the left panel of figure [ fig : cal_result ] , measured line energies for the center or 14@xmath0 offset regions are systematically lower than the mn k@xmath1 line energy by 7.3 ev if they are averaged over four sensors and two detector segments in each pointing . for xis3 , centroid energies are lower than those of the other three chips , however , we confirmed that the spectral analysis of the redshift measurement with and without xis3 gave statistically consistent results ( see section [ sec : resultred ] for more details ) . we then estimate the @xmath270 systematic uncertainty on the energy scale to be @xmath271 ev . this is justified also from the fact that if we calculate the @xmath2 value using the measured mn k@xmath1 line centroid @xmath272 and the @xmath270 statistical error @xmath273 in the xis - i detector , and the mn k@xmath1 line energy in literature @xmath274 . @xmath2 is defined as @xmath275e@xmath276 . with @xmath271 ev , @xmath2 becomes less than 10 for 7 degrees of freedom , which means the probability of @xmath2 exceeding @xmath2 versus the number of degrees of freedom becomes less than 10% : the number of bins is 8 for four sensors and two detector segments , and @xmath277 can be regarded as being consistent with @xmath274 when we consider @xmath271 ev . in the same manner , the systematic errors are obtained as @xmath278 ev , @xmath59 ev , @xmath109 ev , @xmath89 ev for @xmath16 , 34@xmath0 , @xmath18 , and @xmath19 offset regions , respectively . therefore , for the safety s sake , we assign 7 ev for the 1@xmath279 ( 68% confidence level ) gain uncertainty . \(ii ) though calibration sources provide information on the absolute energy scale at the corners of ccd chips , the gain may be dependent on position . this intrachip variation can be effectively studied by comparing line energies of the same sky regions ( which correspond to different detector regions ) on four or three xis sensors ( see also section 3.1 in @xcite ) . to examine this issue , we divided the center region into 16 cells , and fitted the @xmath280 model in the energy range of 5.08.0 kev for each xis sensor . we then estimate the 1@xmath279 systematic error with a reference to obtained redshift , calculating as @xmath281 for each pair of xis sensors , where @xmath282 is the measured line centroid for the xis - i detector in each region , @xmath283 is the mean of the line centroid for all xis - i detectors in each region , and @xmath284 is the @xmath270 statistical error in the xis - i detector . with @xmath28511 ev , @xmath46 becomes less than 17 for 11 degrees of freedom , which means the probability of @xmath46 exceeding @xmath46 versus the number of degrees of freedom becomes less than 10% : the number of bins is 12 for the spectral regions except for the calibration source region in the center region . @xmath284 is @xmath4 10 ev , which is larger than the gain uncertainty : thus , we ignored @xmath284 to avoid underestimating the systematic error . because @xmath284 is ignored in these estimates , the obtained value is an upper limit for intrachip variation within a 68% confidence level . in summary , the intrachip variation is an upper limit and is larger than the gain uncertainty . thus , we concluded from ( i ) and ( ii ) that the 1@xmath279 systematic error within a 68% confidence level is 11 ev . the systematic error is then 18 ev , which corresponds to 818 km s@xmath3 in the line of sight velocity , when we quote a 90@xmath286 confidence level . ( 80mm,60mm)figure10a.eps ( 80mm,60mm)figure10b.eps ( 80mm,60mm)figure11a.eps ( 80mm,60mm)figure11b.eps ( 80 mm , 60mm)figure12a.eps ( 80 mm , 60mm)figure12b.eps as mentioned in subsection [ subsec : single ] and [ subsec : ratios ] , the line - of - sight velocity of icm or redshift can be evaluated using two different spectral models : the apec model or the bremsstrahlung with three gaussian lines . it is then useful to compare these results and see if there is any systematic uncertainty in the redshift measurement due to spectral modeling . in the apec model , the temperature dependence of the fe k@xmath1 line energy is already implemented in the plasma code , which in turn evokes a parameter coupling between the temperature and line energy ( or redshift ) . on the other hand , in the bremsstrahlung@xmath287gaussian model , the line centroid should be determined almost independently of the continuum temperature , although its temperature dependence has to be corrected later . we did this correction by deriving the rest - frame fe k@xmath1 centroid energy as a function of temperature on the basis of spectral simulations , assuming the apec model and xis energy responses . in figure [ fig : redshift ] , the comparison of redshift measured by the two different models is shown for all analysis regions defined in figure [ fig : image ] . because there consistency between the two within their statistical errors , we conclude that either method robustly determines the icm redshift for the present observations . therefore , we will show results of the fits with the apec model because the uncertainties with apec are smaller . the obtained redshifts from the spectral fits with the single temperature ( apec ) model in the energy range of 5.08.0 kev are shown in figure [ fig : redshift ] , where the icm velocity in the line of sight is also calculated as @xmath288 . the resultant redshifts are consistent with the optical redshift in @xcite from the ned data base , within statistical and systematic errors , which corresponds to @xmath2620.0027 in the redshift . the direct measurements of the central energy of the k@xmath1 line of he - like fe also give almost the same redshifts derived from the spectral fits with the single temperature ( apec ) model . to quantitatively derive the redshift , we calculated the mean redshift of the icm in all pointings to be @xmath289 . this is slightly lower than the optical redshift , @xmath290 from the ned database , but within 90@xmath286 error . as shown in table [ tab:1 - 8vapecresults ] , in the center and 14@xmath0 offset regions , the derived highest and lowest redshift are @xmath291 for cell numbers 5 and 8 of the center region , and @xmath292 for cell number 2 of the center region . based on these redshift measurements , the derived icm bulk velocities are @xmath293 km s@xmath3 , and @xmath294 km s@xmath3 . including systematic errors of @xmath295 km s@xmath3 with 90@xmath286 confidence level ( section [ subsec : cal ] ) , there is no significant difference . we estimate the 90@xmath286 upper limit on the velocity gradient to be @xmath296 km s@xmath3 . if we further add systematic and statistical errors in the quadrature , the upper limit of the gas velocity within the 90@xmath286 confidence level becomes @xmath297 @xmath298 2200 km s@xmath3 . for simplicity , assuming that the gas is rigidly rotating at a typical circular velocity of @xmath299 , then @xmath300 km s@xmath3 , which does not exceed the sound velocity of the coma cluster , which is 1500 km s@xmath3 for a temperature of 8 kev . the average redshift in the center region is as follows : @xmath301 in the north - west region ( cell numbers 8 , 9 , 12 , and 13 ) , @xmath302 in the south - west region ( cell numbers 10 , 11 , 14 , and 15 ) , @xmath303 in the north - east region ( cell numbers 0 , 1 , 4 , and 5 ) , and @xmath304 in the south - east region ( cell numbers 2 , 3 , 6 , and 7 ) . we also show the averaged redshift over fov in the upper panel in table [ tab : redshift ] . these results are consistent with each other in the center region , and with the observed redshifts in the 14@xmath0 offset region within statistical and systematic errors . the difference in the icm velocity between the center and 30@xmath0 offset regions is derived as @xmath305 km s@xmath3 , and the difference between the center and 34@xmath0 offset regions is @xmath306 km s@xmath3 . these are also consistent with each other . components of the velocity in the plane of the sky are not measured . if we assume that three components are similar , then the limit on the radial velocity translates into a limit of @xmath307 500 km s@xmath3 , which does not exceed the speed of sound and is consistent with systematic errors . therefore , there were no significant velocity variations on the 10@xmath0 scale within central regions ( center and 14@xmath0 , 30@xmath0 , and 34@xmath0 offset regions ) from our analysis . in the ngc 4839 subcluster region ( 44@xmath0 offset region ) , the observed redshift is [email protected] , which corresponds to 1140@xmath308 km s@xmath3 as the icm bulk velocity without systematic errors . there also seems to be no significant difference between the core of the coma cluster and the ngc 4839 subcluster within the 90% confidence level . as mentioned in subsection [ subsec : cal ] , some sensors have large uncertainties in the determination of the line centroid energy . on xis3 in the center region , the mn k@xmath1 line centroid energy of segment a was different from others . similarly , the segment d of xis1 in the ngc 4839 region was also different . we averaged the spectrum over fov in ( i ) all detectors , and ( ii ) all detectors except those with different mn k@xmath1 line centroid energies to check the spectral shape and measured redshift in the center and ngc 4839 regions ( 44@xmath0 offset region ) . the results are shown in the lower section in table [ tab : redshift ] . for the center region , the derived redshifts were fairly consistent including xis3 or not . on the other hand , for the ngc 4839 subcluster region , the resultant redshift changed slightly from 0.0268 to 0.0287 , although this was also consistent within statistical and systematic errors . as shown in figure [ fig : cal_result ] , the peak of the he - like fe - k@xmath1 line of xis1 is more shifted than that of the other two sensors . the measured redshifts in the 44@xmath0 offset region were [email protected] , [email protected] , and [email protected] for xis0 , 1 , and 3 , respectively . because the measured redshift of the averaged spectra between xis0 and 3 was consistent with the redshift from simultaneous spectral fits , the value 0.0269 would be plausible . in conclusion , there was no significant difference in the icm velocity between the core of the coma cluster and the ngc 4839 subcluster within 90% confidence . ccccc field & x - ray ( @xmath309 ) & optical ( @xmath309 ) & x - ray ( @xmath310 ) & optical ( @xmath310 ) + & & & km s@xmath3 & km s@xmath3 + all pointings & [email protected] & 0.02307 & 6810@xmath262150 & 6917 + center & [email protected] & 0.02286 & 6810@xmath262150 & 6853 + 14@xmath0 offset & [email protected] & 0.02286 & 6900@xmath262300 & 6853 + center and 14@xmath0 offset & [email protected] & 0.02286 & 6810@xmath262150 & 6853 + ngc 4839 & [email protected] & 0.02448 & 8070@xmath262990 & 7339 + center ( xis0123 ) & [email protected] & 0.02286 & 6775@xmath262150 & 6853 + center ( xis012 ) & [email protected] & 0.02286 & 6775@xmath262150 & 6853 + ngc 4839 ( xis013 ) & [email protected] & 0.02448 & 8604@xmath262990 & 7339 + ngc 4839 ( xis03 ) & [email protected] & 0.02448 & 8034@xmath262990 & 7339 + + [ tab : redshift ] ( 80 mm , 60mm)figure13a.eps ( 80 mm , 60mm)figure13b.eps to measure the abundance using the strong fe k lines , we employed the results of spectral fitting in the 5.08.0 kev energy range for the fe abundance . because of the good statistics , the spectra in the lower energy band primarily determine the fitted temperatures . although the spectra are well fitted with a single - temperature model , a small discrepancy between the model and data around the fe - k line can yield a large systematic uncertainty in the derived fe abundances . within the 34@xmath0 region , temperatures and abundances derived from the spectral fitting in the 5.08.0 kev energy range agree well with those from 1.08.0 kev energy ranges within 5% . in the 44@xmath0 and 60@xmath0 offset regions , the fe abundances from the 5.08.0 kev energy range are smaller than those from the 1.08.0 kev range by @xmath40.1 solar . @xmath46 values for the limited energy range of 5.08.0 kev from the former fitting are 10% smaller than those from the latter , although the single - temperature model reproduces the observed spectra at 1.08.0 kev well . therefore , we adopted the fe abundance derived from the spectral fitting in the 5.08.0 kev energy range . figure [ fig : fe ] shows a radial profile and map of the fe abundance of icm with the single temperature ( apec ) model in the 5.08.0 kev energy range . within 30@xmath0 from the x - ray peak , fe abundances are almost constant at @xmath40.4 solar . the derived fe abundances are [email protected] solar in the center region , and [email protected] solar in the 14@xmath0 offset region . the fe abundances in the 30@xmath0 and 34@xmath0 offset regions are slightly lower , @xmath40.3 solar , although it has larger uncertainties . these values agree with recent xmm results @xcite within 1.2 mpc . the fe abundance in the ngc 4839 subcluster region is comparable to that in the central region , and higher than that in the 30@xmath0 and 34@xmath0 offset regions . the average value in the subcluster region is [email protected] solar . this value is also consistent with that in @xcite , although the error bars in their results are much larger than those in our results . x - ray observations of clusters of galaxies can play a major role in determining cosmological parameters , because of the importance of measuring gravitational mass . the ratio of he - like k@xmath1 to h - like k@xmath1 lines of fe is a very steep function of temperature , and has a different temperature dependence from that of the continuum . the continuum emission is affected by the systematic uncertainties in the response and non - thermal emission , while the line ratio does not . thus , by comparing the fitted temperatures between these two methods , the effects of the systematic uncertainty in the response and/or of the non - thermal emission is increased . using the apec plasma code , the temperatures derived from the spectral fitting of the 1.08.0 kev energy range and the line ratio of fe agree within a few percent when we averaged all regions . the systematic uncertainties in temperatures from the line ratio are in the level of the systematic difference between those from apec and mekal , which is about 3% . the consistency of observed temperatures from the continuum and those from line ratios supports the accuracy of suzaku measurements of plasma temperature . in contrast , systematic differences of @xmath4 10% in the cluster temperature among chandra , pn , and mos were reported in relatively high temperature clusters ( @xcite ) . even using the same instrument , the derived temperatures changed systematically by @xmath311% between calibrations @xcite . a systematic uncertainty of several percent in response matrices of a detector can also cause systematic uncertainties in the temperature structure . @xcite found that the temperature structures of icm in the coma cluster derived from multi - temperature fits with pn and mos spectra were different , although the two detectors gave consistent temperatures within several percent . when they allowed higher temperature components over 10 kev in hot clusters including the coma cluster , using mos data , fe abundances from the he - like line increased by several tens of percent , owing to artificial detection of a hot component , while those derived by pn were unchanged . however , with suzaku , we found that the single - temperature model fits the spectra of the coma cluster very well , and we do not need temperature components above 10 kev or below 2 kev to fit the spectra . the multi - temperature components observed by xmm were not needed . in the outer regions of clusters at low surface brightness , uncertainties in the background of chandra and xmm also lead to systematic uncertainties in icm temperatures . some analyses found that icm temperatures decreased outward by half at 0.6@xmath312 ( e.g. , @xcite ; @xcite ) , while others found flatter profiles ( e.g. , @xcite ) . systematic uncertainties in the derived temperatures due to uncertainties in the background with suzaku observations are much smaller , because suzaku xis has a much lower and more stable background . also , previous xmm and chandra results of temperatures in the outer regions have statistical errors more than 20% , while those of suzaku have errors less than 10% . we need precise measurements of icm temperatures to search for a non - thermal hard component as an excess of the thermal component . on the basis of the icm temperature map derived with xmm - pn , @xcite found that the suzaku hxd - pin spectrum was described with the thermal emission from icm , and there was no significant evidence of excess hard emission . the temperatures of the coma cluster derived from suzaku - xis are consistent within error bars with those from the pn detector : the weighted average of temperatures with statistical errors within the center field is close to that derived with xmm - pn within a few percent . the temperatures of offset fields are consistent with those from xmm - pn within error bars . thus , our results support the non - detection of non - thermal hard emissions by @xcite @xcite reported the existence of hot gas at @xmath313 kev in rx j1347.5 - 1145 . they concluded that the gas properties can be explained by a fairly recent ( within the last 0.5 gyr ) collision of two massive ( 5@xmath1210@xmath314 m@xmath315 ) clusters with bullet - like high velocity ( @xmath316 km s@xmath3 ) . in contrast , @xcite found that there was no extremely high temperature gas in the coma cluster in the hxd - pin spectrum up to 70 kev . on the basis of the consistency of derived temperatures from the continuum and the line ratio and the results from the spectral analysis of hxd - pin , we conclude that the amount of very high temperature gas is rather small . from the spatial variation of the derived temperatures , there is no region with extremely hot temperature : the derived temperatures range from 7.0 to 9.0 kev within the center , 14@xmath0 offset , and @xmath17 offset regions . the coma cluster has a cluster - wide synchrotron radio halo . if the electrons in the cluster were accelerated by a very strong shock that occurred owing to a recent merger , we might expect the cluster to contain more very hot gas , although it is difficult to make this argument quantitative . in any case , in the coma cluster , relativistic electrons producing the radio halo were apparently generated without producing a large amount of very hot gas . an example of a process which might do this is the turbulent re - acceleration model ( e.g. , @xcite ) . within the center and offset regions the spatial variation of derived temperatures is relatively small , from 7 kev to 9 kev . there are no extremely hot and cool regions . we did not detect any temperature jump corresponding to a shock in the offset regions . within each region for the spectral analysis , the single - temperature model fit the spectrum . therefore , in the center of the coma cluster , icm is in a nearly relaxed state . although the variation was small , there was some asymmetric temperature structure , like a hot region northwest of the center . if the cluster had been an isolated system collapsing from self - gravity , the temperature structures would be symmetric . the observed asymmetry in the temperature distribution indicates that external effects or interactions , such as merging , have occurred . numerical simulations of the evolution of the cluster ( e.g. , @xcite ; @xcite ) show that nonaxisymmetric structures in the temperature distribution are erased several gyr after each merger . the presence of asymmetric temperature structures suggest that the central region of the coma cluster has experienced a merger , and the lack of hard component shows that its core have relaxed in the last several gyr . the observed cool region in the southeast of the center region coincides with the filamentary structure originating near ngc 4911 and ngc 4921 @xcite . @xcite reported that the temperature of this cold area ranges from 5 to 7 kev . this region mostly contain the stripped subcluster gas , which is thought to be relatively cool in the initial state . the numerical simulations ( e.g. , @xcite ) suggest that the gas would maintain its temperature for several gyr . @xcite showed that even a single subcluster collision with supersonic velocity creates both hot and cool regions in the main cluster . although heating due to collisions occurs , cooling would also occur through the adiabatic expansion of the stripped gas . the measurements of bulk motions in icm are also important to derive the gravitational mass of clusters of galaxies . we constrained the upper limit of icm bulk motions in the center and offset regions on the scale of 117 kpc ( ) and 234 kpc ( 9@xmath0 ) , respectively , which correspond to the scales of regions for spectral accumulation ( see section [ sec : obs ] and subsection [ sec : resultred ] ) . the upper limit of gas bulk motions is @xmath297 @xmath298 1100 km s@xmath3 . as summarized in table [ tab : redshift ] , considering the systematic error of 18 ev , or 818 km s@xmath3 with a 90@xmath286 confidence level ( section [ subsec : cal ] ) , there are no significant bulk motions in the center and offset fields . compared with the sound velocity of the coma cluster , @xmath317 km s@xmath3 for a temperature of @xmath47 = 8 kev , the bulk - motion in the central region of the coma cluster does not exceed the sound velocity . our results do not conflict with the discovery of pressure fluctuations in the center of the coma cluster by @xcite which indicates that at least 10% of the icm pressure is in turbulent form . we have not constrained the turbulence , and the scale of the fluctuation ranges between 29 kpc and 64 kpc , which is smaller than our scale of spectral analysis . the velocity difference between the main cluster and the ngc 4839 subcluster from optical observations is 486 km s@xmath3 . although the best - fit line - of - sight velocity of the subcluster is higher than that of the main cluster by @xmath4 1000 km s@xmath3 , the difference is comparable to statistical and systematic errors . because the temperature of the region is about 5 kev , the sound velocity is @xmath4 1150 km s@xmath3 . considering the error bars , we conclude that the relative line - of - sight gas velocity in the ngc 4839 subgroup region does not greatly exceed the sound velocity . numerical simulations indicate that a major merger between clusters raises the temperature of icm and induces a bulk velocity of the order of 1000 km s@xmath3 ( e.g. , @xcite ; @xcite ) . in the late phase of mergers , turbulent motions develop in a simulation by @xcite . the non - detection of bulk - motions in icm of the coma cluster also indicates that the central region of the cluster is in a somewhat relaxed state . if so , the non - thermal electrons relevant to the radio halo of the coma cluster may have been accelerated by the intracluster turbulence rather than shocks . in conclusion , the line - of - sight gas velocity of the coma cluster in the observed region does not greatly exceed the sound velocity , and the central region of the coma cluster is in a nearly relaxed state . this means that the assumption of hydrostatic equilibrium is approximately valid in calculating the cluster mass of the coma cluster . the fe abundance is nearly constant at @xmath4 0.4 solar within a clustocentric radius of 0.5 mpc , and beyond that distance it decreases with the radius to 0.20.3 solar . this profile is consistent with those measured by xmm ( @xcite ; @xcite ) . suzaku has a lower level of background , and a slightly better energy resolution : therefore , systematic uncertainties in fe abundance are considered to be smaller . differences in abundance profiles of icm in central regions of clusters with and without cool cores have been reported ( @xcite , @xcite , @xcite ) . @xcite showed an abundance difference between the two type of clusters within 0.1@xmath312 , while at 0.10.3@xmath312 , the average values of fe abundances are both @xmath4 0.40.5 solar , and are consistent with each other . the coma cluster is thought to have experienced a major merger in the recent past and subsequently , the central region came to a somewhat relaxed state . during a cluster merger , the mixing of icm could destroy any central fe peak . in the first stage of merging , the fe peak and cool core remain intact as observed in some ongoing merging clusters ( @xcite ; @xcite ) . for example , abell 2256 , an ongoing merging cluster , has an abundance gradient in the direction opposite of the merging direction owing to the remixing of the gas . numerical simulation also support this scenario . @xcite performed hydrodynamical simulations with two different merger types of clusters : cluster 1 has only small merger events , while cluster 2 undergoes a major merger ( mass ratio 1:3 ) . in cluster 1 , the abundance is more than @xmath4 0.4 solar within a radius of 0.1 mpc , and decreases from @xmath4 0.3 solar to @xmath4 0.1 solar with a decrease in radius . the situation is different in cluster 2 . in the coma cluster , the fe abundance is @xmath4 0.4 solar and homogeneous within a radius of 0.5 mpc . this value of the fe abundance is similar to those at 0.20.3@xmath312 in other clusters as shown in figure [ fig : fe ] . therefore , tn the coma cluster , the merger have destroyed the cool core , and mixed the icm completely at least within 0.5 mpc . the coma cluster which was observed with the x - ray imaging spectrometer ( xis ) onboard suzaku , was analyzed by the x - ray satellite in six pointings , centered on the x - ray peak and offset by west , , , , and . because of its low background , suzaku is the most sensitive x - ray satellite for observing k@xmath1 lines of fe in the intracluster medium . after obtaining accurate measurements of fe lines , we studied the temperature structure of the intracluster medium , searched for possible bulk - motions , and measured the fe abundance in the cluster . the spectra of each extracted region were well fitted by the single - temperature apec model , and the two- or three- temperature apec model did not improve @xmath2 . the temperatures derived from the observed ratios of k@xmath1 lines of h - like and he - like fe agree well with those from spectral fittings with the single - temperature apec model . because this line ratio is a strong function of plasma temperature , the observed consistency supports the accuracy of temperature measurements with suzaku , and constrains the temperature structure of icm . the observed values of the central energy of the he - like fe line of the center , , and offset regions are constant within 500 @xmath318 , which corresponds to the calibration error . because relative bulk velocities in the coma cluster are smaller than the sound velocity of the intracluster medium , 1500 km s @xmath3 , we can verify the derived total mass in a cluster on the basis of the hydrostatic icm equilibrium inside the offset region . significant bulk velocities were also not found in the offset region , which corresponds to the ngc 4839 subcluster . the results on the temperature and velocity structure suggest that the core of the coma cluster is in a fairly relaxed state . this is consistent with models in which the non - thermal electrons relevant to the radio halo are accelerated by the intracluster turbulence rather than large - scale shocks . this is also consistent with the fact that numerical simulations show that the turbulence motion is developed in the late phase of mergers . the observed fe abundance of the intracluster medium is almost constant at 0.4 solar inside the offset region , and decreases with radius . the central abundance is slightly lower than that of other clusters or groups . this indicates that central regions of the gas were mixed well during the past merger growth of the cluster . + _ acknowledgements . _ we wish to thank the referee for his / her suggestions on the paper . the authors are grateful to all members of suzaku for their contributions in instrument preparation , spacecraft operation , software development , and in - orbit instrumental calibration . to analyze the data , we used the isas analysis servers provided by isas / c - soda . n. o. acknowledges support by the ministry of education , culture , sports , science and technology of japan , grant - in - aid for scientific research no . cls was supported in part by nasa suzaku grants nnx08az99 g , nnx09ah25 g , and nnx09ah74 g .

we establish a 3-manifold invariant for each finite - dimensional , involutory hopf algebra . if the hopf algebra is the group algebra of a group , the invariant counts homomorphisms from the fundamental group of the manifold to . the invariant can be viewed as a state model on a heegaard diagram or a triangulation of the manifold . the computation of the invariant involves tensor products and contractions of the structure tensors of the algebra . we show that every formal expression involving these tensors corresponds to a unique 3-manifold modulo a well - understood equivalence . this raises the possibility of an algorithm which can determine whether two given 3-manifolds are homeomorphic .

math9201301

this paper describes some new invariants for triangulations and heegaard diagrams of 3-manifolds . specifically , for every finite - dimensional , involutory hopf algebra @xmath1 , we define an invariant @xmath2 for closed , oriented 3-manifolds @xmath3 . the invariants can be extended in various ways to invariants of oriented manifolds with boundary . we will convert a heegaard diagram to an expression in terms of the structure tensors of a hopf algebra . the value of the expression is then the value of the invariant . using index notation , we write @xmath4 and @xmath5 for the multiplication and comultiplication tensors of some involutory hopf algebra , and @xmath6 for the antipode map . we may consider the set of all expressions in terms of these tensors , modulo the axioms of a hopf algebra , which consist of identities which these tensors satisfy . we will show that if two ( prime , closed , oriented ) 3-manifolds produce equivalent formal expressions , then they are homeomorphic . conversely , we may consider the question of determining whether two such expressions are formally equivalent . for example , is the equation : @xmath7 an identity that follows directly from the axioms of an involutory hopf algebra ? the answer is yes if and only if two certain oriented 3-manifolds are homeomorphic ( in this case @xmath8 and @xmath9 ) . this work is part of a new area of mathematics which we will call state model topology , as described in @xcite . as the starting point of state model topology , we are given a topological object , for example a knot , with a combinatorial description , for example a knot projection . we wish to find a state model ( or collection of state models ) which we can assign to all such combinatorial descriptions . we then hope that the partition function of the state model is a topological invariant , i.e. is independent of the chosen combinatorial description . typically , we demonstrate invariance under an appropriate set of local moves on the combinatorial description , for example the three reidemeister moves . ( strictly speaking , the partition function may be a topological `` covariant '' because its value may change in a simple way under some of the moves . ) a _ state model _ @xmath3 consists of a ( commutative ) ring @xmath10 ( usually the complex numbers ) , a bipartite graph @xmath0 , the _ connectivity graph _ , whose vertices are labeled as _ atoms _ and _ interactions _ ; a set @xmath11 for each atom @xmath12 , called the _ state set _ of @xmath12 ; and a function @xmath13 for each interaction @xmath14 ( where @xmath15 are the neighbors of @xmath14 ) , called the _ weight function _ or the _ boltzmann weights _ of @xmath14 . a _ state _ of @xmath3 is a function @xmath16 on the atoms of @xmath3 such that @xmath17 . @xmath18 of a state @xmath16 is defined as the product of the @xmath19 s evaluated at the state @xmath16 when this product converges , and in particular when @xmath0 is finite . finally , the _ partition function _ @xmath20 is defined to be the sum of @xmath18 over all states @xmath16 when this sum converges , and in particular all state sets are finite . ] to obtain a topological state model , we choose a prescription for assigning a connectivity graph and weight functions to each combinatorial description of a topological object . for example , if we are given a projection @xmath21 of a link , we can declare the arcs between crossings to be atoms and the crossings themselves to be the interactions . we choose some @xmath22-element set @xmath23 to be the common state set for all atoms . if the states of the arcs incident to a given crossing are labeled @xmath24,@xmath25,@xmath26 , and @xmath27 as in figure [ f : crossing ] , we may define the weight function @xmath28 of a crossing by : @xmath29 where @xmath30 is chosen so that @xmath31 , and @xmath32 when @xmath33 and 0 otherwise . this state model is a `` link covariant '' called the kauffman bracket , which is essentially the jones polynomial up to normalization . surprisingly , there are many non - trivial topological state models . we may start with the general form of such a state model for some kind of combinatorial description , and interpret combinatorial moves as a list of constraints on the weights of the model . usually there are many more equations than unknowns , and yet the equations have solutions . for example , the jones and homfly polynomials @xcite and most of their variants and generalizations can be defined as topological state models . we can conclude that the constraints on the weights of a topological state model have a deep algebraic structure . indeed , the constraints imposed by invariance under the third reidemeister move are also known as the yang - baxter equations , considered independently in the related area of exactly solved statistical mechanical models @xcite . more recently , many solutions to the yang - baxter equations have been discovered via hopf algebras , or quantum groups , as is summarized in an inspiring paper by drinfeld @xcite . recently , turaev and reshetikhin have extended this analysis to produce a 3-manifold invariant which generalizes the jones polynomial @xcite . the new invariants are closely related to other topological invariants . if we restrict to the special case when @xmath3 is a link complement , we may convert the invariant @xmath2 to a special case of the machinery of drinfeld , turaev , and reshetikhin . this conversion involves a bigger hopf algebra @xmath34 , the quantum double of @xmath1 , which was defined by drinfeld . the following is an instructive example of a topological state model . let @xmath35 be a triangulation of a connected @xmath22-manifold @xmath3 . we orient the edges of @xmath35 . we choose a finite group @xmath0 to be the state set , and we declare the edges of @xmath35 to be the atoms . if we assign @xmath36 to an edge with one orientation , we declare this equivalent to assigning @xmath37 to the same edge with the opposite orientation . we assign interactions to the faces ( 2-simplices ) of @xmath35 . given a state of the model , the weight of a face is 1 if the cyclically - ordered product of the states of the edges ( when the edges are oriented in the same direction around the face ) of the face is the identity and 0 otherwise . we leave it as an exercise that the partition function is simply @xmath38 , where @xmath39 is the number of vertices of @xmath35 . thus , the partition function is an invariant after dividing by @xmath40 . we will see later that this invariant equals @xmath2 when @xmath1 is the group algebra of @xmath0 . vaughan jones introduced me to state model topology and hopf algebras . i am also grateful to my advisor , andrew casson ; rob kirby ; and oleg viro for their encouragement and their knowledge of 3-manifold topology . let @xmath41 be a finite - dimensional vector space over a field @xmath42 . for most of this paper , a tensor will mean an element of some tensor product space @xmath43 , where each @xmath44 is either @xmath41 or @xmath45 . we will adopt index notation for tensors as it is defined in @xcite . in index notation , a tensor @xmath46 is written as @xmath47 , with @xmath22 indices total , where the @xmath42th index is a superscript ( or contravariant index ) if @xmath48 and a subscript ( or covariant index ) if @xmath49 . the indices should be distinct letters which have no meaning independent of their use as a description of the tensor @xmath35 . for example , a vector is written @xmath50 , a dual vector is @xmath51 , a linear operator is @xmath52 , a bilinear form is @xmath53 , and so on . if a tensor @xmath35 has @xmath42 superscripts and @xmath54 subscript , we say that @xmath35 has _ type _ @xmath55 , and _ total type _ @xmath22 . the indices of a tensor can be used for three different operations on tensors : 1 . : : we denote the factor - switching map @xmath56 and its cousins for higher tensor products by permuting the corresponding indices . for example , if @xmath57 , then @xmath36 is a symmetric bilinear form . 2 . : : we write the tensor product of two tensors by juxtaposing the two tensors and giving them distinct indices . for example , one may write @xmath58 if @xmath59 is a linear operator of rank 1 . 3 . : : the canonical map from @xmath60 to @xmath42 which is called the trace map , the contraction map , or the evaluation map is denoted by a repeated index . the index must appear exactly once as a superscript and once as a subscript . used together , contractions and tensor products subsume most of the usual operations in linear algebra . for example , @xmath61 is the value of @xmath28 at @xmath39 , @xmath62 is the operator @xmath59 applied to @xmath39 , @xmath63 is the trace of @xmath59 , and so on . although index notation only allows basis - independent computations with tensors , the notation is motivated by computations in a specific basis . suppose the vector space @xmath41 has dimension @xmath22 and has a given basis . we may interpret the notation @xmath50 to mean the coordinates of the vector @xmath39 , with @xmath24 an integer from 1 to @xmath22 , and in general we may interpret @xmath47 to be the matrix entries of the tensor @xmath35 . in this case a tensor formula in index notation may be interpreted as a formula for the matrix entries of the tensors , and the two formulas are equivalent given the convention that we sum over any repeated indices ( the einstein summation convention ) . we may use the summation convention to arrive at a re - interpretation of certain state models . suppose that @xmath3 is a state model such that every atom is adjacent to precisely two interactions , and suppose for simplicity that there is only one state set @xmath23 . we may declare @xmath23 to be the index set of a basis ( or perhaps @xmath23 is the basis ) of a vector space @xmath41 . we declare each atom @xmath24 to be an index letter . we may then re - interpret the values of each weight function @xmath19 as the matrix entries of a tensor @xmath64 of total type @xmath22 . we arbitrarily declare each of the indices of each @xmath64 to be covariant or contravariant , provided that each atom is used once as a superscript and once as a subscript . in this case the partition function @xmath20 is the result of tensoring together all of the interaction tensors and then contracting all of the ( atom ) indices in the natural manner . thus , we define a _ tensorial state model _ to be a list of tensors over one vector space , together with a formula which is some product of these tensors with all indices contracted . the partition function is defined to be the value of this formula . more generally , we say that @xmath3 is a _ tensorial state model with boundary _ if @xmath3 is a formula which is a product of tensors with some subset of the indices contracted . we call the uncontracted indices the boundary of the model , and the contracted indices the interior . we define the _ partition tensor _ @xmath65 to be the value of the formula . the partition tensor @xmath3 should not be confused with the partition function of @xmath3 . the state models defined in this paper will all be tensorial state models . to conclude this section , we present a more visual version of index notation which we will actually use . to avoid using many different letters for contracted indices , we may draw each tensor with an inward arrow for every subscript and an outward arrow for every superscript . for example , we write the tensor @xmath66 as follows : @xmath67 we denote the contraction of two indices by connecting the corresponding arrows . for example , @xmath68 and @xmath69 are written as : @xmath70 for consistency , we should interpret the strange - looking expression : @xmath71 as the scalar @xmath72 . arrows can and will cross . strictly speaking , the visual notation can be ambiguous . for example , it is not clear if the following denotes @xmath73 or @xmath74 , or perhaps even @xmath75 : @xmath76 we declare that indices should be read off in counter - clockwise order . the meaning of a diagram is still ambiguous if the arrows incident to some tensor are symmetric under rotation . however , all such tensors in this paper will be symmetric under cyclic permutation of their indices , so this ambiguity will be irrelevant to us . a finite - dimensional _ hopf algebra _ is a finite - dimensional vector space @xmath1 ( over some ground field @xmath42 ) such that @xmath1 and @xmath77 are both algebras with unit , and the two algebra structures satisfy certain compatibility axioms . we will describe the axioms with the fancy notation of the previous section . firstly , we may describe the algebra structure on @xmath1 by a _ multiplication tensor _ @xmath3 defined so that @xmath78 is the product of @xmath39 and @xmath28 . @xmath3 satisfies the associativity axiom : @xmath79 similarly , the algebra structure on @xmath77 is described by the _ comultiplication tensor _ @xmath80 which is coassociative : @xmath81 in addition , there are three other tensors @xmath82 , @xmath83 , and @xmath23 called the _ unit _ , the _ counit _ , and the _ antipode _ of @xmath1 . the five tensors together satisfy these axioms : @xmath84 we also assume as an axiom that @xmath23 has an inverse , although this fact follows from the other axioms in the finite - dimensional case @xcite . given @xmath3 and @xmath80 , the other tensors are unique when they exist , so we may say that @xmath3 and @xmath80 are the `` meat '' of a hopf algebra . morphism of hopf algebras _ is defined to be a linear transformation of the underlying vector spaces which commutes with the five structure tensors . since we have not defined infinite - dimensional hopf algebras and will not use them , we will assume that a hopf algebra is finite - dimensional unless explicitly stated otherwise . we adopt the following abbreviations in light of the associativity and coassociativity axioms : @xmath85 the dual vector @xmath35 is called the _ trace _ , and the vector @xmath86 is the _ cotrace_. we will call the last two tensors the _ tracial product _ and _ tracial coproduct_. like the diagrams , the tensors themselves are cyclically symmetric . one motivation for the definition of a hopf algebra is the observation that every group algebra is naturally a hopf algebra . let @xmath0 be a finite group . we may let the elements of @xmath0 be both a basis and a dual basis for a vector space @xmath87 $ ] , and define multiplication and comultiplication by : @xmath88 the antipode is induced by the group inverse . this construction has the nice property that the morphisms between the hopf algebras of two groups correspond to the homomorphisms between the groups themselves . in general , the antipode plays the role of the group inverse in a hopf algebra . however , the antipode need not satisfy @xmath89 . a hopf algebra is called involutory if this equation is satisfied . many important hopf algebras are not involutory , and in particular those used in the analysis of the jones polynomial are non - involutory @xcite . however , we will only obtain results about involutory hopf algebras in this paper . finite - dimensional hopf algebras possess an eight - fold duality . if @xmath1 is a hopf algebra , we obtain a new hopf algebra @xmath90 on the same vector space by reversing multiplication : @xmath91 we will show below that @xmath92 is the antipode of @xmath90 . similarly , we define @xmath93 by reversing comultiplication . finally , we define the _ dual _ of @xmath1 , a hopf algebra structure on the vector space @xmath77 , by switching @xmath3 with @xmath80 and @xmath83 with @xmath82 and reversing all the arrows ( actually , what we are calling @xmath77 is usually defined as @xmath94 , and vice - versa ) : @xmath95 we establish some basic identities involving hopf algebras . the proof of each identity will use the following observation : the tensors : @xmath96 when viewed as vector space endomorphisms of @xmath97 and @xmath98 , are invertible . [ ladder ] we compute : @xmath99 the following identities hold in any hopf algebra : @xmath100 [ basic ] we compute : @xmath101 we will need a few other definitions involving hopf algebras . if @xmath1 is a hopf algebra , a _ left integral _ of a hopf algebra is a dual vector @xmath102 that satisfies : @xmath103 one can similarly define a _ right integral _ , a _ left cointegral _ , and _ right cointegral _ : @xmath104 a hopf algebra always has a non - zero integral and cointegral . by lemma [ basic ] , @xmath83 is an algebra homomorphism . in general , if @xmath12 is a finite - dimensional algebra over @xmath42 and @xmath105 is any homomorphism , then there exists a non - zero element @xmath106 such that @xmath107 for all @xmath108 @xcite . a hopf algebra @xmath1 is called _ semisimple _ if it is semisimple as an algebra , and cosemisimple if @xmath77 is semisimple . equivalently , @xmath1 is semisimple if the tensor : @xmath109 is a non - degenerate bilinear form . we say that @xmath1 has _ invertible dimension _ if @xmath110 in the ground field of @xmath1 . in an involutory hopf algebra , the trace @xmath35 is also a left and right integral @xcite . to prove this , we use the fact that there exists some non - zero integral , and we show that @xmath35 is a multiple of it : the tensor : @xmath111 is always a right integral ( which may be zero ) . [ tracelemma ] let @xmath112 be a non - zero right integral and let @xmath113 be a non - zero left cointegral . we first observe that : @xmath114 therefore : @xmath115 by lemma [ ladder ] . since @xmath112 and @xmath113 are non - zero , the last equation shows that @xmath116 must be non - zero . finally , we obtain : @xmath117 as desired . by virtue of the fact that : @xmath118 we obtain : if @xmath1 is involutory , then : @xmath119 [ tracegen ] thus , lemma [ tracelemma ] demonstrates not only that @xmath35 is a right integral , but also that @xmath1 is semisimple and cosemisimple . moreover , if @xmath1 has invertible dimension , we can express all of the structure tensors of a hopf algebra in terms of tracial product and coproduct ! in the section ( and the rest of the paper ) we will work in the smooth category of manifolds . we will use the term @xmath22-manifold to mean an oriented ( not merely orientable ) and compact @xmath22-manifold which may have boundary and may or may not be connected . we recall some basic elements of 3-manifold topology as described in @xcite . heegaard splitting _ is a decomposition of a closed 3-manifold into two handlebodies of the same genus glued along their boundary ; the boundary is called a _ heegaard surface_. given such a decomposition , we may arbitrarily label the two handlebodies upper and lower . if the handlebodies are labeled , then an orientation of the heegaard surface induces an orientation of the manifold : at a point @xmath120 , we complete a positively - oriented basis @xmath121 of the tangent space of @xmath23 at @xmath120 to a ( positive ) basis @xmath122 of @xmath123 by choosing @xmath124 to point from the upper handlebody to the lower one . if @xmath23 is a surface , we define a _ handlebody diagram _ @xmath27 on @xmath23 to be a prescription for gluing a handlebody to @xmath23 . specifically , it is a family of disjoint circles @xmath125 which divides @xmath23 into planar regions . ( often it is assumed that @xmath125 does not separate @xmath23 , but we do not want this hypothesis . ) we obtain a handlebody @xmath126 from @xmath27 by the following procedure : we start with @xmath127 and glue a 2-handle along a tubular neighborhood of each circle @xmath128 . the boundary of the result is @xmath129 plus a number of spheres . we eliminate the spherical boundary components by gluing in balls to obtain @xmath126 . we say that @xmath130 is a _ heegaard diagram _ of a closed 3-manifold on a surface @xmath23 is a prescription for a heegaard splitting of a manifold @xmath131 , i.e. it is a pair of handlebody diagrams @xmath132 and @xmath133 which describe the upper and the lower handlebodies of the splitting . we assume that @xmath132 and @xmath133 are transverse . we will need to generalize the definition of a heegaard diagram to account for manifolds with boundary , and to relax the constraint that the circles of a handlebody diagram divide the surface of the diagram into planar regions . to this end , we also label the boundary of a 3-manifold @xmath3 by arbitrarily dividing the boundary components of @xmath3 into two disjoint subsets : the upper boundary and the lower boundary . we also define a _ puncture move _ on a 3-manifold @xmath3 with labeled boundary to consist of removing a ball from @xmath3 and labeling the new boundary as upper or lower . we define a _ bordism _ to be an equivalence class of 3-manifolds with labeled boundary under the puncture move . by abuse of terminology , we will not distinguish between a bordism and a representative of a bordism . a _ compression body _ is a manifold which is obtained from @xmath127 by gluing 2-handles on one side , where @xmath23 is a surface ( usually one requires that all spherical boundary components of a compression body are capped by balls , but in the context of bordisms this does not matter ) . a compression diagram is then a generalization of a handlebody diagram : it is a prescription for gluing a compression body to a surface @xmath23 consisting of a collection of disjoint circles . however , the circles are not required to divide the surface into planar regions . we define a generalized heegaard diagram to be a pair of compression diagrams ( an upper diagram and a lower diagram ) on one surface . if we are given a generalized heegaard diagram @xmath130 on @xmath23 , we may form a bordism @xmath3 by gluing together the corresponding compression bodies along @xmath23 . the boundary of the lower compression body becomes the lower boundary of the bordism , and the same for the upper compression body . we will need a set of moves to convert any heegaard diagram of a bordism to any other diagram for the same bordism . also , it may not be obvious that every bordism has a heegaard diagram . we may solve both of these problems using the theory of morse functions , as described in @xcite . we summarize the relevant parts of the theory : if @xmath3 is a bordism , we first choose a riemannian metric for @xmath3 . we choose a morse function @xmath134 on @xmath3 which attains its minimum on the lower boundary of @xmath3 and its maximum on the upper boundary of @xmath3 . using the riemannian metric , we can consider the flow @xmath135 on @xmath3 given by the gradient of @xmath134 . for each fixed point @xmath120 of @xmath135 , we consider the descending manifold of @xmath120 , the set of all points that flow into @xmath120 . the descending manifold is necessarily a disk of some dimension , and in this way we obtain a handle decomposition of @xmath3 . if @xmath134 is in general position , each @xmath22-handles is attached only to @xmath136-handles with @xmath137 ( not merely @xmath138 ) and possibly to the lower boundary . finally , it is easy to convert such a handle decomposition to a generalized heegaard diagram : we let the lower compression body be the union of the 0-handles , the 1-handles , and a tubular neighborhood of the lower boundary , and we let the upper compression body be the complement of the lower compression body . we obtain an upper circle for each 2-handle by intersecting the core ( descending manifold ) of the 2-handle with the heegaard surface , and similarly we intersect the cocore ( ascending manifold ) of each 1-handle with the heegaard surface to obtain the lower circles . since every bordism has a morse function , every bordism has a heegaard diagram . on the other hand , given a heegaard diagram , it is easy to construct a morse function which reproduces it . we can therefore obtain a set of moves on heegaard diagrams from a set of moves on morse functions . the upshot is the following theorem : let @xmath130 be a generalized heegaard diagram on a surface @xmath23 and consider the following moves on @xmath130 : 1 . : : homeomorphism of the diagram . using a homeomorphism from a surface @xmath23 to a surface @xmath35 , we carry @xmath130 to a diagram on @xmath35 . 2 . : : the two - point move . we isotop the lower circles of @xmath130 relative to the upper circles . if this isotopy is in general position , it reduces to a sequence of two - point moves , as shown in figure [ f : isomove ] . 3 . : : sliding one circle past another . suppose @xmath139 and @xmath140 are two circles of @xmath130 , both lower or both upper , and @xmath12 is an arc which connects @xmath139 to @xmath140 but does not cross any other circle . we add a detour to @xmath139 that follows @xmath12 , goes around @xmath140 , and then goes back along @xmath12 , as shown in figure [ f : handlemove ] . 4 . : : creating a trivial circle . we add a contractible circle to @xmath130 which is disjoint from all other circles of @xmath130 . 5 . : : stabilization . we remove a disk from @xmath23 which is disjoint from all circles of @xmath130 and replace it by a punctured torus with one lower circle and one upper circle , as shown in figure [ f : stablemove ] . given a bordism @xmath3 , we can construct a heegaard diagram from a function @xmath134 if @xmath134 has three properties : @xmath134 is a morse function , the descending manifold of each critical point of @xmath134 avoids other critical points of the same degree , and the descending manifolds index 2 critical points are transverse to the ascending manifolds of the index 1 critical points . following @xcite , the space @xmath141 of all smooth functions on @xmath3 has a stratification @xmath142 in which @xmath143 is the set of functions @xmath134 with all three properties . @xmath144 is a codimension 1 subspace of @xmath141 which is divided into three pieces : in each piece , an element @xmath134 fails to have one of the three desired properties . given two elements @xmath145 and @xmath146 of @xmath143 , we choose a path of functions @xmath147 in general position . the path avoids the codimension 2 subspace @xmath148 , and is transverse to @xmath144 . each time it crosses @xmath144 , the effect on the corresponding heegaard diagram is one of the above moves . for example , if @xmath147 fails to be a morse function , then a pair of critical points is created or destroyed , the heegaard diagrams corresponding to @xmath149 and @xmath150 differ by stabilization or creation of a trivial circle . if a descending manifold of a critical point lands on another critical point of the same degree , the corresponding move is sliding one circle past another . finally , if a descending manifold and an ascending manifold fail to be transverse , the effect is a two - point move . in order to say that our state models can be defined on a triangulation of a 3-manifold , we construct a canonical way to convert a triangulation to a heegaard diagram . since a triangulation is a kind of handle decomposition , we can let the lower handlebody be a tubular neighborhood of the 1-skeleton as usual . we add a lower circle for each edge and an upper circle for each face . an intersection between a lower circle and an upper circle corresponds to an ordered pair consisting of an edge and a face that contains the edge . let @xmath130 be a ( generalized ) heegaard diagram on a surface @xmath23 . we assume that @xmath130 is an oriented diagram , i.e. its circles are oriented ( we include reversing the orientation of a circle as a move on oriented heegaard diagrams ) . let @xmath1 be a finite - dimensional involutory hopf algebra of invertible dimension . we define the quantity @xmath151 as follows : to each upper circle @xmath152 , we assign the tensor : @xmath153 where the indices @xmath154 correspond to the crossings on @xmath152 in the order that they are encountered if we travel along @xmath152 in the positively oriented direction . we assign the tensor : @xmath155 to each lower circle in the same way . if , at a given crossing @xmath26 , the tangent vectors of the lower circle and the upper circle , in that order , form a positively oriented basis for @xmath156 at @xmath26 , we contract the tensors assigned to the circle at the index corresponding to @xmath26 . if the vectors form a negatively oriented basis , we interpose the antipode map before contracting : @xmath157 let @xmath158 be the result of this series of tensor products and contractions . ( note that if we choose a basis for @xmath1 , the summations corresponding to the contractions can alternatively be viewed as a sum over states , with @xmath158 the resulting partition function . ) define @xmath159 where @xmath160 is the number of upper circles , @xmath161 is the number of lower circles , and @xmath162 is the genus of @xmath23 . @xmath151 is an invariant of bordisms . in particular , it is an invariant of closed 3-manifolds . we show invariance under the different moves : \1 ) adding a trivial circle . this has the effect of contributing the following factor to @xmath158 : @xmath163 the factor is canceled by the normalization of @xmath151 . \2 ) stabilization . this contributes the following factor to @xmath158 : @xmath164 which is also canceled by normalization . \3 ) orientation reversal . an upper circle corresponds to a factor of @xmath158 of the form : @xmath165 the equality , which follows from lemma [ basic ] , demonstrates invariance . the same argument works for lower circles . \4 ) the two - point move . if the two circles are oriented properly , part of the expression for @xmath158 is of the form : @xmath166 the equality demonstrates invariance . ] \5 ) sliding a circle . we can assume that we are sliding one lower circle past another lower circle ( the corresponding move for upper circles is equivalent by duality ) . we assume as a representative case that both circles have three crossings . figure [ f : slide ] depicts the move after a possible change of orientation of the circles involved . the `` before '' picture corresponds to the following factor of @xmath158 : @xmath167 and the `` after '' picture corresponds to the following replacement for this factor : @xmath168 the following algebra demonstrates the equality of these two expressions : @xmath169 this completes the proof of invariance . the theorem warrants the notation @xmath170 , where @xmath3 is a bordism and @xmath130 is a generalized heegaard diagram for @xmath3 . we first observe that for a closed manifold @xmath3 , @xmath171 , and @xmath172 , where @xmath173 is @xmath3 with the oppose orientation . also , for any two bordisms , @xmath174 , and @xmath175 , @xmath176 , because we can choose a diagram for @xmath177 , the connected sum of @xmath174 and @xmath175 , which is a connected sum of diagrams of @xmath174 and @xmath175 . we will investigate the invariant @xmath2 in the special case that @xmath1 is a group algebra , and in the special case that @xmath3 is a link complement . let @xmath0 be a finite group and let @xmath87 $ ] be its group algebra . as before , view the elements of @xmath0 as a basis for @xmath87 $ ] . in this basis , the matrix entry @xmath178 is 1 when @xmath179 and 0 otherwise , while @xmath180 is 1 when @xmath181 ( as group elements ) , and 0 otherwise . the trace @xmath182 equals @xmath183 when @xmath184 , the identity , and 0 otherwise . the cotrace @xmath185 is 1 for all @xmath24 . thus , the tracial product @xmath186 equals @xmath187 when @xmath188 and 0 otherwise , while the tracial coproduct @xmath189 is 1 when all indices are equal and zero otherwise . let @xmath3 be a closed manifold or a manifold with some upper boundary ( but no lower boundary ) . viewing the invariant @xmath190)$ ] as a state model on a heegaard diagram , we observe that the weight of a state is zero unless for each lower circle the states assigned to all crossings on the circle are equal . thus , we may say that the states are assigned to the lower circles themselves . if the heegaard diagram comes from a triangulation of @xmath3 , the lower circles correspond to edges , so we may say that states are assigned to the edges . meanwhile , the interaction on the upper circles dictates that a state has zero weight unless the product of the group elements around a face is the identity . thus , the state model reduces to the one described in the introduction . let @xmath191 be the complement of ( an open neighborhood of ) a link @xmath59 , where we declare that the boundary of @xmath191 is upper boundary . we wish to reduce @xmath192 to a state model on a projection @xmath21 of @xmath59 ( we assume @xmath21 is a projection of @xmath59 onto the sphere rather than the plane ) . to construct a convenient handle decomposition for @xmath191 , we first consider the tiling of the sphere by squares which is dual to the link projection , as illustrated in figure [ f : squares ] . there will be one 3-handle @xmath193 , and one 0-handle @xmath39 ; imagine @xmath193 as the region above @xmath21 and @xmath39 as a big , fat vertex below @xmath21 . figure [ f : square](a ) shows a piece of @xmath191 corresponding to a given square in figure [ f : squares ] . we add a 1-handle for every edge in the tiling by squares , all attached to the lone 0-handle @xmath152 , as in figure [ f : square](b ) . next , for each square we add a 2-handle which is attached to the 1-handles on the left and right sides in figure [ f : square](c ) . finally , we add a second 2-handle which is attached to all four neighboring 1-handles , as in figure [ f : square](d ) . as before , we convert the handle decomposition to a heegaard diagram by taking the lower handlebody to be a tubular neighborhood of the 1-skeleton . our goal is a model in which states are assigned to the arcs between crossings of @xmath21 and interactions are assigned to the crossings themselves . equivalently , we can orient the link @xmath59 and assign a tensor of type ( 2,2 ) to each crossing , and wherever two crossings are connected , we contract the corresponding tensors . we convert the assembly of handles in figure [ f : square ] to tracial product and coproduct tensors and we modify the coproduct tensors to obtain a tensor of type ( 2,2 ) , to obtain the following replacement rule for a left - handed crossing : @xmath194 using corollary [ tracegen ] , we may simplify this tensor : @xmath195 in @xcite , kirillov and reshetikhin describe a link invariant for any representation of any hopf algebra with a special property called quasi - triangularity . in particular , for any hopf algebra @xmath1 ( involutory or not ) , there is a bigger hopf algebra @xmath34 , called the quantum double of @xmath1 , which can be defined on the vector space @xmath196 and which is always quasi - triangular . the algebras @xmath1 and @xmath77 are contained in @xmath34 by the inclusions @xmath197 and @xmath198 . consider the representation of @xmath34 on itself given by left multiplication , and let @xmath199 be the restriction of this representation to @xmath1 by the inclusion @xmath200 , where the trace @xmath35 is interpreted as an element of @xmath77 . in this context it is easy to show that reshetikhin and turaev s invariant in the special case of the representation @xmath199 of @xmath34 is the same as our invariant @xmath2 . presumably our invariant , in its full generality , is a special case of reshetikhin and turaev s extension of this invariant to 3-manifolds . to make the invariant @xmath2 look good , we define a formal calculus of expressions involving structure tensors of a hopf algebra , or more generally arbitrary expressions involving tensor products and contractions of tensors . we start by defining a contraction category as a natural setting for such computations . a _ contraction category _ is a set @xmath35 , whose elements may be called tensors , with additional structure consisting of two functions @xmath82 and @xmath201 from @xmath35 to @xmath202 , where @xmath203 is called the _ in - degree _ of @xmath30 and @xmath204 is called the _ outdegree _ ( we will also say that @xmath24 has _ type _ @xmath205 ) , and a function @xmath206 , the evaluation function of @xmath35 , where @xmath207 is the set of contraction graphs of @xmath35 . a _ contraction graph _ of @xmath35 is an oriented graph @xmath0 whose edges and vertices are labeled in a certain way . @xmath0 can be a graph in the broadest sense of the word : @xmath0 need not be connected ( and may be empty ) , an edge may go from a vertex to of vertices , an edge may go from a vertex to itself , and there may be edges whose heads or tails ( or both ) are not attached to any vertex . @xmath0 may also have degenerate `` edges '' which are oriented circles without heads or tails at all : @xmath208 the vertices of @xmath0 should be labeled by elements of @xmath35 . at a vertex labeled by @xmath30 , there should be @xmath203 inward edges and @xmath204 outward edges , the inward edges should be numbered from 1 to @xmath203 , and similarly for the outward edges . we also number the free heads of edges of @xmath0 from 1 to @xmath201 for some @xmath201 , and the free tails of edge of @xmath0 from 1 to @xmath82 for some @xmath82 . we call @xmath82 the in - degree of @xmath0 and @xmath201 the outdegree . in short , a contraction graph looks very much like a tensorial expression in arrow notation , where the `` tensors '' are elements of @xmath35 : @xmath209 the evaluation function @xmath86 should map a contraction graph @xmath0 to an element of @xmath35 of the same in - degree and outdegree as @xmath0 . moreover , it should satisfy an axiom of substitution : if @xmath0 and @xmath1 are contraction graphs and @xmath39 is a vertex with the same in - degree and outdegree as @xmath1 , we may define the _ composition _ @xmath210 by replacing the vertex @xmath39 by the graph @xmath1 : @xmath211 the axiom of substitution says that if @xmath212 is the label of @xmath39 , then @xmath213 . we give names to two distinguished tensors which exist in any contraction category . we call the value of the empty graph the _ identity scalar _ ( or one ) , and the value of a circular edge the _ dimension scalar_. if @xmath35 and @xmath214 are contraction categories , a morphism from @xmath35 to @xmath214 is a function from the tensors of @xmath35 to the tensors of @xmath214 which respects in - degree and outdegree and which commutes with the evaluation map . the term `` contraction category '' may seem inappropriate since we have not defined it as a kind of category . however , we can view it a category whose objects are ordered lists of left - pointing arrows and right - pointing arrows ( arrows in the sense of contraction graphs , that is ; we call category - theoretic arrows `` morphisms '' ) , and whose morphisms are contraction graphs with an ordered list of arrows on the left side and another ordered list on the right side : @xmath215 we could modify the definition of a contraction category in natural ways : we could replace occurrences of the word `` set '' by `` class '' , and we could label the arrows by elements of a new set ( or class ) @xmath12 . in its full generality , a contraction category appears to be the same as a strict , symmetric , compact , monoidal category @xcite . as an example of a contraction category , we may define @xmath216 to be the set of all tensors over a finite - dimensional vector space @xmath41 with ground field @xmath42 . in other words , @xmath217 . the evaluation function for contraction graphs may be constructed from the usual contraction and tensor product operations . we see that the value of the dimension scalar is the dimension of @xmath41 . if @xmath35 is any contraction category , one can consider a vector space @xmath41 together with a morphism @xmath218 to be a representation of @xmath35 . a _ hopf contraction category _ is a contraction category with distinguished elements called @xmath3 , @xmath80 , @xmath23 , @xmath83 , and @xmath82 which satisfy the usual identities of a hopf algebra . we may define the _ universal hopf category _ @xmath214 to be the contraction category generated by these five letters with the axioms they satisfy interpreted as relations . that is , the elements of @xmath214 are equivalence classes of contraction graphs on these five letters , where two contraction graphs @xmath0 and @xmath1 are equivalent if one can turn @xmath0 into @xmath1 by using the axioms of a hopf algebra as replacement rules . we can equivalently say that a hopf contraction category is a contraction category @xmath35 together with a morphism from the universal hopf category to @xmath35 . a hopf contraction category which happens to be a vector space is a hopf algebra . we can also define the universal , involutory hopf category and the universal , involutory hopf category of invertible dimension in which trace is a left and right integral , cotrace is a cointegral , and the dimension scalar has an inverse . anticipating the next section , we call the latter category the * heegaard category * , written @xmath219 . one can check that the heegaard category satisfies all of the identities of involutory hopf algebras of invertible dimension presented in section [ hopf ] . we can define the `` invariant '' @xmath220 as before , although its value is not a number , but rather a formal expression which might be as difficult to analyze as the original manifold . we will see that @xmath220 has the surprising property of being a complete invariant on closed , irreducible 3-manifolds . the value of @xmath220 involves factors of the dimension scalar for normalization . for simplicity we pass to the augmented heegaard model @xmath221 , in which we assume that the dimension scalar equals the identity , and we consider @xmath222 . section [ invariant ] gives a procedure for converting a generalized heegaard diagram @xmath130 to a contraction graph over the generating tensors of @xmath221 . we wish to construct an inverse to this procedure . [ inversep ] each element of @xmath221 is realized as @xmath222 for some bordism @xmath3 . let @xmath86 be a graph composed of the generators of @xmath221 . we may re - write @xmath223 , @xmath224 , @xmath225 and @xmath226 using corollary [ tracegen ] : @xmath227 substituting these expressions into @xmath86 , we get a new graph @xmath228 composed of the tracial product and coproduct tensors and the antipode . we build a heegaard diagram using the graph @xmath228 . for each tracial product tensor in @xmath228 , we construct an oriented annulus with an oriented core , which we call a `` lower circle '' , together with a labeled point for each index of the tensor , as in figure [ f : core](a ) . we do the same for each tracial coproduct , where we label the core as an `` upper circle '' . if a product tensor and a coproduct tensor are contracted together , the annuli should be identified , as in figure [ f : core](b ) . the annuli are identified in such a way that their orientations are compatible , and so that at the crossing point the tangent vectors to the lower circle and the upper circle , in that order , should be a positive basis . if a product tensor and a coproduct tensor are contracted together with the antipode map in between , we perform the same gluing operation , except we arrange the tangents to the lower and the upper circle to form a negative basis . the result of gluing these annuli together is a compact , oriented surface @xmath23 which may have many components and many boundary components . to convert @xmath23 to a heegaard diagram @xmath130 on a closed surface , we glue it to a surface @xmath35 which has the same boundary as @xmath23 . the result describes some bordism @xmath3 . by construction , @xmath229 . there is an ambiguity in the construction of the heegaard diagram @xmath130 , because we can choose many different surfaces @xmath35 with a given boundary . we may express this ambiguity by an equivalence relation . suppose that @xmath130 is a heegaard diagram and @xmath86 is a circle on the surface of @xmath130 which is disjoint from all of the upper and lower circles of @xmath130 . then we can cut the surface of @xmath130 along @xmath86 and attach disks along the two resulting boundary components . call this a _ circle move _ on heegaard diagrams . suppose instead that the surface of @xmath130 has component which is a 2-sphere with no upper or lower circles . then we can call the deletion of this component a _ sphere move _ on @xmath130 . proposition [ inversep ] assigns a unique equivalence class ( under the sphere and circle moves ) of heegaard diagrams to each contraction graph @xmath86 . by extension , we can say that two bordisms @xmath174 and @xmath175 differ by an elementary move if there exist diagrams for @xmath174 and @xmath175 which differ by a sphere or a circle move . we can then consider equivalence classes of bordisms . it is easy to check that if @xmath3 is a bordism , then the bordisms produced by proposition [ inversep ] from @xmath222 is equivalent to @xmath3 . one can also check that if two contraction graphs @xmath139 and @xmath140 differ by one of the defining relations of @xmath221 , then the corresponding bordisms are equivalent ( thereby establishing a converse to the proof of invariance in section [ invariant ] ) . thus , @xmath230 if and only if @xmath174 and @xmath175 are equivalent under sphere and circle moves . ] we examine the effect of sphere and circle move on the topology of a bordism @xmath3 . the sphere move corresponds to discarding a component of @xmath3 which is a 3-sphere ; we may call this move on manifolds a * 3-sphere move*. the circle move is more complicated . let @xmath130 be a diagram for @xmath3 and let @xmath23 be the surface of @xmath130 . recall that @xmath3 is obtained by gluing disks and then balls to @xmath231 $ ] . let @xmath86 be a circle on the surface of @xmath130 suitable for a circle move . the circle move corresponds to cutting @xmath3 along the annulus @xmath232 $ ] , and then gluing a 2-handle to each of the two annular scars that result ( see figure [ f : annulusmove ] ) . the operation may produce spherical boundary components , which are irrelevant by the definition of a bordism . we call this operation an _ annulus move _ on @xmath3 . it may happen that one or both of the circles @xmath233 and @xmath234 is trivial in its respective boundary component . in this case the annulus move simplifies @xmath3 and we give it a second name as follows : suppose that only one of the circles @xmath234 and @xmath235 is non - trivial , and assume that it is @xmath234 . we choose an embedded disk @xmath23 in the lower handlebody which bounds @xmath235 . then the circle move corresponds to compressing the boundary of @xmath3 along the disk @xmath236 $ ] . we call this a _ disk move _ on @xmath3 . if both circles are trivial , then we can find a disk @xmath237 in the lower boundary and a disk @xmath238 in the upper boundary so that @xmath239 \cup s_1 $ ] is a 2-sphere , and the circle move on the diagram of @xmath3 corresponds to cutting along this sphere , i.e. the inverse connect - sum operation . we call this a _ 2-sphere move _ on @xmath3 . if a 2-sphere in @xmath3 bounds a ball , then the only effect of the corresponding 2-sphere move is to add a 3-sphere component to @xmath3 , which is reversed by the 3-sphere move . we call such a 2-sphere move _ degenerate_. we will also refer to disk moves and 2-sphere moves as degenerate annulus moves , and to 2-sphere moves as degenerate disk moves . we would like to determine when two bordisms are rendered equivalent by a sequence of annulus moves , disk moves , and non - degenerate 2-sphere moves and the inverse of these moves ( and we apply the 3-sphere move whenever possible ) . we first observe that these three moves are _ destructive _ in the sense that they simplify a bordism , and their inverses are constructive ; only degenerate 2-sphere moves are neither destructive nor constructive . specifically , we define the _ height _ of a bordism @xmath3 to be the ordinal @xmath240 , where @xmath26 , the _ circle number _ of @xmath241 , is the maximum number of disjoint , non - trivial , non - parallel circles in the boundary of @xmath3 , and @xmath16 , the _ kneser number _ of @xmath3 , is the maximum number of disjoint , non - trivial , non - parallel 2-spheres in @xmath3 . ( since @xmath3 is actually an equivalence class of 3-manifolds and some members of this class may have 2-sphere boundary , we must stipulate that a 2-sphere which is parallel to the boundary of a representative of @xmath3 is trivial . ) we may take @xmath242 ; see @xcite for a proof that @xmath16 is finite . clearly , destructive moves decrease the height of a bordism . thus , any sequence of destructive moves , or _ path of destruction _ , is finite . the main result of this section is a uniqueness theorem : if @xmath3 is a bordism , then any two maximal paths of destruction produce the same result . thus if no destructive moves can be performed on two different bordisms @xmath174 and @xmath175 ( e.g. if @xmath174 and @xmath175 are closed and prime ) , then they are either equivalent or @xmath243 . this result is a special case of the jaco - shalen - johannson decomposition theorem , as presented in @xcite and @xcite . we give a self - contained proof here : we outline the plan for a complicated proof by induction . let @xmath3 be a counterexample of minimum height , and consider two possible destructive moves @xmath244 and @xmath245 on @xmath3 , and call the resulting manifolds @xmath246 and @xmath247 . if we can find two more moves @xmath28 and @xmath248 ( which may or may not be degenerate ) such that @xmath249 , then by induction any two maximal paths of destruction beginning with @xmath244 and @xmath245 produce the same result , as indicated in figure [ f : paths](a ) . the moves @xmath244 and @xmath245 involve cutting along surfaces @xmath250 and @xmath251 . if these two surfaces are disjoint , we can directly induct on the height of @xmath3 , because we can perform @xmath244 after @xmath245 and vice - versa , with @xmath252 . if @xmath250 and @xmath251 intersect , we put them in general position and induct on the number of components of their intersection ( each of which is an arc or a circle ) . it suffices to find another destructive move @xmath248 ( possibly after switching @xmath244 and @xmath245 ) such that @xmath253 has fewer components than @xmath254 , and there exist moves @xmath39 and @xmath28 such that @xmath255 , as indicated in figure [ f : paths](b ) . suppose that one of the components of @xmath254 is a circle which is contractible in one of @xmath250 and @xmath251 , say @xmath251 . we may choose an innermost circle @xmath256 inside this one . @xmath86 bounds a disk @xmath257 whose interior does not intersect @xmath250 , and @xmath86 divides @xmath250 into two surfaces @xmath238 and @xmath258 , as shown in figure [ f : innercircle](a ) . the move @xmath244 consists of cutting along @xmath250 and gluing in either 2-handles or balls , and in both cases we can find a disk @xmath259 in the region attached to @xmath250 which makes a sphere with @xmath130 , as shown in figure [ f : innercircle](b ) . we choose two surfaces @xmath260 and @xmath261 which run parallel to @xmath262 and @xmath263 but which do not intersect them or each other , as shown in figure [ f : innercircle](c ) . cutting along @xmath260 and cutting along @xmath261 are valid moves , and cutting along both is equivalent to performing @xmath244 and then cutting along the sphere @xmath264 . therefore at least one of @xmath260 and @xmath261 must be a destructive move , and we may assume it is @xmath260 . at the same time , @xmath260 has less intersection with @xmath251 than @xmath250 does . we let @xmath248 be the move corresponding to @xmath260 . ] from this point on , we suppose that there is no circular component of @xmath254 which is contractible in either @xmath250 or @xmath251 ; in particular , neither @xmath244 nor @xmath245 can be a 2-sphere move . suppose instead that some component is an arc with both endpoints in upper or lower boundary , say upper . this arc , together with part of the boundary of @xmath251 , bounds a disk in @xmath251 . we let @xmath265 be the innermost arc in this disk . @xmath12 , together with part of the boundary of @xmath251 , bounds a disk @xmath257 whose interior does not intersect @xmath250 , and it divides @xmath250 into two surfaces @xmath238 and @xmath258 . we argue as in the previous case , the only difference being that @xmath264 is now a disk instead of a 2-sphere , as shown in figure [ f : innerarc ] . ] thus , we may assume that both @xmath244 and @xmath245 are annulus moves . there are two possibilities : either @xmath250 and @xmath251 intersect in concentric circles which lie between the upper and lower boundary of @xmath250 and @xmath251 , as in figure [ f : annulus](a ) ; or they intersect in radial arcs , each arc connecting the upper and lower boundary of @xmath250 and @xmath251 , as in figure [ f : annulus](b ) . in the first case , let @xmath86 be the circle of @xmath254 which is uppermost in @xmath251 . @xmath86 bounds an annulus @xmath12 in @xmath251 which does not intersect @xmath250 in the interior . it divides @xmath250 into an upper annulus @xmath266 and a lower annulus @xmath267 , as in figure [ f : concentric ] . cutting along @xmath268 is necessarily a destructive move , because its upper boundary is a non - trivial circle . we call this move @xmath248 . the circle @xmath86 bounds a disk @xmath269 in the 2-handle which is attached to @xmath268 in the move @xmath248 , and @xmath86 bounds a disk @xmath259 in the 2-handle attached in the move @xmath244 . we see that the move @xmath248 followed by compressing the disk @xmath270 yields the same result as the move @xmath244 followed by compressing the disk @xmath271 . as before , if we move @xmath268 away from the annulus @xmath12 , it intersects @xmath251 less than @xmath250 does . finally , suppose that @xmath250 and @xmath251 intersect in radial arcs . let @xmath272 be the boundary of a tubular neighborhood of @xmath273 . @xmath272 is the union of disjoint annuli which do not intersect either @xmath250 or @xmath251 . if we perform all of the corresponding annulus moves , then @xmath250 and @xmath251 lie in a component of the resulting bordism which is equivalent to @xmath231 $ ] for some closed surface @xmath23 . any maximal path of destruction removes this component entirely . thus , @xmath250 and @xmath251 are rendered equivalent by the annulus moves represented by @xmath272 . although @xmath220 is a complete formal invariant for closed 3-manifolds , it is not necessarily a complete invariant in the computational sense , because it may be as difficult to distinguish elements of @xmath274 as it is to distinguish 3-manifolds by the standard means . the heegaard category @xmath274 is residually representable . that is , if @xmath24 and @xmath25 are distinct elements of @xmath275 , there exists a vector space @xmath41 and a morphism @xmath276 such that @xmath277 . if this conjecture is true , @xmath220 is a complete invariant in a more meaningful sense . a few special cases of this conjecture might be easy to settle . is there an involutory hopf algebra @xmath1 for which @xmath278 , i.e. @xmath279 ? even if @xmath2 is insensitive to orientation , perhaps there is a hopf algebra that distinguishes @xmath280 from @xmath281 , or @xmath282 from @xmath283 ( correspondingly , @xmath284 from @xmath285 and @xmath286 from @xmath287 ) . the construction given in this paper is almost certainly not as general as it could be . for example , dijkgraaf @xcite mentions the following state model on triangulations which was described to me by vaughan jones : as before , we choose a finite group @xmath0 to be the state set , we assign states to the oriented edges of a triangulation , and we assign an interaction to each face which is 0 unless the product of the states of the face s edges is the identity . however , we also have an interaction for each tetrahedron , which we assume is non - zero whenever the value of the tetrahedron s faces is non - zero . we may view this interaction as a complex - valued function @xmath288 , with @xmath289 , and therefore as a cochain on the canonical triangulation of @xmath290 with coefficients in the multiplicative group @xmath291 . the constraint that the state model is a topological invariant is equivalent to the condition that this cochain is actually a cocycle , and two different cocycles yield the same invariant if they are cohomologous . thus , we get an invariant for every pair consisting of a finite group and an element of @xmath292 . if we choose the trivial cohomology class , we get @xmath293)$ ] . is there a mutual generalization of this invariant and @xmath2 ? can the invariant be extended to non - involutory hopf algebras ? as was mentioned in section [ prop ] , @xmath2 corresponds to a certain link invariant involving the quantum double of @xmath1 . can it be generalized to quasi - triangular hopf algebras other than the quantum double ? peter freyd , david yetter , jim hoste , w. b. r. lickorish , kenneth millett , and adrian ocneanu , _ a new polynomial invariant of knots and links _ , bull . ( n.s . ) * 12 * ( 1985 ) , no . 2 , 183312 . anatol n. kirillov and nicolai yu . reshetikhin , _ representations of the algebra @xmath295 , @xmath296-orthogonal polynomials , and invariants of links _ , infinite - dimensional lie algebras and groups ( luminy - marseille , 1988 ) , world sci . publishing , 1989 , pp .

_ a statistical field theory is developed to explore the density of states and spatial profile of ` tail states ' at the edge of the spectral support of a general class of disordered non - hermitian operators . these states , which are identified with symmetry broken , instanton field configurations of the theory , are closely related to localized sub - gap states recently identified in disordered superconductors . by focusing separately on the problems of a quantum particle propagating in a random imaginary scalar potential , and a random imaginary vector potential , we discuss the methodology of our approach and the universality of the results . finally , we address potential physical applications of our findings . _

cond-mat0110467

over recent years the statistical properties of stochastic non - hermitian operators have come under intense scrutiny @xcite . operators of this kind appear in a number of physical applications ranging from random classical dynamics , and statistical physics , to phase breaking and relaxation in quantum dynamics . for example , a straightforward mapping shows the statistical mechanics of a repulsive polymer chain can be described in terms of the quantum dynamics of a particle subject to a random imaginary scalar potential @xcite . similarly , the statistical mechanics of flux lines in a type - ii superconductor pinned by a background of impurities can be described as the quantum evolution of a particle in a disordered environment subject to an imaginary vector potential @xcite . finally , the diffusion of a classical particle advected by a random velocity field is described by a linear _ non - hermitian _ operator , the `` passive scalar '' equation @xcite . more generally , the dynamics of various classical systems can be expressed in terms of random fokker - planck operators ( for a review see , e.g. , refs . @xcite ) . non - hermitian operators exhibit several qualitatively new phenomena which discriminate their behaviour from those of their random hermitian counterparts . firstly , while the eigenvalues of hermitian operators are real , the spectrum of non - hermitian operators is generically complex , bound by some region of support in the complex plane . the second striking distinction concerns the localization properties of the eigenfunctions . it is well established that a weak random impurity potential brings about the localization of all eigenstates of a hermitian operator in dimensions of two and below @xcite . by contrast , a constant imaginary vector potential is sufficient to delocalize states of the disordered system even in one - dimension ! this phenomenon , which was reported and explained by hatano and nelson @xcite , finds extension to higher dimensions @xcite . beginning with early work on random matrix ensembles @xcite , a variety of techniques have been developed to study the statistical properties of stochastic non - hermitian operators @xcite . some of the techniques have been based on the random matrix theory @xcite , while others have involved the development of perturbative schemes , such as the self - consistent born approximation in the diagrammatic analysis ( e.g. , ref . @xcite ) . in this paper we will be concerned with adapting a third approach which involves the refinement of field theoretic techniques which have proved to be so useful in the description of random hermitian operators . recently , attention has been directed towards the study of ` tail states ' which persist at the edge of the spectral support of the non - hermitian system . indeed , there are already indications @xcite that an important role can be played by those parts of the spectrum which are populated by rare states . now , in stochastic _ hermitian _ systems a small tail of states accumulate below the band edge tightly bound to rare or ` optimal fluctuations ' of the random potential @xcite . these so - called ` lifshitz tail ' states are typically smooth , node - less , localized functions which inhabit regions where the potential is particularly shallow and smooth . by contrast , the character of the complex spectrum in non - hermitian operators allows for the existence of tail states that occupy the entire region which bounds the support of the spectrum . the difference is not incidental : tail states in the non - hermitian system can exhibit features characteristic of the quasi - classical states within the bulk spectrum . to emphasize the point , consider the spectrum of , say , a two - dimensional free quantum particle subject to a complex scalar potential which involves both real _ and _ imaginary components , @xmath0 where @xmath1 . if the imaginary potential @xmath2 is absent , a region of localized tail states accumulate below the band edge ( @xmath3 ) due to optimal configurations of the potential . at energies @xmath4 greatly in excess of the inverse scattering time @xmath5 , bulk states are only ` weakly localized ' with a localization length @xmath6 greatly in excess of scattering mean - free path @xmath7 and wavelength @xmath8 . this is the quasi - classical regime where mechanisms of quantum interference strongly affect the spectral and transport properties of the system @xcite . on this background , let us now suppose that a small imaginary component of the potential is restored . being non - hermitian , the eigenvalues migrate into the complex plane occupying a finite region of support centered on the real axis ( see fig . [ fig : specs ] ) . at the level of the mean - field analysis ( detailed below ) , complex eigenvalues are confined to a sharp region of support in the complex plane . however , as with lifshitz states , optimal fluctuations of the random potential(s ) generate states which lie outside the region of support ( see fig . [ fig : specs ] ) . what is the nature of these states ? are they localized and , if so , over what length scale ? here , in contrast to the hermitian system , tail states must accommodate fast fluctuations of the wave function at the scale of the wavelength of the hermitian system . ) , where the potentials @xmath2 and @xmath9 are both drawn from a gaussian @xmath10-correlated random impurity distribution . here we have included a lattice of @xmath11 @xmath12-points in the momentum space . ] to our knowledge , the problem of tail states in the non - hermitian system was first considered in a recent work by izyumov and simons @xcite . refining an approach developed in refs . @xcite to explore tail states in hermitian systems , the authors introduced a non - perturbative scheme to investigate states at the edge of the support in non - hermitian systems . this instanton technique revealed the existence of tail states at the edge which are localized on a length scale greatly in excess of the wavelength tail states in the non - hermitian system are quasi - classical in nature . building on the work of efetov @xcite , the aim of the present paper is to develop a supersymmetric field theory to explore the spatial profile of the states in the tail region , and to determine the complex density of states ( dos ) in the vicinity of the band edge . in doing so , we will reveal a close correspondence between tail states in the non - hermitian problem and sub - gap states in weakly disordered superconductors . moreover , constraints within the present scheme reveals the universal character of tail states in the non - hermitian system showing that , after a suitable rescaling , the profile of the dos tail depends only on dimensionality and separation from the support . in particular , in contrast to lifshitz states in the hermitian system , properties of the tail states are universal , independent of the nature of the random distribution . to illustrate the generality of our approach we will consider two examples which , according to the classification scheme defined below , belong to different fundamental symmetry . firstly , following our discussion above , we will investigate the properties of tail states in a quantum system involving the propagation of a particle in a random _ scalar _ potential involving both real _ and _ imaginary components ( [ eq : hamil ] ) . secondly , we will investigate the nature of tails states in a quantum system where the particle is subject to a random real scalar potential and a random imaginary _ vector _ potential , @xmath13 ^ 2}{2 m } } + w({{\mathbf r } } ) \ ; . \label{eq : hamiv}\ ] ] as well as the hatano - nelson system @xcite , where @xmath14 , has been studied in the context of random classical dynamics involving classical diffusion in the background of a quenched random velocity field . identifying @xmath15 with the classical diffusion constant @xmath16 , and @xmath17 with a random velocity field @xmath18 , a redefinition of the random potential @xmath19 obtains the classical operator @xmath20 in the quasi - classical approximation , ( i.e. setting @xmath21 as the complex energy , where @xmath22\gg \hbar$ ] ) our goal will be to show that the long - range , low - energy properties of the random systems are described by a supersymmetric non - linear @xmath23-model . at the level of mean - field , the spectrum of both non - hermitian operators will be shown to be bound to a region of support inside the complex plane . by taking into account non - perturbative symmetry breaking instanton field configurations of the action , we will reveal the existence of tail states which extend beyond the region of support predicted by the mean - field . we will determine the localization properties and profile of the dos revealing an intrinsic universality of the present scheme . the paper is organized as follows : after introducing a general scheme to classify linear non - hermitian operators on the basis of their fundamental symmetries , in section [ sec : field ] we formulate a statistical field theory to describe the spectral properties of the random imaginary scalar potential hamiltonian . within this approach , we show that the large scale properties of the spectral support are determined by the saddle - point or mean - field properties of the theory . motivated by a parallel description of the disordered superconducting system , in section [ sec : state ] we show that optimal fluctuations of the random impurity potential induce localized tail states which extend beyond the region of support . within this approach , we determine analytical expressions for the spatial extent of the localized states in the tail region , and determine the scaling of the complex dos as a function of energy . generalizing the scheme to the consideration of the imaginary vector potential hamiltonian in section [ sec : vecto ] , we emphasize the universality of the present scheme . results from both sections are compared with numerical simulations . finally , in section [ sec : concl ] we conclude our discussion . the aim of this section is to prepare a statistical field theory of the weakly disordered non - hermitian system . to be concrete , in the first instance , we will focus on the problem of the imaginary scalar potential . later , we will see that the field theoretic scheme is easily modified to accommodate the constant vector potential perturbation . however , before embarking on this program , it is useful to first contrast general properties of complex linear non - hermitian operators with those of their hermitian counterparts . as well as helping to clarify the idiosyncrasies of the non - hermitian system , we will expose a general scheme in which non - hermitian operators can be classified according to their fundamental symmetries . a general non - hermitian operator @xmath24 is specified by left and right eigenfunctions , @xmath25 where @xmath26 denotes the set of complex eigenvalue . although not a generic feature of non - hermitian operators , throughout we will rely on the assumption that the eigenfunctions form a complete basis set . this is the case if there are no repeated eigenvalues , which will be true for a generic random operator such as those studied here . furthermore we will take the eigenstates to be orthonormal , @xmath27 now as with hermitian operators , spectral ( and localization ) properties of the non - hermitian hamiltonian can be expressed through the complex green function @xmath28 where @xmath29 denotes the complex argument . inserting the resolution of identity , one obtains the spectral decomposition , @xmath30 using the analytical properties of the resolvent , it is straightforward to show that the density of complex eigenvalues @xmath31 is expressible in terms of the complex green function through the identity : @xmath32 the complex green function @xmath33 is , therefore , a non - analytic function everywhere in the complex @xmath21-plane in which the corresponding eigenvalues density is non - vanishing . previous studies have shown that techniques based on standard diagrammatic perturbation theory account only for contributions to @xmath34 which are _ analytic _ in @xmath21 @xcite . to account for all contributions , it is convenient to follow the now standard route @xcite of expressing the green function through an auxiliary hamiltonian which is explicitly hermitian . this is achieved by constructing a matrix hamiltonian , @xmath35 , with the following @xmath36 block structure @xmath37 in this representation , the green function of the non - hermitian operator is expressed as the off - diagonal element of the matrix green function @xcite , @xmath38 where @xmath39 . this ` method of hermitization ' affords a convenient way of classifying the symmetries of a non - hermitian hamiltonian according to the fundamental symmetries of its hermitian counterpart . the utility of this classification will become manifest presently in the construction of the statistical field theory of the non - hermitian system . as an example , let us consider the hamiltonian involving an imaginary scalar potential . applied to this problem , the hermitization procedure leads to the matrix hamiltonian @xmath40 where the pauli matrices @xmath41 act in the @xmath36 sub - space , and @xmath42 . in fact , the matrix structure of @xmath35 bares much in common with the gorkov or bogoliubov - de gennes quasi - particle hamiltonian of a weakly disordered superconductor @xcite . to foster this analogy let us implement the canonical transformation @xmath43 where @xmath44 $ ] . associating the matrix structure with a ` particle / hole ' space , the operator @xmath45 can be identified as a gorkov hamiltonian for a disordered superconductor . here @xmath46 plays the role of the chemical potential , while @xmath47 denotes the bare hamiltonian , and @xmath48 represents the ( in this case real , random ) order parameter . however , taking the analogy further , the energy argument in the gorkov green function corresponds , in this case , to the infinitesimal @xmath49 of the matrix green function @xmath50 evidently , we are interested in the zero energy quasi - particle states of the disordered , time - reversal invariant , superconductor with a random real order parameter . in its hermitian form , the hamiltonian above can be classified according to its fundamental symmetries . according to the classification introduced by zirnbauer @xcite , the matrix construction places a general non - hermitian hamiltonian in the chiral unitary symmetry class denoted aiii the notation adopted by zirnbauer is motivated by that introduced by cartan to classify the @xmath51 symmetric spaces . ] , a general element of which has the form @xmath52 where @xmath53 is an arbitrary complex matrix . however , within this group , there are four subclasses of higher symmetry : class ci ( the time - reversal invariant superconductor ) where matrices @xmath53 are complex symmetric ; class bdi ( chiral orthogonal ) where the elements of @xmath53 are arbitrary and real ; class cii where @xmath53 has an underlying symplectic structure ; and class diii where @xmath53 is complex antisymmetric . as will be clear from the forthcoming analysis , the statistical field theory describing the non - hermitian spectral properties has soft modes associated with each universality class . from our discussion above , the imaginary scalar potential is accommodated in the symmetry class ci . as the second example , let us consider the non - hermitian hamiltonian involving the random imaginary vector potential . in this case , the hermitization procedure leads to the matrix hamiltonian @xmath54 as a result , we find that the system belongs to a different class depending on the value of the imaginary component of the energy argument @xmath55 . along the real axis ( i.e. for @xmath56 ) @xmath35 is real and therefore belongs to the symmetry class bdi . away from the real axis the hamiltonian becomes complex and the symmetry is lowered to class aiii . later we will discuss the physical manifestations of the discrete symmetries in the two limits . with this background , let us now turn to the construction of a field theory to describe statistical correlations of the non - hermitian model . motivated by the correspondence outlined above , the analysis will parallel previous investigations of the weakly disordered superconductor @xcite . the starting point is the generating functional for the single - particle green function . adopting throughout the convention @xmath57 , the latter is expressed as a field integral involving two independent four component supervector fields @xmath58 and @xmath59 with elements having both commuting / anti - commuting and ` particle / hole ' indices : @xmath60 = \int d ( \bar{\psi } , \psi ) \exp \left[i \int d { { \mathbf r } } \ ; \bar{\psi } \left ( i \eta - \hat{\mathcal{h}}\right ) \psi + \int d { { \mathbf r } } \left(\bar{\psi } j + \bar{j } \psi\right)\right ] \ ; .\ ] ] elements of the green function can be generated from the source terms @xmath61 and @xmath62 . for example , from the generating function one can obtain the dos according to the identity : @xmath63 \right|_{j = 0 } \ ; . \label{eq : sourc}\ ] ] to be concrete , let us first consider the properties of the hamiltonian involving the imaginary scalar potential . to exploit the analogy with the superconducting system , it is again convenient to implement the gauge transformation @xmath64 whereupon the generating functional assumes the form @xmath65= \int d ( \bar{\psi } , \psi ) \exp \left[i \int d { { \mathbf r } } \ ; \bar{\psi } ( { { \mathbf r } } ) \left(i \eta - \hat{\mathcal{h}}_\gamma\right ) \psi ( { { \mathbf r } } ) \right ] \ ; , \label{eq : gauge}\ ] ] where @xmath45 is defined above . now , as with the superconductor , the matrix hamiltonian satisfies the particle / hole or ` charge conjugation ' ( @xmath66 ) symmetry @xmath67 the latter is responsible for quantum interference effects which modify the long - range or low - energy properties of the average green function . to accommodate these effects it is convenient to double the field space @xmath68 where , defining the pauli matrix @xmath69 which operates in the charge - conjugation ( @xmath66 ) space , @xmath70 as a result , the generating functional takes the form : @xmath65 = \int d(\bar{\psi},\psi ) \exp\left\{i\int d{{\mathbf r } } \ ; \bar{\psi } ( { { \mathbf r } } ) \left[i \eta \sigma_3^{{\text{\sc{cc } } } } + \left(x - \hat{\zeta}_{\hat{{{\mathbf p } } } } - w\right ) \sigma_3 -\left(y - v\right ) \sigma_2\right ] \psi ( { { \mathbf r}})\right\ } \ ; .\ ] ] with this definition , the fields @xmath71 and @xmath72 are not independent but obey the symmetry relations @xmath73 with @xmath74 , where @xmath75 and @xmath76 are the projectors on the boson - boson and fermion - fermion sector respectively . this completes the construction of the generating functional as a functional field integral . to make further progress , we can draw on the intuition afforded by existing studies of the superconducting system . as a first step towards the construction of an effective low - energy theory it is necessary to subject the generating functional to an ensemble average over the random impurity distribution . for this purpose we will take both the real and imaginary components of the random scalar potential to be gaussian @xmath10-correlated with zero mean , and correlation @xmath77 where @xmath78 represents the unperturbed dos of the hermitian hamiltonian @xmath79 at energy @xmath46 , while @xmath80 and @xmath81 denote , respectively , the mean scattering time of the real and imaginary components of the potential . in the following , we will suppose that the imaginary component of the scattering potential is weak , i.e. @xmath82 . the ensemble averaging over the real random scalar potential @xmath19 induces an interaction of the fields which can be decoupled by the introduction of an @xmath83 component hubbard - stratonovich field @xmath84 , @xmath85 \rangle_w = \int d q \exp \left[\int d { { \mathbf r } } \left({\displaystyle\frac{\pi \nu}{8 \tau } } \operatorname{str}q^2 - { \displaystyle\frac{1}{2 \tau } } \bar{\psi } q \sigma_3 \psi\right ) \right ] \ ; .\ ] ] transforming the fields according to the symmetry , one finds that the supermatrix fields @xmath86 must be subjected to the constraint : @xmath87 similarly , the interaction of the fields induced by the ensemble average over the random imaginary scalar potential @xmath88 can be decoupled by an @xmath83 component hubbard - stratonovich field @xmath89 , satisfying the symmetry constraint , @xmath90 : @xmath91 \rangle_v = \int d p \exp \left[\int d { { \mathbf r } } \left({\displaystyle\frac{\pi \nu}{8 \tau_{n } } } \operatorname{str}p^2 - { \displaystyle\frac{1}{2 \tau_n } } \bar{\psi } p \sigma_2 \psi\right ) \right ] \ ; .\ ] ] finally , an integration over the fields @xmath71 obtains @xmath92 \rangle_{w , v } = \int dq \int dp \exp \left\{{\displaystyle\frac{\pi \nu}{8 } } \int d{{\mathbf r } } \ ; \operatorname{str}\left({\displaystyle\frac{q^2}{\tau } } + { \displaystyle\frac{p^2}{\tau_n}}\right ) - { \displaystyle\frac{1}{2 } } \int d{{\mathbf r } } \ ; \operatorname{str}\langle { { \mathbf r } } | \ln \hat{\mathcal{g}}^{-1 } | { { \mathbf r } } \rangle \right\ } \ ; , \intertext{where } \hat{\mathcal{g}}^{-1 } = i \eta \sigma_3 \otimes \sigma_3^{{\text{\sc{cc } } } } + x - \hat{\zeta}_{\hat{{{\mathbf p } } } } - iy\sigma_1 + { \displaystyle\frac{i}{2 \tau } } q - { \displaystyle\frac{1}{2 \tau_n } } p \sigma_1 \end{gathered}\ ] ] represents the supermatrix green function . further progress is possible only within the a saddle - point approximation . following ref . @xcite , in the quasi - classical approximation @xmath93 $ ] , it is convenient to implement a two - level procedure in which we first formulate an intermediate energy scale action in which the terms in @xmath55 and @xmath94 are treated as small symmetry breaking sources . later , the influence of @xmath2 ( through @xmath95 ) on the complex dos can be explored within a stationary analysis of this reduced action . neglecting the terms in @xmath55 and @xmath95 , a variation of the action with respect to @xmath84 obtains the saddle - point equation @xmath96 in the quasi - classical limit , @xmath97 , taking @xmath84 to be homogeneous in space , this equation can be solved in the pole approximation with @xmath98 . taking into account the analytical properties of the average green function , we identify the solution @xmath99 . however , in the limit @xmath100 , the saddle - point solution is not unique , but expands to span a degenerate manifold of solutions generated by the homogeneous pseudo - unitary transformations @xmath101 compatible with the symmetry of @xmath84 , . fluctuations around the saddle - point can be classified into longitudinal and transverse modes according to whether or not they violate the non - linear constraint @xmath102 . in the quasi - classical limit , the longitudinal fluctuations are rendered massive and do not contribute to the low - energy properties of the generating functional . on the other hand , transverse fluctuations of the fields become soft and must be accommodated . taking into account soft fluctuations @xmath103 , and restoring the perturbations @xmath55 and @xmath95 , a gradient expansion of the action obtains the average generating functional @xmath92 \rangle_{w , v } = \int_{q^2=\mathbb{i } } dq \int dp \exp \left\{{\displaystyle\frac{\pi\nu}{8\tau_n } } \int d { { \mathbf r}}\ ; \operatorname{str}p^2\right\ } \\ \exp\left\{{\displaystyle\frac{\pi \nu}{8 } } \int d{{\mathbf r}}\ ; \operatorname{str}\left[d ( \nabla q)^2 + 4 i \left ( i \eta \sigma_3 \otimes \sigma_3^{{\text{\sc{cc } } } } - iy \sigma_1 - { \displaystyle\frac{p}{2\tau_n } } \sigma_1 \right ) q \right]\right\}\ ; , \end{gathered}\ ] ] where @xmath104 is the diffusion constant . here we have neglected terms higher order in @xmath105 and @xmath106 . in the same approximation @xmath107 , the supermatrix green function takes the form @xcite @xmath108 where @xmath109^{d/2 -1 } j_{d/2 -1 } ( k_f r ) \exp [ - r/2 \ell]$ ] is the friedel function , and @xmath110 . finally , integrating over the supermatrices @xmath95 , we obtain the average generating functional @xmath111 \rangle_{w , v } = \int_{q^2=\mathbb{i } } dq \exp \left\{-s[q]\right\ } \ ; , \intertext{where } s[q ] = -{\displaystyle\frac{\pi \nu}{8 } } \int d { { \mathbf r } } \ ; \operatorname{str}\left[d ( \nabla q)^2 -4 \left(\eta \sigma_3 \otimes \sigma_3^{{\text{\sc{cc } } } } - y \sigma_1\right ) q + { \displaystyle\frac{1}{\tau_n } } \left(\sigma_1 q\right)^2 \right ] \ ; . \label{eq : deeft}\end{gathered}\ ] ] this completes the construction of the intermediate energy scale action for the non - hermitian system . returning to the analogy with the superconductor , we note that the form of the action coincides with that of a disordered superconductor subject to a random real order parameter with an average magnitude @xmath55 and variance set by @xmath94 . to investigate the influence of the imaginary potential on the complex dos , it is necessary to subject this action to a further stationary analysis and obtain the low - energy action . although the soft mode action is stabilized by the quasi - classical parameter @xmath112 , the terms in @xmath55 and @xmath94 render the majority of field fluctuations massive . to obtain the low - energy saddle - point , let us again vary the action with respect to @xmath84 subject to the non - linear constraint , @xmath113 : @xmath114 + { \displaystyle\frac{1}{2 \tau_n } } [ \sigma_1 q \sigma_1 , q ] = 0 \ ; .\ ] ] applying the _ ansatz _ @xmath115 , where @xmath116 , the saddle - point equation assumes the form @xmath117 - { \displaystyle\frac{1}{\tau_n } } \sin 2 \hat{\theta } = 0 \ ; , \label{eq : ineqn}\ ] ] a result reminiscent of the abrikosov - gorkov mean - field equation of the disordered superconductor subject to a pair - breaking perturbation @xcite . taking the solution to be symmetric and homogeneous in space , one obtains @xmath118 - { \displaystyle\frac{\pi}{2 } } \operatorname{sign}(y ) & |y| \tau_n \ge 1 \ ; . \end{cases } \label{eq : mfsol}\ ] ] once again , as @xmath100 , this solution is not unique but expands to span an entire manifold parameterized by transformations @xmath119 , where @xmath120 = 0 $ ] and @xmath121 these represent the soft fluctuations corresponding to the time - reversal invariant superconductor symmetry class ci . substituted back into the action , these fluctuations are described by the corresponding low - energy action @xcite @xmath122 = - { \displaystyle\frac{\pi \nu}{8 } } \int d { { \mathbf r } } \ ; d \left(1 - y^2 \tau_n^2\right ) \operatorname{str}(\nabla \bar{q})^2 \ ; , \ ] ] where @xmath123 . this completes our construction of the field theory describing spectral correlations in the disordered non - hermitian system . in the following section we will use these results to explore the support of the complex dos . for the two - dimensional scalar potential problem . ( a ) in the mean - field approximation , the dos is equal to @xmath124 inside the region @xmath125 of the complex plane @xmath126 and zero outside ; ( b ) @xmath127 versus @xmath55 for a fixed @xmath46 and for two values of the scattering time , @xmath128 : while in the mean - field approximation the dos shows sharp edges , rare realizations of the random potential give rise to tail states ( [ sec : state ] ) . ] making use of eq . , the complex dos is obtained from the generating function as @xmath129 \rangle_q \ ; , \label{eq : nu_gen}\ ] ] where @xmath130}$ ] where @xmath131 $ ] represents the intermediate energy scale action . in the homogeneous saddle - point ( i.e. mean - field ) approximation , the corresponding dos takes the form @xmath132 0 & \quad |y| \tau_n \ge 1\ ; . \end{cases } \label{eq : mfdos}\ ] ] as expected , at the level of the mean - field , the field theory predicts a migration of the dos off the real line and into the complex plane . the density of complex eigenvalues is constant over the region of support set by the effective scattering rate of the non - hermitian potential . reassuringly , the expression for the dos satisfies the sum rule @xmath133 . in the two - dimensional case , the result is illustrated qualitatively in fig . [ fig : fidos ] . as the scattering time @xmath134 does not depend on the real part of the energy @xmath46 , the width of the mean - field spectrum boundary is constant . to complete the analysis , it is necessary to take into account the fluctuations of the fields around the saddle - point . as discussed above , these fluctuations divide into a set which are massive ( on the scale of @xmath55 ) and a massless set . interestingly , one finds that the latter commute with the source and do not contribute to the dos ! indeed , the general profile of the support is born out by numerical simulation . figure [ fig : specs ] shows the amalgamation of data for the two dimensional random system generated from @xmath135 realizations of the complex random scalar potential hamiltonian . however , instead of a sharp cut - off in the support , the data clearly indicate a soft edge with eigenvalues occupying the region outside the mean - field prediction . how can this observation be reconciled with the results of the field theory ? surprisingly , taking into account massive fluctuations of the action perturbatively , the mean - field prediction remains intact . in fact , the tail states that decorate the edge of the support are generated by optimal fluctuations of the _ real _ random impurity potential and are reflected in _ non - perturbative _ instanton configurations of the action . . the two degenerate bounce solutions are qualitatively shown , together with the mean - field starting point @xmath136 and the inversion point @xmath137 . ] to identify corrections to the mean - field dos we can draw on the intuition afforded by recent studies of sub - gap state formation due to optimal fluctuations in the superconducting system @xcite . while the homogeneous solution of the mean - field equation gives rise to a hard edge of the dos support , we will see that inhomogeneous symmetry broken field configurations reflect the influence of rare realizations or ` optimal fluctuations ' of the random scalar potentials which soften the edge ( fig . [ fig : fidos ] ) . here , for simplicity , let us focus on the quasi one - dimensional case , generalizing the discussion to the @xmath138-dimensional system in section [ sec : highd ] . to explore the influence of inhomogeneous field configurations of the action , let us revisit the mean - field equation of motion . operationally , it is convenient to deal not with the saddle - point equation itself , but rather its first integral , @xmath139 , where @xmath140 - { \displaystyle\frac{1}{y \tau_n } } \cos 2\hat{\theta } \ ; , \ ] ] denotes the effective complex ` potential ' . versus the variable @xmath141 for @xmath142 , @xmath143 and @xmath144 . ] to identify the inhomogeneous instanton solution let us first note that , outside the region of support , @xmath145 , the homogeneous mean - field solution takes the form @xmath146 . taking into account the condition that the solution should coincide with @xmath147 at infinity , one can identify a ` bounce ' solution parameterized by @xmath148\operatorname{sign}(y)$ ] , with @xmath141 real , involving the _ real _ potential @xmath149 \operatorname{sign}(y))$ ] with endpoint @xmath137 such that @xmath150 . now integration over the angles @xmath151 is constrained to certain contours @xcite . is the bounce solution accessible to both ? as usual , the contour of integration over the boson - boson field @xmath152 includes the entire real axis , while for the fermion - fermion field @xmath153 runs along the imaginary axis from @xmath154 to @xmath155 . with a smooth deformation of the integration contours , the mean - field saddle - point is accessible to both the angles @xmath151 @xcite . by contrast , the particular bounce solution _ and _ the mean - field solution can be reached simultaneously by a smooth deformation of the integration contour _ only _ for the boson - boson field @xmath152 ( see fig . [ fig : conto ] ) . the particular bounce solution breaks supersymmetry . applying the parameterization for the ` bounce ' @xmath156 \operatorname{sign}(y ) & \theta_{{\text{\sc{ff } } } } = \theta_{{\text{\sc{mf } } } } \ ; , \end{aligned}\ ] ] the first integral takes the form @xmath157 here @xmath158 for @xmath145 , while @xmath136 is purely imaginary for @xmath159 . typical shapes of the potential @xmath160 for different values of @xmath161 are shown in fig . [ fig : poten ] . a bounce solution exists only for values of @xmath161 bigger than one , while for @xmath162 the unique solution is the homogeneous one , @xmath163 . versus @xmath164 in the quasi one - dimensional geometry for two different values of the inverse of the imaginary ` impurity ' concentration , @xmath165 and @xmath166 . the interval @xmath167 $ ] is explicitly plotted for the exact solutions . ] for the quasi one - dimensional system it is possible to derive analytic expressions for both the real instanton action @xmath168 @xmath169 where @xmath170 is the endpoint of the motion , and the bounce solution @xmath171 . in particular , for @xmath172 we have ( fig . [ fig : expsi ] ) @xmath173 at the same time , imposing the boundary conditions @xmath174 , from the first integral one obtains the bounce solution @xmath175^{1/2 } \label{eq : dropl}\end{gathered}\ ] ] defines the extent of the ` droplet ' ( see fig . [ fig : phifu ] ) . in particular , we note that , one approaching the band edge ( @xmath176 ) , @xmath177 . now close to the band edge ( @xmath178 ) , the maximum of the potential @xmath179 converges on the two minima allowing for the development of a perturbative expansion in @xmath180 . setting @xmath181 , @xmath182 in this approximation the action and the bounce solution are given respectively by : @xmath183 ( solid line ) versus @xmath161 in the quasi one - dimensional geometry together with the first term of the expansion in @xmath184 ( dashed line ) . the action vanishes as @xmath185 . in the inset the rescaled dos @xmath186 $ ] versus @xmath187 , for @xmath188 , is plotted . for values of @xmath189 near the critical values @xmath190 , the saddle - point analysis becomes unreliable . here , fluctuations lead to a smooth interpolation between the dos in the tail region , and that coming from the bulk mean - field . ] this completes the estimate of the contribution to the action from a single bounce configuration . however , to complete the analysis it is necessary to explore the influence of fluctuations around the instanton solution . here we sketch the important aspects of the analysis referring for details to the parallel discussion by ref . @xcite in the context of the disordered superconductor . generally , field fluctuations around the instanton solution can be separated into ` radial ' and ` angular ' contributions . the former involve fluctuations of the diagonal elements @xmath151 , while the latter describe rotations including those grassmann transformations which mix the @xmath191 sector . both classes of fluctuations play an important role . dealing first with the angular fluctuations , supersymmetry breaking of the bounce is accompanied by the appearance of a grassmann zero mode separated by an energy gap from higher excitations . this goldstone mode restores the global supersymmetry of the theory . crucially , this mode ensures that the saddle - point respects the normalization condition @xmath192 \rangle_{w , v } = 1 $ ] . associated with radial fluctuations around the bounce , there exists a zero mode due to translational invariance of the solution , and a negative energy mode ( c.f . ref . @xcite ) . combining these contributions , one obtains the following expression for the local complex dos in the tail region , @xmath193 |\chi_0 ( r)|^2 e^{- s_{\text{inst } } } \ ; , \label{eq : fldos}\ ] ] where @xmath194 represents the eigenfunction for the grassmann zero mode and @xmath195 denotes the instanton action . thus , to exponential accuracy , the complex local dos in the tail region becomes non - zero only in the vicinity of the bounce . mechanisms of quantum interference due to optimal potential fluctuations in the non - hermitian system conspire to localize states in the tail regions on a length scale @xmath196 greatly in excess of the wavelength @xmath197 . close to the band edge ( @xmath199 ) , a generalization of the quasi one - dimensional results to higher dimensions can be developed by dimensional analysis . from the expansion for the potential @xmath179 , the bounce configuration is shown to have the scaling form @xmath200 f({{\mathbf r}}/ r_0 ( \kappa ) ) \ ; , \ ] ] which , in turn , implies the instanton action : @xmath201^{(4 - d)/2 } \ ; . \label{eq : ddima}\ ] ] here @xmath202 and @xmath203 represent respectively the dimensionless conductance and average level spacing of the hermitian system ( i.e. with @xmath204 ) , and @xmath205 is a numerical constant ( @xmath206 and @xmath207 ) specifically @xmath208 $ ] . ] . later , in discussing the universality of the results , we will return to consider the zero - dimensional situation . taken together with eq . , this result shows that the complex dos in the tail region varies exponentially with the separation from the mean - field edge . the power law dependence of the exponent depends on dimensionality , presenting a linear dependence on @xmath209 in the two - dimensional system . in this case , moreover , the only dependence in on the real part of the energy @xmath46 occurs through the dimensionless conductance , @xmath210 , so that the width of the exponential tail scales as @xmath211 , as qualitatively appears in fig . [ fig : specs ] . a comparison with the numerical simulation is shown in fig . [ fig : dos_14 ] . the results show a good quantitative agreement with the theory for the tail of the dos . note that close to ( but within ) the edge of the support of the spectrum , the saddle - point approximation used in estimating the mean - field dos becomes unreliable . as shown by the zero - dimensional analysis below , an honest treatment of fluctuations in this region provides a smooth interpolation between the bulk and tail states . = y$ ] for a fixed value of @xmath212 = x$ ] taken from the numerical simulation of fig . [ fig : specs ] . the exponential fit of the data is shown as solid curve . ] to conclude our discussion of tail states in the non - hermitian system , we turn now to an analysis of the random imaginary vector potential hamiltonian . once again , our approach will be to formulate an effective field theory from which the spectral properties can be determined . although , as shown in section [ sec : backg ] , the random vector potential hamiltonian belongs to a different fundamental symmetry from the random scalar potential hamiltonian , we will show below that , with some qualification , the dos in the vicinity of the support exhibits the same universal scaling form . to be concrete , let us consider a @xmath138-dimensional hamiltonian involving an imaginary vector potential field @xmath213 together with a real random scalar potential @xmath19 . once again , let us assume that the scalar potential @xmath19 is drawn from a gaussian @xmath10-correlated impurity distribution . similarly , we will be primarily concerned with a system exhibiting a random vector potential drawn from a gaussian distribution with zero mean @xmath214 and correlation @xmath215 \ ; . \label{eq : corre}\ ] ] here , for reasons that will become clear below , we have introduced an ` envelope ' @xmath216 which limits correlations of the fields at both microscopic and macroscopic length scales . setting @xmath217 @xmath218 is identified as the strength of fluctuations of the incompressible component of the vector field , @xmath219 , while @xmath220 controls the irrotational part , @xmath221 . in the following , it will also be useful to contrast the behaviour of the random vector potential system with that of a _ constant _ vector field @xmath222 the hatano - nelson system @xcite . before turning to the formal analysis of the statistical field theory , which begin with some general considerations which identify certain idiosyncrasies of the constant vector potential system and how they impact upon the existence of tail states in the random vector potential system . previous investigations of the former have revealed unusual localization properties of the random hamiltonian which contrast with those of its hermitian ( i.e. @xmath223 ) counterpart . specifically , it was shown in ref . @xcite that , for the strictly one - dimensional system subject to _ periodic boundary conditions _ at infinity , when the field , @xmath224 , exceeds a critical value @xmath225 corresponding to the ( energy dependent ) localization length @xmath226 of the hermitian system , there is a transition to a delocalized phase . since then , this result has been generalized to the quasi one - dimensional system by kolesnikov and efetov @xcite using techniques which parallel those discussed here . now , following ref . @xcite , a simple argument can be established to describe qualitatively the origin of the transition . for a constant imaginary vector potential , @xmath227 , a similarity transformation of the left and right eigenfunctions @xmath228 and @xmath229 removes from the hamiltonian its dependence on @xmath227 . being now hermitian , the eigenvalues of the hamiltonian must therefore be real and specified by those of the unperturbed random system . however , if periodic boundary conditions are imposed , the similarity transformation must be applied with caution : if the eigenfunctions of the hermitian system are localized on a length scale @xmath230 the similarity transformation is incompatible with the periodic boundary conditions and can not be imposed . here there is a transition from a localized phase where the eigenvalues are real to a delocalized phase where the eigenvalues migrate into the complex plane . now what does this phenomenology tell us about the existence of tail states ? using the same arguments , it is easy to deduce that optimal fluctuations of the impurity potential can _ not _ induce localized tail states in the constant vector potential system : suppose that a tail state appeared at some complex energy outside the band of bulk states . if the state is localized , it must be insensitive to the boundary conditions at infinity and , therefore , its dependence on the constant field can be removed by a similarity transformation . as such , it must therefore be contained within the hermitian theory , and it s eigenvalue must be real . this contradiction strictly eliminates tail states from the infinite constant imaginary vector potential system . ( of course , the hermitian system can and will exhibit low - energy lifshitz band tail states . ) with this background , let us now turn to consider the random vector potential system . once again , for a random field @xmath231 drawn from the gaussian distribution , it is straightforward to confirm that the similarity transformation @xmath232 $ ] and @xmath233 \phi_i^r({{\mathbf r}})$ ] removes from the hamiltonian the dependence on the incompressible component , @xmath234 . therefore , when subject to any _ purely _ irrotational vector field @xmath235 ( with zero average ) , the eigenvalues remain real and coincide with those of the hermitian system . conversely , the incompressible field distribution @xmath236 can not be removed by similarity transformation . such field configurations necessarily generate states with complex eigenvalues . , the flux lines of @xmath237 are approximatively constant , preventing the existence of tail state . ] now , the random distribution with @xmath238 imposes long - ranged , i.e. power - law , correlations in the incompressible field configuration . as well as admitting superdiffusion processes @xcite which dominate the relaxation to equilibrium in the classical system , these long - range correlations impact upon the probability of finding tail states in the non - hermitian system : any random configuration of the incompressible field will involve field lines correlated over arbitrarily long distances . these field lines advect particles in the classical system and are responsible for the characteristic superdiffusion properties of the system . now in such a background , one can identify regions where the field lines are oriented over long distances , say @xmath239 ( see fig . [ fig : qflux ] ) . within this region , the nucleation of tail states with a localization length @xmath240 is prohibited by the same mechanism described above for the hatano - nelson system . clearly , the phenomenology of tail state formation in the superdiffusive system is subtle being sensitive to the arrangement of both the scalar _ and _ vector field . therefore , since we are interested here in exposing the general principles behind tail state formation in the non - hermitian system , to keep our discussion simple , we will limit our considerations to distributions where the correlations of the field lines are limited to a length scale @xmath239 much smaller than the localization length of the incipient tail states , @xmath241 , i.e. introducing a microscopic cut - off @xmath242 , we set @xmath243 where @xmath239 represents the maximum length scale of field correlations . with these considerations in mind , let us now turn to the development of the field theory of the non - hermitian system . following the method outlined for the imaginary scalar potential system , our starting point is the gauge transformed ` hermitized ' matrix hamiltonian . now , under the ` charge - conjugation ' operation , the hamiltonian transforms as @xmath244^\mathsf{t } \sigma_2 = - \hat{\mathcal{h}}_\gamma ( -{{\mathbf h } } ) \ ; .\ ] ] the statistical properties of the non - hermitian system can again be obtained from the generating function by forming an average over the real random potential @xmath245 and introducing a slow field decoupling . in doing so , one obtains the generating functional @xmath246 \rangle_w = \int dq \exp \left [ { \displaystyle\frac{\pi \nu}{8 \tau } } \int d{{\mathbf r } } \ ; \operatorname{str}q^2 - { \displaystyle\frac{1}{2 } } \int d { { \mathbf r } } \ ; \operatorname{str}\langle { { \mathbf r } } | \ln \hat{\mathcal{g}}^{-1 } | { { \mathbf r } } \rangle \right ] \ ; , \intertext{where } \hat{\mathcal{g}}^{-1 } = i \eta \sigma_3 \otimes \sigma_3^{{\text{\sc{cc } } } } + x - { \displaystyle\frac{1}{2 m } } \left[\hat{{{\mathbf p } } } - i { { \mathbf h } } ( { { \mathbf r } } ) \sigma_1 \otimes \sigma_3^{{\text{\sc{cc}}}}\right]^2 - i y \sigma_1 + { \displaystyle\frac{i}{2 \tau } } q({{\mathbf r } } ) \notag\end{gathered}\ ] ] denotes the supermatrix green function . as with the scalar potential , further progress is possible only within a saddle - point approximation . however , following our discussion above , it is now necessary to exercise some caution . in the hermitian system with a real vector potential @xmath247 , the conventional approach @xcite involves subjecting the action to a saddle - point approximation in the absence of the field . the saddle - point manifold @xmath113 is unperturbed by the real vector potential and its effect on the low - energy properties of the system can be accommodated at the level of the gradient expansion . however , as emphasized in ref . @xcite , in the presence of a _ constant _ imaginary vector potential , @xmath248 , depending on the magnitude of @xmath227 , it is possible to identify separate saddle - points of the total effective action . inside the localized phase ( i.e. @xmath249 ) , the dependence on @xmath227 can ( and must ) be removed by a rotation of @xmath84 the counterpart of the similarity transformation . the resulting theory reflects that of the hermitian model and the eigenvalues condense onto the real line . conversely , in the delocalized phase ( i.e. @xmath250 ) , the similarity transformation on @xmath84 is inadmissible . here the dependence of the action on @xmath227 must be developed explicitly , reflecting the fact that the eigenvalues acquire an imaginary component . similarly , for a random vector potential , it is necessary to remove the irrotational component of the field @xmath237 ( i.e. @xmath251 ) explicitly at the level of the saddle - point by subjecting @xmath84 to a similarity transformation @xmath252 as expected , the same transformation leaves the source term for the dos unperturbed . for the two - dimensional system . ( a ) in the mean - field approximation the dos is equal to @xmath253 inside the region @xmath254 of the complex plane @xmath255 and zero otherwise ; ( b ) @xmath256 versus @xmath55 for two values of @xmath257 . the tails of the dos for @xmath258 are shown schematically . ] without the irrotational component of the field , the saddle - point of the total action can be analyzed in the usual way , following a two - stage procedure . as before , taking into account the quasi - classical parameter @xmath259 , one finds that the low - energy fluctuations are confined to the manifold @xmath260 . then , subjecting the action to a gradient expansion keeping @xmath261 and @xmath262 , one obtains the non - linear @xmath23-model action @xcite @xmath263 = -{\displaystyle\frac{\pi \nu}{8 } } \int d { { \mathbf r } } \ ; \operatorname{str}\left[d ( \widetilde{\nabla } q)^2 - 4 \left(\eta \sigma_3 \otimes \sigma_3^{{\text{\sc{cc } } } } - y \sigma_1\right ) q\right ] \ ; , \label{eq : vnlsm}\ ] ] where @xmath264 $ ] represents the covariant derivative , and @xmath265 reflects the incompressible component of the field . now , although the structure of the action compares with that of the imaginary scalar potential field , it exhibits important differences which substantially influences the behaviour of the system : specifically , in addition to the ` diamagnetic ' contribution , i.e. the term proportional to @xmath266 , the action presents a paramagnetic contribution , linear in @xmath237 , which couples to the gradient of @xmath84 . although this term does not affect the homogeneous saddle - point solution , it is responsible for the destabilisation of the tail states in the system with long - range correlations . to see this explicitly , let us return to the consideration of the constant vector potential system . previously , in the qualitative discussion above , it was argued that such a system does not accommodate tail states . how is this phenomenon reflected in the field theory ? in fact , for @xmath267 , the action is still given by the non - linear @xmath23-model . however , in view of the paramagnetic term , to explore the saddle - point structure , it is necessary to adopt a more general parameterization , @xmath268 , where @xmath269 . in doing so , two coupled equations for @xmath151 and @xmath270 are obtained : @xmath271 + 2 y \sin \hat{\theta } = 0 \\ \nabla \cdot \left[\sin^2 \hat{\theta } \left(\nabla \hat{\rho}+ 2 { { \mathbf h}}_0\right ) \right ] = 0 \ ; . \notag\end{gathered}\ ] ] to understand the nature of these equations , let us focus on the quasi one - dimensional system . in this case the second equation can be integrated and substituted in the first one , giving respectively @xmath272 where @xmath273 is a constant fixed by the boundary conditions . from these equations one can identify a homogeneous supersymmetric solution , @xmath274 and @xmath275 where @xmath276 . again , making use of eq . , the homogeneous solution translates to a complex dos , which is flat and non - vanishing only in the interval @xmath277 . now , in the imaginary scalar potential problem , an instanton configuration of the saddle - point equation was signalled by the existence of tail states . what happens in the present case ? since , at infinity , the instanton solution is constrained to be the homogeneous one , the constant @xmath273 is determined by the homogeneous mean - field solution , @xmath278 . therefore , using , one can see that outside the bulk of the spectral support @xmath273 vanishes . therefore , the equation of motion for @xmath151 does not admit the existence of an instanton . as expected from the qualitative discussion above , tails are therefore excluded from the constant imaginary vector potential problem . similarly , in the random vector potential system with long - ranged correlations ( i.e. for @xmath238 ) , we can expect the paramagnetic term to present a similar role . indeed , subjecting the generating functional to an average over the random distribution , one obtains a non - local interaction of the fields . in the present context , such terms have been shown by taras - semchuk and efetov @xcite to reproduce the known renormalisation properties of the superdiffusive system . however , in the present case , we have limited our considerations to correlations of the vector field which extend over a maximum range @xmath239 . then , taking the relevant field configurations of @xmath84 ( i.e. those involving the instanton solution ) to be long - ranged on a scale @xmath241 greatly in excess of @xmath239 , the vector field entering the paramagnetic term can be replaced by its spatial average , i.e. the paramagnetic term can be eliminated . by contrast , the diamagnetic contribution survives spatial averaging . the total action assumes the form @xmath263 = -{\displaystyle\frac{\pi \nu}{8 } } \int d { { \mathbf r } } \ ; \operatorname{str}\left[d ( \nabla q)^2 - 4 \left(\eta \sigma_3 \otimes \sigma_3^{{\text{\sc{cc } } } } - y \sigma_1\right ) q + { \displaystyle\frac{1}{\tau_n } } \left(q \sigma_1 \otimes \sigma_3^{{\text{\sc{cc}}}}\right)^2 \right ] \ ; , \label{eq : vnlsm2}\ ] ] where , setting @xmath279 , @xmath280 is drawn from a gaussian @xmath10-correlated random impurity distribution . here we have included a lattice of @xmath11 @xmath12-points in the momentum space . notice that the constraints on the numerical simulation limit the relevant data to the interval @xmath281<200 $ ] . ] now , as with the imaginary scalar potential , the profile of the bulk dos and tails states can be obtained by subjecting the action to a saddle - point analysis . varying the action with respect to @xmath84 , subject to the non - linear constraint , @xmath113 , one obtains the homogeneous saddle - point equation , @xmath282 + { \displaystyle\frac{1}{2 \tau_n } } [ \sigma_1\otimes\sigma_3^{{\text{\sc{cc } } } } q \sigma_1 \otimes \sigma_3^{{\text{\sc{cc } } } } , q ] = 0 \ ; , \ ] ] then , applying the _ ansatz _ @xmath115 , where @xmath116 , the saddle - point equation coincides with eq . allowing the results of sections [ sec : state ] and [ sec : highd ] to be imported . as a result , we can immediately deduce the homogeneous mean - field solution as well as the inhomogeneous instanton solution of the saddle - point equation . thus , from the homogeneous mean - field solution , one obtains the expression for the dos , i.e. the complex dos is flat and non - vanishing over the interval @xmath283 i.e. , in the two - dimensional system , the support for the dos occupies a wedge of the complex plane with a width that scales in proportion to @xmath284 ( see fig . [ fig : vedos ] ) . furthermore , making use of eqs . and , one obtains complex dos in the tail region , @xmath285^{(4-d)/2}\right\ } \ ; , \ ] ] where , as usual , @xmath202 and @xmath203 represent respectively the dimensionless conductance and average level spacing of the hermitian system . = y$ ] for a fixed value of @xmath212 = x $ ] taken from the numerical simulation of fig . [ fig : specv ] ( triangles ) . a second data set ( circles ) is shown for a value of @xmath286 four times as large . notice that , as expected from the theory , the change in @xmath286 is reflected in the slope of the exponent . the exponential fits of the two data are shown as solid curves . ] in fact , the behaviour of the vector potential system differs from the scalar potential system only in the symmetry of the soft fluctuations around the saddle - point solution . as usual , the homogeneous saddle - point solution @xmath287 is not unique but is spanned by an entire manifold of solutions @xmath288 where for @xmath289 , @xmath290 = 0 = [ t,\sigma_1 \otimes \sigma_3^{{\text{\sc{cc}}}}]$ ] with @xmath291 ( chiral symmetry class aiii in the classification of ref . @xcite ) , while , for @xmath56 , @xmath292 = 0 $ ] ( chiral symmetry class bdi ) . now while the soft ( class aiii ) fluctuations decouple from the dos source ( c.f . the scalar potential hamiltonian ) , the soft ( class bdi ) fluctuations couple . as a result , the complex dos becomes strongly suppressed near @xmath56 while , on the real axis , a finite density of eigenvalues accumulates see below . the amalgamation of data for the eigenvalues of ca . @xmath135 realizations of the imaginary vector potential hamiltonian is shown in fig . [ fig : specv ] . the locus of the edge of the support predicted by the mean - field theory is in good agreement with the numerics , while the depletion of eigenvalues from the interval near @xmath56 and the singular density at @xmath56 are clearly resolved . in particular , a comparison of the complex dos with the numerical simulation shown in fig . [ fig : dos_14 ] shows a good agreement with theory . as a function of @xmath293 for @xmath294 and @xmath295 . ] finally , to conclude our discussion , let us turn to consider the properties of the imaginary vector potential hamiltonian when the system enters the zero - dimensional limit . our analysis above shows that , outside the region of support , the spectrum is characterized by tail states which are localized on a length scale of @xmath296 , . when the size of the droplet region becomes in excess of the system size ( inevitable in a finite system as one approaches the edge of the support , @xmath185 ) , the system enters a zero - dimensional regime where the action is dominated by the zero spatial mode of @xmath84 . here the action collapses onto the zero - dimensional theory analyzed in detail by efetov @xcite . in this limit , we can proceed in two ways : firstly , following the discussion above , we can apply the saddle - point analysis seeking a symmetry broken field configuration . however , in this case , the relative of the bounce solution is now a stationary point of the zero - dimensional potential @xmath297 . alternatively , following efetov @xcite one can undertake an exact evaluation of the zero - dimensional @xmath23-model . a pursuit of the second route obtains the exact formula for the complex dos . the latter can be separated into two contributions @xmath298 , where @xmath299 where @xmath300 with @xmath301 and @xmath302 is the level spacing of the hermitian system . the presence of the anomalous part @xmath303 attests that a finite fraction of eigenvalues remains real . otherwise the component @xmath304 describes the distribution of complex eigenvalues and its form is shown in fig . [ fig : efeto ] . in the region @xmath305 the dos coincides with the constant mean - field value for @xmath306 , while outside this interval one has the following asymptotic expansion : @xmath307 with the exact result at hand , let us compare eq . with the result of the saddle - point analysis from the symmetry broken field configuration . in particular , in the zero - dimensional case , the saddle - point analysis leads to the ` instanton ' action @xmath308 \ ; , \ ] ] where @xmath309 is the minimum of the potential @xmath310 ( fig . [ fig : poten ] ) . evaluating this minimum , one obtains @xmath311 from the expression for the dos , we can conclude that , in the zero - dimensional case , the exponential dependence of the dos coincides with the exact result obtained in ref . to conclude , we have implemented a field theoretic scheme to explore the structure of the dos close to the edge of the support of two linear non - hermitian operators describing a quantum particle subject to a random imaginary scalar and an imaginary vector potential . in doing so , we have provided a general scheme for the symmetry classification of non - hermitian operators . the field theoretic approach is easily generalized to the consideration of higher point spectral correlations of the fields . in the quasi - classical limit , where the real part of the energy is in excess of any other energy scale , the tails are dominated by ` optimal configurations ' of the real random scalar potential . in contrast to band tail states in semi - conductors , these states are quasi - classical in nature being localized on the length scale @xmath312 . as such the profile of the dos and their dimension is universal depending on just a few material parameters and independent of the nature of the impurity distribution . in the particular case of the constant imaginary vector potential , we have argued that tail states of the system are prohibited by the delocalization mechanism of hatano and nelson . m. b. hastings , preprint cond - mat/9909234 . r. a. janik , w. noerenberg , m. a. nowak , g. papp and i. zahed _ phys . e _ * 60 * , 2699 ( 1999 ) . a. v. izyumov and b. d. simons , _ europhys . lett . _ * 45 * , 290 ( 1999 ) . b. d. simons and a. altland , `` mesoscopic physics '' , to be published in the proceedings of the crm summer school _ theoretical physics at the end of the xxth century _ ( banff , canada , 1999 ) , crm series in mathematical physics , springer , berlin ( 2000 ) .

in a recent paper iyama and yoshino consider two interesting examples of isolated singularities over which it is possible to classify the indecomposable maximal cohen - macaulay modules in terms of linear algebra data . in this paper we present two new approaches to these examples . in the first approach we give a relation with cluster categories . in the second approach we use orlov s result on the graded singularity category .

0803.0720

throughout @xmath0 is a field . in @xcite iyama and yoshino consider the following two settings . [ ref-1.1 - 0 ] let @xmath1 $ ] and let @xmath2 be the cyclic group of three elements . consider the action of @xmath3 on @xmath4 via @xmath5 where @xmath6 , @xmath7 . put @xmath8 . [ ref-1.2 - 1 ] let @xmath9 $ ] and let @xmath10 be the cyclic group of two elements . consider the action of @xmath11 on @xmath4 via @xmath12 . put @xmath13 . in both examples iyama and yoshino reduce the classification of maximal cohen - macaulay modules over @xmath14 to the representation theory of certain generalized kronecker quivers . they use this to classify the rigid cohen - macaulay modules over @xmath14 . as predicted by deformation theory , the latter are described by discrete data . the explicit description of the stable category of maximal cohen - macaulay modules over a commutative gorenstein ring ( also known as the singularity category @xcite ) is a problem that has received much attention over the years . this appears to be in general a difficult problem and perhaps the best one can hope for is a reduction to linear algebra , or in other words : the representation theory of quivers . this is precisely what iyama and yoshino have accomplished . the proofs of iyama and yoshino are based on the machinery of mutation in triangulated categories , a general theory developed by them . in the current paper we present two alternative approaches to the examples . hopefully the additional insight obtained in this way may be useful elsewhere . our first approach applies to example [ ref-1.2 - 1 ] and is inspired by the treatment in @xcite of example [ ref-1.1 - 0 ] where the authors used the fact that in this case the stable category @xmath15 of maximal cohen - macaulay @xmath14-modules is a @xmath16-calabi - yau category which has a cluster tilting object whose endomorphism ring is the path algebra @xmath17 of the kronecker quiver with @xmath18 arrows . then they invoke their acyclicity result ( slightly specialized ) : ( * ? ? ? * , thm ) assume that @xmath19 is @xmath0-linear algebraic krull - schmidt @xmath16-calabi - yau category with a cluster tilting object @xmath20 such that @xmath21 is hereditary . then there is an exact equivalence betweem @xmath19 and the orbit category @xmath22)$ ] . from this result they obtain immediately that @xmath15 is the orbit category @xmath23)$ ] . this gives a very satisfactory description of @xmath15 and implies in particular the results by iyama and yoshino . in the first part of this paper we show that example [ ref-1.2 - 1 ] is amenable to a similar approach . iyama and yoshino prove that @xmath15 is a @xmath18-calabi - yau category with a @xmath18-cluster tilting object @xmath20 such that @xmath24 ( * ? ? ? * theorem 9.3 ) . we show that under these circumstances there is an analogue of the acyclicity result of the first author and reiten . [ ref-1.3 - 2 ] ( see [ ref-3.4 - 18 ] ) assume that @xmath19 is @xmath0-linear algebraic krull - schmidt @xmath18-calabi - yau category with a @xmath18-cluster tilting object @xmath20 such that @xmath24 . then there is an exact equivalence of @xmath19 with the orbit category @xmath25)$ ] , @xmath26 , where @xmath27 is the generalized kronecker quiver with @xmath28 arrows and @xmath29 is a natural square root of the auslander - reiten translate of @xmath30 , which on the pre - projective / pre - injective component corresponds to `` moving one place to the left '' . in the second part of this paper , which is logically independent of the first we give yet another approach to the examples [ ref-1.1 - 0],[ref-1.2 - 1 ] based on the following observation which might have independent interest . [ ref-1.4 - 3 ] ( see prop . [ gradedcase ] ) let @xmath31 be a finitely generated commutative graded gorenstein @xmath0-algebra with an isolated singularity . let @xmath32 be the completion of @xmath33 at @xmath34 . let @xmath35 be the stable category of graded maximal cohen - macaulay @xmath33-modules . then the obvious functor @xmath36 induces an equivalence @xmath37 where @xmath38 is the shift functor for the grading . in this proposition the quotient @xmath39 has to be understood as the triangulated / karoubian hull ( as explained in @xcite ) of the naive quotient of @xmath35 by the shift functor @xmath40 . this result is similar in spirit to @xcite which treats the finite representation type case . note however that one of the main results in loc . is that in case of finite representation type case _ every _ indecomposable maximal cohen - macaulay @xmath41-module is gradable . this does not seem to be a formal consequence of proposition [ ref-1.4 - 3 ] . it would be interesting to investigate this further . in [ ref-8 - 57 ] we show that at least rigid cohen - macaulay modules are always gradable so they are automatically in the image of @xmath42 . we expect this to be well known in some form but we have been unable to locate a reference . hence in order to understand @xmath43 it is sufficient to understand @xmath42 . the latter is the graded singularity category @xcite of @xmath33 and by ( * ? ? ? * thm 2.5 ) it is related to @xmath44 where @xmath45 . in examples [ ref-1.1 - 0],[ref-1.2 - 1 ] @xmath14 is the completion of a graded ring @xmath33 which is the veronese of a polynomial ring . hence @xmath46 is simply a projective space . using orlov s results and the existence of exceptional collections on projective space we get very quickly in example [ ref-1.1 - 0 ] @xmath47 and in example [ ref-1.2 - 1 ] @xmath48 ( where here and below @xmath49 actually stands for a quasi - equivalence between the underlying dg - categories ) . finally it suffices to observe that in example [ ref-1.1 - 0 ] we have @xmath50 $ ] and in example [ ref-1.2 - 1 ] we have @xmath51 $ ] ( see [ ref-7 - 51 ] below ) . finally we mention the following interesting side result [ ref-1.5 - 5 ] [ ref-1.5 - 6 ] let @xmath52 be a gorenstein local `` g - ring '' for example @xmath14 may be essentially of finite type over a field with an isolated singularity . then the natural functor @xmath53 is an equivalence up to direct summands . in partular every maximal cohen - macaulay module over @xmath54 is a direct summand of the completion of a maximal cohen - macaulay module over @xmath14 . the original proof ( by the first and the third author ) of this result was unnecessarily complicated . after the paper was put on the arxiv daniel murfet ( who has become the second author ) informed us about the existence of a much nicer proof in the context of singularity categories ( see proposition [ theorem : first_theorem ] ) . the same argument also applies to proposition [ ref-1.4 - 3 ] . so we dropped our original proofs and put the new argument in an appendix to which we refer . meanwhile orlov @xcite has proved ( independently and using different methods ) a very general result which implies in particular proposition [ ref-1.5 - 6 ] . we thank osamu iyama , idun reiten , srikanth iyengar , and amnon neeman for commenting on a preliminary version of this manuscript . the second author thanks kyoji saito , kazushi ueda and osamu iyama for pointing out the properties of the henselization in remark [ remark : henselization ] , and apostolos beligiannis for discussing his work @xcite . we hope most notations are self explanatory but nevertheless we list them here . if @xmath14 is a ring then @xmath55 and @xmath56 denote respectively the category of all left @xmath14-modules and the full subcategory of finitely generated @xmath14-modules . the derived category of all @xmath14-modules is denoted by @xmath57 . if @xmath14 is graded then we use @xmath58 and @xmath59 for the category of graded left modules and its subcategory of finitely generated modules . the shift functor on @xmath58 is denoted by @xmath40 . explicitly @xmath60 . if we want to refer to right modules then we use @xmath61 instead of @xmath14 . if @xmath62 is a scheme then @xmath63 is the category of quasi - coherent @xmath64-modules . if @xmath62 is noetherian then @xmath65 is the category of coherent @xmath64-modules . we are generally very explicit about which categories we use . e.g. we write @xmath66 rather than something like @xmath67 . if @xmath14 is graded and @xmath68 , @xmath69 are graded @xmath14-modules then @xmath70 is the ungraded @xmath71 between @xmath68 and @xmath69 . if we need @xmath71 in the category of graded @xmath14-modules then we write @xmath72 . let @xmath73 be rings . we denote the unbounded derived category of right @xmath74-modules by @xmath75 . we let @xmath76 be the set of triangle equivalences of @xmath77 modulo natural isomorphisms . define @xmath78 as the set of pairs @xmath79 where @xmath20 is a perfect complex generating @xmath80 and @xmath81 is an isomorphism @xmath82 . associated to @xmath83 there is a canonical equivalence @xmath84 such that @xmath85 . it may be constructed either directly @xcite or using dg - algebras @xcite . the induced map @xmath86 is obviously injective ( as it is canonically split ) , but not known to be surjective . below we will informally refer to the elements of @xmath78 as tilting complexes . let @xmath88 be a finite dimensional @xmath0-vector space and let @xmath74 be the path algebra of the quiver @xmath89[]{$w$ } \psfrag{1}[][]{$1 $ } \psfrag{2}[][]{$2 $ } \includegraphics[width=3cm]{beilinson_mini}\ ] ] let @xmath90 be the path algebra of the quiver @xmath91[]{$w^\ast$ } \psfrag{1}[][]{$1 $ } \psfrag{2}[][]{$2 $ } \includegraphics[width=3cm]{beilinson_mini}\ ] ] which we think of as being obtained from by `` inverting the arrows '' and renumbering the vertices @xmath92 . let @xmath93,@xmath94 be respectively the projective , injective and simple right @xmath74-module corresponding to vertex @xmath95 . for @xmath90 we use @xmath96 , @xmath97 , @xmath98 . let @xmath99 be the reflection functor at vertex @xmath95 . recall that if @xmath100 is a representation of @xmath74 then @xmath101 is given by @xmath102 where @xmath103 ( taking into account the renumbered vertices ) . the right derived functor @xmath104 of @xmath105 defines an equivalence @xmath106 . it is obtained from the tilting complex @xmath107\oplus p'_1 $ ] @xcite . one has ( see @xcite ) @xmath108 where @xmath109 is the auslander reiten translate on @xmath75 . assume now that @xmath88 is equipped with an isomorphism @xmath110 . then @xmath111 yields an equivalence @xmath112 , which we denote by the same symbol . we use the same convention for the transpose isomorphism @xmath113 . [ ref-3.2.1 - 11 ] we have @xmath114 as functors @xmath115 . let @xmath100 be a representation of @xmath74 determined by a linear map @xmath116 and put @xmath117 . then one checks that @xmath118 is given by the exact sequence @xmath119 where the first non - trivial map induces the action @xmath120 . similarly if we put @xmath121 then one gets the same sequence @xmath122 where the first non - trivial map again yields the action @xmath123 . thus we have @xmath124 . below we put @xmath125 . [ ref-3.2.2 - 12 ] one has @xmath126 . in particular @xmath127 if and only if @xmath111 is self - adjoint or anti self - adjoint . this is a straightforward verification using lemma [ ref-3.2.1 - 11 ] and . for use below we record @xmath128\\ ai_2&=i_1\end{aligned}\ ] ] we let the notations be as in the previous section , put @xmath129 , @xmath130 $ ] . as @xmath131 is hereditary we have @xmath132\ ] ] inspection reveals that @xmath133\}\ ] ] [ ref-3.3.1 - 15 ] @xmath134 is @xmath18-calabi - yau if and only if @xmath111 is self - adjoint or anti self - adjoint . let @xmath4 be the serre - functor for @xmath131 . being canonical @xmath4 commutes with the auto - equivalence @xmath135 $ ] . hence @xmath4 induces an autoequivalence on @xmath134 which is easily seen to be the serre functor of @xmath134 . in @xmath134 we have @xmath136=(\pi^\ast\circ \pi^{-1})\circ a^2[1]=(\pi^\ast\circ \pi^{-1})[3]$ ] . thus @xmath134 is @xmath18-calabi - yau if and only if @xmath137 is isomorphic to the identity functor . it is easy to see that this is the case if and only is @xmath138 in @xmath139 . [ ref-3.3.2 - 16 ] the object @xmath140 in @xmath134 satisfies @xmath141 for @xmath142\}$ ] one computes @xmath143 thus we find @xmath144)&=\hom_\dscr(p_1,a^{-1}p_1)\\ & = \hom_\dscr(p_1,p_2)\\ & = w\end{aligned}\ ] ] @xmath145 @xmath146)&=\hom_\dscr(p_1,a p_1)\\ & = \hom_\dscr(p_1,i_2[-1])\\ & = 0\end{aligned}\ ] ] and @xmath147)&=\hom_\dscr(p_1,ap_1[1])\\ & = \hom_\dscr(p_1,i_2)\\ & = 0\qed\end{aligned}\ ] ] the following lemma is not used explicitly . the object @xmath140 in @xmath134 has the properties of a @xmath18-cluster tilting object , i.e. if @xmath148 for @xmath149 then @xmath69 is a sum of copies of @xmath140 . assume that @xmath142\}$ ] is such that @xmath150)=\hom_\dscr(p_1,n[2])=0 $ ] . we have to prove @xmath151 . we may rewrite @xmath152)&=\hom_\dscr(p_1[-1],n[1])\\ & = \hom_\dscr(a^{-1}p_1,n[1])\\ & = \hom_\dscr(p_2,n[1])\end{aligned}\ ] ] thus we find @xmath153 . hence @xmath154 . we deduce @xmath155\}$ ] . but if @xmath156 $ ] then @xmath152)&=\hom_\dscr(p_1,i_2[1])\\ & = \hom_\dscr(p_1,ai_2)\\ & = \hom_\dscr(p_1,i_1)\\ & \neq 0\end{aligned}\ ] ] so we are left with the possibility @xmath151 which finishes the proof . let @xmath19 be an algebraic @xmath71-finite krull - schmidt @xmath18-calabi - yau category containing a @xmath18-cluster tilting object @xmath20 such that @xmath157 . [ ref-3.4.1 - 19 ] let @xmath158 . then there exists a distinguished triangle in @xmath19 @xmath159^c\r n[1]\r\ ] ] let @xmath160 be defined ( up to isomorphism ) by the following distinguished triangle for @xmath161 but this would take a lot more space . ] @xmath162^{\ext^2_\tscr(t , n ) } \r n[1 ] \r\ ] ] a quick check reveals that @xmath163 . hence @xmath164 for some @xmath165 . we need to consider the special case @xmath166 $ ] . then the distinguished triangle ( constructed as in the proof ) has the form @xmath167\xrightarrow{\alpha } t[2 ] \xrightarrow{\beta}\ ] ] where @xmath81 is the universal map ( this follows from applying @xmath168 ) . since @xmath169)=k$ ] it follows that @xmath170,@xmath171 are determined up to ( the same ) scalar . this has a surprizing consequence . applying @xmath172 to the triangle we find that @xmath173,t)^{-1}$ ] defines an isomorphism @xmath174 thus @xmath175 comes equipped with an isomorphism @xmath176 which is canonical up to a scalar . in other words we are in the setting of [ ref-3.2 - 8 ] and we now use the notations introduced in sections [ ref-3.2 - 8 ] and [ ref-3.3 - 13 ] . as @xmath165 is obtained from the reflection in vertex @xmath177 , one verifies ( see [ ref-3.2 - 8 ] ) that @xmath165 is associated to the element of @xmath178 given by @xmath179\oplus p_1)$ ] where @xmath180\oplus p_1)$ ] is the composition @xmath181\oplus p_1)\ ] ] since the autoequivalence @xmath165 is a derived functor that commutes with coproducts it is isomorphic to a derived tensor functor @xmath182 for some @xmath183 , by @xcite . as a right @xmath74-module we have @xmath184\oplus p_1 $ ] . now we use the assumption that @xmath131 is algebraic and we proceed more or less as in the appendix to @xcite . by ( 4.3 ) we may assume that @xmath19 is a strict (= closed under isomorphism ) triangulated subcategory of a derived category @xmath185 for some dg - category @xmath186 . we denote by @xmath187 the full subcategory of @xmath188 whose objects are differential graded @xmath189-modules which are in @xmath19 when considered as @xmath186-modules . clearly @xmath190 is triangulated . by ( * lemma a.2.1(a ) ) @xmath20 may be lifted to an object in @xmath187 , which we also denote by @xmath20 . put @xmath191 $ ] . [ ref-3.4.2 - 23 ] one has an isomorphism in @xmath187 @xmath192\ ] ] as objects in @xmath19 we have @xmath193\oplus p_1)\lotimes_c s\\ & = i_2\lotimes_c s[-1]\oplus p_1\lotimes_c s\end{aligned}\ ] ] clearly @xmath194 . to compute @xmath195 we use the resolution @xmath196 tensoring with @xmath4 we get a distinguished triangle @xmath197\r i_2\lotimes_c s\r\ ] ] comparing with we find @xmath198 $ ] . thus , we have indeed an isomorphism @xmath199\ ] ] in @xmath19 . now we check that @xmath200 is @xmath74-equivariant in @xmath19 . the left @xmath74-module structure on @xmath201 is obtained from the ( homotopy ) @xmath74-action on @xmath202\oplus p_1 $ ] as given in . let @xmath203 be an element of @xmath204 . we need to prove that the following diagram is commutative in @xmath19 . @xmath205\lotimes_b s @>\cong > > t[1]\\ @v \pi(\mu)\lotimes_b \id_s vv @vv\mu v\\ p_1 \lotimes_b s @>>\cong > t \end{cd}\ ] ] we write this out in triangles @xmath206 @>\alpha > > t[2]@>\beta > > \\ @v \pi(\mu ) vv @vvv @vv\mu v\\ t @ > > > 0 @ > > > t[1]@>>\id > \end{cd}\ ] ] rotating the triangles we need to prove that the following square is commutative @xmath207 @>\beta[-1 ] > > t^{\ext^{-1}(t , t)}\\ @v\mu vv @vv\pi(\mu)v\\ t @= t \end{cd}\ ] ] this commutivity holds precisely because of the definition of @xmath111 . so @xmath81 is indeed @xmath74-equivariant . but according to ( * ? ? ? * lemma a.2.2 ) , any @xmath74-equivariant morphism in @xmath19 between objects in @xmath187 may be lifted to a morphism in @xmath187 . this finishes the proof . we now have a functor @xmath208 and by lemma [ ref-3.4.2 - 23 ] one finds that @xmath135(?)\lotimes_c t$ ] is isomorphic to @xmath209 . by the universal property of orbit categories @xcite we obtain a triangulated functor @xmath210 which sends @xmath140 to @xmath20 . @xmath211 is an equivalence . we observe that analogues of the distinguished triangles exist in @xmath134 ( with @xmath140 replacing @xmath20 ) . indeed , let @xmath212 . by we have @xmath213\}$ ] . if @xmath214 then @xmath215\cong an$ ] and the analog of is simply the image in @xmath134 of the projective resolution of @xmath216 in @xmath131 ( taking into account that @xmath217 $ ] ) . if @xmath156 $ ] then @xmath215=i_2 $ ] and the analog of is the image in @xmath134 of the projective resolution of @xmath218 in @xmath131 . to prove that @xmath211 is fully faithful we have to prove that @xmath211 induces an isomorphism @xmath219 . using the analogues of we reduce to @xmath220 $ ] . but since @xmath221,n)=\hom_\dscr(p_1[-1],n[-i-1])$ ] we reduce in fact to @xmath222 $ ] . it now suffices to apply @xmath223,-)$ ] to @xmath224^c\r n[1]\r\ ] ] taking into account that @xmath219 is an isomorphism for @xmath225 , @xmath151 , @xmath226 $ ] , @xmath227 $ ] by lemma [ ref-3.3.2 - 16 ] . as a last step we need to prove that @xmath211 is essentially surjective . but this follows from the distinguished triangles together with the fact that @xmath228 . to finish the proof of theorem [ ref-1.3 - 2 ] we observe that since @xmath19 is @xmath18-calabi - yau , so is @xmath134 . hence by lemma [ ref-3.3.1 - 15 ] @xmath111 is either self - adjoint or anti self - adjoint . by lemma [ ref-3.2.2 - 12 ] we deduce @xmath229 and hence we may write @xmath230 . it would be interesting to deduce the fact that @xmath111 is ( anti ) self - adjoint directly from the calabi - yau property of @xmath19 , without going through the construction of @xmath134 first . this would have made our arguments above more elegant . iyama and yoshino also consider @xmath231-calabi - yau categories @xmath19 equipped with a @xmath231-cluster tilting object @xmath20 such that @xmath24 and @xmath232 for @xmath233 . they relate such @xmath19 to the representation theory of the generalized kronecker quiver @xmath234 where @xmath235 . one may show that our techniques are applicable to this case as well and yield @xmath236)$ ] . we thank osamu iyama for bringing this point to our attention . let @xmath237 be a commutative finitely generated graded @xmath0-algebra . as in @xcite we write @xmath238 for the quotient of @xmath239 by the serre subcategory of graded finite length modules . we write @xmath240 for the quotient functor . if @xmath33 is generated in degree one and @xmath45 then by serre s theorem @xcite we have @xmath241 . now assume that @xmath33 is gorenstein . then we have @xmath242 $ ] where @xmath243 is the krull dimension of @xmath14 and @xmath244 . the number @xmath165 is called the gorenstein parameter of @xmath33 ( see ( * ? ? ? * definition 2.1 ) ) . if @xmath33 is a polynomial ring in @xmath28 variables ( of degree one ) then @xmath245 , @xmath246 . for use below we record another incarnation of the gorenstein parameter . let @xmath247 be the graded @xmath0-dual of @xmath33 . then @xmath248\ ] ] where @xmath249 denotes cohomology with support in the ideal @xmath250 . the following is a particular case of ( * ? ? ? * thm 2.5 ) . [ ref-6.1.2 - 44 ] if @xmath251 then there are fully faithful functors @xmath252 such that for @xmath253 there is a semi - orthogonal decomposition @xmath254 hence under the hypotheses of the theorem we obtain in particular that @xmath255 for arbitrary @xmath95 . unfortunately the functors @xmath256 introduced in the previous section are not compatible with @xmath40 . our aim in this section is to understand how @xmath40 acts on the image of @xmath256 . this requires us to dig deeper into orlov s construction which has the unusual feature of depending on the category @xmath257 where @xmath258 are the finitely generated graded @xmath33-modules with non zero components concentrated in degrees @xmath259 . the quotient functor @xmath260 has a right adjoint @xmath261 . its image is denoted by @xmath262 . we let @xmath93 be the graded projective @xmath33-module of rank one generated in degree @xmath95 ( i.e. @xmath263 ) . likewise @xmath264 is the simple @xmath33-module concentrated in degree @xmath95 . as in @xcite we put @xmath265 , @xmath266 and obvious variants with other types of inequality signs . in @xcite it is proved that the image @xmath267 of @xmath256 is the left orthogonal to @xmath268 inside @xmath257 . the identification of @xmath267 with the graded singularity category is through the composition @xmath269 assume @xmath251 . then the relation between @xmath267 , @xmath262 is given by the following semi - orthogonal decompositions @xmath270 this is a refinement of theorem [ ref-6.1.2 - 44 ] . the category @xmath35 comes equipped with the shift functor @xmath40 . we denote the induced endofunctor on @xmath267 by @xmath271 . we will now compute it . for @xmath272 we have @xmath273 where the symbol `` cone '' is to be understood in a functorial sense , for example by computing it on the level of complexes after first replacing @xmath68 by an injective resolution . let @xmath274 . to compute @xmath275 we see by that we have to find @xmath276 such that @xmath277 up to projectives . it is clear we should take @xmath278 now we note that the @xmath279 can be computed in @xmath280 . furthermore since the result lies in @xmath281 we can characterize it uniquely by applying @xmath111 to it . since @xmath111 commutes with @xmath40 we obtain with @xmath282 . let @xmath283 be as above but now assume that @xmath33 has an isolated singularity and let @xmath284 . then by a variant of ( * ? ? ? * thm 8.3 ) we have a canonical graded isomorphism @xmath285)\ ] ] and furthermore an appropriate version of local duality yields @xmath286 in other words we find @xmath287(-a))\ ] ] and hence the serre functor @xmath4 on @xmath288 is given by @xmath289(-a)$ ] . it is customary to write @xmath136 $ ] so that we have the usual formula @xmath290 in this setting we find @xmath291(-a)\ ] ] we remind the reader of the following well - known result . [ ref-6.4.1 - 50 ] let @xmath292 be a polynomial ring in @xmath28 variables of degree one . assume @xmath293 and let @xmath294 be the corresponding veronese subring of @xmath292 . i.e. @xmath295 . then @xmath294 is gorenstein with gorenstein parameter @xmath296 . the gorenstein property is standard . to compute the gorenstein invariant we first let @xmath33 be the `` blown up '' veronese . i.e. @xmath297 let @xmath298 be respectively the gorenstein parameters of @xmath33 and @xmath292 . if @xmath68 is a @xmath292-module write @xmath299 for @xmath300 , considered as graded @xmath33-module . we have @xmath301&=r\gamma_{a_{>0}}(a)\qquad ( \text{see \eqref{ref-6.1 - 43}})\\ & = r\gamma_{a_{>0}}(b)^+\\ & = r\gamma_{b_{>0}}(b)^+\\ & = ( b'(b)[-n])^+\\ & = a'(b)[-n]\end{aligned}\ ] ] in the 3rd equality we have used that local homology is insensitive to finite extensions . we deduce @xmath302 . since @xmath294 is obtained from @xmath33 by dividing the grading by @xmath303 obtain @xmath296 as gorenstein parameter for @xmath294 . in characteristic zero we could have formulated the result for invariant rings of finite subgroups of @xmath304 ( with the same proof ) . however in finite characteristic veronese subrings are not always invariant rings ( consider the case where the characteristic divides @xmath303 ) . let @xmath305 $ ] and @xmath306 . we have @xmath307 . by proposition [ ref-6.4.1 - 50 ] @xmath33 has gorenstein invariant @xmath177 . unfortunately we have to deal with the unpleasant notational problem that the shift functors on @xmath308 coming from @xmath33 and @xmath292 do not coincide . to be consistent with the sections [ ref-6.1 - 42],[ref-6.2 - 45 ] we will denote them respectively by @xmath40 and @xmath309 . thus @xmath310 . note that this choice is rather unconventional . according to theorem [ ref-6.1.2 - 44 ] we have a semi - orthogonal decomposition @xmath311 from the fact that @xmath44 has a strong exceptional collection @xmath312 , @xmath313 . @xmath314 we deduce that there is a semi - orthogonal decomposition @xmath315 in particular @xmath316 defines an equivalence between @xmath317 and the representations of the quiver @xmath318 @xmath91[]{$v$ } \psfrag{1}[][]{$1 $ } \psfrag{2}[][]{$2 $ } \includegraphics[width=3cm]{beilinson_mini}\ ] ] where @xmath319 and where @xmath320 corresponds to the vertex labeled by @xmath95 . by the auslander - reiten translate on @xmath35 is given by @xmath321(-1)$ ] . in other words : the shift functor on @xmath35 is given by @xmath322)^{-1}$ ] . by proposition [ gradedcase ] we find ( using @xmath323 ) @xmath324)\ ] ] which is what we wanted to show . note that this in this example we had no need for the somewhat subtle formula . we use similar conventions as in the previous section , let @xmath325 $ ] and @xmath326 . we have @xmath327 and we denote the corresponding shift functors by @xmath40 , @xmath309 so that @xmath328 . by proposition [ ref-6.4.1 - 50 ] @xmath33 has gorenstein invariant @xmath16 . by theorem [ ref-6.1.2 - 44 ] we have a semi - orthogonal decomposition @xmath329 now @xmath44 has a strong exceptional collection @xmath330 , @xmath331 . @xmath332 , @xmath333 . this sequence is geometric ( * ? ? ? * prop.3.3 ) and hence by every mutation is strongly exceptional ( * ? ? ? we get in particular the following strongly exceptional collection @xmath330 , @xmath332 . @xmath334 , @xmath333 where @xmath335 is defined by the exact sequence @xmath336 where @xmath337 . thus there is a semi - orthogonal decomposition @xmath338 an easy computation yields @xmath339 @xmath340 defines an equivalence between @xmath341 and the representations of the quiver @xmath342 @xmath343[]{$\wedge^2 v$ } \psfrag{1}[][]{$1 $ } \psfrag{2}[][]{$2 $ } \includegraphics[width=3cm]{beilinson_mini}\ ] ] put @xmath344 and choose an arbitrary trivialization @xmath345 . let @xmath176 be the resulting ( self - adjoint ) isomorphism . we are in the setting of [ ref-3.2 - 8 ] and hence can define @xmath29 as acting on the derived category of @xmath342 . we will now compute @xmath346 , @xmath347 . an easy computation yields @xmath348 using the formula we find @xmath349\ ] ] @xmath350\ ] ] where in the second line we have used the dual version of . let @xmath93 be the projective representation of @xmath342 generated in vertex @xmath95 . the endofunctor on @xmath351 induced by @xmath352 will be denoted by the same letter . we will now compute it . from we deduce immediately @xmath353 $ ] . to analyze we note that a suitably shifted slice of the koszul sequence has the form @xmath354 thus @xmath355 corresponds to the cone of @xmath356 which is easily seen to be equal to @xmath357 . if we use our chosen trivialization @xmath345 then we see that at least on objects @xmath352 coincides with @xmath358 $ ] . it is routine to extend this to an isomorphism of functors by starting with a bounded complex of projectives in @xmath359 . by proposition [ gradedcase ] we find ( using @xmath323 ) @xmath360)\ ] ] which is what we wanted to show . we keep notations as in the previous section . since in the iyama - yoshino examples @xmath35 is the derived category of a hereditary category the functor @xmath361 is essentially surjective @xcite and hence @xmath362 is also essentially surjective . in more complicated examples there is no reason however why this should be the case . nevertheless we have the following result which is probably well - known . [ ref-8.1 - 58 ] assume that @xmath0 has characteristic zero . let @xmath237 be a left noetherian graded @xmath0-algebra . put @xmath323 . let @xmath363 be such that @xmath364 . then @xmath68 is the completion of a finitely generated graded @xmath33-module @xmath69 . in the rest of this section we let the notations and hypotheses be as in the statement of the proposition ( in particular @xmath0 has characteristic zero ) . we denote the maximal ideal of @xmath14 by @xmath303 . let @xmath90 be the euler derivation on @xmath33 and @xmath14 . i.e. on @xmath33 we have @xmath365 and we extend @xmath90 to @xmath14 in the obvious way . if @xmath363 then we will define an euler connection as a @xmath0-linear map @xmath366 such that @xmath367 . if @xmath368 for @xmath69 a graded @xmath33-module then @xmath68 has an associated euler connection by extending @xmath369 for @xmath28 a homogeneous element of @xmath69 . [ ref-8.2 - 59 ] let @xmath68 be a finitely generated @xmath14 module . then @xmath68 has an euler connection if and only if @xmath68 is the completion of a finitely generated graded @xmath33-module . we have already explained the easy direction . conversely assume that @xmath68 has an euler connection . for each @xmath28 we have that @xmath370 is finite dimensional and hence it decomposes into generalized eigenspaces for @xmath371 . @xmath372 considering right exact sequences @xmath373 we easily deduce that the multiplicity of a fixed generalized eigenvalue in @xmath370 stabilizes a @xmath374 . thus @xmath375 where @xmath376 is a generalized eigenspace with eigenvalue @xmath170 . we put @xmath377 . then @xmath378 is noetherian since obviously any ascending chain of graded submodules of @xmath378 can be transformed into an ascending chain of submodules in @xmath68 . if particular @xmath378 is finitely generated and we have @xmath379 . now @xmath378 is @xmath0-graded and not @xmath380-graded . but we can decompose @xmath378 along @xmath380-orbits and then by taking suitable shifts we obtain a @xmath380-graded module with the same completion as @xmath378 . let @xmath381 and consider @xmath382 $ ] where @xmath33 acts via @xmath383 . we have a short exact sequence of @xmath33-modules @xmath384\r m\r 0\ ] ] which is split by hypotheses . denote the splitting by @xmath385 . for @xmath386 we have @xmath387 and hence @xmath388 hence @xmath371 is an euler connection and so we may invoke lemma [ ref-8.2 - 59 ] to show that @xmath368 . throughout @xmath389 is a ( commutative ) local noetherian ring , with maximal ideal @xmath390 and residue field @xmath0 . the _ singularity category _ of @xmath33 is the verdier quotient @xmath391 of the bounded derived category of finitely generated @xmath33-modules by the full subcategory of perfect complexes . recall that a functor @xmath392 is an _ equivalence up to direct summands _ if @xmath393 is fully faithful and every object @xmath394 is a direct summand of @xmath395 for some @xmath396 . we say that @xmath33 is a _ g - ring _ if the canonical morphism from @xmath33 to its @xmath390-adic completion @xmath397 is regular @xcite , and that @xmath33 _ has an isolated singularity _ if @xmath398 is regular for every non - maximal prime ideal @xmath399 of @xmath33 . our main result about singularity categories is the following : [ theorem : first_theorem ] let @xmath33 be a local noetherian ring with an isolated singularity , which is also a g - ring e.g. @xmath33 is essentially of finite type over a field . then the canonical functor @xmath400 is an equivalence up to direct summands . when @xmath33 is gorenstein there is an equivalence , due to buchweitz @xcite , between @xmath401 and the stable category of maximal cohen - macaulay @xmath33-modules @xmath402 , so in this case we obtain proposition [ ref-1.5 - 5 ] . we remark that , in general , @xmath403 is not an equivalence ( see e.g. example [ example : not_an_equivalence ] ) . let us outline the proof of the proposition . recall that a _ thick subcategory _ of a triangulated category @xmath404 is a triangulated subcategory closed under retracts . given an object @xmath74 of @xmath404 , we say that an object @xmath62 is _ finitely built _ from @xmath74 if it belongs to the smallest thick subcategory of @xmath404 containing @xmath74 . if every object of @xmath404 has this property , that is , if there are no proper thick subcategories of @xmath404 containing @xmath74 , then @xmath74 is said to _ classically generate _ @xmath404 . the local ring @xmath33 and its completion @xmath32 have the same residue field @xmath0 , and it is not difficult to see that @xmath403 induces an equivalence between the respective subcategories consisting of objects finitely built from @xmath0 . the subtlety lies in showing that , because @xmath33 has an isolated singularity , _ every _ object can be finitely built from @xmath0 . our proof of this fact uses homotopy colimits , which presents a technical problem since @xmath401 lacks infinite coproducts . one approach is to enlarge the category @xmath401 by considering the verdier quotient @xmath405 of the bounded derived category of all @xmath33-modules by the full subcategory of bounded complexes of projective @xmath33-modules . by ( * proposition 1.13 ) the canonical functor @xmath406 is fully faithful , and @xmath407 turns out to contain enough coproducts ( and thus homotopy colimits ) for our purposes . throughout @xmath408 denotes the ( unbounded ) derived category of @xmath33-modules . the next proposition follows immediately from the work of schoutens @xcite , but we give a direct proof in the special case of an isolated singularity . the result also follows from the general result by orlov @xcite and dyckerhoff @xcite has given a proof based on the theory of matrix factorizations in the hypersurface case . we begin with the easy direction . suppose that @xmath401 is classically generated by @xmath0 , and let @xmath409 be a prime ideal . the canonical functor @xmath410 is identically zero , because it sends the generator @xmath0 to zero . the image of this functor contains the residue field @xmath411 , from which we deduce that @xmath412 has finite projective dimension over @xmath398 . hence @xmath398 is regular , and we may conclude that @xmath33 has an isolated singularity . now suppose that @xmath33 has an isolated singularity , and let @xmath68 in @xmath413 be given . the idea is to write @xmath68 as a homotopy colimit , since coproducts in this category are rather subtle . instead , we consider the image under the quotient functor @xmath414 of homotopy colimits in @xmath415 . ] of a sequence of bounded complexes with finite length cohomology ; it follows that @xmath68 is a direct summand of one of the terms in this sequence , from which we conclude that @xmath0 classically generates . first , we set up some notation . given @xmath416 , define complexes @xmath417 } & : = a { \stackrel{a}{\longrightarrow } } a , \text { and } { \operatorname{e}[a ] } : = a { \stackrel{\can}{\longrightarrow } } a[a^{-1}],\end{aligned}\ ] ] both concentrated in degrees zero and one , and observe that the commutative diagram @xmath418_a \ar[r]^1 & a \ar[d]_{a^2 } \ar[r]^1 & a \ar[d]_{a^3 } \ar[r]^1 & \cdots\\ a \ar[r]_a & a \ar[r]_a & a \ar[r]_a & \cdots } \ ] ] is a direct system of complexes @xmath419 } { \longrightarrow}{\straightk[a^2 ] } { \longrightarrow}{\straightk[a^3 ] } { \longrightarrow}\cdots$ ] with colimit @xmath420}$ ] . more generally , given a sequence @xmath421 in @xmath33 , we define @xmath422 } : = \otimes_{j=1}^d { \straightk[a_j]}$ ] and @xmath423 } : = \otimes_{j=1}^d { \operatorname{e}[a_j]}$ ] . setting @xmath424 , there is a canonical isomorphism @xmath423 } \cong \varinjlim_i { \straightk[\bold{a}^i]}$ ] and thus a triangle @xmath425 } \ar[r]^{1-\text{shift } } & \bigoplus_{i \ge 1 } { \straightk[\bold{a}^i ] } \ar[r ] & { \operatorname{e}[\bold{a } ] } \ar[r ] & } \ ] ] in the derived category @xmath408 . this triangle expresses the fact that @xmath423}$ ] is the homotopy colimit of the @xmath426}$ ] in @xmath408 . for background on homotopy colimits , see @xcite . now let @xmath427 be a system of parameters for @xmath33 , and extend the augmentation morphism @xmath423 } { \stackrel{\varepsilon}{\longrightarrow } } a$ ] to a triangle @xmath423 } { \longrightarrow}a { \longrightarrow}{\check{\straightc}[\bold{a } ] } { \longrightarrow}$ ] , where the complex @xmath428 } : = \sigma \ker(\varepsilon)$ ] is given in each degree by @xmath428}^t = \oplus_{i_0 < \cdots < i_t } a[a_{i_0}^{-1},\ldots , a_{i_t}^{-1}]$ ] . tensoring with @xmath68 , we obtain a triangle @xmath429 } \otimes_a m { \longrightarrow}m { \longrightarrow}{\check{\straightc}[\bold{a } ] } \otimes_a m { \longrightarrow}\ ] ] in @xmath408 . since @xmath33 has an isolated singularity , @xmath430 $ ] has finite projective dimension over @xmath431 $ ] , and hence also over @xmath33 , for every sequence of indices @xmath432 in @xmath433 . here we use the fact that @xmath431 $ ] has finite projective dimension as an @xmath33-module \le 1 $ ] , which holds because @xmath434 = a[x]/(ax-1)$ ] . ] . we conclude that @xmath428 } \otimes_a m$ ] is , up to isomorphism in @xmath408 , a bounded complex of projective @xmath33-modules , whence the triangle ( [ eq : dbsing_generated_k1 ] ) gives rise to an isomorphism @xmath423 } \otimes_a m \cong m$ ] in @xmath407 . note that the coproduct @xmath435 } \otimes_a m$ ] is bounded , so tensoring ( [ eq : dbsing_generated_k0 ] ) with @xmath68 yields a triangle in @xmath407 of the form @xmath436 } \otimes_a m \ar[r]^{1-\operatorname{shift } } & \bigoplus_{i \ge 1 } { \straightk[\bold{a}^i ] } \otimes_a m \ar[r ] & m \ar[r ] & . } \ ] ] in what follows , let @xmath437 denote morphism sets in @xmath407 . one can check ( see lemma [ lemma : fg_commutes_coproducts ] below ) that @xmath438 commutes with coproducts coming from @xmath415 via the quotient functor , so applying @xmath438 to ( [ eq : second_triangle ] ) and using the argument of ( * ? ? ? * lemma 2.8 ) we deduce that @xmath439 } \otimes_a m).\ ] ] in particular , the identity @xmath440 corresponds to a split monomorphism @xmath441 } \otimes_a m$ ] in @xmath407 for some @xmath442 . the functor @xmath406 is fully faithful , so @xmath68 is also a direct summand of @xmath443 } \otimes_a m$ ] in @xmath401 . the cohomology modules of @xmath443 } \otimes_a m$ ] have finite length ( @xmath427 is a system of parameters ) so this complex is an iterated extension in @xmath413 of finite direct sums of copies of @xmath0 . it is now clear that any thick subcategory of @xmath401 containing @xmath0 must contain @xmath68 , and since @xmath68 was arbitrary , this completes the proof . we may , without loss of generality , assume that @xmath68 is a bounded above complex of finitely generated projective @xmath33-modules , that @xmath74 is a bounded complex of free @xmath33-modules , and that @xmath200 is a morphism of complexes . let @xmath449 be such that @xmath450 for @xmath451 . the image of @xmath452 is finitely generated , so let @xmath453 be a finite free submodule of @xmath454 with the property that @xmath455 factors as @xmath456 . similarly , let @xmath457 be a finite free submodule of @xmath458 with the property that @xmath459 , where @xmath460 is the differential . then @xmath461 factors as @xmath462 and the differential restricts to a map @xmath463 . proceeding in this way we construct a bounded complex @xmath211 of finite free @xmath33-modules and a factorization @xmath464 , as required . [ lemma : fg_commutes_coproducts ] let @xmath465 be a family of bounded complexes of @xmath33-modules such that there exist @xmath466 with @xmath467 for all @xmath468 $ ] and @xmath469 . then , given @xmath445 , the canonical map @xmath470 is an isomorphism , where @xmath471 denotes the degree - wise coproduct of complexes . by a standard argument , it is enough to prove that any morphism @xmath472 in @xmath407 factors through a finite subcoproduct . such a morphism is defined by a roof @xmath473_f \ar[dr]\\ m & & \oplus_i x_i } \ ] ] in @xmath415 , where the cone of @xmath474 is a bounded complex @xmath475 of projective @xmath33-modules . extending @xmath474 to a triangle @xmath476 in @xmath415 we deduce from lemma [ lemma : simple_factorisation ] that @xmath477 factors as @xmath478 for some @xmath448 . let @xmath479 denote the cone of @xmath480 . from the octahedral axiom applied to the pair @xmath481 we obtain a commutative diagram in @xmath415 of the form @xmath482_h \ar[dr]\\ y \ar[r]_f & m } \ ] ] where the cone of @xmath483 belongs to @xmath484 . the upshot is that the morphism in @xmath407 represented by the roof in ( [ eq : fg_commutes_coproducts_roof ] ) may also be represented by a roof with @xmath485 ( replace @xmath160 with @xmath486 ) . in this case @xmath160 is compact in @xmath415 by ( * ? ? ? * proposition 6.15 ) , so the morphism @xmath487 in the roof factors through a finite subcoproduct , which implies that @xmath488 factors through a finite subcoproduct in @xmath407 . to begin with , let @xmath33 denote an arbitrary local noetherian ring , and consider the canonical functor @xmath489 restriction of scalars defines a functor @xmath490 which sends a bounded complex of projective @xmath32-modules to a bounded complex of flat @xmath33-modules . since flat @xmath33-modules have finite projective dimension by ( * ? ? ? * part ii , corollary 3.2.7 ) , there is an induced functor @xmath491 right adjoint to @xmath492 . the unit of this adjunction is the canonical morphism @xmath493 which is obviously an isomorphism on @xmath0 , and thus also an isomorphism on the smallest thick subcategory @xmath494 of @xmath407 containing @xmath0 . by a standard argument of category theory , the restriction of @xmath492 to @xmath494 is fully faithful . in particular , @xmath403 induces an equivalence of the smallest triangulated subcategory of @xmath401 containing @xmath0 with the smallest triangulated subcategory of @xmath495 containing @xmath0 . now we assume that @xmath33 is a g - ring with an isolated singularity . the ( only ) reason for assuming that @xmath33 is a g - ring is that this guarantees that the completion @xmath32 has an isolated singularity ( * ? ? ? * lemma 2.7 ) . by proposition [ theorem : dbsing_generated_k ] the subcategory @xmath494 includes the image of @xmath401 under the canonical embedding @xmath496 , from which we infer that @xmath403 is fully faithful . it follows from a second application of proposition [ theorem : dbsing_generated_k ] that the thick closure of @xmath401 in @xmath495 is all of @xmath495 . since the thick closure of a triangulated subcategory is just the class of all direct summands of objects in the subcategory ( * ? ? ? * remark 2.1.39 ) , @xmath403 is an equivalence up to direct summands . it is easy to construct examples where @xmath403 is not an equivalence . it suffices to give a cohen - macauly module over the completion of a gorenstein local ring @xmath41 which is not _ extended _ from @xmath33 , i.e. which is not of the form @xmath497 for a cohen - macaulay @xmath33-module . [ example : not_an_equivalence ] let @xmath498_{(x , y)}/(x^3 + x^2 - y^2)$ ] be the local ring of a node , so the completion of @xmath33 is isomorphic to the reduced ring @xmath499/(uv)$ ] . this is a singularity of type @xmath500 and by ( * ? ? ? * ( 9.9 ) ) there are up , to isomorphism , exactly three indecomposable maximal cohen - macaulay @xmath4-modules , which are @xmath501 clearly @xmath502 , whence @xmath503 in @xmath504 . since @xmath505 are minimal prime ideals , @xmath506 and @xmath507 are fields , and it follows from a result of levy and odenthal ( * ? * theorem 6.2 ) that a finitely generated @xmath4-module @xmath68 is extended if and only if @xmath508 . hence @xmath399 and @xmath509 are not extended , and thus not in the essential image of @xmath403 , but their direct sum @xmath510 is extended . this corresponds to the fact that the nodal curve is irreducible , while the curve @xmath511 has two irreducible components . another argument that @xmath512 in @xmath504 belongs to the essential image of @xmath403 uses k - theory : simply apply ( * ? ? ? * corollary 4.5.12 ) . note that @xmath513 is a system of parameters for @xmath4 . it follows from the proof of proposition [ theorem : dbsing_generated_k ] that @xmath399 is a direct summand in @xmath504 of @xmath514 } \otimes { \mathfrak{p}}$ ] for some @xmath515 . in fact , @xmath514 } \otimes { \mathfrak{p } } = { \mathfrak{p } } { \stackrel{u^n}{\longrightarrow } } { \mathfrak{p}}$ ] is quasi - isomorphic to @xmath516 , and @xmath399 is a direct summand of @xmath517 in @xmath504 . to see this , observe that there is a triangle in the derived category @xmath518 and @xmath519 is zero in @xmath504 ( as it factors via @xmath4 ) so we may conclude that @xmath520 in @xmath504 . since @xmath521 is isomorphic as an @xmath4-module to the residue field @xmath522 , we see for a third time that @xmath523 is in the essential image of @xmath403 . [ remark : henselization ] denoting by @xmath524 the henselization of @xmath33 , the ring homomorphisms @xmath525 give rise to a factorization of @xmath403 as the composite @xmath526 where @xmath527 and @xmath528 . in the situation of proposition [ theorem : first_theorem ] , @xmath529 is an equivalence : up to a shift , every object of @xmath495 is a finitely generated module @xmath68 free on the punctured spectrum , and by elkik s theorem ( * ? ? ? * thorme 3 ) such modules can be descended to the henselization ; that is , there exists a finitely generated @xmath524-module @xmath69 such that @xmath530 . in particular , @xmath403 is an honest equivalence ( not just up to direct summands ) when @xmath33 is henselian . now we give the proof of proposition [ ref-1.4 - 3 ] . in @xcite krause produces an embedding @xmath531 , where @xmath532 is the homotopy category of @xmath533 of acyclic complexes of injective @xmath33-modules . this category is compactly generated , and @xmath203 induces an equivalence up to direct summands between @xmath401 and the full subcategory of compact objects in @xmath532 . the embedding @xmath203 produces a dg - enhancement for @xmath401 where for @xmath534 we put @xmath535 if @xmath33 is a noetherian @xmath380-graded ring ( not necessarily commutative ) then we may define the graded singularity category @xmath536 in the obvious way . since @xmath536 has an analogous dg - enhancement as @xmath401 we may define the orbit category @xmath537 ( see @xcite ) . by construction @xmath537 is a triangulated category ( with a dg - enhancement ) equipped with an exact functor @xmath538 such that @xmath537 is classically generated by its essential image and such that for @xmath539 we have @xmath540 forgetting the grading yields an exact functor @xmath541 which makes the shift @xmath542 isomorphic to the identity functor . hence by the universal property of orbit categories @xmath393 factors canonically through @xmath543 we then use the well - known formula @xmath548 and the corresponding formula in the graded case . this reduces us to proving @xmath549 this follows easily by replacing @xmath68 by a projective resolution . the third functor is an equivalence because of proposition [ theorem : first_theorem ] . the second functor is an equivalence because of @xcite . finally in lemma [ fff ] we have shown that @xmath544 is fully faithful . so we have to show that it is essentially surjective . this is clear by proposition [ theorem : dbsing_generated_k ] since @xmath0 lies in the essential image of @xmath544 . a. beligiannis , _ the homological theory of contravariantly finite subcategories : auslander - buchweitz contexts , gorenstein categories and ( co-)stabilization _ , comm . algebra * 28 * ( 2000 ) , no . 10 , 45474596 . buchweitz , d. eisenbud , and j. herzog , _ cohen - macaulay modules on quadrics _ , singularities , representation of algebras , and vector bundles ( lambrecht , 1985 ) ( berlin ) , springer , berlin , 1987 , pp . 58116 . r. elkik , _ solution dquations au - dessus danneaux hensliens _ , quelques problmes de modules ( sm . , cole norm . paris , 1971 - 1972 ) , soc . france , paris , 1974 , pp . astrisque , no . 16 . p. gabriel , _ auslander - reiten sequences and representation - finite algebras _ , representation theory , i ( proc . workshop , carleton univ . , ottawa , ont . , 1979 ) ( berlin ) , lecture notes in math . , vol . 831 , springer , berlin , pp . 171 . a. neeman , _ the connection between the @xmath553-theory localization theorem of thomason , trobaugh and yao and the smashing subcategories of bousfield and ravenel _ , ann . sci . cole norm . ( 4 ) * 25 * ( 1992 ) , no . 5 , 547566 .

we discuss the phenomenology of supersymmetric models in which supersymmetry breaking terms are induced by the super - weyl anomaly . such a scenario is envisioned to arise when supersymmetry breaking takes place in another world , i.e. , on another brane . we review the anomaly - mediated framework and study in detail the minimal anomaly - mediated model parametrized by only parameters : , , , and . the renormalization group equations exhibit a novel `` focus point '' ( as opposed to fixed point ) behavior , which allows squark and slepton masses far above their usual naturalness bounds . we present the superparticle spectrum and highlight several implications for high energy colliders . three lightest supersymmetric particle ( lsp ) candidates exist : the wino , the stau , and the tau sneutrino . for the wino lsp scenario , light wino triplets with the smallest possible mass splittings are preferred ; such winos are within reach of run ii tevatron searches . finally , we study a variety of sensitive low energy probes , including , the anomalous magnetic moment of the muon , and the electric dipole moments of the electron and neutron . # 1#2#3jhep * # 1 * , # 3 ( 19#2 ) # 1#2#3eur . phys . j. * c#1 * , # 3 ( 19#2 ) # 1#2#3nucl . phys . * b#1 * , # 3 ( 19#2 ) # 1#2#3nucl . phys . . suppl . * # 1 * , # 3 ( 19#2 ) # 1#2#3phys . lett . * b#1 * , # 3 ( 19#2 ) # 1#2#3phys . lett . * # 1b * , # 3 ( 19#2 ) # 1#2#3phys . rev . * d#1 * , # 3 ( 19#2 ) # 1#2#3phys . rev . lett . * # 1 * , # 3 ( 19#2 ) # 1#2#3phys . rep . * # 1 * , # 3 ( 19#2 ) # 1#2#3ann . rev . astron . astrophys . * # 1 * , # 3 ( 19#2 ) # 1#2#3ann . rev . nucl . part . sci . * # 1 * , # 3 ( 19#2 ) # 1#2#3mod . . lett . * a#1 * , # 3 ( 19#2 ) # 1#2#3zeit . fr physik * c#1 * , # 3 ( 19#2 ) # 1#2#3ap . j. * # 1 * , # 3 ( 19#2 ) # 1#2#3ann . phys . * # 1 * , # 3 ( 19#2 ) # 1#2#3rev . mod . phys . * # 1 * , # 3 ( 19#2 ) # 1#2#3comm . math . phys . * # 1 * , # 3 ( 19#2 ) epsf.tex ( # 1 width # 2)=#2

hep-ph9907319

signals from supersymmetry ( susy ) are important targets for particle physics experiments . these signals range from the direct discovery of supersymmetric particles at high energy colliders to indirect signals at lower energy experiments through measurements of flavor - changing processes , magnetic and electric dipole moments , and so on . the set of possible signals and the promise of individual experiments for susy searches depend strongly on what model of susy breaking is assumed . it is therefore important to understand the characteristic features and predictions of well - motivated susy breaking scenarios . probably the most well - known scenario is that of susy breaking in the supergravity framework , i.e. , `` gravity - mediated '' susy breaking . in this framework , susy breaking originates in a hidden sector and is transmitted to the observable sector though planck scale - suppressed operators . in particular , soft masses for squarks , sleptons , and higgs bosons are induced by direct khler interactions between hidden and observable sector fields . unfortunately , these khler interactions are not , in general , flavor - diagonal . squark and slepton mass matrices therefore typically have large flavor mixings , and these induce unacceptably large flavor - changing processes , such as @xmath6-@xmath7 mixing and @xmath8 @xcite . these difficulties , together commonly referred to as the susy flavor problem , may be avoided if the khler potential is somehow constrained to be flavor - diagonal . gauge - mediated susy breaking @xcite is one proposal for solving this problem . recently the mechanism of `` anomaly - mediated '' susy breaking has been proposed as a possibility for generating ( approximately ) flavor - diagonal squark and slepton mass matrices @xcite . in this scenario , susy is again broken in a hidden sector , but it is now transmitted to the observable sector dominantly via the super - weyl anomaly @xcite . gaugino and scalar masses are then related to the scale dependence of the gauge and matter kinetic functions . for first and second generation fields , whose yukawa couplings are negligible , wavefunction renormalization is almost completely determined by gauge interactions . their anomaly - mediated soft scalar masses are thus almost diagonal , and the susy flavor problem is solved . note that this solution requires that the anomaly - mediated terms be the dominant contributions to the susy breaking parameters . this possibility may be realized , for example , if susy breaking takes place in a different world , i.e. , on a brane different from the 3-brane of our world , and direct khler couplings are thereby suppressed @xcite . as will be discussed below , the expressions for anomaly - mediated susy breaking terms are scale - invariant . thus , they are completely determined by the known low energy gauge and yukawa couplings and an overall mass scale @xmath1 . anomaly - mediated susy breaking is therefore highly predictive , with fixed mass ratios motivating distinctive experimental signals , such as macroscopic tracks from highly degenerate wino - like lightest supersymmetric particles ( lsps ) @xcite . unfortunately , one such prediction , assuming minimal particle content , is that sleptons are tachyons . several possible solutions to this problem have already been proposed @xcite . we will adopt a phenomenological approach , first taken in ref . @xcite , and assume that the anomaly - mediated scalar masses are supplemented by an additional universal contribution @xmath9 . for large enough @xmath2 , the slepton squared masses are positive . along with the requirement of proper electroweak symmetry breaking , this defines the minimal anomaly - mediated model in terms of only 3 + 1 parameters : @xmath1 , @xmath2 , @xmath3 , and @xmath4 , where @xmath3 is the ratio of higgs vacuum expectation values ( vevs ) , and @xmath10 is the higgsino mass parameter . the simplicity of this model allows one to thoroughly examine all of parameter space . in this paper , we present a detailed study of the phenomenology of the minimal anomaly - mediated model . we begin in sec . [ sec : amsb ] with a brief discussion of the mechanism of anomaly - mediated susy breaking . in sec . [ sec : mam ] we review the tachyonic slepton problem and the universal @xmath2 `` solution , '' and present in detail the minimal anomaly - mediated model described above . the universal scalar mass @xmath2 breaks the simple scale invariance of expressions for soft terms . however , this breaking is rather minimal , in a sense to be explained , and the minimal anomaly - mediated model inherits several simple properties from the pure anomaly - mediated case . the naturalness of this model is examined in sec . [ sec : naturalness ] . we find that the minimal anomaly - mediated model exhibits a novel renormalization group ( rg ) `` focus point '' ( as opposed to fixed point ) behavior , which allows slepton and squark masses to be well above their usual naturalness bounds . the title `` supernatural supersymmetry '' derives from this feature and the envisioned other - worldly susy breaking . we then turn in sec . [ sec : highenergy ] to high - energy experimental implications . we explore the parameter space and find a variety of interesting features , including 3 possible lsp candidates : a degenerate triplet of winos , the lighter stau @xmath11 , and the tau sneutrino @xmath12 . the wino lsp scenario is realized in a large fraction of parameter space and has important new implications for both collider physics @xcite and cosmology @xcite . we find that naturalness and electroweak symmetry breaking favor light winos with the smallest possible mass splittings , i.e. , the ideal region of parameter space for wino searches and within the discovery reach of run ii of the tevatron . while anomaly - mediated models have the virtue that they predict very little flavor - changing in the first and second generations , they are not therefore automatically safe from all low - energy probes . in sec . [ sec : lowenergy ] we analyze several sensitive low - energy processes : @xmath5 , which probes flavor - changing in the third generation , and three important flavor-_conserving _ observables , the anomalous magnetic dipole moment of the muon , and the electric dipole moments of the electron and neutron . our conclusions and final remarks are collected in sec . [ sec : conclusions ] . in the appendix , we present expressions for anomaly - mediated susy breaking terms in a general supersymmetric theory and also the full flavor - dependent expressions for the specific case of the minimal anomaly - mediated model . in supergravity , susy breaking parameters always receive anomaly - mediated contributions . however , in the usual gravity - mediated susy breaking scenario , susy breaking masses also arise from direct interactions of observable sector fields with hidden sector susy breaking fields . such contributions are usually comparable to the gravitino mass , and so anomaly - mediated contributions , which are loop - suppressed relative to the gravitino mass , are sub - leading . however , in a model with no direct coupling between observable and hidden sectors , the anomaly - mediated terms can be the dominant contributions . in this paper , we assume that this is the case , and that the anomaly - mediated terms are ( one of ) the leading contributions to the susy breaking parameters . this is realized , for example , in the `` sequestered sector '' model of ref . @xcite , where the susy breaking sector and the observable sector are assumed to lie on different branes , thereby suppressing direct observable sector - hidden sector couplings . in global susy , the ( loop - corrected ) effective lagrangian may be written as @xmath13 w^\alpha w_\alpha + { \rm h.c . } \nonumber \\ & & + \int d^4\theta z_\phi ( \box , \lambda_{\rm cut}^*\lambda_{\rm cut } ) \phi^ * \phi \nonumber \\ & & + \int d^2\theta y \phi^3 + { \rm h.c . } + \cdots \ , \label{lglobal}\end{aligned}\ ] ] where @xmath14 and @xmath15 are the gauge field strength and chiral superfields , respectively . here @xmath16 is the @xmath17-function coefficient for the gauge coupling constant @xmath18 , @xmath19 is the wavefunction renormalization factor of @xmath15 , @xmath20 is the yukawa coupling constant , and @xmath21 is the cut - off of the theory . however , once we consider local susy , i.e. , supergravity , this expression is modified . the most important modification for our argument results from the fact that , in global susy , @xmath22 is given by @xmath23 . in supergravity , @xmath24 becomes a dynamical field and is part of the supergravity multiplet . @xmath22 must therefore be promoted to an object compatible with supergravity . the complete expression for @xmath22 is complicated . however , since we are interested only in the susy breaking terms , our task is simplified . perhaps the easiest prescription for deriving the susy breaking terms is to introduce the compensator superfield @xmath25 , whose vev is given by @xmath26 here @xmath27 is proportional to the vev of an auxiliary field in the supergravity multiplet and is of order the gravitino mass after susy breaking . with this compensator field , all of the terms relevant for calculating the anomaly - mediated susy breaking parameters are contained in the lagrangian @xcite @xmath28 because @xmath22 appears in eq . ( [ lglobal ] ) only through terms @xmath29 and @xmath30 , the replacement of @xmath22 by its supergravity generalization is effectively carried out by the replacement @xmath31 @xcite . expanding the above lagrangian with the vev of @xmath25 given in eq . ( [ < phi > ] ) , and solving the equation of motion for the auxiliary component of @xmath15 , the anomaly - mediated contributions to the gaugino mass @xmath32 , scalar squared mass @xmath33 , and trilinear scalar coupling @xmath34 are @xmath35 where @xmath36 here @xmath16 and @xmath37 are to be evaluated with the supersymmetric field content present at the appropriate scale . in the above formulae , indices have been suppressed . the full expressions for general chiral superfield content may be found in the appendix . one important feature of this result is that the formulae for the anomaly - mediated susy breaking parameters are rg - invariant @xcite . the anomaly - induced masses are given as functions of the gauge and yukawa coupling constants , as shown in eqs . ( [ m_g(amsb ) ] ) ( [ a(amsb ) ] ) , and the @xmath17-functions for the individual susy breaking parameters agree with the @xmath17-functions of the right - hand sides whose scale dependences are determined through the gauge and yukawa coupling rg equations . as described in the previous section , in pure anomaly - mediated susy breaking , soft terms are determined by rg - invariant expressions involving the gauge and yukawa couplings . the soft terms are therefore completely fixed by the low energy values of these couplings and an overall scale @xmath1 . if a scalar has negligible yukawa interactions , its squared mass is determined by gauge coupling contributions @xmath38 , where the sum is over all gauge groups under which the scalar is charged , and ( positive ) constants have been omitted ( see appendix ) . from this form , we see that sleptons , which interact only with non - asymptotically free groups ( @xmath39 ) , have negative squared masses . tachyonic sleptons are the most glaring problem of the anomaly - mediated scenario . several mechanisms for solving the tachyonic slepton problem have been proposed . additional positive contributions to slepton squared masses may arise from bulk contributions @xcite , gauge - mediated - like contributions @xcite , new yukawa interactions @xcite , or non - decoupling higher order threshold effects @xcite . here , we adopt a simple phenomenological approach @xcite : we assume an additional , universal , non - anomaly - mediated contribution @xmath9 to all scalars at the grand unified theory ( gut ) scale @xmath40 . the resulting boundary conditions , @xmath41 define the minimal anomaly - mediated model . for large enough @xmath9 , slepton squared masses are therefore positive , and the tachyonic slepton problem is averted . such a universal term may be produced by bulk interactions @xcite , but is certainly not a feature common to all anomaly - mediated scenarios . the extent to which the following results depend on this assumption will be addressed in sec . [ sec : conclusions ] . the addition of a non - anomaly - mediated term destroys the feature of rg invariance . however , the rg evolution of the resulting model nevertheless inherits some of the simplicity of the original pure anomaly - mediated relations . schematically , scalar masses @xmath42 satisfy the one - loop rg equations @xmath43 \ , \label{rge}\ ] ] where @xmath44 , positive numerical coefficients have been omitted , and the sum is over all chiral fields @xmath45 interacting with @xmath46 through the yukawa coupling @xmath20 . letting @xmath47 , where @xmath48 thus , at one - loop , the deviations from pure anomaly - mediated relations satisfy simple evolution equations that depend only on the deviations themselves . for scalars with negligible yukawa couplings , such as the first and second generation squarks and sleptons , the deviation @xmath49 is a constant of rg evolution . for them , @xmath50 is simply an additive constant , and the weak scale result for @xmath51 is independent of the scale at which @xmath52 is generated . for fields interacting through large yukawa couplings such as the top yukawa coupling , the deviations @xmath52 evolve ; however , this evolution is simply analyzed . we will see an important consequence of this evolution for naturalness in sec . [ sec : naturalness ] . we will assume that the boundary conditions of eq . ( [ bc ] ) are given at @xmath53 . the susy breaking parameters are then evolved with one - loop rg equations to the superparticle mass scale @xmath54 , which we have approximated to be the squark mass scale . for the gaugino mass parameters , we also include the largest next - to - leading order corrections from @xmath55 and @xmath56 given in ref . @xcite . all parameters of the theory are then specified , except for @xmath10 , the higgsino mass parameter , and @xmath57 , the soft bilinear higgs coupling . we do not specify the mechanism for generating these parameters , but assume that they are constrained so that electroweak symmetry is properly broken . given the other soft parameters at @xmath54 , the higgs potential is determined by @xmath10 and @xmath57 , or alternatively , by the fermi constant @xmath58^{-1 } \simeq 1.17\times 10^{-5}~{\rm gev}^{-2 } $ ] ( or , equivalently , the @xmath59 mass ) and @xmath60 . it is more convenient to use the latter two as inputs ; @xmath10 and @xmath57 are then fixed so that the higgs potential has a proper minimum with correct @xmath61 and @xmath3 . we minimize the higgs potential at one - loop , including radiative corrections from third generation quarks and squarks @xcite , but neglecting radiative corrections from other particles . in fact , the constraint of proper electroweak symmetry breaking does not determine the sign of the @xmath10 parameter . is a complex parameter , and its phase can not be determined from the radiative breaking condition . in sec . [ sec : edm ] , we consider the implications of complex @xmath10 . however , in the rest of the paper , we assume that @xmath10 is real . in the anomaly - mediated framework , there are several models in which all cp - violating phases in susy parameters are absent @xcite . ] the entire parameter space of the minimal anomaly - mediated model is therefore specified by 3 + 1 parameters : @xmath62 supersymmetric theories are considered natural from the point of view of the gauge hierarchy problem if the electroweak scale is not unusually sensitive to small variations in the underlying parameters . there are a variety of prescriptions for quantifying naturalness with varying degrees of sophistication @xcite . for the present purposes , we simply consider a set of parameters to be natural if no large cancellations occur in the determination of the electroweak scale . at tree - level , the relevant condition is @xmath63 where @xmath64 and @xmath65 are the soft susy breaking masses for up- and down - type scalar higgses . naturalness then requires that @xmath66 as determined from electroweak symmetry breaking not be too far above the electroweak scale . a typical requirement is @xmath67 . in fig . [ fig : mu ] we present values of @xmath10 in the @xmath68 plane for three representative values of @xmath3 : 3 ( low ) , 10 ( moderate ) , and 30 ( high ) . we have chosen @xmath69 to avoid constraints from @xmath5 at large @xmath3 ( see sec . [ sec : bsg ] ) , but similar @xmath66 are found for @xmath70 . the parameter @xmath1 is not phenomenologically transparent , and so on the right - hand axis , we also give approximate values of the wino mass @xmath71 , using @xmath72 . the value of @xmath66 rises with increasing @xmath1 , as expected . irrespective of @xmath2 and @xmath3 , @xmath73 implies @xmath74 . such a restriction is encouraging for searches for degenerate winos at upcoming runs of the tevatron , as will be discussed more fully in sec . [ sec : charginos ] . the special case of @xmath75 corresponds to pure anomaly - mediated susy breaking . in this case , the expressions for soft susy breaking terms are rg - invariant and the soft masses may be evaluated at any scale , including a low ( tev ) scale . based on this observation , it has been argued that , since the stop masses do not enter the determination of @xmath76 with large logarithms through rg evolution , stop masses of @xmath77 or even higher are consistent with naturalness @xcite . this is contradicted by fig . [ fig : mu ] : for @xmath75 , as will be seen in sec . [ sec : squarks ] , stop masses of @xmath77 require very large @xmath1 corresponding to values of @xmath66 above 2 tev . stop masses of 2 tev are therefore as unnatural in _ pure _ anomaly - mediated susy breaking as they are in more conventional gravity - mediated scenarios , such as minimal supergravity . this applies to all cases where the pure anomaly - mediated relations are approximately valid for squark and higgs soft masses , and includes models in which a mechanism for avoiding tachyonic sleptons is invoked which does not disturb the squark and higgs masses . for the minimal anomaly - mediated model with @xmath78 , however , the squark and higgs masses are explicitly modified , and the argument above does not apply . it is exactly in this case , where the soft susy masses are _ not _ rg - invariant , that there is the possibility that heavy squarks can be consistent with naturalness , and we will see that , in fact , this is realized by a novel mechanism for large @xmath2 . in fig . [ fig : mu ] , for @xmath79 , an upper bound on @xmath66 implies an upper bound on @xmath2 . however , for moderate and large @xmath3 , the contours of constant @xmath66 are extremely insensitive to @xmath2 , and so large squark and slepton masses are consistent with naturalness in the large @xmath2 regime . to @xmath2 is implicit in fig . 1 ; its implications for naturalness were not noted . ] this behavior may be understood first by noting that , for moderate and large @xmath3 , eq . ( [ naturalness ] ) implies that @xmath10 depends sensitively on @xmath64 only . the rg evolution of @xmath64 is most easily understood by letting @xmath80 , where @xmath81 is the pure anomaly - mediated value , and similarly for all other scalar masses . the deviations @xmath52 satisfy simple rg equations , as discussed in sec . [ sec : mam ] . for @xmath3 not extremely large , the only large yukawa is the top yukawa @xmath82 , and @xmath64 is determined by the system of rg equations @xmath83 where @xmath84 and @xmath85 denote the third generation squark su(2 ) doublet and up - type singlet representations , respectively . such systems of rg equations are easily solved by decomposing arbitrary initial conditions into components parallel to the eigenvectors of the evolution matrix , which then evolve independently @xcite . in the present case , the solution with initial condition @xmath86 is @xmath87 -\frac{m_0 ^ 2}{2 } \left ( \begin{array}{c } 1 \\ 0 \\ -1 \end{array } \right ) \ .\ ] ] for @xmath88 and @xmath82 such that @xmath89 = 1/3 $ ] , @xmath90 , i.e. , @xmath64 assumes its pure anomaly - mediated value for any @xmath2 . the rg evolution of @xmath64 is shown for several values of @xmath2 in fig . [ fig : run ] . as expected , the rg curves intersect at a single point where @xmath64 is independent of @xmath2 ; we will call this a `` focus point . '' remarkably , however , the focus point occurs near the weak scale for @xmath82 corresponding to the physical top mass of @xmath91 . thus the weak scale value of @xmath64 is nearly its pure anomaly - mediated value for all values of @xmath2 . note that this behavior applies only to @xmath64 ; no other scalar mass has a focus point behavior . the focus point is not a fixed point ; for example , below the focus point , the rg curves diverge again . the position of the focus point depends on @xmath82 , and we must check the sensitivity to variations in @xmath82 . in fig . [ fig : run ] we show also the behavior for @xmath82 corresponding to @xmath92 gev . the exact weak scale value of @xmath64 depends on @xmath82 and , when the focus point is not exactly at the weak scale , also on @xmath2 . however , for top quark masses near the physical one , the focus point remains within a couple of decades of the weak scale , and the sensitivity to variations in @xmath2 is always suppressed . this is demonstrated in fig . [ fig : naturalness ] , where values of @xmath10 are given in the @xmath93 plane . even for @xmath94 and @xmath95 , we find that @xmath96 lies naturally below 2 @xmath97 . an interesting question is whether @xmath2 can be bigger than the weak scale by a loop factor without compromising naturalness . if this were the case , there would be no need to appeal to a sequestered sector to eliminate tree - level scalar masses . however , @xmath2 can not be arbitrarily large . in fig . [ fig : naturalness ] , we see that the requirement of proper electroweak symmetry breaking implies @xmath98 . in any case , a similar bound would follow from requiring that one - loop finite corrections to the higgs squared mass parameter , which are proportional to @xmath99 , not introduce large fine - tunings . the maximum allowed @xmath9 is thus roughly an order of magnitude below @xmath100 . thus , while it is possible to eliminate the sequestered seor mechanism for direct khler interaction suppression , it is still required that the tree - level scalar squared mass @xmath9 be suppressed by an order of magnitude relative to its `` natural '' value @xmath101 . nevertheless , given that we have no understanding of the source of @xmath2 , it is at least somewhat reassuring that it may be far above the weak scale without incurring a fine - tuning penalty . a direct consequence of this is that the minimal anomaly - mediated model is a model that naturally accommodates multi - tev sleptons and squarks . as we will see below , this has important phenomenological consequences both for high energy colliders and low energy probes . having defined the minimal anomaly - mediated model in sec . [ sec : mam ] and explored the natural range of its fundamental parameters in sec . [ sec : naturalness ] , we now consider the resulting masses and mixings of the superpartners . the lightest supersymmetric particles are either a degenerate triplet of charginos and neutralinos , the lighter stau @xmath11 , or the tau sneutrino @xmath12 . we begin by considering these , and conclude with a discussion of the squark spectrum . we do not discuss the gluino and heavy higgses in detail . however , their masses are given in eq . ( [ gauginoratios ] ) and figs . [ fig : bsgp ] and [ fig : bsgn ] , respectively . charginos and neutralinos are mixtures of gauginos and higgsinos . their composition is determined by @xmath71 , @xmath102 , @xmath10 , and @xmath3 at tree - level . inserting the values of the gauge coupling constants at @xmath103 in eq . ( [ m_g(amsb ) ] ) , and including the largest next - to - leading corrections as described in sec . [ sec : mam ] , we find @xmath104 typical values of @xmath105 allowed by radiative electroweak symmetry breaking in the minimal anomaly - mediated model are given in fig . [ fig : mum2 ] . combined with the anomaly - mediated relation @xmath106 , fig . [ fig : mum2 ] implies @xmath107 with substantial hierarchies in these parameters throughout parameter space . the chargino and neutralino mass eigenstates are therefore well - approximated by pure gaugino and pure higgsino states with masses @xmath108 and the lightest of these is always a highly degenerate triplet of winos . in much of parameter space , as we will see in sec . [ sec : sleptons ] , these winos are the lsps . the possibility of searching for supersymmetry in the wino lsp scenario has been the subject of much recent attention @xcite . the detection of wino lsps poses novel experimental challenges . neutral winos pass through collider detectors without interacting . charged winos are detectable in principle , but are typically highly degenerate with neutral winos , with @xmath109 and corresponding decay lengths @xmath110 cm @xcite . they therefore decay to invisible neutral winos and extremely soft pions before reaching the muon chambers , thereby escaping both conventional searches based on energetic decay products and searches for long - lived charged particles that produce hits in the muon chamber . [ fig : mum2 ] , however , has two important and encouraging implications for wino lsp searches . first , as noted in sec . [ sec : naturalness ] , naturalness bounds on @xmath66 imply stringent bounds on @xmath71 . from fig . [ fig : mum2 ] , for example , we find that @xmath111 implies @xmath112 . continuing searches at lep @xcite , although limited kinematically to the region @xmath113 , will be able to probe a significant fraction of this parameter region . in addition , such limits on the wino mass imply large cross sections at the tevatron . for @xmath114 and @xmath115 , the wino pair production rate is @xmath116 fb , and if a jet with @xmath117 and @xmath118 is required for triggering , the associated production rate is @xmath119 fb @xcite . such cross sections imply hundreds of wino pairs produced at the upcoming run ii , and tens of wino pairs produced in association with jets . second , the region of @xmath105 space favored in fig . [ fig : mum2 ] is the far gaugino region , where @xmath120 is minimized . for the parameters of fig . [ fig : mum2 ] , @xmath121 , corresponding to decay lengths of @xmath122 cm . ( see ref . thus , a significant fraction of winos will pass through several vertex detector layers . when produced in association with a jet for triggering , such winos will be discovered off - line as high @xmath123 tracks with no associated calorimeter or muon chamber activity . such a signal should be spectacular and background - free . this possibility is discussed in detail in ref . @xcite , where an integrated luminosity of 2 @xmath124 is shown to probe the entire region discussed here with @xmath125 . it is exciting that run ii of the tevatron will either discover wino lsps or exclude most of the natural region of parameter space in this model . slepton masses and mixings are given by the mass matrix @xmath126 in the basis @xmath127 , and sneutrino masses are given by @xmath128 where @xmath129 and @xmath130 are the soft susy breaking masses . in anomaly - mediated models , as discussed in ref . @xcite , if both @xmath129 and @xmath130 receive the same @xmath9 contribution , the diagonal entries of the slepton mass matrix are accidentally highly degenerate . the anomaly - mediated boundary conditions imply ( see the appendix ) @xmath131 + m_z^2 \left[2\sin^2\theta_w - \frac{1}{2 } \right ] \cos 2\beta \ . \label{sleptondiff}\ ] ] for @xmath132 , @xmath133 , and both bracketed expressions are extremely small . this accidental degeneracy implies that same - flavor sleptons may be highly degenerate . the physical mass splitting for staus is given in fig . [ fig : staudm ] . for low @xmath3 ( and , by implication , for all @xmath3 for selectrons and smuons ) , degeneracies of order 10 gev or less are found throughout the parameter region . for large @xmath3 , however , large yukawa effects dilute the degeneracy significantly . equation ( [ sleptondiff ] ) also implies that even small off - diagonal entries may lead to large mixing . the left - right mixing for staus is given in fig . [ fig : staulr ] . throughout parameter space , and even for low @xmath3 , the stau mixing is nearly maximal . in fact , even smuon mixing may be significant for large @xmath3 and low @xmath1 , it too is almost maximal . nearly degenerate and highly - mixed same flavor sleptons are a distinctive feature of the minimal anomaly - mediated model and distinguish it from other gravity- and gauge - mediated models , where , typically , @xmath134 . these features may be precisely tested by measurements of slepton masses @xcite and mixings @xcite at future colliders . the lighter stau @xmath11 is always the lightest charged slepton , and it therefore plays an important phenomenological role . the @xmath11 mass is displayed in fig . [ fig : staum ] . for low @xmath2 , @xmath11 is either tachyonic or excluded by experimental bounds . the current bounds are fairly complicated in this model , since the mass ordering and mass splittings between @xmath11 , the winos , and the sneutrinos vary throughout the parameter space . for staus decaying to neutralinos with a mass splitting greater than 15 gev , combined lep analyses of the @xmath135 data yield the bound @xmath136 @xcite , but this drops to near the lep i limit of 45 gev as the mass splitting goes to zero . however , for stable staus , combined lep analyses of data up to @xmath137 imply @xmath138 @xcite . the light shaded region of fig . [ fig : staum ] is excluded by @xmath139 and represents a rough summary of these bounds . in the remaining region , the bounds @xmath140 @xcite , @xmath141 @xcite , and @xmath142 @xcite are always satisfied . in the following , we will include the excluded shaded region in plots of observables that involve sleptons . for quantities such as squark masses or rates for @xmath5 , we omit this , as such quantities are well - defined even for small @xmath2 , and in fact , the @xmath75 axis gives their values in anomaly - mediated models where the slepton mass problem is fixed without changing the squark and higgs masses . for large @xmath2 , @xmath143 , and the wino is the lsp . this is the case in the unshaded region of fig . [ fig : staum ] . the experimental implications of the wino lsp scenario have been discussed above in sec . [ sec : charginos ] . finally , there exists an intermediate @xmath2 region , in which the lsp is either the @xmath11 or the @xmath144 . in the @xmath11 lsp scenario ( the dark shaded region of fig . [ fig : staum ] ) , the stau may be found at both lep and the tevatron through its spectacular anomalous @xmath123 and time - of - flight signatures @xcite . at the tevatron , for example , for @xmath145 , @xmath146 fb , and so a significant fraction of the stau lsp parameter space may be explored .. ) this scenario therefore requires some mechanism for diluting the stau density , such as late inflation with a low reheating temperature @xcite . ] in the case of the sneutrino lsp ( the blackened region of fig . [ fig : staum ] ) , there are many possible experimental signatures . while this region appears only for a limited range of susy parameters , superparticles tend to be relatively light in this region , with @xmath147 and @xmath148 , and so it is amenable to study at lep . in this region , the slepton mass ordering is always @xmath149 and the wino triplet may appear anywhere between the sneutrinos and @xmath150 . typically , though not always , the only kinematically accessible superparticles at lep are the sneutrinos , @xmath11 and the winos . the two possible mass orderings and dominant decay modes in each scenario are then @xmath151 in anomaly - mediated susy breaking , squarks are universally very heavy , as their masses receive contributions from the strong coupling . the gauge coupling contribution to scalar squared masses is of the form @xmath152 , where @xmath153 is the one - loop @xmath17-function coefficient ( see appendix ) , and so the strong coupling contribution completely overwhelms those of the su(2 ) and u(1 ) couplings . squark masses for the first two generations are therefore both flavor- and chirality - blind ; we find that the @xmath154 , @xmath155 , @xmath156 , and @xmath157 , and their second generation counterparts are all degenerate to within @xmath158 gev throughout parameter space . the first and second generation squark masses are given in fig . [ fig : sq ] . the squarks are hierarchically heavier than winos and sleptons for low @xmath2 , and their mass increases as @xmath2 increases . for @xmath159 , the squark mass is above 2 tev . thus , the focus point naturalness behavior discussed in sec . [ sec : naturalness ] , which allows such large @xmath2 , has important phenomenological consequences . direct detection of 2 tev squarks is likely to be impossible at the lhc or nlc , and must wait for even higher energy hadron or muon colliders . note , however , that some superparticles , notably the gauginos , can not evade detection at the lhc and nlc . unlike the squarks of the first two generations , the masses of third generation squarks @xmath160 , @xmath161 , @xmath162 , and ( for large @xmath3 ) @xmath163 receive significant contributions from large yukawa couplings . these are shown in figs . [ fig : stop ] and [ fig : sbot ] for small and large values of @xmath3 . yukawa couplings always reduce the masses and their effect may be large . for example , @xmath164 may be reduced by as much as 40% relative to the first and second generation squark masses . at the lhc , therefore , stops and sbottoms may be produced in much larger numbers than the other squarks , adding to the importance of @xmath16-tagging . as in the case of sleptons , third generation squarks may have large left - right mixing . for @xmath165 , left - right mixing in both the stops and sbottoms is large , and is nearly maximal for low @xmath2 . for @xmath79 , sbottom mixing is negligible , but stop mixing may still be as large as @xmath166 . anomaly - mediated supersymmetry breaking naturally suppresses flavor - violation in the first and second generations , but not all low energy constraints are therefore trivially satisfied . in particular , since anomaly - mediated soft terms depend on yukawa couplings , non - trivial flavor mixing involving third generation squarks can be expected . we first study the flavor - changing process @xmath5 , which is well - known for being sensitive to third generation flavor violation . we then consider magnetic and electric dipole moments , observables that are flavor - conserving , but are nevertheless highly sensitive to susy effects . [ [ sec : bsg ] ] in the standard model , the flavor - changing transition @xmath5 is mediated by a @xmath167 boson at one - loop . in supersymmetric theories , @xmath5 receives additional one - loop contributions from charged higgs- , chargino- , gluino- , and neutralino - mediated processes . the charged higgs contribution depends only on the charged higgs mass and @xmath3 , interferes constructively with the standard model amplitude , and is known to be large even for charged higgs masses beyond current direct experimental bounds . the supersymmetric contributions may also be large for some ranges of susy parameters . thus , @xmath5 provides an important probe of all supersymmetric models , including those that are typically safe from other flavor - violating constraints . in the well - studied cases of minimal supergravity and gauge - mediated susy breaking @xcite , the chargino- and , to a lesser extent , gluino - mediated contributions may be significant for large @xmath3 . neutralino contributions are always negligible . for @xmath168 ( in our conventions ) , these contributions are constructive and so , for large @xmath3 , positive @xmath10 is favored . in the present case of anomaly - mediated susy breaking , several new features arise . first , in contrast to the case of minimal supergravity and gauge - mediation where squark mixing arises only through rg evolution , flavor violation in the squark sector is present even in the boundary conditions ( and receives additional contributions from rg evolution ) . more importantly , the signs of the parameter @xmath169 and the gluino mass @xmath170 are opposite to those of minimal supergravity and gauge - mediation . the leading contributions for large @xmath3 in the mass insertion approximation from charginos and gluinos are given in fig . [ fig : feynmanbsg ] . for large @xmath3 , the amplitudes @xmath171 and @xmath172 are both opposite in sign relative to their values in minimal supergravity and gauge - mediation . @xmath173 may be calculated by first matching the full supersymmetric theory on to the effective hamiltonian @xmath174 at the electroweak scale @xmath175 . in the basis where the current and mass eigenstates are identified for @xmath176 , @xmath177 , and @xmath178 , supersymmetry contributes dominantly to the wilson coefficients @xmath179 and @xmath180 of the magnetic and chromomagnetic dipole operators @xmath181 ( contributions to operators with chirality opposite to those above are suppressed by @xmath182 and are negligible . ) we use next - to - leading order ( nlo ) matching conditions for the standard model @xcite and charged higgs @xcite contributions . the remaining supersymmetric contributions are included at leading order @xcite . some classes of nlo supersymmetric contributions have also been calculated @xcite ; however , a full nlo calculation is not yet available . for the present purposes , where we will be scanning over susy parameter space , the leading order results are sufficient . note that the inclusion of some , but not all , nlo effects is formally inconsistent , but by doing so , we are effectively assuming that the nlo corrections in a given renormalization scheme are numerically small . the wilson coefficients @xmath183 at the weak scale are then evolved down to a low energy scale @xmath184 of order @xmath185 , where matrix elements are evaluated using the resulting effective operators . the nlo anomalous dimension matrix is now known @xcite , as are the nlo matrix elements @xcite and the leading order qed and electroweak radiative corrections @xcite . these have been incorporated in the analysis of ref . @xcite , where a simple form for @xmath173 in terms of weak scale wilson coefficients is presented . the exact parametrization depends on the choice of @xmath184 and the photon energy cutoff @xmath186 . we choose @xmath187 and @xmath188 . the susy branching fraction is then given by @xcite @xmath189 where @xmath190 are the fractional deviations from standard model amplitudes : @xmath191 for the standard model value , we take @xcite @xmath192 where the theoretical error includes uncertainties from scale dependence and standard model input parameters . the most stringent experimental bounds are @xmath193 which may be combined in a weighted average of @xcite @xmath194 bounds on susy parameter space are extremely sensitive to the treatment of errors . with this in mind , however , to guide the eye in the figures below , we also include bounds from eq . ( [ bsgtotalexp ] ) with @xmath195 experimental errors : @xmath196 similar bounds would follow from combining 1@xmath197 experimental and theoretical errors linearly . given a set of parameters @xmath1 , @xmath2 , @xmath3 , and @xmath4 , we may now determine @xmath198 , assuming the central value of eq . ( [ bsgsm ] ) . in fig . [ fig : bsghplus ] we plot @xmath199 as a function of @xmath200 , for three representative values of @xmath3 , fixed choice of @xmath4 , and scanning over the remaining parameters @xmath1 and @xmath2 . the solid lines show the value when only the charged higgs diagram is included . as in minimal supergravity and gauge - mediated models , the neutralino diagrams are negligible , but the chargino and , to a lesser extent , gluino diagrams may be substantial , especially for large @xmath3 . in contrast to these other susy models , however , as a result of the sign flips in @xmath169 and @xmath170 noted above , both chargino and gluino contributions _ enhance _ the standard model prediction for @xmath70 . the parameter space with @xmath70 is thus highly constrained , and requires large charged higgs masses , especially for large @xmath3 . for example , for @xmath165 , the upper bound of eq . ( [ bsgbound ] ) implies @xmath201 , significantly more stringent than the bound @xmath202 that would apply in the absence of chargino and gluino contributions . for @xmath203 , the supersymmetric contributions may cancel the charged higgs contribution , and the parameter space is constrained only for very low @xmath1 and @xmath2 , where the destructive susy contributions push @xmath198 _ below _ experimental bounds . in figs . [ fig : bsgp ] and [ fig : bsgn ] we plot @xmath199 in the @xmath68 plane for various values of @xmath3 and @xmath4 . regions excluded by eq . ( [ bsgbound ] ) are shaded ; for @xmath204 and large @xmath3 , this includes much of the parameter space with light sleptons and light winos . while anomaly - mediated susy breaking does not contribute substantially to flavor - violating observables involving the first and second generations , it may give significant contributions to flavor-_conserving _ observables involving the first and second generations . it is well - known that susy loops may give a sizable contribution to the muon magnetic dipole moment ( mdm ) @xcite . the susy contribution to the muon mdm is from smuon - neutralino and sneutrino - chargino loop diagrams . since these superparticles may have masses comparable to the electroweak scale , these contributions may be comparable to , or even larger than , electroweak contributions from @xmath167- and @xmath59-boson diagrams . the on - going brookhaven e821 experiment @xcite is expected to measure the muon mdm with an accuracy of @xmath205 , which is about a few times smaller than the electroweak contribution to the muon mdm . therefore , the brookhaven e821 experiment will provide an important constraint on susy models . in general , the muon anomalous mdm is given by the coefficient of the `` magnetic moment - type '' operator @xmath206 where the anomalous magnetic moment @xmath207 is related to the muon @xmath208 by @xmath209 . as suggested from the structure of the operator , diagrams for the muon anomalous mdm require a left - right muon transition . in susy diagrams , this transition may occur through a chirality flip along the external muon line , through left - right mixing in the smuon mass matrix , or through the interaction of a muon and smuon with a higgsino . in the latter two cases , the diagrams are proportional to the muon yukawa coupling constant and are therefore enhanced for large @xmath3 . these diagrams also include gaugino mass insertions . as a result , in the large @xmath3 limit , the muon anomalous mdm is given by @xmath210 where the @xmath211 functions ( see the last reference in ref . @xcite ) are typically @xmath212 , with @xmath54 being the mass scale of the superparticles in the loop . for large @xmath3 , then , the susy contribution @xmath213 is approximately proportional to @xmath3 and may be much larger than the electroweak contribution . results for the susy contribution to the muon mdm @xmath213 in the minimal anomaly - mediated model are given in fig . [ fig : amu ] . both @xmath3 enhanced and unenhanced contributions were included by using the mass eigenstate bases of squarks , sleptons , neutralinos , and charginos . the susy contribution to the muon mdm is typically @xmath214 , and is enhanced for large @xmath3 . furthermore , heavier superparticles suppress @xmath213 , as expected . experimentally , the muon anomalous mdm is currently constrained to be @xcite @xmath215 and hence the anomaly - mediated susy contribution is usually smaller than the present experimental accuracy , unless @xmath3 is very large . however , as mentioned above , in the near future , the brookhaven e821 experiment will improve the measurement , with a projected error of @xmath205 . if this is realized , some anomaly may be seen in the muon mdm in the anomaly - mediated susy breaking scenario , particularly for moderate or large values of @xmath3 . in general , parameters in susy models are complex , and ( some combinations of ) their phases are physical . in the anomaly - mediated susy breaking scenario , most of the susy breaking parameters are proportional to the single parameter @xmath1 , and so many of the phases can be rotated away . in particular , the gaugino mass parameters and the @xmath34 parameters can be made real simultaneously . however , even in anomaly - mediated susy breaking , a physical phase may exist in the @xmath10 and @xmath57 parameters since their origins are not well - understood . in our analysis , we have not assumed any relation between @xmath10 and @xmath57 , and have simply constrained them so that electroweak symmetry is properly broken . in this approach , one physical phase remains , which is given by @xmath216 if this phase is non - vanishing , electric dipole moments ( edms ) are generated . as is known from general analyses , the edms of the electron and neutron may be extremely large unless @xmath217 is suppressed @xcite . to determine the constraints on this phase in the anomaly - mediated framework , we calculate the electron and neutron edms with the minimal anomaly - mediated model mass spectrum . the edm @xmath218 of a fermion @xmath219 is given by the effective electric dipole interaction @xmath220 which becomes @xmath221 in the non - relativistic limit . the calculation of the electron edm is similar to that of the muon anomalous mdm , since the structure of the feynman diagrams is almost identical . if the slepton masses are flavor universal , @xmath207 and @xmath222 are approximately related by , @xmath207 is proportional to @xmath223 in the large @xmath3 limit . ] @xmath224 therefore , the electron edm is also proportional to @xmath3 . the calculation of the up and down quark edms is also straightforward , given the susy model parameters . the only major difference from the electron edm is the contribution from the squark - gluino diagram . however , in calculating the neutron edm , we must adopt some model for the structure of the neutron . we use the simplest model , i.e. , the non - relativistic quark model . the neutron edm is then given by @xmath225 since @xmath226 is also proportional to @xmath3 , the neutron edm is also enhanced for large @xmath3 . figures [ fig : edme ] and [ fig : edmn ] show the edms of the electron and neutron , respectively , in the minimal anomaly - mediated model . the edms are proportional to @xmath227 . in these plots , we assume maximal cp violation , i.e. , @xmath228 . currently , there is no experimental result which suggests a non - vanishing edm , and experimental constraints on the edms are very stringent . for the electron edm , using @xmath229 cm @xcite , we obtain the constraint @xmath230 where the right - hand side is the upper bound on @xmath222 at 90% c.l . for the neutron , @xmath231 is constrained to be @xcite @xmath232 the naturalness arguments of sec . [ sec : naturalness ] play an important part in evaluating the sensitivity of the edms . for @xmath222 and small @xmath3 , while very large effects are possible , @xmath222 may be within the experimental bounds even for @xmath233 close to 1 without violating the condition @xmath111 . for moderate and large @xmath3 , @xmath222 becomes much larger , and the physical phase @xmath234 is constrained to be @xmath235 for @xmath236 . however , for such @xmath3 , the naturalness bound on @xmath2 is also relaxed , and reasonably large @xmath237 phases are possible in natural regions of parameter space where @xmath222 is suppressed by slepton masses of a few tev . thus , while large effects comparable to current bounds are predicted in much of parameter space , constraints from @xmath222 may also be satisfied by superpartner decoupling in the minimal anomaly - mediated model . for @xmath231 , similar conclusions hold . in fact , the constraints from @xmath231 on the cp - violating phases are more easily satisfied , and @xmath222 appears to be the more stringent constraint at present . in our discussion , as noted above , we have not assumed a specific model for the @xmath10 and @xmath57 parameters , and hence we regarded @xmath234 as a free parameter . however , several mechanisms have been proposed to generate @xmath10 and @xmath57 in which @xmath227 vanishes @xcite . in those scenarios , of course , @xmath222 and @xmath231 vanish , and the edm constraints are automatically satisfied . in this study we have analyzed a model of `` supernatural supersymmetry , '' in which squarks and sleptons may be much heavier than their typical naturalness limits , and susy is broken in another world . susy breaking is then communicated to our world dominantly via anomaly - mediation , and we have considered in detail a model in which tachyonic sleptons are avoided by a non - anomaly - mediated universal scalar mass @xmath2 . the novel naturalness properties of this model are a result of a `` focus point '' behavior in the rg evolution of @xmath64 , such that its weak scale value is highly insensitive to @xmath2 . naturalness bounds on superparticle masses are therefore highly variable and differ from naive expectations . naturalness places strong bounds on gaugino masses , and wino masses @xmath238 are preferred . on the other hand , for moderate and large values of @xmath3 , multi - tev values of @xmath2 , and therefore slepton and squark masses , are natural . a number of spectacular collider signals are possible . the possibility of a highly degenerate triplet of wino lsps has recently attracted a great deal of attention @xcite . in the minimal anomaly - mediated scenario , we find that winos are not only the lsps in much of parameter space , but are typically light , with mass @xmath239 , and extraordinarily degenerate , with charged wino decay lengths of several centimeters . such wino characteristics are ideal for tevatron searches , where winos may appear as vertex detector track stubs in monojet events . the prospects for discovery at the tevatron in run ii or iii are highly promising @xcite . in the remaining parameter space , the lsp is either the lighter stau , or the tau sneutrino . in the @xmath11 lsp scenario , the @xmath11 is typically lighter than 200 gev and is stable . it may be found in searches for stable charged massive particles at both lep @xcite and the tevatron @xcite . in the @xmath12 lsp scenario , the winos , @xmath11 and sneutrinos are all @xmath240 . in both scenarios , ongoing searches at lep and the tevatron will be able to probe substantial portions of the relevant parameter space . the minimal anomaly - mediated model also has a number of other features that distinguish it from other models . in addition to characteristic gaugino mass ratios , these include highly degenerate same - flavor sleptons , and large left - right mixing . if susy is discovered , measurements of slepton masses and mixings will provide strong evidence for or against the minimal model and its assumption of an additional universal slepton mass . we have also considered a variety of low energy observables that are sensitive probes of anomaly - mediated parameter space . effects on the flavor - changing process @xmath5 may be large , and significant regions of parameter space for large @xmath3 and @xmath70 are already excluded . the anomalous magnetic moment of the muon may also be affected at levels soon to be probed by experiment . finally , the electron and neutron electric dipole moments provide rather strong constraints on the cp - violation phase @xmath241 in much of parameter space , but even for large @xmath3 , @xmath237 phases are still be allowed for multi - tev @xmath2 at its focus point naturalness limit . it is interesting to note that positive signals in these low energy experiments may not only provide evidence for susy , but may also exclude some supersymmetric interpretations and favor others . for example , the signs of the susy contributions to @xmath5 and @xmath242 are determined by @xmath243 and @xmath169 are correlated , as they are in anomaly - mediation , and , through rg evolution , in gauge - mediated models and minimal supergravity . ] and @xmath244 , respectively . a large anomalous measurement of @xmath213 would imply large @xmath3 , and , given the current bounds on @xmath5 , a preferred sign for @xmath245 . the sign of the @xmath213 anomaly then determines @xmath246 . for example , assuming a susy interpretation , a large negative anomalous mdm measurement would imply @xmath247 , and would favor anomaly - mediated models over virtually all other well - motivated models . finally , as stated in sec . [ sec : mam ] , the assumption of a universal scalar mass contribution , while possibly generated by bulk contributions @xcite , does not hold generally in anomaly - mediated scenarios . several features presented above depend on various parts of this assumption , and we therefore close with a brief discussion of these dependences . the naturalness properties described above , and , in particular , the focus point behavior , results from the fact that the non - anomaly - mediated piece is identical for @xmath64 , @xmath248 , and @xmath249 . while the focus point mechanism as implemented here relies on this subset of the universal boundary conditions , a variety of other boundary conditions also have similar properties , for any @xmath250 , also leads to focus point behavior . ] , and it would be interesting to explore applications of the focus point mechanism in other settings . the accidental degeneracy of left- and right - handed sleptons , and the possibility for large left - right mixings , holds only if both left- and right - handed sleptons receive the same non - anomaly - mediated contribution . measurement of large left - right smuon mixing , along with confirmation of anomaly - mediated gaugino mass parameters , for example , would therefore be strong evidence for anomaly - mediation with a universal slepton mass contribution . finally , the low energy observables discussed are sensitive quantitatively to either the hadronic or leptonic superpartner spectrum . however , qualitative results , such as the stringency of constraints for large @xmath3 , can be expected to remain valid for a variety of anomaly - mediated models , as long as the attractive flavor properties of anomaly - mediation are preserved in these models and they do not have new large sources of flavor violation . we are grateful to greg anderson , jon bagger , toru goto , erich poppitz , lisa randall , yael shadmi , yuri shirman , and frank wilczek for helpful correspondence and conversations . the work of jlf was supported by the department of energy under contract de fg0290er40542 and through the generosity of frank and peggy taplin . the work of tm was supported by the national science foundation under grant phy9513835 and a marvin l. goldberger membership . in this appendix , we present the leading order soft supersymmetry breaking terms , first for a general anomaly - mediated supersymmetric theory , and then for the minimal anomaly - mediated model . consider a supersymmetric theory with simple gauge group @xmath251 . the anomaly - mediated boundary conditions are completely specified in terms of the gauge coupling @xmath18 , supersymmetric yukawa couplings in the convention that the soft supersymmetry - breaking terms are @xmath253 the leading order anomaly - mediated soft supersymmetry breaking terms are @xmath254 where @xmath255 here @xmath256 , @xmath257 , and the one - loop @xmath17-function coefficient is @xmath258 , where @xmath259 is the quadratic casimir invariant for representation @xmath260 , and @xmath261 is the total dynkin index summed over all the chiral superfields . in terms of the matter field wavefunction @xmath59 , @xmath262 . for minimal field content , anomaly - mediated gaugino masses are given as @xmath263 where @xmath264 in the gut normalization . furthermore , with the superpotential @xmath265 the flavor - dependent wavefunction factors are @xmath266 where the yukawa couplings @xmath267 are @xmath268 matrices in generation space . 16\pi^2 \dot{{\mbox{\boldmath $ y_d$ } } } & = & { \mbox{\boldmath $ y_d$ } } \left [ 3 { \:\!\text{tr}\:\!}({\mbox{\boldmath $ y_d$ } } { \mbox{\boldmath $ y_d$}}^\dagger)+ 3 { \:\!\text{tr}\:\!}({\mbox{\boldmath $ y_e$ } } { \mbox{\boldmath $ y_e$}}^\dagger ) + 3{\mbox{\boldmath $ y_d$}}^\dagger{\mbox{\boldmath $ y_d$ } } + { \mbox{\boldmath $ y_u$}}^\dagger { \mbox{\boldmath $ y_u$ } } -\frac{16}{3}g_3 ^ 2 - 3 g_2 ^ 2 - \frac{7}{15}g_1 ^ 2 \right ] \\ 16\pi^2 \dot{{\mbox{\boldmath $ y_e$ } } } & = & { \mbox{\boldmath $ y_e$ } } \left [ 3 { \:\!\text{tr}\:\!}({\mbox{\boldmath $ y_d$ } } { \mbox{\boldmath $ y_d$}}^\dagger ) + { \:\!\text{tr}\:\!}({\mbox{\boldmath $ y_e$ } } { \mbox{\boldmath $ y_e$}}^\dagger ) + 3 { \mbox{\boldmath $ y_e$}}^\dagger { \mbox{\boldmath $ y_e$ } } - 3 g_2 ^ 2 - \frac{9}{5}g_1 ^ 2 \right ] \ .\end{aligned}\ ] ] see , for example , m. wise , in _ proceedings of tasi 87 _ , edited by r. slansky and g. west ( world scientific , singapore , 1988 ) ; j. bagger , in _ proceedings of tasi 95 _ , edited by d. soper ( world scientific , singapore , 1996 ) , pp . 109 - 162 ; j. l. feng , c. kolda , and n. polonsky , ; j. bagger , j. l. feng , and n. polonsky , hep - ph/9905292 . the susy working group , j. bagger _ et al . _ , in _ proceedings of the 1996 dpf / dpb summer study on new directions for high energy physics ( snowmass 96 ) _ , june 25 july 12 , 1996 , pp . 642654 , hep - ph/9612359 . d. stuart , private communication ; k. hoffman for the cdf and dcollaborations , talk presented at hep 97 , jerusalem , israel , august 1926 , 1997 , hep - ex/9712032 ; d. cutts and g. landsberg , hep - ph/9904396 . cleo collaboration , s. glenn et al . conf 9817 , ichep98 1011 ; j. alexander , in the proceedings of the 29th international conference on high - energy physics ( ichep 98 ) , vancouver , canada , 23 - 29 july 1998 . see , for example , j. lopez , d. v. nanopoulos , and x. wang , ; u. chattopadhyay and p. nath , ; m. carena , g. f. giudice , and c. e. m. wagner , ; t. moroi , ; erratum , * d56 * ( 1997 ) 4424 ; and references therein .

_ _ the three complex form factors entering the vertex are calculated to in the framework of a chiral effective theory with explicit(1232 ) degrees of freedom included . it is shown that the low behavior of the form factors is governed by , loop effects . predictions are given for the-dependence of the three transition - multipoles . furthermore , the role of presently unknown low energy constants that affect the values of the multipole ratios and is elucidated . plus 2 mm minus 2 mm epsf.sty epsfig.sty plus 2 mm minus 2 mm # 1 ( # 1)eq . ( [ # 1 ] ) 5_5 # 10=-.025em0 - 0 .05em0 - 0 -.025em.0433em0

hep-ph9810426

the electromagnetic transition of the @xmath2(1232 ) resonance to the nucleon @xmath9 is of particular interest @xcite as far as our understanding of the structure of the latter is concerned . historically ( _ e.g. _ @xcite ) , this reaction raised a lot of interest because it allowed one to probe the issue of whether the nucleon or its first resonance is `` deformed''the reason being that apart from the dominant magnetic dipole ( @xmath10 ) transition electromagnetic selection rules also allow an electric ( @xmath11 ) and a coulomb ( @xmath12 ) quadruple transition , which vanish in simple models of the nucleon with spherical symmetry . accordingly , the amount of deformation can be quantified by the multipole ratios emr@xmath13 and cmr@xmath14 , which acquire a four - momentum ( squared ) dependence in the case of virtual photons @xmath15 . by the late 1990s there is nearly uniform consensus in the physics community that indeed there exist a small quadrupole components in the electromagnetic @xmath16 transition @xcite . in the case of real photons one nowadays believes an re@xmath17\approx-1/\%\ldots -4\%$ ] , but a more precise determination of this fundamental property of the nucleon has been surprisingly elusive and hotly contested throughout the past decade , both among theorists and among experimentalists . it is our hope that the ongoing experiments @xcite of electroproduction of @xmath2(1232 ) with the resulting better information on @xmath3-dependence of the @xmath16-transition form factors , as well as on emr@xmath18 , cmr@xmath18 will lead to clear picture of the underlying physics and enable us to identify the relevant degrees of freedom for several different regimes of momentum transfer . a dramatic change of the physics underlying the electromagnetic @xmath16 transition is very much expected from _ perturbative _ qcd , which predicts for `` large '' @xmath19 that emr@xmath20 . at which finite @xmath21 the crossover from a negative to a positive emr should happen and whether this point is kinematically accessible at present / future electron scattering machines is a further issue of current debate and interest , _ e.g. _ @xcite . > from a theorist s perspective , treatments of the electromagnetic @xmath16 transition may be grouped into two categories : 1 . calculations of the electromagnetic @xmath22 vertex with its associated 3 complex form factors _ per se _ , using different theoretical frameworks that aspire to fundamentally - based descriptions of the @xmath23 system . the chiral bag model(_e.g . _ @xcite ) , quark models with meson exchange currents ( _ e.g _ @xcite ) , lattice gauge theory ( _ e.g. _ @xcite ) and effective chiral lagrangians ( _ e.g. _ @xcite ) constitute such attempts . direct theoretical treatments @xcite of the full scattering processes ( _ e.g. _ @xmath24 ) in the @xmath2(1232 ) resonance region , either based on phenomenological lagrangians supplemented with a method of choice to unitarize the amplitudes or dispersion relations . a point of strong contention therein is the issue of separation of background vs. resonance contributions . for a recent summary of the status of the resulting emr / cmr extractions we refer to the talk by workman @xcite . in the present work we calculate the @xmath2(1232 ) to @xmath25 radiative transition @xmath9 in the region of small ( _ i.e. _ @xmath26 gev@xmath27 ) photon virtuality , utilizing a recently developed effective chiral lagrangian approach @xcite that systematically incorporates the spontaneous and the explicit breaking of the chiral symmetry of qcd . a small scale @xmath28 denoting , collectively , small momenta , the pion mass and the delta - nucleon mass splitting is used to establish a systematic power - counting , thus telling us precisely which diagrams / vertices have to be included if we want to calculate up to a certain order in @xmath29 . this approach allows for an efficient inclusion of @xmath2(1232 ) degrees of freedom consistent with the underlying chiral symmetry of qcd and is referred to as the `` small scale expansion '' ( sse ) @xcite , constituting a phenomenological extension of heavy baryon chiral perturbation theory @xcite . the formalism can be used to _ calculate _ both the vertex as well as full scattering amplitudes here we focus on the former . clearly , as we are dealing with a _ low energy _ effective theory , the @xmath3 dependence of a given physical quantity can be trusted only at low valuesgev@xmath27 regime @xcite . ] . accordingly , our final applications will be discussed in this spirit . the most general form of the @xmath0 radiative decay amplitude that complies with lorentz covariance , gauge invariance and parity conservation is described by three form factors . we begin our discussion from the widely used form @xmath30u^\m_\d(p_\delta ) \ .\end{aligned}\ ] ] here @xmath31 is the nucleon mass , @xmath32 denotes the relativistic four - momentum of the nucleon , delta and @xmath33 , @xmath34 are the photon momentum and polarization vectors , respectively . the delta is described via a rarita - schwinger spinor @xmath35 with free lorentz index @xmath36 . the @xmath10 , @xmath11 and @xmath12 multipoles allowed in @xmath37 electromagnetic transitions can be cast as linear combinations of the form factors @xmath38 and @xmath39 @xcite . in this work we study the radiative vertex to @xmath1 in the above mentioned sse formalism . this constitutes the first order where pion - nucleon and pion - delta loop graphs enter the vertex . as will be shown later , an @xmath1 calculation entails corrections up to @xmath40 , with @xmath41 the chiral symmetry breaking scale . one of the main tasks during this calculation has been the consistent matching between the results of our perturbative calculation and the most general vertex parameterization as given for example in eq.([eq1 ] ) . in order to provide for a stringent test of our new predictions with experiment we note that a complementary calculation of the full pion photoproduction amplitude in the @xmath2(1232 ) region to the same order in @xmath42 is in progress @xcite . there have been previous analyses of the radiative transition in a similar theoretical approach by butler et al . @xcite and by napsuciale and lucio @xcite . our work differs from the aforementioned references in the following aspects : * the most crucial difference is that we address the _ form factors _ and not just the real - photon point . this entails ( a ) a @xmath3 dependence in our expressions and ( b ) the presence of an additional form factor ( @xmath39 ) , or , equivalently , our calculation yields the cmr@xmath18 , in addition to the emr@xmath18 . * sse systematically keeps track of @xmath43 ( i.e. , relativistic ) corrections to lower order couplings . these corrections have not been included in previous analyses @xcite . we also find that our identification of the form factors differs from the one used in @xcite at the real photon point . this article is structured as follows . in the next section we briefly review the essentials of the `` small scale expansion '' and its power counting . in section iii , we discuss the most general form of the pauli - reduced transition amplitude which is consistent with the @xmath44 calculation in the sse . we then proceed , in section iv , to calculate the loop as well as chiral counterterm contributions . the resulting expressions for the form factors are identified in section v , while those for the multipoles in section vi . numerical values for the @xmath45 using presently available ( theoretical ) input are furnished in section vii . in the same section we give numerical as well as analytical results for ( complex ) slopes of the three form factors . we summarize and offer our perspective for future efforts in the concluding section . finally , we devote an appendix to the discussion of technical matters . qcd being a strongly coupled theory at low ( of the order of 1 gev ) energies , renders traditional perturbative expansions in the coupling constant inadequate . chiral perturbation theory offers an alternative perturbative expansion , namely one that is realized in terms of the external momenta involved in a given physical process . the original strategy was based on the notion that at low energies an effective theory of qcd will involve only the nearly massless ( i.e. the goldstone bosons : pions etc . ) degrees of freedom @xcite . accordingly , chiral perturbation theory has been very successful with respect to applications in the meson sector . with the inclusion of baryons , however , the systematic power counting of chpt fails , since baryon masses @xmath46 gev ) can not join the set of the expansion parameters @xmath47 as they are by no means small and remain finite in the chiral limit . a systematic power counting can , nevertheless , be defined through a splitting of the nucleonic field degrees of freedom into heavy - light modes and integrating out the former . the cost of this procedure is to burden the effective description with additional , higher order contact interactions . generalizing recent developments in heavy quark effective theories ( see , e.g. , @xcite,@xcite ) heavy fermion methods were first applied to baryon chiral perturbation theory by jenkins and manohar @xcite . the basic premise is the adoption of a non - relativistic mode of description , which entails a restriction to four - velocities of the form @xmath48 . on an operational level this means that all momentum dependence in the theory is only governed by ( non - relativistic ) soft momenta @xmath49 , defined via @xmath50 where @xmath51 is a typical nucleon relativistic four - momentum and @xmath52 corresponds to the nucleon mass in the chiral limit . the range of validity of the resulting effective theory demands that each component @xmath53 , with the chiral symmetry breaking scale being @xmath54gev . this approach is commonly referred to as heavy baryon chiral perturbation theory ( hbchpt ) . in the construction of the effective theory we will follow the systematics laid out in @xcite . the general philosophy is to take the fully relativistic theory as the starting point and then perform a systematic non - relativistic reduction . this procedure automatically guarantees the proper @xmath43 corrections to the couplings in higher order lagrangian terms . alternatively , one could start from a general lagrangian within the non - relativistic framework , but then has to implement the so called `` reparameterization invariance '' . the latter approach is , for example , quite common in the field of heavy quark effective theories @xcite . we now briefly sketch the derivation of the ( non - relativistic ) chiral lagrangians for matter fields . for details we refer the extensive literature of reviews ( _ e.g. _ @xcite,@xcite,@xcite ) we start from the chiral relativistic su(2 ) lagrangian for nucleons @xmath55 with the relativistic nucleon isospinor field @xmath56 and @xmath57 being a string of general nucleon - nucleon transition matrices @xmath58 of increasing chiral power @xmath59 @xcite . for example , to leading order one obtains the well - known structure @xmath60 where @xmath61 denotes the chiral covariant derivative . the parameter @xmath62 corresponds to the axial vector coupling constant ( in the chiral limit ) and the chiral tensor @xmath63 describes the coupling of an odd number of pions with nucleon . for more details we refer to @xcite . the second step is a redefinition of the relativistic nucleon fields via @xmath64 with the velocity - dependent projection operators @xmath65 . @xmath25 is typically called the `` light '' field , whereas @xmath66 is commonly referred to as the `` heavy '' field . the relativistic lagrangian eq.([eq : rellag ] ) then takes the general form @xmath67 where the matrices @xmath68 are defined via @xmath69 we note that each matrix consists of an infinite string of terms of increasing chiral power @xmath59 , analogous to eq.([eq : gamma ] ) start with chiral power @xmath70 , while @xmath71 begins with @xmath72 . this appearance of a large mass term @xmath73 in @xmath71 is also the reason why the @xmath66 fields are denoted `` heavy '' and ultimately get integrated out . ] . now one shifts the fields and integrates out the heavy component @xmath66 . the resulting ( non - relativistic ) effective lagrangian reads @xmath74 for the case of su(2 ) the explicit form of this effective lagrangian for spin 1/2 nucleons has been worked out up to @xmath75 by several groups ( _ e.g. _ @xcite ) . generalizations to even higher orders are under way @xcite . the important point to note is that the inverse of matrix @xmath76 is calculated perturbatively , which confines the resulting effective lagrangian to the non - relativistic regime . hbchpt as described above , has been quite successful for scattering processes off a single nucleon near threshold ( for reviews see @xcite ) . at higher energies , however , the contribution from nucleon resonances like @xmath2(1232 ) can no longer be parameterized via higher order nucleon - nucleon couplings in @xmath77 , etc . at some point , i.e. once explicit propagation of a nucleon- or meson - resonance has to be included , the above described contact interaction approach breaks down . so if one is interested in kinematic conditions of such dynamic resonance contributions , or in investigations into the low energy structure of nucleon resonances , it becomes mandatory to include low lying resonances as explicit degrees of freedom in the effective lagrangian eq.([eq : leff ] ) . in particular this means the inclusion of the spin 3/2 nucleon resonance @xmath2(1232 ) in the case of an su(2 ) analysis . the first efforts towards this direction were performed by jenkins and manohar @xcite . in the present work , however , we follow a specific generalization of the construction method of ref.@xcite ( as outlined above ) , which is called the `` small scale expansion '' ( _ sse _ ) @xcite . the main difference to hbchpt lies in the fact that _ the chiral power counting is modified in a phenomenologically inspired fashion_. in the _ sse _ approach one expands in the small scale @xmath78 , where @xmath79 corresponds to the delta - nucleon mass splitting in the chiral limit , whereas in hbchpt one expands in the quantity @xmath80 counts as a quantity of order @xmath81 . this is formally correct but can lead to poor convergence properties in the perturbation series . for more details we refer to the discussion in @xcite regarding the spin - polarizabilities of the nucleon . the chiral power counting of hbchpt as an expansion of all quantities in a power series governed by @xmath82 is taken over by _ sse _ as an expansion of all quantities in a power series governed by @xmath83 . in the following we only briefly discuss the construction of the relevant su(2 ) lagrangians with explicit pion , nucleon and delta degrees of freedom and their couplings to arbitrary external fields . for details we refer the interested reader to @xcite . the starting point this time is a set of coupled relativistic su(2 ) lagrangians of relativistic nucleon and delta fields @xmath84 @xmath85 where each baryon matrix @xmath86 is a string of terms of increasing power in @xmath83 , analogous to eq.([eq : gamma ] ) . for the spin 3/2 , isospin 3/2 delta field we utilize a rarita - schwinger - isospurion notation , as laid out in the appendices of ref.@xcite . in the rarita - schwinger formulation for spin 3/2 fields one encounters the well - known redundancy of degrees of freedom , which calls for six projection operators before one is able to isolate the `` light '' spin 3/2 component @xmath87 from the `` heavy '' one @xmath88 : @xmath89 with @xmath90 the `` heavy '' spin 3/2 field @xmath88 is a five - component object . an explicit representation can be found in @xcite but is not needed here . suppressing all spin - isospin indices , the coupled set of relativistic lagrangians in eq.([eq : coupled ] ) can then be written as @xmath91 now one integrates out the `` heavy '' nucleon and `` heavy '' delta components @xmath92 and finds the resulting ( non - relativistic ) effective lagrangian with explicit nucleon and delta fields : @xmath93,\ ] ] with @xmath94 and @xmath95 the explicit form of the effective lagrangian eq.([ck12 ] ) in sse has been worked out to @xmath96 in ref.@xcite . for @xmath70 one finds @xmath97 in addition to the structures already discussed in section [ s2 ] one encounters the pauli - lubanski spin vector @xmath98 and the @xmath99 bare coupling parameters @xmath100 . furthermore , the chiral tensors @xmath101 parameterize the coupling of an odd number of pions to an @xmath16 , and a @xmath102 transition current , whereas @xmath103 corresponds to the chiral covariant derivative acting on a spin 3/2 - isospin 3/2 field . with @xmath104 = 1 , 2 , 3 we denote explicit isospin indices and @xmath105 corresponds to the isospin 3/2 projector . for more details we refer to @xcite . at @xmath96 we are only interested in @xmath106 vertices , as will be discussed in section [ born ] . the relevant relativistic @xmath107 @xmath106 contact terms can be parameterized in terms of the low energy constants ( lecs ) @xmath108 , @xmath109 as follows @xmath110 \gamma_\rho\g5 f^{\nu\rho}_{i\hspace{0.01in}+ } \psi _ n-\frac{b_6}{2m_0 ^ 2 } \bar\psi^\mu_i \left [ g_{\mu\nu } + y_6 \gamma_\mu\gamma_\nu \right ] \g5 f^{\nu\rho}_{i\hspace{0.01in}+ } d_\rho\psi _ n+h.c . \ , \ ] ] where @xmath111 $ ] denotes the isovector photon component and @xmath112 are the so - called off - shell coupling constants . performing the transition to the heavy baryon fields for the vertices of interest to our calculation one finds @xcite @xmath113 which is independent of the couplings @xmath114 . in the radiative decay of the @xmath2(1232 ) the fact that there is no s - wave multipole allowed precludes the possibility of having an @xmath115 @xmath106 vertex . the parameter @xmath108 therefore denotes the leading @xmath116 coupling and carries the main strength of the m1 @xmath106 transition @xcite . while the desired analysis for @xmath96 is quite simple , the situation at @xmath75 is more complex . in fact , there is no agreement in the literature about number and/or structure of the subleading @xmath106 vertices ( _ e.g. _ compare refs . @xcite ) . in the next subsection we shall explicitly construct the @xmath75 @xmath106 vertices according to the sse approach . referring to the sse master - formula eqs.([ck12]-[ck14 ] ) , the relevant lagrangian @xmath117 contains terms which have two separate origins the dimension six contact interactions contributing to @xmath118 and possible @xmath43 corrections related to the @xmath119 @xmath120 vertices . the latter are given by @xmath121 where the superscripts denote the sse dimension . the pertinent coupling matrices can be found in @xcite , _ except _ for the structures denoted as @xmath122 , @xmath123 which only start contributing at @xmath1 . we therefore construct these matrices , but only for the vertices of interest in our calculation . using eq.([rel1 ] ) and the orthogonality properties of the light - heavy projection operators @xcite we find ( dots signify terms that are irrelevant to the present analysis ) @xmath124 and ( for the @xmath108 part ) @xmath125 \g5 f_{\nu\rho}^+ \ , \ ] ] where we have suppressed isospin indices . in a similar manner one obtains the contribution to @xmath126 associated with @xmath109 . the complete 1/m corrections at @xmath127 which had to be constructed via eq.([corr ] ) then read @xmath128 interestingly , one observes that the off - shell couplings @xmath129 drop out in the final result . finally we have to determine the dimension six @xmath116 counterterms contained in @xmath118 as mandated by eq.([ck12 ] ) . once more we have to start from that part of the most general relativistic @xmath127 lagrangian that contains all allowed @xmath116 vertices at this order , in which the chiral tensor @xmath130 \right\ } $ ] plays a dominant part . at a first glance the following structures are possible candidates : @xmath131 , @xmath132 , @xmath133 , @xmath134 , @xmath135 , where the free index @xmath36 is to be contracted with the corresponding index of the rarita - schwinger delta spinor in the initial state . it is straightforward to verify that only the first and third quantities will survive in @xmath118 and can be shown to give identical contributions up to higher order terms . accordingly we find @xmath136 \right\ } + \frac{i e_1}{2 m_0}\ ; \frac{\delta_0}{m_0 } \;f_{+\mu\nu}^{\ ; i}\;s^\nu+ ... \equiv a_{n\delta\gamma}^{(3)}+ ... ,\ ] ] where @xmath137 , @xmath138 are now identified as the 2 @xmath116 low energy constants of @xmath127 . ) , proportional to @xmath138 , appears in the `` small scale expansion '' due to the fact that the delta - nucleon mass splitting @xmath139 counts as a quantity of @xmath115 . therefore one has to construct additional relativistic contact terms which are products of lower order interactions and the new small scale @xmath139 . for a discussion of this issue see ref . ] this set is sufficient for the renormalization of the process @xmath140 , a more detailed discussion of the construction of these counterterms can be found in @xcite gathering the above results we determine the relevant sse @xmath116 lagrangian to @xmath127 : @xmath141 t + h.c.\nonumber \\ & = & \frac{d_1}{4 m_{0}^2}\ ; \bar{n}g^{\mu\nu}v^\alpha s^\beta f_{\nu\beta\alpha}^{i\,+ } t_{\mu}^i+ \frac{i e_1}{2 m_0}\ ; \frac{\delta_0}{m_0 } \bar{n } \;f_{+\mu\nu}^{\ ; i}\;s^\nu \;t_{i}^\mu \nonumber \\ & & -{b_1 + 2 b_6 \over 4 m^2_0 } \bar n ( s\dt d ) v^\rho f_{\mu\rho}^{i+}t_i^\mu + \frac{b_1 - 2b_6}{4m_0 ^ 2 } \bar n f_{\rho\mu + } ^k v^\rho\xi^{kj}_{3/2 } s\cdot d^{ji}t^\mu_i \nonumber \\ & & + \frac{b_1}{2m_0 ^ 2 } \bar nf^k_{\rho\beta + } s^\rho v^\beta\xi_{3/2}^{kj } d_\mu^{ji}t^\mu_i + h.c . \label{eq : lag3}\end{aligned}\ ] ] since the leading order lagrangian counts as @xmath143 and , from the general exposition in the previous section follows that each new order in the sse contributes a power @xmath144 , the calculation of the radiative delta decay to @xmath145 will include all terms up to @xmath146 . thus , in order to compare our calculation with eq.([eq1 ] ) we must find the most general form of the amplitude consistent with an @xmath146 calculation . in our case it is convenient to use the @xmath2(1232 ) rest frame where a pauli reduction of the amplitude allows one to identify the various multipoles . as far as dirac spinors are concerned the situation with the pauli reduction is well - known . for the delta field , on the other hand , we think that is worth providing our conventions . we start from a typical ( _ e.g. _ @xcite ) rarita - schwinger representation of the @xmath2(1232 ) spinor with the subsidiary conditions @xmath147 a specific representation consistent with these constraints is the following @xmath148 with @xmath149 , @xmath150 and @xmath151 the energy , mass and three momentum of the delta particle , respectively and @xmath152 the four - vector given by @xmath153,\ ] ] where @xmath154 denotes the spin @xmath155 transition matrices @xcite . next , we have to expand the various terms entering eq.([eq1 ] ) in powers of @xmath43 and restrict ourselves up to @xmath146 accuracy . using the analogue of the four - component notation of the sse @xmath156 with @xmath157 , we finally arrive at the following expression @xmath158 \right . \nonumber \\ & & + ( s\cdot q ) \epsilon_\mu \ \left [ -\frac{g_1(q^2)}{m_n}-\frac{\delta}{2m_n^2}g_1{(0)}+\frac{\delta}{4m_n^2 } g_2{(0)}+{\cal o}(1/m_n^3 ) \right ] \nonumber \\ & & + ( s\cdot q ) ( v\cdot\epsilon ) q_\mu \ \left [ \frac { ( g_1{(0 ) } - { 1\over 2 } g_2{(0 ) } ) } { 2m_n^2 } + { \cal o}(1/m_n^3 ) \right ] \nonumber \\ & & \left . + ( s\cdot q ) ( q \cdot \epsilon ) q_\mu \ \left [ 0 + { \cal o}(1/m_n^3 ) \right ] \right\ } u_{v,\delta}^{\mu}(0)\ , \end{aligned}\ ] ] which has been evaluated in the rest - frame of the delta particle . some comments regarding the above equation are appropriate : 1 . to the order we are working the energy @xmath159 of the outgoing photon four - momentum is @xmath160 = @xmath161 . . = @xmath162 + @xmath163 . 2 . for @xmath164 eq.([exp ] ) is not in agreement with the analogous expression of ref . there , the nucleon mass dependence of the dirac spinors apparently was not taken into account . 3 . eq.([exp ] ) has been derived under the assumption that all the form factors are quantities of @xmath165 at leading order . it is inconsistent with our subsequent loop results . this can be most easily seen by the structure proportional to @xmath166 . according to the @xmath43 expansion this structure should not exist to @xmath1 . however , as it will turn out , the loop diagrams of fig . 1 produce a non - trivial @xmath3 dependence which scales as @xmath167 . we conclude that the popular form of the relativistic isovector nucleon - delta transition current as given in eq.([eq1 ] ) , is not ideally suited for microscopic calculations of these form factors which rely on @xmath43 expansions involving a chiral power counting . we find it , therefore , advantageous form factor which behaves as @xmath168 at @xmath164 , _ i.e. _ @xmath169 . however , such a result would violate the basic premise of non - relativistic formalisms like hbchpt or sse that the large scale @xmath170 should _ always _ appear as @xmath171 in order to ensure a consistent power counting . ] , in view of the above remarks , to define a new set of form factors , by rescaling the previous ones . we propose the following definition @xmath172 which leads to the relativistic @xmath0 transition matrix element @xmath173 u^\m_\d(p_\delta ) \ .\end{aligned}\ ] ] eq.([eq1neu ] ) will serve as the reference point for future comparisons with other calculations . the advantage offered by this parameterization can be seen best when performing a @xmath43 expansion of eq.([eq1neu ] ) , once more restricted to @xmath146 : @xmath174 \right . \nonumber \\ & & + ( s\cdot q ) \epsilon_\mu \left [ -\frac{g_1(q^2)}{m_n}-\frac{\delta g_1(0)}{2m_n^2}+\frac{\delta g_2(q^2)}{4m_n^2 } + \frac{q^2}{4m_n^2\delta}g_3(q^2)+{\cal o}(1/m_n^3 ) \right ] \nonumber \\ & & + ( s\cdot q ) ( v\cdot\epsilon ) q_\mu \left [ \frac{g_1(0)}{2m_n^2 } - \frac{g_2(q^2)}{4m_n^2 } + { \cal o}(1/m_n^3 ) \right ] \nonumber \\ & & \left . + ( s\cdot q ) ( q \cdot \epsilon ) q_\mu \left [ -\frac{g_3(q^2)}{4m_n^2\delta } + { \cal o}(1/m_n^3 ) \right ] \right\ } u_{v,\delta}^{\mu}(0)\ .\end{aligned}\ ] ] the above formula serves as the basic connection between our ( non - relativistic ) results calculated via _ sse _ and the desired ( relativistic ) transition current eq.([eq1neu ] ) . one can easily see that to @xmath127 one is sensitive to the leading and subleading behavior of the form factor @xmath175 , whereas the form factors @xmath176 start out at @xmath177 and therefore only allow us to determine them to leading order . one can also check that eq.([exp2 ] ) is gauge - invariant to @xmath146 and allows for a non - vanishing @xmath3-dependent contribution in each of the four independent structures , consistent with the _ sse _ calculation to @xmath1 . we now move to the details of the calculation . in the present subsection we discuss the born contribution to the amplitude . the relevant expressions resulting from each @xmath22 tree level vertex have the following explicit forms with the physical values @xmath178 . ] @xmath179u_{v,\delta}^{\mu}(0 ) \nonumber \\ i{\cal m}^{e_1}_{\delta\rightarrow n\gamma^*}= & -\sqrt{\frac{2}{3 } } e\;\bar{u}_v(r_n ) & \left[(s\cdot \epsilon\;q_\mu - s\cdot q\ ; \epsilon_\mu ) \frac{e_1\delta}{2m_n^2}\right]u_{v,\delta}^{\mu}(0 ) \nonumber \\ i{\cal m}^{d_1}_{\delta\rightarrow n\gamma^*}= & -\sqrt{\frac{2}{3 } } e\;\bar{u}_v(r_n ) & \left [ -s\cdot\epsilon\;q_\mu\frac{d_1\delta}{4m_n^2}+\epsilon_0\;s\cdot q\;q_\mu\frac{d_1}{4m_n^2}\right]u_{v,\delta}^{\mu}(0 ) \nonumber \\ i{\cal m}^{b_1\hspace{0.05in}rc}_{\delta\rightarrow n\gamma^*}= & -\sqrt{\frac{2}{3 } } e\;\bar{u}_v(r_n ) & \left[s\cdot q\ ; \epsilon_\mu\frac{b_1\delta}{4m_n^2 } + \epsilon_0\;s\cdot q\;q_\mu\frac{-b_1}{4m_n^2}\right]u_{v,\delta}^{\mu}(0 ) \nonumber \\ i{\cal m}^{b_6\hspace{0.05in}rc}_{\delta\rightarrow n\gamma^*}= & -\sqrt{\frac{2}{3 } } e\;\bar{u}_v(r_n ) & \left[s\cdot q\ ; \epsilon_\mu\frac{b_6\delta}{2m_n^2 } + \epsilon_0\;s\cdot q\;q_\mu\frac{-b_6}{2m_n^2}\right]u_{v,\delta}^{\mu}(0),\end{aligned}\ ] ] where the superscripts tabulate the relevant vertices , in relation to the lecs , while @xmath180 stands for relativistic correction " . we mention that , the second relativistic correction lagrangian term in eq.([rel3 ] ) gives _ zero _ contribution to our form factors . to see this , note that the derivatives acting on the delta field bring down factors @xmath181 and @xmath182 where @xmath183 is the @xmath2(1232 ) soft momentum . in the @xmath2(1232 ) rest frame , where our calculation is performed one finds @xmath184 . both @xmath185 and @xmath182 therefore vanish due to the light delta constraints . we do anticipate , on the other hand , that this operator will contribute to the resonant born diagram in pion photo - production where the intermediate @xmath2(1232 ) will also contain off - shell components of the delta field . the overall born contribution , therefore , is @xmath186u_{v,\delta}^{\mu}(0 ) . \label{eq : born}\end{aligned}\ ] ] in the next subsection we will turn our attention to contributions from loop corrections to the transition vertex . to @xmath1 one can draw 10 independent loop - diagrams for the @xmath142 transition ( see fig . [ fig0 ] ) . however , due to the constraints @xmath187 = @xmath188 = @xmath189 = 0 satisfied by the on - shell spin 3/2 isospin 3/2 spinor @xmath190 only diagrams 1i , 1j in fig . 1 survive . their contribution can be formally written as @xmath191 \nonumber \\ + & ( s\cdot q ) \epsilon_\mu & \left [ g_a g_n + \chi g_1 g_\delta \right ] \nonumber \\ + & ( s\cdot q ) q_\mu & \bigl ( \epsilon_0 \left [ g_a j_n + \chi g_1 j_\delta \right ] \bigr . \nonumber \\ & & + \bigl . ( q\cdot\epsilon ) \left [ g_a h_n + \chi g_1 h_\delta \right ] \bigr ) \bigr\ } u_{v,\delta}^{\mu}(0 ) \ .\end{aligned}\ ] ] here the @xmath25 and @xmath2 labels on the functions @xmath192 denote the intermediate baryon propagator in diagrams ( i ) and ( j ) of fig . 1 , respectively , while the isospin factor @xmath193 takes the value @xmath194 . the loop functions are defined via @xmath195\nonumber \\ g_n(t ) & = & 2 \int_0 ^ 1 d x \ x { a^{301}(\delta x , m^2_t)\over d } \nonumber \\ g_\delta(t ) & = & -2 \int_0 ^ 1 dx \left[x { a^{301}(-\delta x , m^2_t)\over d } + 2 ( 1-x ) { a^{301}(-\delta x , m^2_t)\over d(d-1 ) } \right]\nonumber \\ j_n(t ) & = & \int_0 ^ 1 dx \ x(1-x ) a^{310}(\delta x , m^2_t ) \nonumber \\ j_\delta(t ) & = & \int_0 ^ 1 dx\ x(1-x ) { d-3\over d-1}a^{310}(-\delta x , m^2_t ) \nonumber \\ h_n(t ) & = & \int_0 ^ 1 dx\ x(1-x ) ( 1 - 2x ) a^{300}(\delta x , m^2_t ) \nonumber \\ h_\delta(t ) & = & \int_0 ^ 1 dx\ x(1-x ) ( 2x-1 ) { d-3\over d-1}a^{300}(-\delta x , m^2_t ) \ , \end{aligned}\ ] ] where the @xmath196 dependence arises implicitly through @xmath197 . explicit representations for the @xmath198 are given in the appendix . the @xmath177 sensitivity of our calculation is saturated by the two inverse powers of the quantity @xmath199 which is of the order of the chiral symmetry breaking scale @xmath200 . we close the present subsection with a brief discussion concerning renormalization issues . it is straightforward , using eq.([functions ] ) and eq.([exp1 ] ) , to see that eq.([hbxpt ] ) for the loop contributions to the decay amplitude of the @xmath2(1232 ) is a _ complex _ quantity and contains several divergent pieces . the lower - order couplings @xmath108 , @xmath109 only possess a finite part , whereas the infinities encountered in the loop amplitudes are absorbed by the infinite parts of the @xmath1 counterterms , @xmath109 couplings . ] @xmath138 , @xmath137 @xmath201 with @xcite @xmath202 .\ ] ] the two pertinent beta - functions @xmath203 can be found from eq.([functions ] ) and the expressions in appendix a. it is straightforward to obtain the results @xmath204 at the moment , however , we have no information about the numerical size of their finite parts @xmath205 at the chosen renormalization scale @xmath206 , as well as about the magnitude of the coupling @xmath109 . we note that this information will be available in the near future as several calculations regarding different scattering processes in the delta resonance region using sse are underway , which will allow for a systematic extraction of the relevant higher order @xmath16-couplings . in section vii we will nevertheless give some numerical estimates for linear combinations of these couplings based on calculations for @xmath16-transition multipoles performed in theoretical frameworks outside sse . gauge invariance requires that , upon setting @xmath207 , the amplitude must vanish . the general amplitudes eq.([exp ] ) and eq.([exp2 ] ) satisfy this requirement explicitly ( recall that @xmath208 ) . in the case of our loop calculation this requires @xmath209 separately for @xmath210 and @xmath211 . this provides for a stringent test of our calculation . from the explicit formulae eq.([functions ] ) and eq.([exp1 ] ) one can show that the constraints of eq.([con ] ) , are indeed satisfied . this is most easily seen by partial integration of the @xmath212 part of @xmath213 . combining the loop result , eq.([hbxpt ] ) , with the born contributions , eq.([eq : born ] ) , we obtain the total @xmath145 amplitude . this is to be compared with the general form of the amplitude , eq.([exp2 ] ) . the identification of the three form factors is straightforward . at this point we have no information about the magnitude of the counterterms @xmath214 . however , the important observation is that the unknown counterterms only affect the _ overall normalization _ of the 3 form factors @xmath215 and leave the _ @xmath3-dependence unaffected_. we therefore separate the real photon point @xmath164 from each form factor and write @xmath216 with @xmath217 . it is then straightforward to arrive at the results of table i. comparing the _ relativistic _ form of the coupling structure proportional to @xmath108 , eq.([rel1 ] ) , with our general _ relativistic _ amplitude , eq.([eq1neu ] ) , one expects that only the @xmath218 form factor receives contributions from this coupling , which also holds in the non - relativistic sse formalism . analogously , the coupling @xmath109 in table [ tab1 ] is shown only to contribute to the form factor @xmath219 , again as expected from its relativistic analogue in eq.([rel1 ] ) . furthermore , the results of table [ tab1 ] are , as they should be , gauge - independent . this happens automatically for the chiral counterterm contributions . for the loop contributions , however , gauge invariance can be explicitly demonstrated via the use of eq.([con ] ) . the results of table [ tab1 ] can now be compared with the ones in ref.@xcite where only the real - photon case is considered ( no @xmath220 ) . we find a difference from our result originating in eq.([exp ] ) . in particular , there is no contribution to @xmath219 from the @xmath108 counterterm . this happens because , in @xcite , the coefficient of @xmath218 in the @xmath221 term is @xmath222 rather than @xmath223 . in section [ numerics ] we derive an estimate for the unknown couplings . we will then be in position to describe both the @xmath3 dependence and the overall normalization of the three form factors . once again we stress that _ chiral symmetry demands that these form factors are complex quantities _ , which is due to the on - shell @xmath4 intermediate state of loop diagram ( i ) in fig.1 . in the relativistic case the identification of the transition multipoles m1(@xmath3),e2(@xmath3),c2(@xmath3 ) in terms of the `` dirac''-type form factors @xmath224 is obtained @xcite by reading off the spin - space tensor properties of the pauli - reduced most general amplitude , eq.([eq1neu ] ) , in the @xmath225 gauge . the explicit expressions read ) one easily verifies that they agree with the corresponding expressions given in @xcite . we are grateful to rick davidson for valuable support in this point . ] @xmath226 \right.\nonumber \\ & & \left . -g_2(q^2)\frac{m_\delta}{2m_n}(m_\delta^2-m_n^2-q^2 ) -g_3(q^2)q^2\frac{m_\delta}{\delta } \right\ } \nonumber \\ e2(q^2)&=&-c_\delta \left\ { g_1(q^2)(m_\delta^2-m_n^2+q^2)-g_2(q^2)\frac{m_\delta}{2m_n}(m_\delta^2-m_n^2-q^2 ) -g_3(q^2)q^2\frac{m_\delta}{\delta } \right\ } \nonumber \\ c2(q^2)&=&-c_\delta 2m_\delta |\vec{q}| \left\{g_1(q^2)-g_2(q^2)\frac{e_n}{2m_n}-g_3(q^2)\frac{\omega}{2\delta}\right\},\end{aligned}\ ] ] with @xmath227 we now need to match these general expressions to the @xmath146 accuracy of our results for the @xmath228 form factors obtained via sse in the previous section . we may proceed in two ways : 1 . use the same reasoning as in the relativistic case , _ i.e. _ , obtain the multipoles from the spin tensor structure of the appropriate pauli - reduced expression , eq.([exp2 ] ) . 2 . expand the relativistic formulae for the multipoles to @xmath146 . in both cases we obtain @xmath229 \nonumber \\ e2(q^2 ) & = & { e\sqrt{\delta^2-q^2 } \over 6 \sqrt{\delta}}\ \left [ - { \delta\over 2 m^2_n } g_1{(0 ) } + { \delta \over 4 m^2_n } g_2(q^2 ) + { q^2 \over { 4 \delta m^2_n } } g_3(q^2)+{\cal o}(1/m_n^3 ) \right ] \nonumber\\ c2(q^2 ) & = & { e(\delta^2-q^2 ) \over 6 \sqrt{\delta } } \ \left [ -{1\over { 2 m^2_n } } g_1{(0 ) } + { 1 \over { 4 m^2_n } } g_2(q^2 ) + { 1 \over { 4 m^2_n } } g_3(q^2)+{\cal o}(1/m_n^3 ) \right ] \ .\end{aligned}\ ] ] > from eq.([emr ] ) one can directly see that the @xmath127 prediction for m1 is sensitive to subleading order , whereas the prediction for the quadrupole multipoles e2 , c2 only gives us the leading result . one also observes the peculiar behavior that _ in the case of the e2- , c2-form factors their @xmath3-dependence is not suppressed by powers of @xmath230 with respect to the @xmath164 point _ ! according to our knowledge this situation has not been observed before in chiral calculations of form factors , e.g. compare @xcite which we discuss in section [ predictg ] . ] . its formal origin lies in the fact that for quadrupole transitions it is not possible to write down a counter term at a lower order than @xmath127 , i.e. the same order where loops start contributing . _ keeping the above caveat in mind that eq.([emr ] ) only represents the leading behavior for the e2- , c2-form factors , the chiral structure underlying eq.([emr ] ) leads us to the expectation that the e2-,c2-form factors only show a weak suppression at low @xmath3 with respect to their @xmath164 point and that the @xmath3-evolution could be very different from the corresponding ( dipole behavior of the ) nucleon sachs form factors _ @xmath231 , e.g. @xcite . on the other hand , the m1 form factor shows the expected 1/m suppression of its radius with respect to the photon point quite analogous to the sse calculation of the dipole form factor @xmath232 in @xcite . this question will be addressed in more detail in a future communication once precision information on the relevant coupling constants is available . we further note that the quadrupole multipoles in eq.([emr ] ) satisfy the long wavelength ( @xmath233 ) constraint @xmath234 . to provide numerical results we need to fix the coupling constants entering the calculation . for our @xmath127 calculation it is appropriate to use @xmath235 mev , @xmath236 mev , @xmath237 mev , @xmath238 mev , @xmath239 , @xmath240 @xcite . however , the four @xmath241 couplings @xmath242 as well as the leading @xmath243 coupling @xmath244 are only poorly known at the moment . until we have more information on these couplings we therefore proceed _ semi - empirically _ and evaluate the _ exact @xmath127 sse results of sections ii to vi _ with phenomenological input from other calculations . in particular , for the leading m1 coupling we use @xmath245 as in ref.@xcite @xmath116 vertex in eq.([eq : a2 ] ) , in order to be consistent with the sse formalism paper ref.@xcite . the sign of b1 is chosen to give a positive result for the @xmath107 m1 @xmath116 transition multipole at @xmath164 , consistent with the previous analyses @xcite . ] , although this is not entirely consistent with our @xmath127 calculation as this particular strength for the coupling only holds to @xmath107 and undergoes an as yet undetermined renormalization in @xmath127 . the remaining three @xmath116 couplings are still undetermined @xcite , for now we find that we can parameterize them via _ 2 independent linear combinations @xmath246 _ and fix them from refs.@xcite in the next section . finally , due to lack of more precise information at the moment , we use the su(6 ) quark model estimate @xmath247 for the @xmath243 coupling in eq.([lagrangian ] ) , which should be accurate to 25% ( _ e.g. _ @xcite ) . we also note that the so called off - shell parameters @xmath112 of eq.([rel1 ] ) decouple from our results . first , we note that the unknown couplings , @xmath248 . ] @xmath249 only show up via two independent linear combinations in our calculation . we therefore introduce the new set of parameters @xmath250 we now want to utilize existing phenomenological analyses for @xmath16 transition multipoles in order to fix these two parameters . most of the existing work relies on parameterizations of pion photoproduction in the delta region and has focused on the determination of the transition multipoles at the so called k - matrix pole position , e.g. see @xcite . unfortunately , this wealth of information is not directly transferable onto our problem because the k - matrix approach by construction only gives the imaginary part of the transition amplitude at the kinematical point where the real part vanishes . for our calculation , however , we need information of the transition multipoles at the t - matrix pole , where @xmath251 can be complex quantities even at the resonance position . this information has recently become available @xcite . our fitting procedure will refer to either the values of the real part of the transition multipoles @xmath252 at @xmath164 or to the ratio e2/m1 at @xmath164 . we shall adopt the first choice for the rpi , vpi results @xcite and the second one for the mainz results @xcite . the two parameter sets determined by fitting to rem1(0 ) , ree2(0 ) , with re denoting the real part , of the rpi , vpi data will be designated a and b respectively while a further subscript taking the values 1 or 2 will denote the two fits used in reference @xcite . specifically , using @xmath253 , @xmath254 we obtain @xmath255 while using @xmath256 , @xmath257 we find @xmath258 we now consider the vpi group results . the values @xmath259 , @xmath260 give @xmath261 and finally the values @xmath256 , @xmath262 produce @xmath263 the next step is to use the speed - plot analysis . ] of the mainz group @xcite . this time we need our expression @xmath264 from eq.([emr ] ) and table [ tab1 ] . we end up with a new parameter set , denoted by c @xmath265 all the fits produce `` natural size '' values for the parameters @xmath266 . based on the above parameter sets we may determine an average value and a relevant error for c2 , c3 . we arrive at the values @xmath267 having fixed @xmath246 in the previous subsection we can now give the following predictions for the average values of the electromagnetic form factors and their relevant errors @xmath268 for the ratios we determine @xmath269 we note that the imaginary parts of the multipole form factors ( as well as the imaginary parts of the form factors @xmath270 @xmath271 ) are the same for all fits and so we are not in a position to produce error bars for them . the reason is simply that the imaginary part of the transition amplitude stems exclusively from loop contributions which do not depend on @xmath272 . of course the same is not true for the ratios of the multipoles since in this case real and imaginary parts of the numerator and the denominator are mixed in order to obtain the real and imaginary part of the ratio . we also note that our results appear much less sensitive to @xmath273 compared with their dependence on @xmath274 . the reason is that this first parameter appears only in the @xmath218 form factor and its contibution with respect to the other ones is suppressed by a power of the ratio @xmath275 . in figures [ fig3 ] , [ fig4 ] , [ fig5 ] the @xmath276dependence of both the real and the imaginary parts of @xmath218 , @xmath219 , @xmath220 is depicted , using the c , @xmath277 and @xmath278 sets , respectively , for the values of the parameters @xmath273 , @xmath274 . as for the real part of the form factor @xmath220 we do not want to imply that it changes its sign near @xmath279 . given that @xmath280 approaches quickly to zero even small corrections from higher orders can change the result for @xmath281 @xmath282 . the resulting @xmath276dependence of the multipole form factors @xmath10 , @xmath11 , @xmath12 is shown in figures [ fig6 ] , [ fig7 ] , [ fig8 ] and [ new ] respectively . finally the last remaining figures show the real and the imaginary parts of the ratios emr , cmr . we close the present section with a remark concerning the range of @xmath3 , for which we believe our results to be meaningful . as we have already mentioned in the previous paragraph , the _ imaginary _ part of the amplitudes is determined entirely by a 1-loop diagram related to @xmath283-scattering at the relatively large kinematic energy of approximately @xmath284 mev . we therefore do not necessarily expect that the @xmath127 calculation leads to a good description of the _ imaginary parts _ and point to the _ possibility _ of large higher order corrections in the _ imaginary parts _ of the @xmath215 form factors ! this issue can only be resolved by a future @xmath285 calculation . though enjoying less fame than the corresponding transition multipoles of the previous section , the three @xmath16 transition form factors @xmath215 are actually better suited for a chiral analysis the reason being that to @xmath127 the complete @xmath3-dependence is solely governed by @xmath115 parameters , whereas the poorly known couplings @xmath286 only enter in the @xmath164 normalization of the form factors . utilizing @xmath246 from the previous section one fixes @xmath287 we note that all three imaginary parts and the real part of @xmath288 do not depend on the couplings @xmath286 and are given by the much better known leading order parameters of sse . we now proceed to give predictions for the slopes parameters in the discussion . ] @xmath289 which can be directly calculated from eqs.(42 - 44 ) . defining @xmath290 and @xmath291 = @xmath292 we find @xmath293 \ .\end{aligned}\ ] ] one can see immediately that @xmath294 is suppressed by a power of @xmath295 compared to the real photon point @xmath164 . this behavior can also be observed in the @xmath296 dirac form factor of the nucleon ( e.g. @xcite ) . an analogous result is expected for the radius of the m1(@xmath3 ) form factor , see the discussion in section [ trm ] . for the slope parameters corresponding to @xmath219 and @xmath220 , @xmath297-differentiation of the @xmath298 , @xmath66 functions is needed , resulting in two divergent contributions ( one non - integrable singularity resulting from differentiating under the integral sign and one divergent surface term from differentiating w.r.t . the integration limit ) . treating the integral as a principal - valued one the infinities are shown to cancel out , leaving @xmath299 \right ] \\ \rho^2_3 & = & -2 \pi { \cal n}\frac{m_n}{\delta } \left [ 8\m + { 16\over 3}\m^3 - ( 1+{15\over 2}\m^2 ) \left ( { \pi\over 2 } - i\log({1\over\m } + \sqrt{{1\over\m^2}-1}\right ) -i\sqrt{1-\m^2}\left({19\over 6 } + { 16\over 3}\m^2\right ) \right ] \nonumber .\end{aligned}\ ] ] it is important to note that this is a quite peculiar result . one can see immediately that these two slope parameters scale as @xmath300 , i.e. they show no suppression by a heavy scale with respect to the @xmath164 point ! this behavior is completely analogous to the one expected for @xmath301 ( cf . section [ trm ] ) , but has not yet been observed in chiral calculations of baryon form factors . the first moment with respect to @xmath3 of these `` quadrupole - like '' form factors @xmath302 is entirely given by the small scales @xmath303 and leading order coupling constants allowing , _ in principle _ , a large variation of the form factors at low @xmath3 compared to their @xmath164 point . however , we want to emphasize that eqs.([radii],[radii2 ] ) only represent the leading results for the three slope parameters , more can not be learned from a @xmath127 calculation . at present , we have no way of knowing the size of the @xmath285 corrections , but in principle it is possible to calculate them at a later stage and the subleading corrections to the @xmath304 vertex , as these couplings would enter the new loop diagrams of @xmath285 . ] . finally , we can obtain the chiral limit ( @xmath305 ) behavior of the slope parameters from eqs.([radii ] , [ radii2 ] ) , @xmath306 + { { \rm o}(\m ) } \nonumber\\ \rho^2_2 & = & -{4\over 3 } i\pi{\cal n}\frac{m_n}{\delta } + { { \rm o}(\m ) } \nonumber\\ \rho^2_3 & = & - 2\pi { \cal n } \frac{m_n}{\delta } \left [ -{\pi\over 2 } + i ( \log 2 - \log\m - { 19\over 6 } ) \right ] + { \rm o}(\m)\ .\end{aligned}\ ] ] thus , the real part of @xmath307 diverges logarithmically in the chiral limit while its imaginary part remains finite . on the other hand , the real part of @xmath308 vanishes altogether in the chiral limit , while its imaginary part remains finite . as for @xmath309 , it is the real part that remains finite in the chiral limit while its imaginary part diverges logarithmically . the numerical values for the slope parameters are also collected in table [ tab2 ] and since they take contributions only from the loops are independent from the values of @xmath273 , @xmath274 . the main features of our investigation can be summarized as follows : * the @xmath0 amplitude has been calculated to @xmath127 in the `` small scale expansion '' formalism . the pertinent counterterm contributions , relevant 1/m corrections to lower order coupling and all loop diagrams allowed to this order have been analyzed systematically . it was found that the @xmath3-dependence of the transition is generated by two @xmath310loop diagrams and that two counterterms are needed to renormalize the result . * we have obtained an estimate for presently poorly known couplings from existing phenomenological analyses and then given predictions for the complex transition multipoles @xmath311 , @xmath312 , @xmath313 as well as the complex multipole ratios cmr(0 ) , emr(@xmath3 ) , cmr(@xmath3 ) . while matching the @xmath127 results to the corresponding expressions of the multipoles as functions of @xmath3 we found that the radii of the quadrupole form factors are not suppressed by a large scale compared to the @xmath164 point , indicating a possibly rapid variation at low @xmath3 . further study of this interesting result will continue once more precise information on the necessary couplings is available . * three appropriately defined electromagnetic @xmath16 transition form factors @xmath314 have been identified from the @xmath127 @xmath0 calculation . the longitudinal form factor @xmath280 had to be rescaled compared to existing definitions in the literature in order to achieve consistency with the power - counting of sse . @xmath175 could be found to subleading order , whereas for @xmath302 we obtained the leading result . * the three @xmath215 form factors are found to be complex quantities due to the @xmath4-loop diagram ( i ) in fig.1 . their entire @xmath3dependence is controlled by relatively well - known @xmath115 parameters making the @xmath215 form factors the prefered testing ground of chiral symmetry in the radiative @xmath16-transition . the corresponding slope parameters have been calculated and their chiral limit behavior has been discussed . the numerical results of this work have to be considered preliminary due to the presently poor state of knowledge on coupling constants in the sse formalism . however , this situation is going to improve in the near future and we will at a later point revisit our @xmath127 analytical results and try to improve upon the numerical accuracy of our predictions . * acknowledgments * we would like to thank v. bernard , r. davidson , u .- g . meiner , m. neubert , c.n . papanicolas and r. workman for valuable discussions . c. n. k. also wishes to acknowledge several interesting exchanges with a. bernstein . the research of c. n. k. and g. c. g. has been supported in part by the eu programme training and mobility of researchers " , network hadronic physics with high energy electromagnetic probes " , contract erb fmrx - ct96 - 0008 . finally , the research of g.p . has been financed under erbfmbict 961491 return marie curie fellowship . t. r. h. would like to thank the national institute for nuclear theory ( int ) in seattle for its hospitality , where part of this work was done . the @xmath198 functions used in eq.([hbxpt ] ) are defined as euclidean integrals in dimensional regularization @xmath315 to calculate them , first of all , we express the denominator using the schwinger proper time representation @xmath316 with @xmath317 the well - known @xmath318 function . the constraint @xmath319 is explicitly fulfilled for the @xmath320 @xmath321 case while for @xmath320 @xmath322 we analytically continue our expressions . finally using the tabulated integrals @xmath323 @xmath324dt=\frac{2^{1-\beta-\frac{\nu}{2}}\gamma(\beta ) } { \gamma(\frac{1}{2}\nu+\frac{1}{2}\beta+\frac{1}{2})}(z+k)^ { -\frac{\beta}{2}}f\left(\frac{\nu}{2},\frac{\beta}{2};\frac { \nu+\beta+1}{2};\frac{z - k}{z+k}\right)\ ] ] with @xmath325 the parabolic cylinder function and f(a , b;c;x ) the hypergeometric function , we find explicitly @xmath326 where the factor l carries the infinity in dimensional regularization and is defined in eq . ( [ divfactor ] ) . it is clear that @xmath327 is finite while @xmath328 and @xmath329 diverge . the integrals in eq.([functions ] ) are split between @xmath330 $ ] ( trigonometric branch ) and @xmath331 $ ] ( log branch ) where @xmath332 is the large root of @xmath333 , @xmath334 with @xmath335 to the order we work . in the calculation the functions are defined with @xmath336 for the @xmath337 intermediate state loop diagram and @xmath338 for the @xmath339 one . this entails ( cf . eq.([exp1 ] ) ) that the logarithms have negative arguments ( and therefore there is an absorptive piece of the amplitude ) for the @xmath337 diagram alone , as expected from general considerations ( @xmath340 for @xmath210 only ) . the imaginary parts are computed analytically . p. bartsch et al . , c2/m1 in @xmath343 via @xmath344 , talk given at baryons 98 , to be published in the proceedings ; m.o . distler et al . , `` recent measurements of the @xmath345 response functions at bates '' , _ ibid . _ ; r.w . gothe et al . , `` the @xmath346 transition at low momentum transfer '' , _ ibid._. bates - oops collaboration , c.n . papanicolas , spokesperson ; mami - a1 collaboration , r. neuhausen , spokesperson ; elsa - elan collaboration , b. schoch , spokesperson ; bnl - legs collaboration , a. sandorfi , spokesperson .

we review the current status of andreev reflection spectroscopy on the heavy fermions , mostly focusing on the case of cecoin , a heavy - fermion superconductor with a critical temperature of 2.3 k. this is a well - established technique to investigate superconducting order parameters via measurements of the differential conductance from nanoscale metallic junctions . andreev reflection is clearly observed in cecoin as in other heavy - fermion superconductors . considering the large mismatch in fermi velocities , this observation seemingly appears to disagree with the blonder - tinkham - klapwijk ( btk ) theory . the measured andreev signal is highly reduced to the order of maximum 13% compared to the theoretically predicted value ( 100% ) . the background conductance exhibits a systematic evolution in its asymmetry over a wide temperature range from above the heavy fermion coherence temperature down to well below the superconducting transition temperature . analysis of the conductance spectra using the extended btk model provides a qualitative measure for the superconducting order parameter symmetry , which is determined to be-wave in cecoin . it is found that existing models do not quantitatively account for the data , which we attribute to the intrinsic properties of the heavy fermions . a substantial body of experimental data and extensive theoretical analysis point to the existence of two fluid components in cecoin and other heavy - fermion compounds . a phenomenological model is proposed employing a fano interference effect between two conductance channels in order to explain both the conductance asymmetry and the reduced andreev signal . this model appears plausible not only because it provides good fits to the data but also because it is highly likely that the electrical conduction occurs via two channels , one into the heavy electron liquid and the other into the conduction electron continuum . further experimental and theoretical investigations will shed new light on the mechanism of how the coherent heavy - electron liquid emerges out of the kondo lattice , a prototypical strongly correlated electron system . unresolved issues and future directions are also discussed .

0902.1009

it is well understood that the resistance minimum observed in metallic alloys diluted with small amount of magnetic impurities ( e.g. , au - fe or cu - fe ) is caused by the spin - spin scattering between conduction electrons and local moments @xcite . at temperatures approaching absolute zero , the magnetic moments of impurity atoms are screened by the conduction electron spins , prohibiting the divergence of the resistance due to the logarithmic scattering rate . this single impurity kondo effect has been extensively studied in a variety of condensed matter systems @xcite . if localized magnetic moments form a dense ordered array , i.e. , kondo lattice , the underlying physics becomes much more complicated and its understanding has been a great challenge to theorists and experimentalists . according to doniach s kondo lattice model @xcite , the fate of the local moments in the absence of direct dipolar interaction depends on the relative strength of two energy scales : one for kondo interaction ( @xmath3 ) ; and the other for ruderman - kittel - kasuya - yoshida ( rkky ) interaction ( @xmath4 ) , where @xmath5 is the boltzmann constant , @xmath6 is the bandwidth , @xmath7 is the exchange coupling and @xmath8(0 ) is the conduction electron density of states ( dos ) . the ground state will be an antiferromagnetically ordered one if @xmath9 , whereas a fermi liquid of heavy electrons emerges in the opposite case . this heavy electron liquid , also called a kondo liquid , does not exhibit spontaneous magnetism . many theoretical and experimental efforts have been made towards a microscopic understanding of such phase diagrams @xcite , particularly near quantum critical points @xcite . the microscopic mechanism for the formation of heavy electron bands remains unknown as the topic of the most fundamental importance to this field . a key question is how the localized f - electrons embedded in a quantum sea of conduction electrons acquire the itinerancy over the system . one scenario is that the f - electrons become delocalized via hybridization with the conduction electrons : a highly dynamical and collective process . this collective compensation of kondo lattice spins by conduction electron spins forms new bloch states and results into a narrow renormalized band of heavy electrons @xcite . the majority of heavy electron materials are based on some rare earth ( mostly ce and yb ) or actinide ( mostly u ) elements that have valence electrons in the highly localized 4f or 5f orbitals . the overlap between neighboring f orbitals is negligible since they are highly confined to the nuclei over a length scale less than the interatomic spacing . among dozens of ce - based heavy - fermion compounds discovered so far , the 1 - 1 - 5 family cemin@xmath0 ( m = co , rh , ir ) have added to new avenues for the study of rich and novel physics including quantum criticality , the fulde - ferrell - larkin - ovchinnikov ( fflo ) phase transition and the interplay between superconductivity and magnetism ( see @xcite and references therein ) . the ce 1 - 1 - 5 family is a subgroup of a larger class : ce@xmath10m@xmath11in@xmath12 . here , the ce atoms occupy basal planes of a tetragonal crystal structure . measurements of field - oscillatory magnetization ( de haas - van alphen effect ) in cecoin@xmath0 have identified multiple bands crossing the fermi level @xcite . major fermi surfaces consist of warped cylinders for heavy electron and hole bands and small ellipsoidal pockets for light holes @xcite . the resistivity maximum that signifies the beginning of heavy fermion coherence ( @xmath13 * ) occurs around 45 k and a rapid drop is followed with decreasing temperature @xcite . the emerging heavy electron liquid is found to disobey the landau theory of fermi liquid as evidenced by the @xmath13-linear ( below @xmath1 20 k ) temperature dependence of the resistivity and the ln@xmath13 dependence of the electronic specific heat coefficient , to name a few @xcite . this non - fermi liquid behaviour implies proximity to a quantum critical point @xcite . superconductivity in cecoin@xmath0 sets in below 2.3 k and the large specific heat jump at @xmath14 indicates that heavy electrons participate in the cooper pairing @xcite . most of known heavy - fermion superconductors exhibit unconventional pairing symmetry and gap structure @xcite . for cecoin@xmath0 , the power law dependences of electronic and heat transport and the thermodynamic properties point to the existence of line nodes on the fermi surface @xcite . magnetic field - angle dependent thermal conductivity @xcite and specific heat @xcite measurements exhibit four - fold oscillations in the basal plane , further supporting that the gap has @xmath15-wave symmetry @xcite . however , these results are seemingly contradictory for the precise node locations . thermal conductivity indicates @xmath2 @xcite and specific heat supports @xmath16 @xcite . there have been theoretical analyses which explain the controversy @xcite . not only is the interpretation of these experiments complex @xcite , but also they intrinsically can not provide phase information of the order parameter . an unambiguous determination of the pairing state between @xmath16 and @xmath2 is of crucial importance to elucidating the pairing mechanism ( e.g. , see @xcite ) . tunneling spectroscopy as a probe of the superconducting order parameter has proven extremely useful and thus has been extensively adopted for both conventional @xcite and unconventional @xcite superconductors . this technique was instrumental in confirming the predictions of the bardeen - cooper - schrieffer ( bcs ) theory of superconductivity in conventional systems by measuring the quasiparticle dos and phonon modes mediating the cooper pairing ( @xcite and references therein ) . however , such a direct measurement of the order parameters in heavy - fermion superconductors has proven much more difficult compared to conventional superconductors and even to other unconventional superconductors , where planar tunneling and scanning tunneling spectroscopy ( sts ) have played key roles . this is due to their unique materials properties such as extremely low @xmath14 , short coherence length , high reactivity , dependence of superconducting properties on the structural disorder , difficulty of thin film growth and non - cleavability . as an alternative technique , andreev reflection ( ar ) measurements based on point - contact methods have been frequently adopted @xcite . this technique , called point - contact spectroscopy ( pcs ) , has been successful to some extent but also has raised some issues including non - spectroscopic effects and how ar is possible at a normal - metal / heavy - fermion superconductor interface . the non - spectroscopic effects can originate from a contact being in the non - ballistic regime due to short mean free path , inter - grain josephson coupling , etc , as will be discussed below . we review the current status of experimental and theoretical studies of ar in heavy - fermion systems , focusing on cecoin@xmath0 . in section 2 , basic theories on the charge transport in normal - metal / superconductor ( n / s ) junctions are reviewed . basics of ar measurements using pcs technique are described in section 3 and technical issues in pcs are discussed in sections 4 . section 5 is devoted to the discussion whether an observation of a zero - bias conductance peak can be taken as a signature of the sign change in the order parameter . in section 6 , our experimental results are presented in terms of the superconducting order parameter symmetry . how ar is possible in heavy fermions is discussed in section 7 . possible origins of the conductance asymmetry , a common feature in all our measurements on the ce 1 - 1 - 5 family , are addressed in section 8 . our conductance modeling based on a possible fano resonance in the kondo lattice is described in section 9 . unresolved issues and future directions are presented in section 10 . finally , conclusive remarks will follow . while trying to explain an unexpected increase of thermal resistance in the intermediate state of type - i superconductors , in 1964 andreev @xcite discovered that there must be an additional scattering at interfaces between normal and superconducting domains . its microscopic process was precisely derived by solving the boundary condition problem using bogoliubov - de gennes equations . the hidden scattering is retroreflection of an electron as a hole as it approaches an n / s interface from the n side . this particle - hole conversion process , called andreev reflection , is essentially a scattering off a superconducting pair potential , during which energy , momentum and spin are conserved . the missing electronic charge is transferred as a cooper pair into the superconductor , doubling the conductance within the energy gap , @xmath17 . a schematic of this process is depicted in figure [ fig : ar ] . if the normal metal is replaced with a ferromagnet , ar is suppressed since the dos for the opposite spin band is severely reduced ( see figure [ fig : ar](b ) with unbalanced spin bands in n ) . this phenomenon has been extensively utilized to measure the spin polarization of ferromagnetic materials @xcite . a schematic of andreev reflection process . ( a ) a real space picture . n and s denote a normal - metal and superconducting electrode , respectively . ( b ) a picture in the e ( energy ) vs. dos space when the s is biased positively . the dashed line indicates the fermi level . an electron incoming from the n side , impinging on the n / s interface , is retroreflected as a hole and get paired with another electron with opposite spin and momentum , transferring as a cooper pair into the superconductor over a region @xmath18 ( superconducting coherence length ) near the interface . note that energy , momentum , spin and charge ( including a cooper pair ) are conserved in this process . ] at about the same time as andreev formulated the occurrence of ar , de gennes and saint - james @xcite considered a similar boundary condition problem in an n / s bilayer system and discovered that the quasiparticle dos is modified by the induced bound states . although such ar - related effects had been known for a long time ( e.g. , @xcite ) , ar was not considered as a direct measure of the superconducting energy gap until a theory came out in 1982 . while studying the charge transport properties of n / s and s / n / s junctions , blonder , tinkham and klapwijk ( btk ) came up with an elegant theoretical formulation to explain various characteristics in a unified model @xcite . they considered a boundary condition problem with a delta - function potential barrier at an n / s interface , @xmath19 , and the quasiparticle energy in the superconductor determined by the bogoliubov - de gennes equations . for an incoming electron impinging on the n / s interface from the n side , there are four possible trajectories : normal reflection , ar as a hole , transmission into the same branch and transmission with a branch crossing . in the btk theory , the differential conductance across an n / s interface is given by the following formula . @xmath20 , \label{eq : btk}\ ] ] where @xmath21 is the dos at the fermi level , @xmath22 is the electronic charge , @xmath23 is the fermi velocity , @xmath24 is the junction area , @xmath25 is the fermi distribution function , @xmath26 is the ar probability and @xmath27 is the normal reflection probability . a remarkable aspect of this model is that it can reproduce characteristic conductance features from ar to tunneling as a function of single parameter , @xmath28 , the dimensionless barrier strength . this is demonstrated in figure [ fig : sbtk ] for an @xmath29-wave superconductor as calculated by ( [ eq : btk ] ) . for a purely metallic junction ( @xmath30 ) , the subgap conductance is doubled with respect to the normal state value , which is due to 100% probability of ar for @xmath31 . with increasing @xmath32 , a double - peak structure develops and the conductance at zero bias decreases , both features due to the energy dependence of ar probability . for a large @xmath32 value , say , @xmath33 , ar is almost suppressed and the conductance shape is dominated by single particle tunneling . in this limit , the tunneling conductance maps out the quasiparticle dos which , in the strong coupling case , may contain information on the bosonic modes mediating the cooper pairing @xcite . since ar is a scattering off a pair potential , it is sensitive to the phase coherency of the cooper pairs @xcite . thus , ar measurements on unconventional superconductors may give different information than tunneling spectroscopy . this is an important issue that has been debated actively in the case of high-@xmath14 cuprate superconductors @xcite . ( a ) current vs. voltage and ( b ) normalized di / dv vs. voltage characteristics calculated by the btk formula for a normal - metal / s - wave superconductor junction at @xmath13=0 . the dotted line in ( a ) is an asymptote to the @xmath32=1.5 curve . @xmath34 , where @xmath35 is the normal state resistance of a purely metallic junction . in the ar regime , an excess current is observed , which is the intercept obtained when the high - bias part of an i - v curve is extrapolated to the y - axis . note that the btk theory reproduces the transitional behaviour from ar to tunneling with a single parameter , @xmath32 , the dimensionless barrier strength . ] in an effort to analyze the conductance data obtained from nb / cu point - contact junctions , blonder and tinkham introduced a second term into the barrier strength parameter as follows @xcite . @xmath36^{1/2 } , \label{eq : barrierz}\ ] ] where @xmath37 , the ratio of fermi velocities in the normal - metal ( @xmath38 ) and superconducting ( @xmath39 ) electrodes . they observed that the fitting parameter @xmath40 ranged @xmath41 and interpreted that the minimum value of 0.3 is consistent with the @xmath40 value calculated by ( [ eq : barrierz ] ) using @xmath42 for nb / cu and assuming no insulating barrier ( @xmath32=0 ) . our experimental data on dozens of nb / au ( @xmath43 @xcite ) point - contact junctions over a range of sample - tip pressures have shown that @xmath44 . this indicates that it may not be justified to extract the fermi velocity mismatch contribution from the obtained @xmath40 values , contrary to their interpretation and to the claim by deutscher and nozires @xcite that one can determine the mass enhancement factor by comparing velocities obtained from pcs with known values . regarding this issue , lukic considered generalized boundary conditions in the btk problem and argued that the effective mass is just a parameterization of the unknown microscopic parameters @xcite . in heavy - fermion materials , the electronic mass is highly enhanced ( by an order of @xmath45 ) with a correspondingly reduced fermi velocity . according to ( [ eq : barrierz ] ) , a junction with a heavy fermion as one electrode and a conventional metal as the other is inherently in the tunneling regime ( e.g. , @xmath46 for au / cecoin@xmath0 ) and thus ar can not occur . however , ar has been frequently observed in such junctions as we present our data below and as observed by others ( @xcite and references therein ) . deutscher and nozires addressed this discrepancy by proposing that the boundary conditions are not affected by the mass enhancement factor @xcite . these issues are further discussed in section 7 . detailed reviews on this topic have been published , e.g. , @xcite . here we only present fundamental concepts and elements of pcs . let us begin by considering electrical conduction across a metallic contact between two identical metal electrodes . assume the two metals are connected only through a constriction of radius @xmath47 and otherwise electrically isolated . if a bias voltage is applied across the junction , the resulting conductance will depend on how the electric field distributes in the contact region . here , the most relevant length scales are the electronic mean free paths , elastic ( @xmath48 ) and inelastic ( @xmath49 ) , in comparison to the contact radius @xmath47 . if @xmath50 , electrons gain or lose energy only when crossing the interface and their velocity change is proportional to the applied voltage . this gives rise to an ohmic current - voltage characteristic and the contact resistance is given by the following formula , known as sharvin resistance @xcite : @xmath51 where @xmath52 is the resistivity of the metal . this limit ( @xmath50 ) is called ballistic or sharvin limit . note that the resistance depends only on the contact area since the product @xmath53 is constant in the simplest drude picture . in the opposite limit , namely , if @xmath54 , the electron distribution would be that as in the bulk . then , the point - contact resistance is given by the formula for a bulk sample of length @xmath55 and diameter @xmath56 @xcite , except for a geometric factor : @xmath57 this limit ( @xmath54 ) is called the thermal or maxwell limit and a local joule heating can predominate the electrical transport process . by solving boltzmann transport equations for a point - contact junction with an arbitrary value for @xmath58 , wexler derived a generalized formula for the point - contact resistance as follows @xcite . @xmath59 , \label{eq : wexler}\ ] ] where @xmath60 is a smooth function of @xmath61 . the intermediate region between ballistic and thermal regimes is called the diffusive regime , where @xmath62 and electrons undergo multiple elastic scattering across the constriction . in the case of an n / s point - contact junction , this may reduce the ar signal . in figure [ fig : pcsregime ] , the three regimes are indicated in relation to the contact size and mean free paths . an extreme limit , not relevant to the pcs discussed in this paper , can occur if the contact size becomes comparable to the fermi wavelength , in which case quantum size effects dominate @xcite . this regime is reached in atomic size quantum point contacts @xcite . schematic of the pcs regimes . @xmath47 is the radius of a point contact and @xmath48 ( @xmath49 ) is the elastic ( inelastic ) mean free path of electrons in the electrodes . ballistic : @xmath50 . diffusive : @xmath62 . thermal : @xmath54 . ] if a point - contact junction is in the ballistic limit , energy - dependent quasiparticle scatterings appear as nonlinearities in the current - voltage characteristics , analysis of which gives spectroscopic information of the material under study . what makes this pcs possible is that the contact area remains cold due to a large mean free path compared with the contact size and any energy dissipation occurs away from the junction . in contrast , in the thermal regime , energy dissipation occurs within the contact volume , causing local heating . this pcs technique was first demonstrated by yanson in 1974 in his second harmonic measurements of micro - shorted thin film tunnel junctions @xcite , from which he obtained phonon spectrum in pb , similar to that obtained in the tunneling measurements by mcmillan and rowell @xcite . shortly after , jansen and coworkers developed a more controllable method to make point - contact junctions @xcite . this technique , called the ` spear - anvil ' method , has been adopted most widely for pcs because of its simplicity . during its early days , pcs was mostly utilized to investigate phonon spectra in metals and alloys @xcite . since the btk theory came out @xcite to explain the experimental data nicely @xcite , pcs has been frequently adopted as an alternative tool to planar tunneling or sts for the study of superconducting gap structures of many novel and unconventional superconductors owing to its versatility and technical simplicity @xcite . in the majority of pcs experiments on superconductors , the differential conductance vs. bias voltage data have been analyzed successfully using the btk model to determine their superconducting order parameters . there exist many reports in the literature where the pcs data may not necessarily be spectroscopic . as discussed above , for purely spectroscopic measurements , the junction must be in the ballistic regime . a first check of this is to estimate junction size using ( [ eq : wexler ] ) and compare with known values for the mean free paths . because in general point - contact junctions are made by bringing two metals into mechanical contact on the nanoscale , data can also be affected by factors other than the nominal contact size , such as contact pressure @xcite and geometry @xcite , meaning that the nominal contact size being in the ballistic limit alone can not guarantee the spectroscopic nature of the data . the most important diagnostic is the reproducibility and consistency of the conductance spectra point to point , sample to sample and along different crystallographic orientations , as we discuss in the next three sections . like for other class of superconductors , the spear - anvil method has been employed widely for the pcs investigation of heavy - fermion superconductors . their much lower @xmath14 s ( typically less than 1 - 2 k ) make its implementation complicated because of the inevitable involvement of a @xmath63he or dilution refrigerator . in these cases , a purely mechanical approach of the tip to the sample based on the differential screw mechanism , which is widely adopted for @xmath64he systems , is not suitable . instead , a combination of mechanical and piezoelectric adjustments is more common , where a coarse approach is made by a fine - pitched screw and a fine adjustment is made by driving piezoelectric elements @xcite . although mechanically cut wires ( typically au or pt - ir ) have been used , electrochemically polished tips provide better control over the tip size and surface morphology . this is an important technical issue since it could affect the contact geometry and thus the conductance spectra . it is also crucial to prepare smooth and clean sample surfaces in order to obtain spectroscopic data . almost all materials investigated by pcs have surface layers that are not the same as in bulk . these layers could be precipitates from growth , surface reconstruction or due to oxide growth when exposed to air . in some cases , these layers can be penetrated by the metal tip during the tip approach , enabling one to probe bulk characteristics . otherwise , the surface layer can act as a potential barrier , proximity or degraded layer . if the surface layer is thin , the potential barrier case will still provide spectroscopic data which can be analyzed by the btk theory , but approaching , or in , the tunneling limit . in the other two cases , obtaining spectroscopic data is more difficult , particularly if the surface layer is thick on the order of the coherence length or longer . it is ideal if a superconductor under study can be prepared such that it has flat and smooth surfaces along all major crystallographic orientations . most unconventional superconductors are first produced in bulk ( poly- or single - crystalline ) forms . pcs on polycrystals can reveal some information on the gap values but the momentum direction is not identified or defined well , meaning that data analysis and interpretation might not provide crystallographic orientation - dependent information . in addition , because of a granular structure , measured data are vulnerable to artificial effects such as intergrain josephson coupling @xcite . thus , single crystalline samples are highly desirable . a drawback , particularly in anisotropic materials , is that crystals are often grown as platelets only along the preferred surface orientations . thin film samples can be grown along desired directions by adjusting growth conditions but heavy - fermion compounds are known to be extremely difficult to synthesize in thin film form . for pcs , we rely on the high - quality single crystals that have been grown and the preferred as - grown surface of a single crystal is first investigated . for measurements along the other directions , broken edges are frequently used . preparation of surfaces by polishing is rarely adopted due to the possibility of surface degradation during the polishing process . however , this approach for sample preparation has been successful for the ar study of cecoin@xmath0 @xcite . the heavy - fermion compound cecoin@xmath0 has a tetragonal crystal structure with the ce atoms occupying the basal plane . many experimental reports have pointed to a @xmath15-wave symmetry of the cooper pairing state @xcite . considering its known fermi surfaces @xcite , the gap nodes are supposed to form lines along the @xmath65-axis , similarly to the case of cuprate superconductors . from the group theoretical point of view on the possible pairing state , there is a correspondence between real and momentum space regarding the gap node directions in @xmath15-wave superconductors @xcite . therefore , it is essential to make pcs measurements into the @xmath66 plane to probe along nodal and antinodal directions . more specifically , measurements into ( 100 ) and ( 110 ) surfaces should be made to identify the location of nodes . the largest face ( typically ranges a few mm ) of an as - grown cecoin@xmath0 crystal is perpendicular to the @xmath65-axis . thus , this surface is naturally chosen for pcs along the @xmath65-axis . since the thickness of single crystals ranges only a few hundred @xmath67 s , it is necessary to devise a method to hold a sample securely such that its in - plane direction is aligned with the tip axis . this can be achieved by embedding a crystal into a mold made of low - temperature epoxy and cutting and polishing it @xcite . cecoin@xmath0 crystals are grown by the flux method using excess indium flux @xcite . because of the nature of this method , samples can contain indium precipitates , which may affect the pcs measurements . to eliminate this possibility , one slightly etches crystals using hydrochloric acid , noting that cecoin@xmath0 is also dissolved although at much slower rate than indium . microscopically rough surfaces can therefore be produced , causing spurious effects . again , to ensure the spectroscopic nature of the conductance spectra , reproducibility is the most important diagnostic . in particular , data should be compared along all major crystallographic orientations . in general , considering the extreme cleanness of cecoin@xmath0 ( the electronic mean free path ranges several @xmath68 s at low temperature @xcite ) , it is highly feasible to perform pcs measurements in the ballistic regime . three independent groups have reported results of pcs on cecoin@xmath0 . et al_. @xcite first reported two different types of conductance spectra obtained from point contacts along the @xmath65-axis : one with double - peak and zero - bias dip structure , and the other with a zero - bias peak and surrounding dips . the latter feature was interpreted as hinting at an unconventional symmetry ( @xmath15-wave was implied ) . et al_. @xcite also reported two different types of spectra from nominally @xmath65-axis contacts : one with a zero - bias peak and hump structure , and the other with a double hump structure . these authors claimed both @xmath15-wave symmetry and multiple order parameter components . the validity of the claim for @xmath15-wave symmetry based simply on the observation of a zero - bias conductance peak ( zbcp ) has been questioned @xcite because many other origins for a zbcp are also known . figure [ fig : dbtk ] shows normalized di / dv curves calculated for a normal - metal/@xmath15-wave superconductor junction using the kernel developed by tanaka and kashiwaya _ et al . _ @xcite as a function of @xmath40 , the effective barrier strength in ( [ eq : barrierz ] ) . here , it is assumed that the injected electrons have a momentum distribution over the whole range of angle ( 90 @xmath69 ) around the junction normal . in practice , in the tunneling limit , tunneling electrons may have a much narrower momentum distribution , known as tunneling cone effect @xcite . in the limit of a 90 degree tunneling cone , there is no difference between the @xmath65-axis , nodal or antinodal directions as there is an integration over the full half space of momentum , as shown in figures [ fig : dbtk ] . for an antinodal or @xmath65-axis junction , the cusp - like feature and linear slope near zero bias for purely metallic ( @xmath70 ) and tunnel junctions ( @xmath71 ) , respectively , exhibit the characteristic dos of a @xmath15-wave superconductor this is just a qualitative difference from the @xmath29-wave case , where a flat conductance shape is seen due to a fully isotropic opening of the gap . as @xmath40 is increased from zero , a double - peak and zero - bias dip structure develops , which is also similar to the @xmath29-wave case ( see figure [ fig : sbtk](b ) ) . an interesting case is the nodal direction with a finite @xmath40 . the conductance shape varies in a strikingly different way , as displayed in figure [ fig : dbtk](b ) . beginning with the same shape for @xmath70 as for the antinodal junction , the conductance curve becomes narrower and sharper with increasing @xmath40 , forming a zbcp instead of a dip . this zbcp originates from andreev bound states ( abs ) formed on the nodal surface due to a sign change of the @xmath15-wave order parameter around the fermi surface @xcite . the zbcp in tunnel junctions on hole - doped cuprate superconductors , well known to be @xmath15-wave , has been frequently observed @xcite and well understood theoretically @xcite . the abs occurs due to the constructive interference between an incoming electron and an andreev - reflected hole and can be formed at zero energy on a nodal surface of a @xmath15-wave superconductor because of the additional phase shift of @xmath72 @xcite . the same shape for a nodal junction as for an antinodal junction when @xmath70 can be understood as due to smearing of abs in the absence of barrier @xcite . normalized di / dv vs. voltage characteristics calculated by the extended btk formula for a normal - metal/@xmath15-wave superconductor junction at @xmath13=0 . ( a ) the junction normal is along the antinodal direction . the conductance behaviour is qualitatively similar to the @xmath29-wave case except for the cusp - like ( ar ) and v - shape ( tunneling ) features near zero bias , which are due to a @xmath15-wave superconducting dos . ( b ) the junction normal is along the nodal direction . completely different behaviours are observed due to the surface andreev bound states . ] it has been frequently claimed that a zbcp observed from pcs is a signature of abs , thus , evidence for a sign change of the superconducting order parameter . however , as pointed out by us @xcite , and sheet and raychaudhuri @xcite , just an observation of a zbcp does not corroborate such claims . a zbcp can also arise from other physical mechanisms without requiring a sign change such as local heating @xcite , intergrain josephson coupling @xcite , etc . therefore , it is essential to perform diagnostic measurements and analyses in order to prove that a zbcp is a real signature of abs . in the high-@xmath14 cuprate tunnel junctions , it is well established that abs - originated zbcps are split either spontaneously or by applied magnetic field @xcite . one of the broadly accepted explanations is the doppler shift of the bound state energy due to coupling to the superfluid momentum @xcite . the amplitude of an abs - originated zbcp is predicted to follow 1/@xmath13 dependence , in contrast with the ln@xmath13 behaviour of kondo scattering effect in tunnel junctions . in the case of metallic junctions as in pcs , it is proposed that the critical magnetic field over which the zbcp can be split is much stronger @xcite . another important diagnostic measurement , particularly necessary for metallic junctions because the splitting may not be observed for the reasons above is to compare conductance spectra along different crystallographic directions . in the case of cecoin@xmath0 , given that @xmath15-wave symmetry is most likely , measurements along both ( 100 ) and ( 110 ) directions are essential to differentiate between @xmath2- and @xmath16-wave symmetry . neither of the reports by goll _ _ @xcite and by rourke _ provided such diagnostic measurements , so origins of their observed zbcps remain unclear . _ s claim on the existence of multiple order parameters relies on their observation of multiple hump structures in the differential conductance data . however , such multiple hump or peak and dip structures have been sometimes observed in pcs study of a known single gap superconductor . to prove that the hump positions can be taken as gap values for multiple order parameters , diagnostic measurements including crystallographic orientation and temperature dependences should be made along with reproducibility . in general , considering the band structure and fermi surface topology @xcite , it is possible that the superconductivity in cecoin@xmath0 has multiband nature , but it is a different issue to prove that measured pcs data actually reflect such features . we have reported differential conductance spectra from au / cecoin@xmath0 point - contact junctions along all three crystallographic directions @xcite . figure [ fig : pcs001](a ) displays temperature evolution of the conductance along the @xmath65-axis over a wide temperature range . at high temperatures , the conductance curves are symmetric and flat , which is characteristic of simple metallic junctions over a small bias region . with decreasing temperature , an asymmetry in the di / dv curves is developed . this conductance asymmetry appears to begin below the heavy fermion coherence temperature @xmath73 ( @xmath1 45 k ) @xcite and increases with decreasing temperature down to @xmath14 ( 2.3 k ) , below which it remains constant , as shown in figure [ fig : pcs001](a ) . as @xmath14 is crossed , the conductance near zero - bias begins to be enhanced and this enhancement increases with decreasing temperature . in the literature , it has been frequently adopted to analyze conductance data by normalizing them with respect to those just above @xmath14 . noting that our data below @xmath14 exhibit nearly the same background shape , we normalized them against the data taken at 2.6 k. the result is plotted in figure [ fig : pcs001](b ) . the temperature and bias voltage dependences and the shape of the conductance curves ( btk - like ) near zero bias indicate the enhancement is due to ar . however , the magnitude of the ar conductance ( @xmath1 13% ) is greatly suppressed in comparison to the theoretical prediction ( 100% ) @xcite . ( a ) differential conductance spectra of a ( 001 ) cecoin@xmath0 point - contact junction . note the systematic evolution of the background conductance asymmetry and the enhancement of conductance near zero bias due to andreev reflection . ( b ) conductance curves normalized by the 2.6 k data . there is no dip in the raw data in ( a ) , so the shallow dips around @xmath74 1 mv at 400 mk in ( b ) are due to the normalization , not caused by local heating . the magnitude of ar conductance is on the order of @xmath1 13% . after @xcite . ] our attempt to fit the normalized conductance spectra on the ( 001 ) cecoin@xmath0 junction using the extended btk model by tanaka and kashiwaya _ et al . _ @xcite revealed that this model could not fully account for the data @xcite . the best - fit curves deviate from the data substantially ( figure [ fig : sbtkfit](a ) ) and the temperature dependence of the fitting parameter @xmath75 , smearing factor @xcite , basically tracks that of the gap , which is unphysical ( figure [ fig : sbtkfit](b ) ) @xcite . the large deviation around gap edges implies that the failure is not solely due to the large suppression of ar but also due to a large shift of the spectral weight @xcite . this situation remains unchanged no matter which modified btk model we adopt for @xmath29- or @xmath15-wave symmetry considering the fermi surface mismatch in the two electrodes and the possible breakdown of andreev approximation @xcite . all these results point to a necessity to develop a new theoretical framework to understand how ar is possible at all and why it is observed so reduced in heavy - fermion / conventional - metal interfaces . even without the full theoretical framework , pcs can be used as a spectroscopic probe of the order parameter symmetry as we discuss below . ( a ) normalized di / dv ( @xmath76 ) of a ( 001 ) cecoin@xmath0 junction in figure [ fig : pcs001](b ) and best fit curves ( @xmath77 ) by the @xmath29-wave btk model . ( b ) energy gap ( @xmath17 ) and quasiparticle lifetime broadening factor ( @xmath75 ) obtained from the @xmath29-wave fit . @xmath17 predicted by the bcs theory ( @xmath78 ) and @xmath75 ( @xmath79 ) obtained from the zero - bias conductance fit . note that the temperature dependence of @xmath75 is unphysical . after @xcite . ] we now address the superconducting order parameter symmetry in cecoin@xmath0 . as seen in the simulation in figure [ fig : dbtk ] , a cusp - like shape is predicted for @xmath15-wave symmetry if there is no barrier ( @xmath70 ) . the zero effective barrier strength condition is difficult , if not impossible , to achieve in conventional - metal / heavy - fermion - superconductor point - contact junctions because of some existing surface layer or even some effect of the fermi velocity mismatch . we therefore assume @xmath40 is always finite . in this case , a cusp - like or peaked structure , even for small finite @xmath40 , would never be observed in the antinodal direction . there always exist smearing effects due to finite temperature , finite quasiparticle lifetime and depairing . thus , the interpretation of a flat conductance shape as we observed in ( 001 ) and ( 100 ) cecoin@xmath0 junctions is ambiguous even between @xmath29- and @xmath15-wave symmetry . this was reported earlier for ( 001 ) junctions @xcite , where a @xmath15-wave fit gives slightly better results . therefore , it is essential to obtain conductance data in the @xmath66-plane to determine the pairing symmetry . figures [ fig : dwevidence](a ) & ( b ) compare such data along the ( 100 ) and ( 110 ) directions . while both spectra exhibit similar background asymmetry as seen in the ( 001 ) junction , their detailed shapes in the sub - gap region are quite different from each other . namely , the ( 100 ) data look flat , similarly to the ( 001 ) data , whereas the ( 110 ) data are cusp - like . a comparison of these experimental observations with theoretically predicted behaviours gives strong evidence for @xmath15-wave symmetry , in fact @xmath2 @xcite . figures [ fig : dwevidence](c ) & ( d ) show calculated curves in the small-@xmath40 limit using the @xmath15-wave btk model @xcite . the calculated conductance curves at @xmath70 are identical , as discussed earlier . for finite @xmath40 , they develop completely different features : double - peak and dip for the antinodal junction vs. narrower and shaper peak at zero bias for the nodal junction . we conclude that the ( 100 ) data are consistent with antinodal junction behaviour with @xmath80 and the ( 110 ) data represent nodal junction behaviour with a smaller @xmath40 . therefore , the superconducting order parameter symmetry in cecoin@xmath0 is @xmath2-wave , not @xmath16-wave . this spectroscopic result resolves the controversy of the node locations @xcite . this definitive determination of the pairing symmetry narrows down possible candidates for the bosonic mode involved in the microscopic pairing mechanism @xcite . it is notable that ar measurements along different crystallographic directions provide spectroscopic evidence for the order parameter symmetry . this is possible via detecting the sign change albeit abs are smeared to finite energy due to the higher junction transparency compared to tunnel junctions . therefore , like abs tunneling spectroscopy , pcs can provide phase - sensitive information in unconventional superconductors @xcite . as for the anisotropy in the order parameter amplitude , the conductance widths are similar for the ( 100 ) and ( 110 ) junctions probably due to the large tunneling cone effect . this is an issue to be addressed more rigorously in the future . a cusp - like feature similar to that in our data is observed in goll _ _ s measurements into the @xmath66-plane of cecoin@xmath0 but the precise crystallographic orientation was not reported @xcite . ( a ) and ( b ) are normalized di / dv of the ( 100 ) and ( 110 ) cecoin@xmath0 junctions , respectively . ( c ) and ( d ) are normalized di / dv curves calculated by the @xmath15-wave btk formula ( @xmath75=0 and @xmath13=0 ) for antinodal and nodal junctions , respectively . ( e ) and ( f ) are magnetic field ( parallel to the @xmath65-axis ) dependence for the ( 100 ) junction at 400 mk and the ( 110 ) junction at 420 mk , respectively . after @xcite . ] as discussed previously , abs are known to split spontaneously or by applied magnetic field . this splitting occurs due to the doppler shift effect in the bound state energy arising from a coupling of the quasiparticle velocity ( @xmath81 ) to the superfluid momentum ( @xmath82 ) , @xmath83 . according to tanaka _ @xcite , the critical field for splitting of abs is proportional to the junction transparency . point - contact junctions generally have high transparency , so splitting of abs would only be accomplished at very high field . this is completely consistent with our field - dependent measurements , as shown in figures [ fig : dwevidence](e ) and ( f ) . other possibilities for the non - splitting of abs have also been suggested , including the tunneling cone effect @xcite and atomic scale disorder @xcite in cuprate tunnel junctions . our conductance data on cecoin@xmath0 is a clear example for ar in heavy fermions . also , there have been many reports on the observation of ar in other heavy - fermion superconductors @xcite . as mentioned previously , this experimental observation is in conflict with what the btk theory predicts as in ( [ eq : barrierz ] ) : ar should be completely suppressed due to the large mismatch in the fermi velocities of quasiparticles . thus , how ar is possible in heavy fermions is a longstanding issue . deutscher and nozires @xcite addressed this issue and proposed a scenario that the velocity ( @xmath84 ) of injected or extracted particles are not the same as the quasiparticle velocity ( @xmath23 ) as seen in specific heat or coherence length measurements . namely , @xmath85 where @xmath86 is the mass renormalization factor and @xmath87 is the mass enhancement factor @xcite . in heavy fermions , @xmath86 is very small . lukic considered generalized boundary conditions in the btk problem and proposed that one should not interpret the parameter @xmath40 as representing materials properties but just as parameterization of the unknown microscopic details of the interface @xcite . even though these arguments provide a way to explain the discrepancy between the btk theory and experimental observations , they do not account for why ar is strongly reduced in heavy fermions . it is our conjecture that this issue might be intimately connected to why the conductance shows an asymmetry , so we focus on this issue in the next two sections . in our pcs data on cecoin@xmath0 , there are two prominent and common features that are not seen in other class of superconductors : highly suppressed ar and background conductance asymmetry . it is noted that these features have also been observed in other reports on the pcs of heavy - fermion superconductors @xcite . although our qualitative analysis of the point - contact conductance spectra using the @xmath15-wave btk model gives spectroscopic evidence for the superconducting order parameter symmetry in cecoin@xmath0 , it does not give us precise information on the gap amplitude . as discussed previously , existing models taking the fermi surface mismatch and possible breakdown of andreev approximation in heavy fermions into consideration do not quantitatively account for our data . this is an indication that a crucial element is missing in the currently available models . this situation is in striking contrast to those cases in the other class of superconductors , where the btk models have been so successfully applied @xcite . we attribute the origin of this failure to the intrinsic properties of heavy fermions , more specifically , the coexistence of multiple electronic components @xcite . in figure [ fig : asymmvst](a ) , it is found that the temperature dependence of the conductance asymmetry qualitatively follows that of the spectral weight for the coherent heavy electron liquid reported by nakatsuji , pines and fisk @xcite . also , there is a systematic evolution of this background conductance peak position as shown in figure [ fig : asymmvst](b ) : it moves from the far - negative bias side towards zero bias with decreasing temperature . yang and pines @xcite recently reported a universal scaling behavior between various experimental data including our conductance asymmetry in figure [ fig : asymmvst](a ) and the spectral weight for the emergent heavy - electron liquid . ( a ) temperature dependence of the background conductance asymmetry , defined as the ratio between conductance values at 2 mv and at + 2 mv from the ( 001 ) cecoin@xmath0 data in figure [ fig : pcs001 ] . @xmath73 is the heavy fermion coherence temperature . the inset is a semi - logarithmic plot of the same data . ( b ) temperature - dependent evolution of the normal state conductance . after @xcite . ] according to harrison s theorem @xcite , the electronic dos of a simple metal can not be measured using tunnel junctions since it cancels out the velocity factor in the conductance kernel . this theorem appears to hold in many cases where the single particle picture is valid . applying this argument to point - contact junctions consisting of simple metals , the electronic dos is not expected to be measured from such measurements . the conductance of a metallic contact is given by @xmath88 where @xmath89 is the velocity , @xmath90 is the electronic dos of the counter - electrode , @xmath25 is the fermi distribution function and @xmath91 is the differential solid angle @xcite . over a small bias region , the conductance is expected to be flat . beyond that region , it is expected and observed to be curved downward due to increased scattering , say , with phonons , as the energy increases @xcite , and this shape is not attributed to the dos . the downward curvature in metallic junctions is contrary to the upward curvature in tunnel junctions @xcite . now , the question is whether harrison s theorem still holds in the case where the many - body interaction is not negligible , as in heavy fermions . the observed conductance asymmetry in our data on cecoin@xmath0 might be strongly tied to this issue . as discussed in detail in @xcite , the conductance asymmetry can not be explained by the models proposed in the literature . certainly the asymmetry indicates it is easier to remove electrons from than to add to the heavy - fermion electrode cecoin@xmath0 . however , a simple argument that it is more difficult to add than to remove an electron from the f - orbital does not work because the f - electrons do not remain localized , but are itinerant , in heavy - fermion compounds . the argument that competing orders cause a conductance asymmetry @xcite does nt appear relevant to our case because it is observed in all three members of the ce - based 1 - 1 - 5 heavy fermions @xcite , no matter how far they are from the competing region . the persistency of conductance asymmetry in magnetic field up to 9 t ( see figures [ fig : dwevidence](e ) & ( f ) ) rules out a connection to the argument based on non - fermi liquid behaviours @xcite . in the past , the conductance asymmetry observed in pcs of heavy fermions was frequently attributed to the large seebeck coefficients of heavy fermions @xcite combined with contact s being in the thermal regime . however , this model does nt agree with our experimental data on 1 - 1 - 5 heavy fermions @xcite . a simple counter argument to this scenario is why the asymmetry persists in the superconducting state as seen in figure [ fig : pcs001](a ) , where the seebeck effect disappears @xcite and , thus , the asymmetry should , too . anders and gloos proposed that both the conductance asymmetry and reduced ar in heavy - fermion superconductors could be caused by the strongly energy - dependent quasiparticle scattering and dos of heavy fermions @xcite . this theory does account for our observed conductance features but , since the calculations are done in a green s function formalism , the microscopic physical mechanism is not obvious . nowack and klug @xcite used boltzmann transport equations to deal with electronic scattering by an energy - dependent dos of heavy fermions . in this model , the asymmetry arises from the dos centered at a finite energy . we note that the dos effect of heavy fermions is invoked in both models . we have proposed @xcite a two - channel conductance model based on the two - fluid picture @xcite and originally assuming a lorentzian form of the heavy electron dos . here , the two channels are into the superconducting heavy electron liquid and a possible normal conducting light electron liquid @xcite . it was assumed that the two channels are independent of each other . in theory it was argued that superconductivity should be induced to the light normal electron liquid due to the proximity effect @xcite but this issue is not yet settled . our experimental data were found to accurately fit to this model at lowest temperatures but the data could nt be fit with the model at higher temperatures ; the fit failing more severely with increasing temperature . this is technically because the background conductance shape is not strictly lorentzian . instead , it is observed @xcite to resemble a fano line shape @xcite , implying that the two conductance channels need to be entangled instead of being independent . motivated by our two - channel conductance model , arajo and sacramento recently formulated a btk model adopting the two - fluid picture for heavy fermions @xcite . in this model , ar is reduced essentially due to the existence of a normal conduction channel , similarly to our model , but they claim that the two channels should be allowed to mix via an interference term instead of being independent . the fano resonance @xcite , originally discovered in electron - helium inelastic scattering cross sections , has been frequently observed in a variety of condensed matter and other physical systems . in essence , it is a manifestation of an interference effect between waves that have passed through two different paths containing discrete and continuum states , respectively . this interference causes an asymmetry in the resonance spectra whose line shapes can be reproduced by the fano formula @xcite . recently , this phenomenon has been a subject of intensive investigations in quantum dot experiments @xcite and sts of single adatoms and molecules @xcite . for example , kobayashi _ et al . _ observed systematic evolution of the asymmetry in the differential conductance across a mesoscopic ring containing a quantum dot in one arm and a continuum in the other arm @xcite . investigated conductance behaviours of single adatoms deposited onto metallic substrates using stm @xcite . the measured differential conductance showed featureless shapes from non - magnetic atoms , whereas highly asymmetric curves from magnetic atoms . their proposed conductance formula including the fano resonance term successfully accounts for the measured data @xcite . here , the interference occurs between tunneling electrons into the conduction band and into the local d orbital of an adatom . the original fano formula is given as : @xmath92 where @xmath93 is the fano parameter , @xmath94 and @xmath95 and @xmath96 are the resonance energy and width , respectively . using this formula , fano lines are simulated as a function of @xmath93 , as plotted in figure [ fig : fanofit](a ) . in the extreme limit of @xmath97 , the line becomes symmetric and represents a lorentzian resonance . in the opposite limit , i.e. , @xmath98 , it corresponds to an antiresonance . in between these extremes , the line is asymmetric with a peak positioned at @xmath99 . adopting a similar formula to that used by madhavan _ @xcite , we were able to fit the data and extract useful information @xcite . we set up a model that the differential conductance is determined by the following equation containing the fano formula ( [ eq : fano ] ) as the kernel . @xmath100 where @xmath101 is a prefactor for the fano resonance contribution and @xmath102 is a constant conductance . an example of the fit using this formula is shown in figure [ fig : fanofit ] . here , the used fit parameters are : @xmath103 , @xmath104 mev , @xmath105 mev , @xmath101 = 0.0061 @xmath106 , and @xmath102 = 0.164 @xmath106 . in our previous paper @xcite , we already reported a fano resonance fit to the normalized conductance data over a narrower bias region using the following parameter values : @xmath107 , @xmath108 mev , @xmath109 mev , @xmath101 = 0.03646 , and @xmath102 = 0.8458 . from these two fits , it is clearly seen that the parameter values vary widely depending on the voltage range for fitting . therefore , it is highly desirable to obtain and fit conductance data over a wide enough voltage range . ( a ) fano line shapes as a function of the fano parameter , @xmath93 . ( b ) fano fit to the data on ( 001 ) cecoin@xmath0 at 2.43 k. fitting parameters used are @xmath103 , @xmath104 mev , @xmath105 mev , @xmath101 = 0.0061 @xmath106 , and @xmath102 = 0.164 @xmath106 . after @xcite . ] one of the unresolved issues in our previous fit using a lorentzian dos @xcite was why the resonance energy is located below the fermi level . theoretically , the kondo resonance in ce - based kondo lattice systems is known to occur at or above the fermi level , which is also observed in photoemission measurements on several heavy - fermion systems @xcite ( cf . see @xcite for the kondo resonance energy position in single impurity stm experiments ) . in the current fano fit , it is seen that our data can be reproduced by the fano formula with the resonance energy above the fermi level , in agreement with theoretical predictions and other experiments . this was possible with the choice of a negative value for @xmath93 , which forces the resonance energy , @xmath95 , to be positive . in general , @xmath93 is a complex number since the fano resonance arises from an interference effect , as discussed above . thus , the negative @xmath93 value we obtained can be understood as due to the phase factor in the interference . similar analysis using negative @xmath93 was also reported in some single impurity stm experiments @xcite . recently yang @xcite reported results from an extensive fit using a modified fano formula to our previously reported data @xcite over the @xmath74 50 mv range at two temperatures ( 1.47 k and 20.62 k ) and data over the @xmath74 4 mv range at several temperatures between 2.5 k and 60 k. the calculated curves fit to the data well , but the implications of the extracted parameter values remain to be investigated as their values depend on the range of voltage used in the fitting , as demonstrated above . also , the development of a microscopic model beyond phenomenological arguments is needed , as was done for the single impurity sts study @xcite . we discuss relevant issues in the following . one might raise questions about the validity of our analysis based on a fano effect . first , how is it possible to observe a fano resonance from point - contact ( not tunneling ) measurements and from the kondo lattice systems ? regarding this issue , we note that there is an example @xcite where continuous evolution of the fano resonance across a tunneling - metallic junction boundary is observed from stm measurements on single adatoms . this implies that it may also be possible in pcs . however , one should note a configurational difference between the kondo lattice and single adatoms since in the latter the fano ( kondo ) resonance is widely known to be sensitive to the environment around the impurity atom . second , what is the origin of the interference ? in the two - fluid picture of heavy fermions @xcite , there are two components in the electronic spectra in cecoin@xmath0 : one is the heavy electron liquid and the other is conduction electrons that do not participate in the hybridization . then , one can imagine that the interference occurs between these two channels @xcite . according to the theoretical derivation by madhavan _ @xcite , the fano parameter @xmath93 is given by the ratio of two matrix elements : @xmath110 , where @xmath111 is the coupling to a local orbital either direct or indirect via hybridization and @xmath112 is the coupling to the conduction electron continuum . it is to be investigated further which of the two terms for @xmath111 is dominant in pcs of the kondo lattice system . as seen in the above fit , there exists a large background term , @xmath102 , which might indicate that there exists substantial contribution from the channel into the conduction band without causing the interference . the fano line shape in our data persists over a wide temperature range including both normal and superconducting states , implying that the interference must be persistent over a wide temperature range . third , how can we combine a fano resonance with ar or tunneling in heavy - fermion superconductors ? as seen in figure [ fig : pcs001 ] , the background conductance asymmetry affects the shape of the ar conductance . thus , a successful microscopic theory should also explain this connection . in regards to this issue , we note that flint _ recently proposed a theory on the ar into a composite paired superconductor @xcite . they claimed that it is possible for electrons to cotunnel into the kondo lattice and this would provide an enhanced contribution to ar with a fano resonant structure . it is interesting to look into the similarities and differences between their fano resonant structure and the asymmetric conductance shape in our model . _ boundary conditions in the btk problem . _ although the theory by deutscher and nozires @xcite explains how ar is possible in heavy fermions , what role is played by the fermi velocity mismatch is not clear . related to this topic , arajo and sacramento @xcite reported recently that the argument by deutscher and nozires can be proved explicitly by solving this problem for a normal - metal / heavy - fermion - superconductor interface . quantitative measure of the fermi velocity mismatch effect in pcs might be achieved by measuring ar conductance from a ballistic junction involving a pair with well - matched fermi velocities , e.g. , au / nb , but without any oxide layer formed at the interface . _ tunneling cone effect . _ in order to detect abs unambiguously , making measurements in the tunneling limit will provide more flexible diagnostics . thin film growth of cecoin@xmath0 has been attempted but the quality is not yet good enough for all - thin - film based tunnel junctions . as an alternative , it is worthwhile trying to deposit an artificial tunnel barrier onto single crystals . _ ar in heavy - fermion - metal / conventional superconductor . _ despite some success reported in cecoin@xmath0/nb point contacts , the magnitude of ar signal in such junctions is not well determined due to the intervening oxide layer on the nb tip @xcite . further measurements on such junctions but without oxide interface layers will provide important clues to many issues on ar in heavy fermions . a related experiment will be exploring whether proximity effect exists in such junctions @xcite . the fact that ar is frequently observed in heavy fermions implies that proximity effect should exist because it is well known that ar is at the heart of this phenomenon @xcite . _ fano effect in kondo lattice systems . _ although our proposed model has shown some evidence for this effect , it is not yet studied theoretically whether such effect could exist and , if so , is measurable by pcs . also , it is an open question how to apply this picture to explain the reduced ar as well as the conductance asymmetry . microscopic derivation of this effect may lead to further insightful investigations of the kondo lattice physics . in conclusion , pcs on cecoin@xmath0 has not only provided spectroscopic information on the superconducting order parameter but also clarified several experimental and theoretical issues . sets of reproducible conductance spectra were obtained as a function of temperature , magnetic field and crystallographic orientation . a qualitative analysis of the data based on the extended btk model shows that the order parameter in cecoin@xmath0 has a @xmath2 symmetry . detailed analysis of the conductance asymmetry led us to a phenomenological model based on a possible fano effect in this kondo lattice system . further experimental and theoretical investigations of this model will lead us to more deepened understanding of the kondo lattice physics . we greatly acknowledge longstanding collaborations with e. d. bauer , j. l. sarrao and j. d. thompson at los alamos national laboratory , c. capan and z. fisk at the university of california - irvine and p. c. canfield at the iowa state university . we are grateful to j. e. elenewski , v. lukic , a. j. leggett , m. stone , y .- f . yang , d. pines , r. flint and p. coleman for theoretical discussions and p. w. anderson for bringing up the issue of harrison s theorem . this work was supported by the u.s . national science foundation under the grant no . dmr 07 - 06013 and by the u.s . department of energy , division of materials sciences under award no . defg02 - 07er46453 , through the frederick seitz materials research laboratory and the center for microanalysis of materials at the university of illinois at urbana - champaign . 122 coleman p 2007 heavy fermions : electrons at the edge of magnetism _ handbook of magnetism and advanced magnetic materials _ ed h kronmller h and s parkin vol 1 _ fundamentals and theory _ ( john wiley and sons , ltd ) thompson j d , nicklas m , bianchi a , movshovich r , llobet a , bao w , malinowski a , hundley m f , moreno n o , pagliuso p g , sarrao j l , nakatsuji s , fisk z , borth r , lengyel e , oeschler n , sparn g and steglich f 2003 _ physica _ b * 329 - 333 * 446 shishido h , settai r , aoki d , ikeda s , nakawaki h , nakamura n , iizuka t , inada y , sugiyama k , takeuchi t , kindo k , kobayashi t c , haga y , harima h , aoki y , namiki t , sato h and onuki y 2002 _ j. phys . jpn . _ * 71 * 162 zasadzinski j 2003 tunneling spectroscopy of conventional and unconventional superconductors _ the physics of superconductors _ ed k h bennemann and j b ketterson vol 1 _ conventional and high-@xmath14 superconductors _ ( berlin : springer - verlag ) annett j f , goldenfeld n d and leggett a j 1996 experimental constrains on the pairing state of the cuprate superconductors : an emerging consensus _ physical properties of high temperature superconductors _ ed d m ginsberg vol v ( singapore : world scientific )

we present solutions to local connectivity problems in matrix representations of the form^{n } ) \to a_{n,\varepsilon } \leftarrow c_{\varepsilon}(\mathbb{t}^{2})$ ] for any $ ] and any integer , where and where denotes the * soft torus*. we solve the connectivity problems by introducing the so called toroidal matrix links , which can be interpreted as normal contractive matrix analogies of free homotopies in differential algebraic topology . in order to deal with the locality constraints , we have combined some techniques introduced in this document with some techniques from matrix geometry , combinatorial optimization , classification and representation theory of c-algebras .

1605.06590

in this document we study the solvability of some local connectivity problems via constrained normal matrix homotopies in c@xmath5-representations of the form @xmath6 for a fixed but arbitrary integer @xmath7 and any integer @xmath2 . in particular we study local normal matrix homotopies which preserve commutativity and also satisfy some additional constraints , like being rectifiable or piecewise analytic . we build on some homotopic techniques introduced initially by bratteli , elliot , evans and kishimoto in @xcite and generalized by lin in @xcite and @xcite . we combine the homotopic techniques with some techniques introduced here and some other techniques from matrix geometry and noncommutative topology developed by loring @xcite , shulman @xcite , bhatia @xcite , chu @xcite , brockett @xcite , choi @xcite , effros @xcite , exel @xcite , eilers @xcite , elsner @xcite , pryde @xcite , mcintosh @xcite and ricker @xcite , to construct the so called _ toroidal matrix links _ , which we use to obtain the main theorems presented in section [ main_results ] , and which consist on local connectivity results in matrix representations of the form [ first_representation ] and also of the form @xmath8^{n } ) . \label{second_representation}\ ] ] toroidal matrix links can be interpreted as noncommutative analogies of free homotopies in algebraic topology and topological deformation theory , they are introduced in section [ matrix_varieties_and_links ] together with some other matrix geometrical objects . in [ c_0_technique ] we present a connectivity technique which provides us with very important imformation on the local uniform connectivity in matrix representations of the form @xmath9 . given @xmath10 , a function @xmath11 and two matrices @xmath12 in a set @xmath13 such that @xmath14 , by a * _ @xmath15-local matrix homotopy _ * between @xmath16 and @xmath17 , we mean a matrix path @xmath18,m_n)$ ] such that @xmath19 , @xmath20 , @xmath21 and @xmath22 for each @xmath23 $ ] . we write @xmath24 to denote that there is a @xmath25-local matrix homotopy betweeen @xmath16 and @xmath17 . the motivation and inspiration to study local normal matrix homotopies which preserve commutativity in c@xmath5-representations of the form [ first_representation ] and [ second_representation ] , came from mathematical physics @xcite and matrix approximation theory @xcite . the problems from mathematical physics which motivated this study are inverse spectral problems , which consist on finding for a certain set of matrices @xmath26 which _ approximately satisfy _ a set of polynomial constraints @xmath27 @xmath28 on @xmath29 nc - variables , a set of _ nearby matrices _ @xmath30 which approximate @xmath26 and exactly satisfy the constraints @xmath31 . the problems from matrix approximation theory that we considered for this study , are of the type that can be reduced to the study of the solvability conditions for approximate and exact joint diagonalization problems for @xmath29-tuples of normal matrix contractions . since the problems which motivated the research reported in this document can be restated in terms of the study local piecewise analytic connectivity in matrix representations of the form @xmath32 and @xmath33^{n})$ ] , we studied several variations of problems of the form . [ first_problem ] given @xmath34 , is there @xmath10 such that the following conditions hold ? for any integer @xmath2 , some prescribed sequence of linear compressions @xmath35 for some @xmath36 , and any two families of @xmath29 pairwise commuting normal contractions @xmath37 and @xmath38 in @xmath39 which satisfy the constraints @xmath40 , @xmath41 , there are two families of @xmath29 pairwise commuting normal contractions @xmath42 and @xmath43 in @xmath44 which satisfy the relations : @xmath45 , @xmath46 and @xmath47 , @xmath41 . moreover , there are @xmath29 peicewise analytic @xmath25-local homotopies of normal contractions @xmath48,m_{mn})$ ] between the corresponding pairs @xmath49 , @xmath50 in @xmath44 , which satisfy the relations @xmath51 , for each @xmath52 and each @xmath53 . by solving problem * p.*[first_problem ] we learned about the local connectivity of arbitrary @xmath54-close @xmath29-tuples of pairwise commuting normal contractions @xmath26 and @xmath55 @xmath56 in @xmath39 , which was the main motivation of the research reported here . we also obtained some results concerning to the geometric structure of the joint spectra ( in the sense of @xcite ) of the @xmath29-tuples . for a given @xmath10 , the study of the solvability conditions of problems of the form * p.*[first_problem ] provided us with geometric information about local deformations of particular representations of the form @xmath57 and @xmath58 , where @xmath59 are pairwise commuting unitary matrices such that @xmath60 . by local deformations we mean a family @xmath61}\subseteq m_n$ ] of abelian c@xmath5-algebras , with @xmath62 and where @xmath63,{\mathbb{u}(n)})$ ] are @xmath15-local matrix homotopies between @xmath64 and @xmath65 for some function @xmath11 . the main results are presented in [ main_results ] , in section [ main_results_piecewise_analytic ] we use toroidal matrix links to obtain some local piecewise analytic connectivity results which are non - uniform in dimension . in section [ main_lifted_results ] we derive an uniform approximate connectivity technique via matrix homotopy lifting and in section [ compressible_matrix_sets ] we present a connectivity lemma that can be used to derive some uniform connectivity results betweeen matrix representations of finite sets of universal algebraic contractions , some of the details of these constructions will be presented in @xcite . given two elements @xmath12 in a c@xmath5-algebra @xmath66 , we will write @xmath67 $ ] and @xmath68(y)}$ ] to denote the operations @xmath67:=xy - yx$ ] and @xmath68(y)}:={xyx^*}$ ] . given any c@xmath5-algebra @xmath66 and any element @xmath16 in @xmath69 , we will denote by @xmath70}$ ] the operation defined by the expression @xmath71}\\ \left ( \begin{array}{cccc } x_{11 } & x_{12 } & \cdots & x_{1n}\\ x_{21 } & x_{22 } & \cdots & x_{2n}\\ \vdots & \vdots & \ddots & \vdots \\ x_{n1 } & x_{n2 } & \cdots & x_{nn } \end{array } \right ) & \mapsto & \left ( \begin{array}{cccc } x_{11 } & 0 & \cdots & 0\\ 0 & x_{22 } & \cdots & 0\\ \vdots & \vdots & \ddots & \vdots \\ 0 & 0 & \cdots & x_{nn } \end{array } \right).\end{aligned}\ ] ] given a c@xmath5-algebra @xmath66 , we write @xmath72 , @xmath73 and @xmath74 to denote the sets of normal , hermitian and unitary elements in @xmath66 respectively . we will write @xmath75 , @xmath76 and @xmath77 instead of @xmath78 , @xmath79 and @xmath80 . a normal element @xmath81 in a c@xmath5-algebra @xmath66 is called a partial unitary if the element @xmath82 is an otrhogonal projection in @xmath66 , i.e. @xmath83 satisfies the relations @xmath84 , we denote by @xmath85 the set of partial unitaries in @xmath66 and we write @xmath86 instead of @xmath87 . we write @xmath88 , @xmath89 , @xmath90 and @xmath91 to denote the sets @xmath92 $ ] , @xmath93 $ ] , @xmath94 and @xmath95 . for some arbitrary matrix set @xmath96 and some arbitrary compact set @xmath97 , we will write @xmath98 to denote the subset of elements in @xmath99 described by the expression , @xmath100 for instance we can write @xmath101 to denote the set of nomal contractions . we will denote by @xmath102 the c@xmath5-algebra described by @xmath103 in this document we write @xmath104 to denote the identity matrix in @xmath39 . the symbol @xmath105 will be used to denote the diagonal matrices @xmath106}.\ ] ] we will write @xmath107 and @xmath108 to denote the unitary matrices defined by @xmath109}\end{aligned}\ ] ] and @xmath110 the unitary matrices @xmath107 and @xmath108 are related , by the equation @xmath111 where @xmath112 is the discrete fourier transform ( dft ) unitary matrix . given an abstract object ( group or c@xmath5-algebra ) @xmath66 we write @xmath113 to denote the operation consisting on taking the free product of @xmath29 copies of @xmath66 . given a linear mapping @xmath114 , with @xmath115 , and given a set @xmath96 , we say that @xmath116 * locally preserves * @xmath99 if there is @xmath117 such that @xmath118 , if in particular @xmath119 we say that @xmath116 * locally preserves normality*. the linear compression @xmath120 defined by @xmath121 locally preserves normality with respect to the set @xmath122 . the linear map @xmath123 with @xmath2 and @xmath124}$ ] , locally preserves commutativity with respect to the set @xmath125 . using the same notation as pryde in @xcite , let @xmath126 denote the clifford algebra over @xmath127 with generators @xmath128 and relations @xmath129 for @xmath130 and @xmath131 . then @xmath126 is an associative algebra of dimension @xmath132 . let @xmath133 denote the set @xmath134 . then the elements @xmath135 form a basis when @xmath136 and @xmath137 . elements of @xmath126 are denoted by @xmath138 where @xmath139 . under the inner product @xmath140 , @xmath126 becomes a hilbert space with orthonormal basis @xmath141 . the _ clifford operator _ of @xmath29 elements @xmath142 is the operator defined in @xmath143 by @xmath144 each element @xmath145 acts on elements @xmath146 by @xmath147 . so @xmath148 . by @xmath149 we will mean the operator norm of @xmath150 as an element of @xmath151 . as observed by elsner in @xcite we have that @xmath152 it is often convenient to have @xmath29-tuples ( or @xmath153-tuples ) of matrices with real spectra . for this purpose we use the following construction , initiated by mcintosh and pryde . if @xmath154 is a @xmath29-tuple of @xmath155 by @xmath155 matrices then we can always decompose @xmath156 in the form @xmath157 where the @xmath158 all have real spectra . we write @xmath159 and call @xmath160 a partition of @xmath161 . if the @xmath158 all commute we say that @xmath160 is a commuting partition , and if the @xmath158 are simultaneously triangularizable @xmath160 is a triangularizable partition . if the @xmath158 are all semisimple ( diagonalizable ) then @xmath160 is called a semisimple partition . we say that @xmath29 normal matrices @xmath142 are _ simultaneously diagonalizable _ if there is a unitary matrix @xmath162 such that @xmath163 is diagonal for each @xmath164 . in this case , for @xmath165 , let @xmath166 the @xmath167 element of @xmath168 , and set @xmath169 . the set @xmath170 is called the joint spectrum of @xmath26 . we will write @xmath171 to denote the @xmath172-component of @xmath173 , in other words we will have that @xmath174}.\ ] ] the following theorem was proved in mcintosh , pryde and ricker @xcite . [ pryde_joint_spectral_variation ] let @xmath154 and @xmath175 @xmath176 be @xmath29-tuples of commuting @xmath155 by @xmath155 normal matrices . there exists a permutation @xmath177 of the index set @xmath178 such that @xmath179 for all @xmath180 . in this theorem , @xmath181 is an explicit constant depending only on @xmath29 defined in ( * ? ? ? * ( 2.4 ) ) . the following lemma has been proved by h. lin in @xcite . [ lin_toroidal_links_0 ] for any @xmath34 and @xmath182 , there exists @xmath10 satisfying the following : suppose that @xmath66 is a unital c@xmath5-algebra and @xmath183 is a unitary such that @xmath184 contains an arc with length @xmath185 . suppose that @xmath186 with @xmath187 such that @xmath188 then there is a self - adjoint element @xmath189 such that @xmath190 , @xmath191 for all @xmath192 . if , furthermore , @xmath193 is a projection , we have @xmath194 the following lemma was proved by h. lin in @xcite using l.[lin_toroidal_links_0 ] , since for any integer @xmath2 and any @xmath195 , we will have that @xmath184 contains an arc with length at least @xmath196 . [ lin_toroidal_links ] let @xmath34 , @xmath2 be an integer and @xmath197 . there exists @xmath10 satisfying the following : for any finite set @xmath198 with @xmath199 for all @xmath200 , and a unitary @xmath201 such that @xmath202 there exists a continuous path of unitaries @xmath203 with @xmath204 and @xmath205 such that @xmath206 morover , @xmath207 ( * the obstruction @xmath208 . * ) [ loring_bott_index_definition ] given two unitaries in a @xmath209-simple real rank zero c@xmath5-algebra @xmath66 that almost commute , the obstruction @xmath208 is the bott element associated to the two unitaries as defined by loring in @xcite . it is defined whenever @xmath210 , where @xmath211 is a universal constant . it is defined as the @xmath212-class @xmath213- \left [ \left ( \begin{array}{cc } 1 & 0\\ 0 & 0 \end{array } \right ) \right],\ ] ] where @xmath214 is a self - adjoint element of @xmath215 of the form @xmath216 where @xmath217,@xmath218,@xmath219 are certain universal real - valued continuous functions on @xmath90 . for details on the subject of @xmath116-theory for @xmath220-algebras the reader is referred to @xcite . as observed by bratteli , elliot , evans and kishimoto in @xcite , given a pair @xmath221 we have that the obstruction @xmath208 needs to vanish in order to be able to solve the problem @xmath222 by deforming @xmath223 to @xmath224 continuously in @xmath74 , when @xmath225 . let us denote by @xmath226 a universal separable hilbert space , by @xmath227 the c@xmath5-algebra of bounded operators on @xmath226 , and for any given @xmath228 let us denote by @xmath229 the closed @xmath230-ball in @xmath99 defined by @xmath231 . given some @xmath232 and a set @xmath233 of normed polynomial relations on the @xmath29-set @xmath234 of nc - variables , we will call the set @xmath235}$ ] described by @xmath236}:=\{{x_1,\ldots , x_n}|\mathcal{r}({x_1,\ldots , x_n})\ } \label{nc_set_def}\ ] ] with @xmath237 , a noncommutative semialgebraic set . as an example of normed @xmath238-polynomial relations we can consider the set @xmath239 . given a nc - semialgebraic set @xmath235}$ ] , we will use the symbol @xmath240}$ ] to denote the universal c@xmath5-algebra @xmath241}:={c^*\left\langle { x_1,\ldots , x_n } \left | \mathcal{r}({x_1,\ldots , x_n } ) \right . \right\rangle } , \label{environment_c_star_def}\ ] ] which we call the environment c@xmath5-algebra of @xmath235}$ ] . for details on universal c@xmath5-algebras described in terms of generators and relations the reader is referred to @xcite . [ matrix_variety ] given @xmath242 , a system of @xmath243 polynomials @xmath244 in @xmath29 nc - variables @xmath245 and a real number @xmath246 , a particular matrix representation of the noncommutative semialgebraic set @xmath247 described by @xmath248 will be called a * @xmath249-semialgebraic matrix variety * ( @xmath249-smv ) , if @xmath250 we can refer to the set as a * matrix variety*. as a first example , we will have that the set @xmath251 @xmath252 is a matrix variety defined by the set with one @xmath238-polynomial relation @xmath253 . if for some @xmath10 , we set now @xmath254\|\leq \delta\}$ ] , the set @xmath255 is a matrix semialgebraic variety defined by the set with one normed @xmath238-polynomial relation @xmath256 . other example of a matrix semialgebraic variety , that has been useful to understand the geometric nature of the problems solved in this document , is described by the matrix set @xmath257 , defined for some given @xmath258 and any two normal contractions @xmath16 and @xmath17 in @xmath39 , by the expression @xmath259\|=0\\ \|z - y\|\leq \delta \end{array } \right . \right\}.\ ] ] a finsler manifold is a pair @xmath260 where @xmath261 is a manifold and @xmath262 is a function ( called a finsler norm ) such that * @xmath263 is smooth on @xmath264 , * @xmath265 with equality if and only if @xmath266 , * @xmath267 = @xmath268 for all @xmath269 , * @xmath270 for all @xmath271 at the same tangent space with @xmath272 . given a finsler manifold @xmath260 , the length of any rectifiable curve @xmath273 \to m$ ] is given by the length functional @xmath274 = \int_a^b f(\gamma(t),\partial_t \gamma(t))\,dt,\ ] ] where @xmath275 is the finsler norm on each tangent space @xmath276 . the pair @xmath277 is a finsler manifold , where @xmath278 denotes the set of normal matrices @xmath278 ( of any size ) and @xmath279 denotes the operator norm . given a piecewise-@xmath280 matrix path @xmath281\to \mathcal{n}$ ] , we define its curvature @xmath282 $ ] to be @xmath283:=\frac{1}{\|\partial_t\gamma(t)\|}\left\|\partial_t \left(\frac{\partial_t\gamma(t)}{\|\partial_t\gamma(t)\|}\right)\right\|.\ ] ] given @xmath2 , a mapping @xmath284 @xmath285 will be called a matrix flow in this document . if we have in addition that @xmath286 for every @xmath287 , we say that the matrix flow is isospectral . given two matrices @xmath16 and @xmath17 in @xmath39 and a matrix flow @xmath288 such that @xmath289 and @xmath290 , we say that the corresponding path @xmath291 is a solvent path for the interpolation problem @xmath292 . given two matrix paths @xmath293,m_n)$ ] we write @xmath294 to denote the concatenation of @xmath161 and @xmath295 , which is the matrix path defined in terms of @xmath161 and @xmath295 by the expression , @xmath296 ( * @xmath297 * ) given a matrix path @xmath298 in @xmath39 we will write @xmath299 to denote the length of @xmath298 with respect to the operator norm which is defined by the expression @xmath300 where the supremum is taken over all partitions of @xmath88 as @xmath301 . if the function @xmath302 is a piecewise @xmath303 function , then @xmath304 ( * @xmath279-flatness * ) a set @xmath305 of @xmath39 is said to be @xmath279-flat if any two points @xmath306 can be connected by a path @xmath307 such that @xmath308 . * * [ matrix_links ] given any two normal contractions @xmath12 in @xmath39 , a toroidal matrix link is any piecewise analytic normal path @xmath309 $ ] induced by a locally normal piecewise analytic matrix flow @xmath310 with @xmath311 , together with a locally normal compression @xmath312 with relative lifting map @xmath313 , which satisfy the interpolating conditions @xmath314=x$ ] and @xmath315=y$ ] together with the constraints @xmath316\|\leq 1 $ ] for each @xmath192 . in the particular case where @xmath317=0 $ ] for each @xmath192 , whenever @xmath67=0 $ ] , we call @xmath318 a toral matrix link . the _ * curved nature * _ of the matrix varieties ( _ * as finsler sub - manifolds of @xmath278 * _ ) whose local connectivity is studied in this document , induces an obstruction to local connectivity via entirely _ * flat * _ toroidal matrix links in general . the toroidal matrix links @xmath319,\mathcal{n})$ ] we have used to solve the connectivity problems which motivated this study satisfy the constraint @xmath320 \leq \frac{2}{\ell_{\|\cdot\|}(t ) } , \:\ : \forall t\in \mathbf{t}.\ ] ] given some fixed but arbitrary @xmath321 , using the operation @xmath322 one can define the mapping @xmath323 , determined by the expression . @xmath324(x)}\label{matrix_torus_definitions_2 } \\ ( w , x)&\mapsto&\{({\mathrm{diag}_{n}\left [ wxw^ * \right]})_{k , k}\}_{1\leq k\leq n } \label{matrix_torus_definitions_3}\end{aligned}\ ] ] it is clear that @xmath325(x ) } \right]}={\mathrm{diag}_{n}\left [ wxw^ * \right]}$ ] and that @xmath326(x ) } \right]}={\mathrm{diag}_{n}\left [ x \right]}$ ] , because of this when @xmath327 we will write @xmath328 instead of @xmath329(x)}$ ] . given a matrix flow @xmath330 , one can identify @xmath161 with the set of flow lines in @xmath331 determined by @xmath332 . the geometric picture determined by the mapping cylinder @xmath333 will be called the embedded matrix mapping cylinder relative to the flow @xmath161 . we can think of the embedded matrix mapping cylinder in topological terms as a deformation described by the expression @xmath334 , which is defined as @xmath335:= \frac{(z_1\times \mathbb{i } ) \sqcup z_2}{z_1 \rightsquigarrow_{x_1 } z_2 } , \label{matrix_mapping_cylinder}\ ] ] where @xmath336 and @xmath337 are some prescribed matrix varieties such that @xmath338 and @xmath339 . [ m3_graphical_example ] let us set @xmath340 where @xmath341 . for some prescribed @xmath342 , we can obtain a graphical example of a particular geometric picture of the computation of the embedded matrix mapping cylinder relative to the interpolating flow @xmath343 which solves the problem @xmath344 . let us set @xmath345=0\},\\ z_2&:=&\{z\in{\mathbb{u}(3)}|[w_3\hat{u}_3w_3^*,z]=0\}.\end{aligned}\ ] ] using projective methods , we can trace specific flow lines along the matrix flows corresponding to the dynamical deformation @xmath346 $ ] , which solve the interpolation problem @xmath344 . a particular ( approximate ) geometric picture of the matrix deformation induced by the toral matrix link @xmath347 in @xmath348 , projected in @xmath331 for each @xmath192 via @xmath349 is presented in figures f.[mapping_cylinder_1]-f.[mapping_cylinder_3 ] . } ( \hat{u}_3)$ ] in @xmath348 . ] in @xmath348 . ] in @xmath348 . ] alternative methods to trace particular flow lines on mapping cylinders can be otained using matrix homotopies , this can be done using similar methods to the ones implemented in @xcite . given a mat - rix algebra @xmath66 @xmath350 , a universal c@xmath5-algebra @xmath351 for which there is a matrix representation @xmath352 such that @xmath353 , will be called an * environment algebra * for @xmath66 . let us consider the universal c@xmath5-algebras @xmath354 , @xmath355 , @xmath356 , @xmath357 , @xmath358 and @xmath359 , defined in terms of generators and relations by the expressions . @xmath360 @xmath361 @xmath362 @xmath363 @xmath364 @xmath365 let us consider now a local matrix representation result that we will use later in the construction of particular representation schemes . [ c_star_f2_inf_surj ] for every integer @xmath2 , there are @xmath366 such that the diagram @xmath367 \ar@{->>}[r ] & c^*\braket{(\mathbb{z}/n)^{\ast 2 } } \ar@{->>}[r ] & c^*_n(u_n , v_n ) \ar@{=}[d]\\ c^*\braket{\mathbb{z}/n\ast \mathbb{z}/2 } \ar@{->>}[r ] & c^*_n(s_2,v_n ) \ar@{=}[r ] & m_n } \ ] ] commutes , where @xmath368 , @xmath369 and @xmath370 are unitary elements in @xmath39 . since we have that @xmath371 , by universality of the c@xmath5-representa - tions @xmath372 and by the structural properties of @xmath39 , it is enough to find for any @xmath373 , up to unitary congruence in @xmath39 , three unitaries @xmath374 such that @xmath375 and @xmath376 , this can be done by taking for any @xmath377 the orthogonal projection @xmath378}\in\mathbb{h}(n)$ ] and the matrix @xmath379 , setting @xmath380 and @xmath381 for @xmath382 and @xmath383 for @xmath384 , by functional calculus and direct computations it is easy to verify that @xmath374 for every @xmath377 , and that @xmath385 , it is also easy to verify that the system of matrix units @xmath386 and @xmath369 can be expressed as words in @xmath387 for every @xmath377 , it is also clear that @xmath388 and hence , @xmath389 can be written as linear combinations of words in @xmath390 , we will then have that @xmath391 and @xmath392 by the universal properties of @xmath393 and @xmath394 respectively , since it can be easily verified that @xmath395 from these facts and the universal property of @xmath396 , the result follows . [ pre_localization ] it can be seen that for any matrix c@xmath5-subalgebra @xmath397 , there is @xmath10 such that both @xmath398 and @xmath357 are environment algebras of @xmath66 . it can also be seen that for any abelian c@xmath5-subalgebra @xmath399 , @xmath355 is an environment algebra of @xmath400 . [ nc_controlled_topology ] given @xmath10 , a function @xmath401 , a finite set of functions @xmath402 and two unitary matrices @xmath403 such that @xmath225 , we say that the set @xmath263 is @xmath54-controlled by @xmath404 $ ] if the diagram , @xmath405 \ar[d]_{\mathrm{ad}[v ] } & \{u\ } \ar[d]_{\mathrm{ad}[v ] } \ar[l]^{\imath } \ar[rd]^{f}_{\approx_{\varepsilon(\delta ) } } & \\ & c^*(vuv^ * ) \ar[lu ] & \{vuv^*\ } \ar[l]^{\imath } \ar[r]_{f } & { \mathcal{n}(n)}({{\mathbb{d}^{2 } } } ) } \ ] ] commutes up to an error @xmath15 for each @xmath406 . the @xmath220-homomorphism @xmath407 allows us to see that the soft torus @xmath357 provides an environment algebra for any @xmath54-controlled set of matrix functions . [ joint_spectral_variation_inequality_2 ] given @xmath34 there is @xmath408 such that , for any @xmath409 families of @xmath29 pairwise commuting normal matrices @xmath410 and @xmath411 which satisfy the constraints @xmath412 for each @xmath41 , there is a @xmath220-homomorphism @xmath413 such that @xmath414 , @xmath415=0 $ ] and @xmath416 , for each @xmath41 . by changing basis if necessary , we can assume that @xmath411 are diagonal matrices . from t.[pryde_joint_spectral_variation ] we will have that there is a permutation @xmath177 of the index set @xmath178 such that for each @xmath165 we have that @xmath417 using [ clifford_bound ] and as a consequence of [ joint_spectral_variation_inequality ] we can find a permutation matrix @xmath418 such that @xmath419}\mathcal{t}-{\mathrm{diag}_{}\left [ \lambda(y_j ) \right]}\|&\leq & e_{n,0}\|{\mathrm{cliff}(x_1-y_1,\ldots , x_n - y_n)}\|\nonumber \\ & \leq & e_{n,0}n\delta , \:\ : 1\leq j\leq n. \label{existence_of_nu_reaching_joint_diagonalizer } \end{aligned}\ ] ] let us set @xmath420 . for the matrices @xmath410 there is a unitary joint diagonalizer @xmath421 such that @xmath422}w^*=x_j$ ] , @xmath41 , @xmath423}w^*-\mathcal{t}^*{\mathrm{diag}_{}\left [ \lambda(x_j ) \right]}\mathcal{t}\| & \leq&\|w{\mathrm{diag}_{}\left [ \lambda(x_j ) \right]}w^*-y_j\|\nonumber \\ & & + \|y_j-\mathcal{t}^*{\mathrm{diag}_{}\left [ \lambda(x_j ) \right]}\mathcal{t}\| \nonumber\\ & \leq&(1+c_n)\|x_j - y_j\|\leq ( 1+c_n)\delta . \label{controlled_jmp_inequalities } \end{aligned}\ ] ] if we set @xmath424 and @xmath425 , we will have that by [ existence_of_nu_reaching_joint_diagonalizer ] and [ controlled_jmp_inequalities ] the inner @xmath220-automorphism @xmath426 $ ] satisfies the constraints in the statement of this lemma , and we are done . [ isospectral_approximants_remark ] the @xmath220-automorphism @xmath413 from l.[joint_spectral_variation_inequality_2 ] is called an isospectral approximant for the two @xmath29-tuples @xmath410 and @xmath411 . if @xmath427 $ ] for some @xmath321 , then we will have that its inverse @xmath428 will be given by the expression @xmath429 $ ] . the constant @xmath430 in the proof of l.[joint_spectral_variation_inequality_2 ] depends only on the number @xmath29 of matrices in each family . it does not depend on the matrix size . in this section we will present some piecewise analytic local connectivity results in matrix representations of the form @xmath32 and @xmath431 . [ local_connection_of_n_tuples_of_normal_contractions ] given @xmath34 and any @xmath373 , there is @xmath10 such that , for any @xmath153 normal contractions @xmath410 and @xmath432 @xmath433 in @xmath39 which satisfy the relations @xmath434=[y_j , y_k]=0 , \:\:\ : 1\leq j , k\leq n,\\ \|x_j - y_j\|\leq \delta , \:\:\ : 1\leq j\leq n , \end{array } \right.\ ] ] there exist @xmath29 toral matrix links @xmath435 in @xmath39 , which solve the problems @xmath436 and satisfy the constraints @xmath437 for each @xmath52 and each @xmath438 . moreover , @xmath439 . by l.[lin_toroidal_links_0 ] , l.[lin_toroidal_links ] and l.[joint_spectral_variation_inequality_2 ] we will have that given @xmath34 , there are @xmath440 and an isospectral approximant @xmath427 $ ] ( with @xmath441 ) for @xmath410 and @xmath411 such that , @xmath442 and @xmath415=0 $ ] for each @xmath41 , we will also have that there is a unitary path @xmath443 which is defined by the expression @xmath444 for each @xmath192 , where @xmath445 is a hermitian matrix such that @xmath446 and @xmath447\|\leq \varepsilon/2 $ ] for each @xmath41 , and that is defined by @xmath448 , for some function @xmath449 $ ] , and where @xmath450 , with @xmath451 chosen in such a way that @xmath452 contains an arc of length @xmath185 ( with @xmath453 ) . moreover , we can choose @xmath54 and @xmath454 in such a way that the path @xmath455 satisfies the inequalities @xmath456\|\leq \varepsilon/2 $ ] for each @xmath23 $ ] and each @xmath41 . it can be seen that the paths @xmath457(x_j)}$ ] will solve the problem @xmath458 for each @xmath41 . let us set @xmath459 , we can now construct @xmath29 toroidal matrix links of the form @xmath460 which solve the problems @xmath461 , locally preserve normality and commutativity and satisfy the @xmath279-distance constraints @xmath462 together with the @xmath279-length constraints @xmath463(x_j)\|dt+\|\psi({x}_j)-y_j\|\\ & = & \|[h_w,\psi({x}_j)]\|+\|\psi({x}_j)-y_j\|\\ & \leq&\frac{\varepsilon}{2}+\nu\leq \varepsilon , \end{aligned}\ ] ] which hold whenever @xmath412 , @xmath41 , and we are done . it can be noticed that the solvent matrix links @xmath464 @xmath465 whose existence is stated in t.[local_connection_of_n_tuples_of_normal_contractions ] are factored in the form @xmath466 , we call @xmath467 and @xmath468 the _ * curved * _ and _ * flat * _ factors of @xmath469 respectively . we will derive now , some corollaries of the proof of t.[local_connection_of_n_tuples_of_normal_contractions ] . [ local_hermitian_interpolants ] given @xmath34 and any integer @xmath2 , there is @xmath10 such that , for any @xmath153 hermitian contractions @xmath410 and @xmath411 in @xmath39 which satisfy the relations @xmath434=[y_j , y_k]=0 , \:\:\ : 1\leq j , k\leq n,\\ \|x_j - y_j\|\leq \delta , \:\:\ : 1\leq j\leq n , \end{array } \right.\ ] ] there exist @xmath29 toral matrix links @xmath435 in @xmath39 , which solve the problems @xmath436 and satisfy the constraints @xmath470 for each @xmath52 and each @xmath438 . moreover , @xmath439 . since for any @xmath471 , any pair of hermitian matrices @xmath472 and any partial unitary @xmath473 , we have that @xmath474 and @xmath475 are also in @xmath76 , the result follows as a consequence of l.[joint_spectral_variation_inequality_2 ] and t.[local_connection_of_n_tuples_of_normal_contractions ] . [ local_connectivity_n_tuples_commuting_unitaries ] given any @xmath246 and any integer @xmath2 , there is @xmath258 such that given any @xmath153 unitary matrices @xmath476 , @xmath477 in @xmath478 which satisfy the relations @xmath479=[v_{j},v_{k}]=0,\\ \|u_{k}-v_{k}\|\leq \delta , \end{array } \right . \end{aligned}\ ] ] for each @xmath480 , there are toral matrix links @xmath481 in @xmath478 which solve the interpolation problems @xmath482 and also satisfy the relations @xmath483 for each @xmath192 and each @xmath52 . moreover , @xmath484 , @xmath41 . since for any @xmath220-automorphisms @xmath413 we have that @xmath485 , and since any two commuting unitaries @xmath486 and @xmath487 can connected by a flat unitary path @xmath488 , for @xmath489 . we will have that the result can be derived using a similar argument to the one implemented in the proof of t.[local_connection_of_n_tuples_of_normal_contractions ] . let us denote by @xmath490 the matrix compression @xmath491 defined by the mapping @xmath492 let us write @xmath493 to denote the @xmath220-homomorphism defined by the expression @xmath494 . [ std_dilations ] given a @xmath220-automorphism @xmath495 $ ] ( with @xmath321 ) in @xmath39 , we will denote by @xmath496}$ ] the @xmath220-automorphism in @xmath497 defined by the expression @xmath496}:=\mathrm{ad}[{\mathbb{1}}_2\otimes w]=\mathrm{ad}[w\oplus w]$ ] . we call @xmath496}$ ] a standard dilation of @xmath413 . [ z2_dilations ] given a @xmath220-automorphism @xmath495 $ ] ( with @xmath321 ) in @xmath39 , we will denote by @xmath498}$ ] the @xmath220-automorphism in @xmath497 defined by the expression @xmath498}:=\mathrm{ad}[(\sigma_2\otimes { \mathbb{1}}_n)(w^*\oplus w)]$ ] . we call @xmath498}$ ] a @xmath499-dilation of @xmath413 . [ local_matching_remark ] it can be seen that @xmath500 for any @xmath501 , it can also be seen that @xmath502}(\imath_2(x)))=\kappa(\psi^{[s]}(\imath_2(x)))$ ] . [ lifted_connectivity ] [ lifted_local_connection_of_n_tuples_of_normal_contractions ] given @xmath34 , there is @xmath10 such that , for any @xmath153 normal contractions @xmath410 and @xmath411 in @xmath39 which satisfy the relations @xmath434=[y_j , y_k]=0 , \:\:\ : 1\leq j , k\leq n,\\ \|x_j - y_j\|\leq \delta , \:\:\ : 1\leq j\leq n , \end{array } \right.\ ] ] there is a @xmath220-homomorphism @xmath503 and @xmath29 toral matrix links @xmath504 @xmath505 in @xmath506 , which solve the problems @xmath507 and satisfy the constraints @xmath508 for each @xmath52 and each @xmath438 . moreover , @xmath509 . by l.[joint_spectral_variation_inequality_2 ] we will have that given @xmath34 , there are @xmath510 and an isospectral approximant @xmath427 $ ] ( with @xmath441 ) for @xmath410 and @xmath411 such that , @xmath442 . by setting @xmath511}\circ \imath_2\circ \psi$ ] , by d.[z2_dilations ] , d.[std_dilations ] and r.[local_matching_remark ] it can be seen that @xmath503 is a @xmath220-homomorphism such that @xmath512 , for each @xmath41 . since @xmath513}:=\mathrm{ad}[\hat{w}_s]$ ] with @xmath514 and since @xmath515 , we will have that @xmath516 can be represented as @xmath517 for any @xmath2 . if we set @xmath518}(\imath_2({x}_j))$ ] , @xmath41 , we also have that there is a unitary path @xmath519 with @xmath520 , which satisfies the conditions @xmath521 , @xmath522 , together with the normed estimates @xmath523 for each @xmath41 and each @xmath53 . moreover , for each @xmath41 we have that the paths @xmath524(\tilde{x}_j)}$ ] satisfy the normed estimates @xmath525(\tilde{x}_j)})\|dt,\\ & = & \frac{\pi}{2}\|\hat{w}_s\tilde{x}_j-\tilde{x}_j\hat{w}_s\| \leq \nu.\end{aligned}\ ] ] for each @xmath41 , we can now use the flat paths @xmath526 together with the previously described curved paths @xmath467 to construct the solvent toral matrix links @xmath527,m_{2n})$ ] we are looking for , and which can be defined by @xmath460 for each @xmath41 , and we are done . it can be seen that by using the technique implemented in the proof of t.[lifted_local_connection_of_n_tuples_of_normal_contractions ] one can obtain lifted versions of c.[local_connectivity_n_tuples_commuting_unitaries ] and c.[local_hermitian_interpolants ] . [ detection_methods ] as a consequence of t.[lifted_local_connection_of_n_tuples_of_normal_contractions ] we can derive simple detection methods to identify families of pairwise commuting matrices in @xmath39 that can be connected uniformly via piecewise analytic toral matrix links . the existence of these detection methods raises some interesting questions for further studies . we can interpret t.[lifted_local_connection_of_n_tuples_of_normal_contractions ] as an existence theorem of solutions to lifted connectivity problems defined on matrix representations of the form @xmath528 & c^*(\hat{u}_s,\hat{v } ) \ar[r ] & m_{2n } \ar[d ] \\ c^*\langle\mathbb{f}_2\rangle \ar[r ] \ar[ru ] & c_{\delta}(\mathbb{t}^{2 } ) \ar[r ] & c^*(u , v ) \ar[u ] \ar[r ] & m_n } , \ ] ] with @xmath529 and @xmath530 . some further applications of t.[lifted_local_connection_of_n_tuples_of_normal_contractions ] to approximation of matrix words and norm behavior will be presented in @xcite . using t.[lifted_local_connection_of_n_tuples_of_normal_contractions ] we can solve all connectivity problems ( together with their softened versions ) in @xmath39 that can be reduced to connectiviy problems of the form @xmath531 in @xmath101 , with @xmath532 and @xmath533 . for each @xmath534 $ ] , we can use the previously described symmetries and @xmath535 to interpret @xmath536 as matrix analogies of the _ klein bottle_. by a _ softened matrix klein bottle _ we mean that the symmetries are softened , in particular we can consider the connectiviy problems @xmath531 and @xmath537 in @xmath101 subject to the normed constraints @xmath538 , @xmath539 and @xmath533 . the details regarding to the solvability of these local connectivity problems will be addressed in future communications . the technique presented in this section can be used to solve local connectivity problems in matrix representations of the form @xmath541 uniformly via @xmath540-unitary paths . suppose @xmath542 and @xmath543 are unitary matrices in @xmath544 for @xmath545 and @xmath546 and we define @xmath547 and @xmath548 for @xmath545 or @xmath546 the @xmath549-algebra generated by @xmath542 and @xmath543 is abelian , so select a masa @xmath550 in each case . let @xmath551,\mathbf{m}_{n}(\mathbb{c})\right)\right|x(0)\in c_{0}\mbox { and } x(1)\in c_{1}\right\ } .\ ] ] [ lem : tsr_one]the @xmath549-algebra @xmath552 has stable rank one . starting with @xmath161 continuous with @xmath553 in @xmath554 at the endpoints , we can adjust this by a small amount , leaving the endpoints in @xmath554 , to get @xmath161 piece - wise linear , with the endpoints of every linear segment having no spectral multiplicity and being invertible . using kato s theory of analytic paths , we can get a piece - wise continuous unitary @xmath542 and piece - wise analytic scalar paths @xmath555 so that the new path @xmath556 satisfies @xmath557u_{t}^{*}.\ ] ] there may be finitely may places where @xmath558 is not invertible . these places will be in the interior of the segment so in an open interval where @xmath542 is continuous . a small deformation of some of the @xmath559 will take the path through invertibles . we have not moved the endpoints in the second adjustment so the constructed element is in @xmath552 and close to @xmath161 . the endpoint - restriction map @xmath560 induces an injection on @xmath561 . the kernel of @xmath562 is @xmath563,\mathbf{m}_{n}(\mathbb{c})\right)$ ] which has trivial @xmath561-group . so this result follows from the exactness of the usual six - term sequence in @xmath116-theory . [ lem : bott_trivial ] given unitaries @xmath486 and @xmath487 in @xmath552 , with @xmath564\right\vert \nu_0 $ ] as if d.[loring_bott_index_definition ] ( so the bott index makes sense ) , @xmath565 is the trivial element of @xmath566 . by the previous lemma , we need only calculate @xmath567 . these unitaries are in a commutative @xmath549-algebra so they have trivial bott index . [ c_0_unitary_interpolants ] given @xmath568 , there exists @xmath10 so that for all @xmath155 , given unitary matrices @xmath569 @xmath570 @xmath571 @xmath572 in @xmath544 with @xmath573 , @xmath574 , @xmath575 and @xmath576 , then there exists continuous paths @xmath542 and @xmath543 between the given pairs of unitaries with each @xmath542 and @xmath543 unitary , and with @xmath577 , @xmath578 and @xmath579 for all @xmath580 . the paths @xmath542 and @xmath543 defined in equations [ eq : u_t ] and [ eq : v_t ] will be almost commuting unitary elements of @xmath552 . by lemma [ lem : tsr_one ] we may apply ( * ? ? ? * theorem 8.1.1 ) regarding approximating in @xmath552 by commuting unitaries . lemma [ lem : bott_trivial ] tells us there is no invariant to worry about , so we can find @xmath581 and @xmath582 close of @xmath542 and @xmath543 that are commuting continuous paths of unitaries with @xmath581 and @xmath582 in @xmath554 for @xmath583 . the unitary elements in the commutative @xmath554 are locally connected , so we can find a short path from @xmath584 and @xmath585 to @xmath586 and @xmath587 , and likewise at the other end . concatenating , we get a paths of commuting unitary matrices from @xmath584 and @xmath585 to @xmath588 and @xmath572 so that at every point we are close to some pair @xmath589 . these then are all close to @xmath584 and @xmath585 . by combining t.[lifted_local_connection_of_n_tuples_of_normal_contractions ] , c.[local_connectivity_n_tuples_commuting_unitaries ] and t.[c_0_unitary_interpolants ] it can be seen that . [ analytic_approximants ] given @xmath568 , there exists @xmath10 so that for all @xmath155 , given unitary matrices @xmath569 @xmath570 @xmath571 @xmath572 in @xmath544 with @xmath573 , @xmath574 , @xmath575 and @xmath576 , there exist continuous ( interpolants ) paths @xmath542 and @xmath543 in @xmath497 which solve the problems @xmath590 and @xmath591 with each @xmath542 and @xmath543 unitary , and with @xmath577 , @xmath592 and @xmath593 for all @xmath580 . there are also a c@xmath5-homomorphism @xmath594 such that @xmath595 and two piecewise analytic unitary pairwise commuting paths @xmath596,m_{2n})$ ] which solve the problems @xmath597 , @xmath598 with @xmath599 for each @xmath53 . moreover , @xmath600 and @xmath601 . given @xmath602 , we can now consider an alternative approach to the local connectivity problem involving two @xmath29-sets of pairwise commuting normal matrix contractions @xmath26 and @xmath603 such that @xmath604 for each @xmath41 . the approach that we will consider in this section consists of considering the existence of a normal contraction @xmath605 such that @xmath606 , and which also satisfies the constraint @xmath607 for some @xmath41 . a matrix @xmath605 which satisfies the previous conditions will be called a _ * nearby generator * _ for @xmath26 , it can be seen that for any @xmath608 one can find a flat analytic path @xmath609,{\mathcal{m}_\infty})$ ] that performs the deformation @xmath610 , where @xmath605 is a nearby generator for @xmath26 . given any joint isospectral approximant @xmath413 with respect to the families normal contractions described in the previous paragraph , along the lines of the program that we have used to derive the connectivity results t.[local_connection_of_n_tuples_of_normal_contractions ] and t.[lifted_local_connection_of_n_tuples_of_normal_contractions ] , we can use l.[joint_spectral_variation_inequality_2 ] to find a c@xmath5-automorphism which solves the extension problem described by the diagram , @xmath611^{\hat{\psi}}\\ c^*({x_1,\ldots , x_n } ) \ar@{^{(}->}[ru ] \ar[r]_{\psi } & c^*({y_1,\ldots , y_n } ) } \label{njc_lifting}\ ] ] and satisfies the relations @xmath612 for each @xmath41 together with the normed constraints @xmath613 we refer to the c@xmath5-automorphism @xmath614 in [ njc_lifting ] as a * compression * of @xmath413 or a * compressive joint isospectral approximant * ( * cjia * ) for the @xmath29-sets of normal contractions . let us now consider a special type of inner c@xmath5-automorphisms that can be described as follows . given @xmath602 and two @xmath29-sets of pairwise commuting normal contractions @xmath26 and @xmath603 in @xmath102 such that @xmath40 , @xmath41 , a joint isospectral approximant @xmath413 of the @xmath29-sets is said to be * uniformly compressible * if there are a nearby generator @xmath605 for @xmath26 , a compression @xmath615 $ ] ( with @xmath616 ) of @xmath413 and a unitary @xmath617 such that @xmath618 . we refer to the @xmath153 normal contractions @xmath26 and @xmath603 for which there exists a uniformly compressible jia ( * ucjia * ) as uniformly jointly compressible ( * ujc * ) . [ ujc_connectivity ] given @xmath34 , there is @xmath10 , such that for any two @xmath29-sets of * ujc * pairwise commuting normal contractions @xmath26 and @xmath603 in @xmath102 such that @xmath40 for each @xmath41 , we will have that there are @xmath29 toral matrix links @xmath48,{\mathcal{m}_\infty})$ ] that solve the interpolation problem @xmath619 , for each @xmath41 . since the @xmath29-sets of pairwise commuting normal contractions @xmath26 and @xmath603 are * ujc * , we have that given @xmath620 , there are a normal contraction @xmath621 which commutes with each @xmath156 together with a * ucjia * @xmath622 $ ] for some @xmath616 and a unitary @xmath617 such that @xmath623 let us set @xmath624 , as a consequence of the inequality [ compresion_inequality_1 ] we will have that there is a hermitian matrix @xmath625 in @xmath102 such that @xmath626 . by using [ compresion_inequality_1 ] again , it can be seen that we can now use the curved paths @xmath627(x_j)}$ ] to solve the problems @xmath628 , and then we can solve the problems @xmath629 using the flat paths @xmath630 . we can construct the solvent toral matrix links by setting @xmath631 for each @xmath41 . this completes the proof . the detection matrix representations of universal c@xmath5-algebras that can be connected uniformly via piecewise analytic paths induces interesting problems which are topological / k - theoretical and computational in nature . motivated by the @xmath540-connectivity technique , we consider that the use of t.[local_connection_of_n_tuples_of_normal_contractions ] , c.[local_connectivity_n_tuples_commuting_unitaries ] and t.[lifted_local_connection_of_n_tuples_of_normal_contractions ] and l.[ujc_connectivity ] to study local matrix connectivity in c@xmath5-representations of the form @xmath632 \ar[r ] & m_{2n } & m_{n } \ar[l ] & c_{\varepsilon}({{\mathbb{t}^{2 } } } ) \ar[l ] \ar@/^1pc/[ll ] } \ ] ] will present interesting challenges and questions that will be the subject of future study . in particular we are interested in the application of t.[lifted_local_connection_of_n_tuples_of_normal_contractions ] , c.[local_connectivity_n_tuples_commuting_unitaries ] and l.[ujc_connectivity ] to the study of the question . is @xmath633 rfd ? ( this is equivalent to * connes s embedding problem*. ) a better understanding of the geometric and approximate combinatorial nature of toroidal matrix links would provide a mutually benefitial interaction between matrix flows in the sense of brockett @xcite and chu @xcite , topologically controlled linear algebra in the sense of freedman and press @xcite and matrix geometric deformations in the sense of loring @xcite . this also may provide some novel generic numerical methods to study and compute normal matrix compressions , sparse representations and dimensionality reduction of large scale matrices . using a similar approach we plan to use t.[lifted_local_connection_of_n_tuples_of_normal_contractions ] and l.[ujc_connectivity ] to answer some questions in topologically controlled linear algebra in the sense of @xcite , raised by m. h. freedman . the construction and generalization of detection methods like the ones mentioned in the remark r.[detection_methods ] of theorem t.[lifted_local_connection_of_n_tuples_of_normal_contractions ] together with their implications on inverse spectral problems , will be the subject of future studies . in particular we will use toral matrix links to study the local deformation properties of matrix representations of the form @xmath634 ( where @xmath635 denotes the standard flip ) via _ softened matrix klein bottles_. these problems are related to spectral decomposition problems with _ spectral symmetry _ in quantum theory and to _ deformation theory _ for c@xmath5-algebras in the sense of loring @xcite . we will also use toroidal matrix links to study the local connectivity of some soft group c@xmath5-algebras in the sense of farsi @xcite . some generalizations of t.[lifted_local_connection_of_n_tuples_of_normal_contractions ] and particular applications of l.[ujc_connectivity ] to the study of matrix equations on words will also be the subject of future study . in particular , the combination of toroidal matrix links with some matrix lifting techniques along the same lines of the proof of t.[lifted_local_connection_of_n_tuples_of_normal_contractions ] combined with l.[ujc_connectivity ] , seem also promising on the solvability of some conjectures studied numerically on @xcite . both authors are very grateful with the erwin schrdinger institute for mathematical physics of the university of vienna , for the outstanding hospitality during our visit to participate in the research program on topological phases of quantum matter in august of 2014 . much of the research reported in this document was carried out while we were visiting the institute . the second author wants to thank moody t. chu for his warm hospitality during his visit to the department of mathematics at north carolina state , for precise comments , challenging questions , and for sharing interesting conjectures and problems with him . he also wants to thank alexandru chirvasitu for several interesting questions and comments that have been very helpful for the preparation of this document .

high resolution far ultraviolet spectra of the early - type stars and were obtained to measure the interstellar deuterium abundances in these directions . the observations were made with the interstellar medium absorption profile spectrograph ( imaps ) during the _ orfeus - spas ii _ mission in 1996 . imaps spectra cover the wavelength range 9301150 with . the interstellar features are resolved and cleanly separated from interstellar in the and profiles of both sight lines , and also in the profile of . the profiles were modeled using a velocity template derived from several lines in the imaps spectra recorded at higher signal - to - noise . to find the best column density , we minimized for model profiles that included not only the as a free parameter , but also the effects of several potential sources of systematic error which were allowed to vary as free parameters . column densities were measured by analyzing absorption profiles in a large number of high dispersion spectra for each of these stars and applying this same -minimization technique . ultimately we found that d / h for and for , values that contrast markedly with d / h derived in 1 for ( the stated errors are 90% confidence limits ) . evidently , the atomic d / h ratio in the ism , averaged over path lengths of 250 to 500 pc , exhibits significant spatial variability . furthermore , the observed spatial variations in d / h do not appear to be anticorrelated with n / h , one measure of heavy element abundances . we briefly discuss some hypotheses to explain the d / h spatial variability . within the framework of standard big bang nucleosynthesis , the large value of d / h found toward is equivalent to a cosmic baryon density of , which we regard as an upper limit since there is no correction for the destruction of deuterium in stars . 1paper i # 110^#1 3 cm^-3 2h

astro-ph0007238

the abundance ratio of atomic deuterium to hydrogen ( d / h ) in interstellar gas is widely regarded as an important tracer of galactic chemical evolution ( audouze & tinsley 1974 ; boesgaard & steigman 1985 ; tosi et al . 1998 ) and a key discriminant of the cosmic baryon - to - photon ratio @xmath5 in big bang nucleosynthesis ( bbn , walker et al . 1991 ) . a standard interpretation is that d should not be produced in significant quantities in astrophysical sites other than the big bang ( epstein , latimer , & schramm 1976 ) . the generally accepted viewpoint is that while d is produced primordially , its destruction takes place when some of the gas is cycled through stars . the uncertainties surrounding this process represent a stumbling block in arriving at the primordial value . for this reason , a measurement of d / h is often regarded as a lower limit to the primordial ratio , and this may be translated into an upper limit to @xmath6 ( for bbn , larger d / h implies lower @xmath6 ) . rogerson & york ( 1973 ) made the first measurement of the atomic d / h abundance ratio in the interstellar medium ( ism ) on the line of sight toward @xmath7 cen . measurements with the satellite toward an additional 14 stars ( @xmath8 pc ) found that d / h values cluster around @xmath9 , but with a dispersion that , in some cases , seemed to exceed the stated uncertainties ( see review by vidal - madjar & gry 1984 ) . while a simple interpretation of the data suggested that the d / h measurements revealed spatial variations , the reality of these differences has been difficult to substantiate due to concerns stemming from the somewhat inadequate resolution of ( @xmath10 fwhm ) for this purpose . using the goddard high resolution spectrograph ( ghrs ) and space telescope imaging spectrograph ( stis ) on the _ hubble space telescope _ ( ) , several measurements of d / h in the local interstellar medium ( lism , @xmath11 pc ) have been made with observations of ( see lemoine et al . 1999 for a review ) . linsky et al . ( 1993 , 1995 ) found d / h @xmath12 toward capella . vidal - madjar et al . ( 1998 ) reported evidence for a factor of @xmath13 difference in d / h between two components on the line of sight toward the white dwarf g191-b2b ( @xmath14 pc ) . sahu et al . ( 1999 ) re - evaluated the ghrs data as well as new stis echelle spectra of g191-b2b and concluded that , as a result of improved instrumental characterization ( echelle scattered light correction near ) , d / h values in both components appeared to be consistent with the usual lism value . questions remain , however . for example , howk & sembach ( 2000 ) found a different stis echelle scattered light correction , one that agrees with the ghrs profile ( vidal - madjar 2000 ) . the resolution of the conflicting conclusions about the g191-b2b sightline may be provided by observations of higher lyman series lines ( and ) by the _ far ultraviolet spectroscopic explorer _ ( , moos et al . 2000 ) . the interstellar medium absorption profile spectrograph ( imaps ) provides a new window to study deuterium in the galaxy by virtue of its wavelength coverage and very high spectral resolution ( @xmath0 ) which alleviate many of the previous obstacles to accurate d / h measurements . prior to the imaps deuterium studies , d / h measurements in our galaxy after the mission have been confined to the local ism , since the saturated core of envelopes the feature for @xmath15 and higher lyman lines lie beyond the reach of . the goal of the imaps deuterium program is to obtain high - quality measurements of d / h on sightlines toward bright ob stars beyond the lism . in the first paper of this series , jenkins et al . ( 1999a , hereafter 1 ) measured the d / h abundance ratio toward using imaps spectra . they found d / h @xmath16 , a value that is much lower than that found in the lism . 1 also showed that the low abundance of d toward is _ not _ accompanied by an overabundance of n and o relative to h , as would be expected if the gas had been more thoroughly cycled through stellar interiors . we note that two other deuterium measurements using the tbingen echelle spectrograph ( resolution @xmath17 ) on the mission have been recently reported : glz et al . ( 1998 ) found d / h @xmath18 toward bd+28@xmath194211 , an o - type subdwarf @xmath20100 pc away , and bluhm et al . ( 1999 ) measured d / h @xmath21 toward bd+39@xmath193226 , a sdo star located @xmath22 pc from the sun . in this paper we analyze the lines of sight toward two well - known hot stars , orum ( hd 68273 ) and pis ( hd 66811 ) . they are among the brightest stars in the sky at 1000 , well studied by , have column densities @xmath23 , and hence are prime d / h targets for imaps . these stars were observed under an guest investigator program . lcccccccc & 66811 & 25398 & -471 & o5 iaf & 2.25 & 211 & -18.04 & @xmath24 & 68273 & 26281 & -770 & wc8+o8 & 1.78 & & -17.48 & @xmath25 the sightlines to and have been studied extensively (: morton & bhavsar 1979 ; sahu 1992 ; fitzpatrick & spitzer 1994 ; : morton & dinerstein 1976 ; morton 1978 ) , including d / h measurements with observations from the satellite . york & rogerson ( 1976 ) found d / h@xmath26 toward . vidal - madjar et al . ( 1977 ) studied the sightline toward and found that solutions consistent with the data permit a wide range of d / h values ( @xmath27 d / h @xmath28 ) . and are the most luminous stars in the vela - puppis region . their role in powering the gum nebula and their relationship to the vela ob2 association , vela supernova remnants , and various structures in the ism in vela- puppis have been analyzed by sahu ( 1992 ) . table [ stardata ] summarizes some basic properties of the two stars . the distances listed were derived from _ data by schaerer , schmutz & grenon ( 1997 ) , placing significantly closer than earlier estimates . is the closest ( and brightest ) wolf - rayet star and consequently it has been studied intensively for more than a century ( van der hucht et al . nevertheless , due to the complexity of the stellar system , its fundamental stellar parameters have been continuously debated and revised ( as recently as 1997 when several papers were published , e.g. , schaerer et al . 1997 ; schmutz et al . 1997 ) . is a double - lined spectroscopic binary star composed of a wr star ( wc8 ) and a late o star ( o8 iii ) of comparable brightness . the complex properties of this system are evident in the significant changes in the uv spectrum , and this behavior will be addressed in [ gveln(hi ) ] when we discuss our determination of the interstellar hydrogen column density through measurements of the absorption feature . is an extremely luminous , massive star ( o4 iaf ) and the brightest o - type star in the sky . it is a prototype for o - star spectral properties and variability ( morton & underhill 1977 ; massa et al . 1995 ) and theoretical understanding of radiation - driven winds in massive stars ( e.g. puldrach et al . is believed to be a single star . the high velocity resolution provided by imaps opens the way for more accurate line profile measurements in the far ultraviolet spectra of bright stars . our observations of and described in [ obs ] build on our earlier results for described in 1 . we follow with discussions of how we derived column densities ( [ n(di ) ] ) , column densities ( [ n(hi ) ] ) , and values for d / h and n / h ( [ d / h ] ) . the paper concludes with a discussion in [ discuss ] of the possible implications of the differences in d / h for the three stars covered in this paper and 1 . the spectra analyzed in this paper were obtained with imaps during the mission ( sts-80 ) in late 1996 ( see hurwitz et al . 1998 ) . the design of imaps and its performance on earlier flights have been described in detail by jenkins et al . ( 1988 , 1996 ) . in 1 we described the in - flight performance of imaps during the 1996 mission . here we only summarize the instrument s principal characteristics . imaps is an objective - grating echelle spectrograph designed to record the far - uv spectrum of bright stars at high spectral resolution . the optical design consists of a wire grid collimator to reject off - axis light from stars other than the target , an echelle grating with a 63 blaze angle , and a parabolic cross disperser . the spectral format is imaged on a solid kbr photocathode , whose electrons are magnetically focussed on a windowless , back - illuminated ccd with a 320x256 pixel format . the nominal wavelength range of 9301150 is obtained in four selectable tilts that span the free spectral range of the echelle grating . the gratings were coated with lif over aluminum , providing excellent throughput longward of 1000 . although the reflectivity of lif drops substantially shortward of 1000 , imaps achieved a useful throughput even in the 930 980 region . the spectral resolution in imaps spectra obtained during the 1996 flight was approximately @xmath29 , or @xmath30 . the telluric lines ( e.g. 950.112 near - see fig [ gvlylines ] ) have fwhm @xmath31 , but these lines are probably partly resolved in imaps spectra ( see jenkins et al . 1999b ) . and were chosen for this imaps guest investigator program because they were available in 1996 november , have a flux near @xmath32 , have @xmath33 , and @xmath23 . was observed by imaps on 1996 november 27 14:3415:11 and 16:0016:47 ut for a total of 4224 s. was observed on 1996 november 26 18:4419:23 and 20:1620:57 ut also for a total of 4224 s. the exposure time at each echelle grating position ranged from 817.6 s to 1226.4 s. these exposure times were chosen to obtain spectra with good signal - to - noise ( s / n ) ratios near the cores of the ( 949.485 ) and ( 937.548 ) lines . typically , we found s / n=1015 in the local continuum near the and interstellar features . the data reduction and analysis of the and spectra and the determination of the column densities were identical to those described in 1 for the line of sight toward . imaps spectra were analyzed to determine the total ( [ gveln(di ) ] and [ zpupn(di ) ] ) . a large number of _ international ultraviolet explorer _ ( ) high resolution spectra were obtained from the national space science data center archive to determine the total column density ( ) in [ n(hi ) ] . we note that given that @xmath34 on these sightlines and that d / h @xmath35 , the gas that gives rise to the damping wings is the same gas responsible for the and features . in [ niprof ] we show that is a good tracer of and on these sight lines . the line of sight column density per unit velocity [ , see 1 ] for , defined by a range of lines recorded in the imaps spectra , provided a velocity template for modelling the profiles . the high s / n for the lines helped to constrain the model profiles that gave an acceptable fit to the lower s / n lines . this can prevent noise in the profiles from giving arbitrarily large or small at specific velocities . we did not decompose the velocity profiles of or into separate ( blended ) gaussian components the d / h ratios determined here were based on total column densities for each sight line . using the method outlined in 1 , we corrected for the effects of the weak lines 937.652 and 2 lyman 140 @xmath36(2 ) 949.351 . the line is located at @xmath37 in and at @xmath38 in on the heliocentric velocity scale whose zero point is the laboratory wavelength of ( see figs . [ gvlylines ] and [ zplylines ] ) . the 2 line is located at @xmath39 in and at @xmath40 in on the heliocentric velocity scale . the strength and shape of these features were computed from other transitions of the same species recorded at longer wavelengths ( and higher s / n ) in the imaps spectra of and . the computed central residual intensities of these features are nearly identical for the two stars : 0.66 for 937.6 and 0.95 for 2 949.3 . the other potentially contaminating lines noted in table 1 of 1 can safely be ignored . the possible contamination of the order by scattered light from an adjacent order that contains the strong absorption lines of the 954 multiplet was evaluated in the same manner as described in 1 . the analysis found that any residual contamination of the spectrum in the vicinity of by the pattern of saturated features was @xmath41% of the continuum in both and . amplitudes for this correction larger than @xmath42% caused unacceptably bad deviations in the bottom of the profile . given the general noise characteristics of the profiles , this is a negligible effect . for the case of , there is another potential contamination source . a companion star , @xmath43 vel , is located 42 to the southwest of . its light should be accepted along with that from , since imaps is an objective - grating spectrograph without an entrance slit to reject unwanted sources . ( the grid collimator rejects light coming from more than 1 from the axis , however . ) @xmath43 vel is 2.5 magnitudes fainter than in @xmath44 and has a spectral classification of b1@xmath45iv ( hoffleit & jaschek , 1982 ) . even though @xmath43 vel is cooler than , its flux is not diminished much below the peak of the planck distribution at our wavelengths of interest , so we expect its intensity at a given wavelength to be not more than about a factor of 10 fainter than . fortunately , when the observations were taken the roll angle of the spacecraft about the optical axis of imaps ( governed by the position of the sun in the sky ) was such that the faint spectrum of @xmath43 vel was displaced along the cross dispersion direction toward the long - wavelength part of the format . as a consequence , any spectral segment in the spectrum of had light superposed on it from shorter wavelengths in @xmath43 vel ( the separation was slightly less than the distance between 3 echelle orders ) . the rapid decline in the sensitivity of imaps toward shorter wavelengths thus amplified the factor of 10 disparity in relative fluxes at any given position on the format . therefore , the contamination of the spectrum by light from @xmath43 vel is negligible . as a final note , on visual inspection we are unable to see any ghost - like spectrum of @xmath43 vel on top of the spectrum of . the lyman line profiles for and are shown in figures [ gvlylines ] and [ zplylines ] . these profiles have been corrected for and 2 line absorption , as described above . the background levels near the lines were determined from the broad , saturated cores of the adjacent profiles . as explained in 1 , we compared resolution - degraded forms of the imaps profiles with those recorded by to test the proposition that the cores of the lines indeed represented the zero - intensity levels in the vicinity of the adjacent features . this was done to check that we were not being misled by an effect from possible strong , broad wings extending away from the main peak in the instrumental profile of imaps . we concluded that indeed the cores of the profiles provided very good estimates of the zero levels in the vicinity of the lines . the best answers for and the deviations permitted by the data were evaluated by minimizing ( see lampton , margon & bowyer 1976 and bevington & robinson 1992 for details ) when all of our unknown parameters were allowed to vary the ten free parameters for are ( 1 ) , ( 2 ) the gas temperature @xmath46 , ( 3 - 8 ) the continuum slopes , y - intercepts , and background levels for and , ( 9 ) a shift in the velocity zero point between the lines and the template , and ( 10 ) a coefficient for scaling the contamination signal in ( see 3.2 in paper i ) . the gas temperature is an important parameter because the line can be broadened significantly compared to the template by thermal motions ; see 4.1 in 1 for details . there are three additional free parameters for because of the addition of to the analysis . note that the linear continuum fitting ( with specific velocity limits as given in [ gveln(di ) ] and [ zpupn(di ) ] ) is an integral part of the evaluation , i.e. , the deviations of the continuum levels are considered , in addition to the behavior inside the profiles . the same holds for the zero level as defined by the bottom of the adjacent profile . we used powell s method ( press et al . 1992 , p. 406 ) to find the minimum . we then set confidence limits by increasing ( or decreasing ) and @xmath46 with the other parameters freely varying until increased by the appropriate amount for the confidence limit of interest with two useful parameters , and @xmath46 . we also used different initial values to establish that the minimum is unique . in 1 we discussed the benefits of obtaining a velocity profile based on high quality data for and , two species that are not significantly depleted in the ism and which have very similar ionization balances to those of and ( ferlet 1981 ; york et al . 1983 ) . this profile information is helpful in constraining the variety of possible interpretations that would produce acceptable fits with the d profiles . for , we used imaps spectra of the 10 lines in the multiplets at 952.4 , 953.8 , 954.1 and 1134.7 . for all of the lines except those in the 952.4 multiplet , the background level was easily defined because the stronger lines were saturated . of course , these lines were useful only for defining the behavior of the gas at velocities somewhat removed from the line core . since the lines in the multiplet at 952.4 were not saturated , we had to determine the background level by a different method . for both and , there are u1 scans of this weak multiplet available in the archive of spectra recorded by the satellite ( rogerson et al . 1973 ) . after comparing our imaps spectrum degraded to the resolution of with the actual scans , we determined the adjustments to the background levels that were needed for the imaps spectra of this multiplet . the backgrounds caused by grating scatter in the spectra were taken from the observed count rates in the bottoms of the hydrogen and lines , and these levels were subtracted off before the comparison was made . as was done for our analysis in 1 , we adopted a method developed by jenkins & peimbert ( 1997 ) to create a composite profile for the column density of as a function of velocity from the 10 lines in the four multiplets , with the weak lines in each case defining the main part of the profile and the strong lines outlining the exact shape of the profile s wings , well away from the saturated part of the line . for places where there was overlap in the useful portions of the lines , there was satisfactory agreement . the @xmath47-values of goldbach et al . ( 1992 ) were adopted for our analysis . figure [ ni_navfig ] shows the profiles derived for toward and . in 1 we showed that the profiles are unlikely to be contaminated by additional contributions from telluric absorption . is @xmath201% of in the earth s atmosphere at the altitude of imaps ( @xmath20305 km ) during the mission . even for our strongest transition used to derive the far wings of for ( 1134.98 ) , the telluric absorption should be only one third as strong as the 950.112 feature present in figures [ gvlylines ] and [ zplylines ] , and at a heliocentric velocity of @xmath48 it would be buried in the saturated portion of the profile . consequently , telluric has a negligible effect on the interstellar profiles . is the best tracer of in the ism since the ionization potential of ( 13.56 ev ) is nearly identical to that of . the ionization potential of ( 14.53 ev ) is only slightly greater that that of . furthermore , both o and n are coupled to h by resonant charge exchange reactions . as discussed by sofia & jenkins ( 1998 ) and jenkins et al . ( 2000 ) , closely follows and unless @xmath49 . in the imaps wavelength band there is no suitable set of transitions to completely define a velocity profile template . the available lines are highly saturated or are not detectable , either because they are buried in the core of a lyman series line or because they are too weak . in 1 we made use of an archival spectrum of the very weak intersystem transition of at 1355.6 for . there is no such spectrum available for , and the measurement of this line in the spectrum of taken by fitzpatrick & spitzer ( 1994 ) is too noisy to adequately define the shape of the profile . therefore , we were forced to use the next best option , , to define the velocity profile template . although we do not have a complete profile for , we tested the assumption that traces by comparing selected portions of the profile derived from available line profiles in . first , we computed from the wings of 1039.230 . second , we used the ghrs spectrum of 1355.6 ( fitzpatrick & spitzer 1994 ) to define in the core of the line . the computed was scaled by the difference in the n and o solar abundances ( @xmath50 dex , anders & grevesse 1989 ) . the results , shown in fig [ ni_navfig ] , demonstrate that the wings and core of for and are in excellent agreement . consequently , we have high confidence that the profile for provides an accurate model for the velocity distribution of . the continuum near the line shown in figure [ gvlylines ] was determined in the velocity range @xmath51 on the blue side of the line and @xmath52 on the red side . for , the continuum limits were @xmath53 and @xmath54 . the background level was determined from the core of the two lines in the velocity range @xmath55 . the profiles were fit in the velocity range @xmath56 for and @xmath57 for . with a spacing of independent velocity samples of @xmath58 , this resulted in 276 degrees of freedom in the analysis to fit the 10 free parameters described in [ x2method ] . by adjusting our estimate for the noise , which was only known to limited accuracy beforehand , we achieved a minimum value for equal to 247.0 at a column density of @xmath59 . for 276 degrees of freedom , there was a 90% chance that we would have found this value of or greater . with this 90% confidence level , we arrived at a conservatively high estimate for the noise . it then followed that this noise level , which is perhaps higher than the real noise , gave a conservatively large confidence interval for , as determined by how rapidly the values of deviated from the minimum value . with the noise level having been set in the manner just described , we explored for 90% and 99% confidence limits for ( i.e. , @xmath60 and 2.58@xmath61 deviations ) , which correspond to ( min)+4.6 and ( min)+9.2 , respectively . table [ gvdtab ] lists and @xmath46 for the best value and these limits . the @xmath6290% confidence limits on the model profiles ( _ cross - hatched regions _ ) for and are compared with the observed deuterium profiles in figure [ gvdfits ] . cccc minimum at the 99% confidence limit & 0.96 & 6630 & 255.6 minimum at the 90% confidence limit & 1.00 & 6530 & 251.6 best & 1.12 & 6070 & 247.0 maximum at the 90% confidence limit & 1.27 & 5280 & 251.7 maximum at the 99% confidence limit & 1.34 & 4870 & 256.1 cccc minimum at the 99% confidence limit & 1.07 & 10600 & 397.7 minimum at the 90% confidence limit & 1.13 & 10400 & 393.1 best & 1.30 & 9550 & 388.5 maximum at the 90% confidence limit & 1.49 & 8500 & 393.1 maximum at the 99% confidence limit & 1.58 & 7900 & 397.7 in figure [ gvdfits ] we also show with a _ dashed line _ the expected shape and depth of the profile for = @xmath63 . this corresponds to a d / h ratio of @xmath64 , assuming the column density derived below in [ n(hi ) ] . for this case , ( min ) = 52.4 , which is clearly an unacceptable fit . the _ dot - dashed line _ in figure [ gvdfits ] corresponds to the profile for = @xmath65 , the column density necessary to achieve d / h=@xmath66 , the value found toward in 1 . a similar demonstration for the measurement of is be given in [ d / h ] . the measurement of toward used the same methodology described in the previous section with the following modifications . in addition to and we were able to include the profile of in the analysis . the blue and red regions for the linear continuum fits were @xmath51 and @xmath67 for , @xmath68 and @xmath69 for , and @xmath70 and @xmath71 for . the background levels were determined from the saturated cores of the lines over these velocity limits : @xmath72 for and , @xmath73 for . the model profiles were fit over the range @xmath74 for , @xmath75 for , and @xmath76 for . this resulted in 426 degrees of freedom to simultaneously fit the 13 free parameters described in [ x2method ] . the minimum is 388.5 and corresponds to the most probable value of for , namely @xmath77 . the results of the analysis are summarized in table [ zpdtab ] , including the 90% and 99% confidence limits . the same conservative treatment of the noise described for was followed for . figure [ zpdfits ] shows the observed lyman line profiles and the range in the best fit model profiles and continua fits allowed by the 90% confidence limits . the primary objective of this imaps program is to determine d / h ratios of sufficient accuracy to test for spatial inhomogeneities . to achieve this goal , we must strive for a precision in that is as good as ( or preferably better than ) that of the deuterium measurement . we must also understand the magnitude and nature of the uncertainties in and so that we can realistically assess the uncertainty in d / h for comparison to other measurements in the ism . for these reasons , we decided to conduct our own investigation of the column densities toward and , even though a number of measurements of toward these stars are already available in the literature ( e.g. jenkins 1971 ; york & rogerson 1976 ; vidal - madjar et al . 1977 ; bohlin , savage , & drake 1978 ; shull & van steenberg 1985 ; diplas & savage 1994 ) . we described our method for determining in 1 ; here we provide a summary . in [ zpupn(hi ) ] and [ gveln(hi ) ] we discuss details particular to the and sight lines and present the results . we used the line to determine the total . since and are on the flat part of the curve of growth , high velocity gas could influence their profiles . is immune from this potential problem . due to the great breadth of the lorentzian wings ( @xmath78 see the velocity scale in figure [ gvhiblowup ] ) , absorption due to any high velocity interstellar gas ( @xmath79 ) that may be present is confined to the saturated core of the profile . in principle , some high velocity , which is too widely dispersed in velocity to show up in the profiles , could contribute to the damping wings . however , this gas is not detected in the strong lines 1134.98 and 1039.23 observed with imaps . it is clear that such high velocity gas , if present , would have a low column density and a negligible effect on the damping wings . since the line is not covered by imaps , was not observed by the spectrographs , and the available scans suffer from several problems ( see 1 ) , we analyzed high dispersion observations of to determine . furthermore , these stars were observed many times with over the course of many years , and this offers an opportunity to evaluate potential sources of systematic error . for example , is a spectroscopic binary , and the large database allowed a search for systematic changes in the derived as a function of orbital phase . after screening and retrieving the data of interest and correcting for interorder scattered light ( see 1 ) , we used an approach first used by jenkins ( 1971 ) to constrain the column density , i.e. , we determined the which provides the best fit to the profile with the optical depth @xmath80 at a given wavelength @xmath81 calculated from the expression @xmath82 where @xmath83 is the line center at the velocity centroid of the hydrogen ( note that the useful portion of the profile is entirely due to the lorentzian wings and the effects of instrumental and doppler broadening can be neglected ) . as was done for , we also determined the important free parameters which can be adjusted to fit the absorption profile , then we found the set of parameters that minimized using powell s method . we set confidence limits on the column as described in 3.1 with only one parameter of interest , . the other ( uninteresting ) free parameters we selected for fitting the profile were three coefficients which specify a second - order polynomial fit to the continuum , and an additive correction to the flux zero point . the continuum was fit to the spectrum in several windows covering the range 1185 to 1276 . despite our use of the bianchi & bohlin ( 1984 ) correction for interorder scattered light , in many cases inspection of the flat - bottomed , saturated portion of the profile showed that the zero intensity level was not quite correct . a zero point shift in the flux scale , one of the terms for evaluating , was determined from a region within the saturated core . for both stars , the zero point of the wavelength scale was set by placing the 1200 multiplet in agreement with the imaps profiles . the profile was then fit only to the red wing because of the presence of strong stellar features superposed on the blue wing . however , the uncontaminated portion of the blue wing of was checked for consistency with the fit to the red wing and found to be in good agreement . in 1 , considerable attention was paid to the spectroscopic binary nature of and whether or not this could cause systematic errors in . a similar analysis of is presented in [ gveln(hi ) ] . the determination of toward was less complex than the other stars observed by imaps to study d / h . as far as is known , is not a spectroscopic binary star . it is an o4 supergiant , so the stellar absorption line makes a negligible contribution to the absorption profile . the star is possibly a non - radial pulsator with very weakly variable stellar absorption lines with a period of 8.5 hr ( reid & howarth 1996 ) , and the variability of the stellar wind p - cygni profiles is well known and studied ( e.g. , prinja et al . 1992 ; howarth , prinja , & massa 1995 ) and shows significant power at 19.2 hr and 5.2 day periods . both of these sources of spectral variations are expected to have little impact on the interstellar column density derived from , but this can be checked given the large number of observations and good temporal sampling . there are more than 200 high - dispersion observations of in the archive , primarily because the star was selected for intensive observing programs to study stellar wind variability in massive stars . we concentrated on the spectra obtained in three specific multi - day observing runs in 1986 , 1989 , and 1995 . in 1995 was observed continuously for 16 days ( dubbed the mega campaign , massa et al . we omitted three observations from these observing sessions because the archival data were corrupted or unavailable . this left 189 observations for measurement of . all of the observations were obtained with the large aperture and the signal - to - noise ratios are comparable ; the data set is more uniform than the data used for measuring toward and . lccccc 1986 & 11 & 9.23@xmath85 & 0.47@xmath85 & 0.14@xmath86 & 0.69 1989 & 31 & 9.42@xmath85 & 0.43@xmath85 & 0.08@xmath86 & 0.56 1995 & 147 & 8.88@xmath85 & 0.46@xmath86 & 0.04@xmath85 & 0.69 adopted value & 189 & 9.18@xmath85 & & & the values derived from each individual observation are plotted in figure [ zphifig ] , along with the 1@xmath61 uncertainties , versus the observation date . table [ zpnhi ] summarizes the mean column density @xmath87@xmath88 and the root mean square ( rms ) dispersion @xmath61 obtained from the 1986 , 1989 , and 1995 data sets analyzed separately , and @xmath87@xmath88 for each data set is indicated with a heavy dashed line in figure [ zphifig ] . we also list in table [ zpnhi ] the formal error in the mean @xmath89 ( here @xmath90 is the number of measurements ) and the reduced for the three data sets . in 1 we discussed the potential sources of systematic error in the derivation of from spectra , and we suggested that the real uncertainty in the mean is likely to be greater than the formal estimate . the very large data set for shown in figure [ zphifig ] appears to confirm that this is indeed the case for this star . from this figure one can see that the column densities derived from the 1995 data are systematically lower than the column densities derived from the 1989 observing run . the mean of the 1995 data set is lower than the mean of the 1989 data by @xmath91 ( 0.0256 dex ) , and this difference is substantially greater than the formal error estimates ( @xmath92 ) . this systematic error could be the result of secular changes in the stellar mass loss or circumstellar environment , a previously unrecognized stellar variability with a long period , or instrumental effects . or perhaps some aspect of the interstellar sight line changed during this period . since the imaps observations for were made almost two years after the mega campaign , it is possible that a similar systematic error is present in any estimate of that we choose for comparison with the imaps . the exact magnitude of this systematic uncertainty is difficult to estimate , but we find that obtained from additional large aperture spectra of taken at other epochs are in good agreement with the estimates in table [ zpnhi ] . this indicates that the systematic error is not much larger than 0.0256 dex . for example , large aperture observations taken in 1988 may , 1989 december , and 1991 march , combined with the three @xmath87@xmath88 values for the three data sets in table [ zpnhi ] , yield a mean = @xmath93 . we conclude that 0.0256 dex is a conservative estimate of the @xmath94 uncertainty in . we therefore adopt the unweighted mean of the three @xmath87@xmath88 values in table [ zpnhi ] as the best estimate of toward and 0.0256 dex as its @xmath94 uncertainty : @xmath95 . we elected to use a mean of the @xmath87@xmath88 values in table [ zpnhi ] instead of a straight mean of all of the individual measurements shown in fig . [ zphifig ] because the latter would give excessive weight to the 1995 data set due to the much larger number of measurements obtained during that observing campaign . as noted in 1 , is a complex stellar system with a strongly variable uv stellar spectrum . figure [ gvhisamples ] shows two examples of the high - dispersion spectra of that we have used to derive . these examples were selected to illustrate the variability of the p - cygni emission lines in the vicinity of , which are weakest at orbital phase @xmath200.5 . figure [ gvhisamples ] also shows the complex and variable stellar absorption superposed on much of the blue wing of the interstellar profile . this variability was a source of concern for this paper . can we derive reliable column densities despite the complex spectral changes occurring in this star ? due to the stellar absorption , the blue wing of was not useful for constraining . fortunately , the red wing of was clean and stable as a function of orbital phase . this can be seen by comparing panel ( a ) to panel ( b ) in figure [ gvhisamples ] . in this paper , we used only the red wing for fitting the profile , and the resultant fits generally look quite good ( see figures [ gvhisamples ] and [ gvhiblowup ] ) . we examined as a function of orbital phase of the spectroscopic binary to check for systematic errors in the derived interstellar column density due to the complicated variability of the stellar spectrum . there are considerable discrepancies in the orbital elements published by various groups ( stickland & lloyd 1990 ; schmutz et al . 1997 , and references therein ) . we adopt the orbital elements derived by schmutz et al . ( 1997 ) who find a period of @xmath96 d with velocity semiamplitudes of @xmath97 and @xmath98 . in addition to the usual intrinsic variability of wr stars , it is believed that the spectrum may also change as a result of periodic absorption of the o star component by the wr wind ( stickland & lloyd 1990 ) . after screening the high - dispersion observations of , we were left with 42 spectra , 34 small aperture observations obtained early in the mission and 8 later observations obtained with the large aperture . in figure [ gvhifig ] we plot the column densities derived from all of the observations as a function of the phase of the spectroscopic binary , and table [ g2nhi ] summarizes the @xmath87@xmath88 , @xmath61 , and @xmath92 derived from the large aperture data only , the small aperture data only , and all of the data combined . while one might argue that a trend is apparent in figure [ gvhifig ] with a minimum in the derived values at orbital phase @xmath20 0.5 , this is a marginal result at best , and we do not believe that it should be taken too seriously . on the contrary , the generally good agreement of the column densities at various phases despite the dramatic changes in the stellar spectrum ( see figure [ gvhisamples ] ) is encouraging , and we conclude from figure [ gvhifig ] that the data provide a good determination of @xmath87@xmath88 . since the small and large aperture measurements are in good agreement , we adopted the results from the combined data set . systematic errors probably cause the true uncertainty in @xmath87@xmath88 to exceed the formal error estimate provided in table [ g2nhi ] . drawing from our experience with the more extensive data set for , we conservatively adopt 0.0256 dex as the @xmath94 uncertainty in toward as well : @xmath99 . lccccc lg . aperture & 8 & 5.35@xmath85 & 0.51@xmath85 & 0.18@xmath85 & 1.15 sm . aperture & 34 & 5.10@xmath85 & 0.48@xmath85 & 0.08@xmath85 & 1.71 all data & 42 & 5.13@xmath85 & 0.49@xmath85 & 0.08@xmath85 & 1.62 combining the results for and column densities derived in the previous sections , we determined the atomic d / h abundance ratio for and . toward we found that d / h @xmath100 ( the errors are 90% confidence limits ) . this result is a large departure from the value of d / h usually attributed as `` typical '' in the ism of the galaxy ( @xmath101 ) . a large value for the average d / h ratio implies that even larger values could exist in individual components if other components have lower d / h values closer to the lism ratio . although the previous d / h measurement for ( @xmath102 , york & rogerson 1976 ) appears to be in close agreement with the imaps result , we consider this to be coincidental in view of the magnitude of the uncertainties in the measurement . for we find @xmath103 , a value slightly below the lower limit derived from spectra by vidal - madjar et al . they found that a large range in ( @xmath104 ) was possible for this sight line due to its complexity and a lack of adequate constraints . since the value of derived by vidal - madjar et al . ( 1977 ) for is consistent with ours , their large range in d / h must be a result of the uncertainty . much more is known now about the complexity of this sightline than at the time of the study ( e.g. welty , morton & hobbs 1996 ) . more importantly , the imaps spectra have sufficient spectral resolution that the structure of the sightline in neutral hydrogen can be defined by the profile for and used to constrain the determination of . thus with higher spectral resolution and better knowledge of the velocity structure along the sightline , a more accurate value of and d / h is derived . we find d / h @xmath105 . the , , and d / h results for the three stars studied by imaps are summarized in table [ abund3 ] , where all errors are presented as 90% confidence limits . cccc @xmath106 & @xmath107 & @xmath108 & @xmath109 @xmath90 ( ) @xmath106 & @xmath110 & @xmath111 & @xmath112 @xmath113 & @xmath114 & @xmath115 & @xmath116 d / h @xmath117 & @xmath118 & @xmath119 & @xmath120 n / h @xmath117 & @xmath121 & @xmath122 & @xmath123 d / n & @xmath124 & @xmath125 & @xmath126 we now address a critical question , namely , is there sufficient uncertainty in the and measurements to reconcile the d / h abundance ratio with the general result observed in the local ism or with the d / h derived in 1 for the sight line to ? figure [ gvhiblowup]a shows with dotted and dashed lines the expected appearance the profile would have if our measurement of is correct but the were made high enough so that d / h @xmath127 . in figure [ gvhiblowup ] , the dotted line represents the profile that would be needed for the most probable value for , while the dashed lines represent the 90% confidence limits . for comparison , we also show in figure [ gvhiblowup]a the best - fitting profile ( solid line ) derived in the previous section for this particular observation . figure [ gvhiblowup]b is an analogous plot with forced to take on a value so that d / h @xmath128 as observed toward ( 1 ) . from figure [ gvhiblowup ] it is clear that forcing to make d / h @xmath127 , assuming from [ gveln(di ) ] , produces a very poor ( and unacceptable ) fit to the profile , and the fit becomes ridiculous if d / h @xmath129 . we have assumed the best - fit continuum ( solid line ) to compute the profiles shown in figure [ gvhiblowup ] . we have also explored whether or not the d / h ratios can be reconciled by adjusting the continuum placement , and we found that absurd continuum placements and shapes must be assumed to make d / h @xmath127 or ( even worse ) @xmath130 . as demonstrated at the end of [ gveln(di ) ] , the changes in needed to make d / h @xmath127 or @xmath129 , assuming our most probable value of , gave a large mismatch with the data , which were clearly unacceptable . we conclude that spatial inhomogeneities in d / h in the ism within @xmath20500 pc of the sun have high statistical significance , especially when contrasting the d / h ratios in the directions of and , which were measured with the same instrument and techniques . the column density was computed by integrating the column density profiles shown in figure [ ni_navfig ] . the results for the three targets are listed in table [ abund3 ] . this table also contains the resulting n / h and d / n abundance ratios . we examined the potential error in by examining the errors in the portions of the various profiles used to construct . the @xmath94 uncertainty in for and is conservatively estimated to be 5% since the profiles were all of high quality . there may be some question about the structure in the core of for . the uncertainty in for was estimated by supposing the spike in at @xmath131 is due largely to noise fluctuations in the core of 952.5 , the weakest line we detect on this sightline . if we truncate the profile at the level of the secondary maximum at @xmath132 [ @xmath133 , we find that the area of the spike is @xmath134 , or 4.7% of the total . we then suppose that the total uncertainty in for is @xmath135 times this , or 1@xmath136 , to allow for the possibility that there may be other such fluctuations . there are two principal conclusions of the imaps d / h program . first , the atomic d / h ratio in the ism , averaged over path lengths of 250 to 500 pc , exhibits significant spatial variability . differences in the atomic d / h ratio on long pathlengths in the ism have been suspected for many years ( vidal - madjar et al . 1978 ; vidal - madjar & gry 1984 ) , but not until now have data of sufficient quality been available to evaluate and reduce statistical and systematic errors to levels where these differences are unequivocal . second , we find no support for the simple picture that variations in d / h anticorrelate with those of n / h , i.e. , one measure of how much the gas has been processed through stellar interiors . figure [ dnfig ] shows the relationship between d / h and n / h for the three sightlines studied by imaps plus the white dwarf g191@xmath137b2b ( vidal - madjar et al . 1998 ; sahu et al . we point out that some elements are systematically removed from the gas phase as they are incorporated into interstellar dust ( savage & sembach 1996 ) , but the abundance of n does not seem to be appreciably altered by this effect ( meyer et al . . beyond the effects from depletions onto dust , spatial variations in interstellar gas abundances can arise as a natural consequence of galactic chemical evolution and the changing influences of different stellar populations . the lack of an anticorrelation between n / h and d / h depicted in fig . [ dnfig ] indicates that the variability of d / h is not just a consequence of different mixing ratios of material with differing levels of stellar processing , as we might anticipate , for instance , from the variable addition of metal - poor , infalling gas from the galactic halo ( meyer et al . we may need to go further and draw a distinction between contributions from stars that simply destroy deuterium and those that both destroy deuterium and enrich the medium with additional nitrogen . that is , we could envision some stars cycling material only through their shallow layers that are only hot enough to burn deuterium , while others eject material from much deeper layers where the synthesis of heavier elements has taken place . this additional level of complexity could explain the behavior that we observed . global models of galactic chemical evolution ( audouze & tinsley 1974 ; tosi 1988a , b ; dearborn , steigman & tosi 1996 ; scully et al . 1996 ; tosi et al . 1998 ) describe the destruction of d during stellar formation , evolution , and eventual mass loss . these models predict variations in d / h , n , and o abundances that are manifested as abundance gradients on a scale of @xmath1381 kpc . however , the predicted trends in galactic abundances may not accurately represent what is observable in the diffuse ism . tenorio - tagle ( 1996 ) showed that the chemical enrichment of the diffuse ism by ob associations , including supernovae from massive stars , is a slow process . following a supernova explosion , chemically enriched ejecta remain clumpy and not well mixed with the diffuse ism it encounters until it is incorporated into new star forming regions . this is a result of very long time scales for diffusion between different parcels of gas in the warm ( @xmath139k ) and cold ( @xmath140k ) phases of the ism . the diffusion time scale for enriched gas to thoroughly mix with the warm diffuse ism can be very long ( @xmath141 yr - tenorio - tagle 1996 ) . on the spatial scale sampled by the imaps observations , different sight lines may encounter regions with very different dynamical and chemical histories . differential galactic rotation and random cloud motions are expected to stir the diffuse ism and chemically enriched parcels of gas , but these parcels retain their distinct chemical properties until they are disrupted through photoevaporation , most likely by the formation of new massive stars . tenorio - tagle finds that diffusion is efficient only for the hot phase of the ism ( @xmath142 yr ) , which accounts for a very small fraction of the total diffuse ism . only after the enriched gas and diffuse gas are highly ionized would the chemically enriched gas from the earlier generation of stars quickly diffuse into the ambient ism . thus , the time scale for mixing interstellar gases with different processing histories can be much longer than the chemical evolution time scale . however , it is possible that interstellar turbulence and its secondary phenomena may accelerate the mixing rate . since the distribution of star forming regions ( ob associations ) shows large inhomogeneities on scales @xmath143 kpc , their corresponding chemical enrichment of the ism may be expected to be nonuniform as well . this is perhaps revealed indirectly by variations in region abundances ( peimbert 1999 ) and solar - type stars at similar galactocentric radii ( edvardsson , et al . 1993 ) . several processes unrelated to stellar nucleosynthesis may , under the right circumstances , alter the atomic d / h ratio of some parcels of interstellar gas ( see lemoine et al . 1999 for a review ) . d may be incorporated into hd ( watson 1973 ) , but the fraction of molecular gas on our sight lines is very low ( see 1 ) . differential radiation pressure on d and h ( vidal - madjar et al . 1978 ; bruston et al . 1981 ) may lead to a separation of d in some clouds near strong radiation fields . adsorption of d onto dust grains ( jura 1982 ) may deplete d from the gas phase . bauschlicher ( 1998 ) found that reactions of h and d with polycyclic aromatic hydrocarbon ( pah ) cations might systematically provide some d enrichment in pahs . a very different perspective has been offered by mullan & linsky ( 1999 ) , who suggested that significant quantities of d may be formed in stellar flares from m dwarf stars and ejected into interstellar space . some or all of these processes could be at work in the diffuse ism and alter the atomic d / h ratio on individual sight lines , independent of the degree of chemical enrichment from stellar evolution . however , they have not yet been demonstrated to be quantitatively significant . observational and theoretical tests of the efficiency and applicability of these processes are needed to better understand the mechanisms affecting the d / h ratio in the diffuse ism . while the d / h ratios derived from imaps spectra are in the general range expected from galactic chemical evolution models , a factor of three variation in the mean d / h ratio on path lengths of several hundred pc is unexpected . the apparent lack of an anti - correlation of the d / h abundance ratio with the metallicity of the gas and its variability on smaller than expected scales suggests that other processes in the galaxy may be masking more general chemical evolution trends . this may pose a problem for deriving a `` primordial '' d / h by extrapolating back from d / h measurements in the milky way to extragalactic absorbers at higher and higher redshifts , or to even to evaluate `` primordial '' d / h directly from high redshift observations , until we understand the reasons for these differences . the spatial variations found in this study underline the importance of high - quality d / h determinations . d / h measurements in more distant regions of the galaxy are needed to determine whether the properties of the gas within 500 pc of the sun are representative of the galactic disk . observations with the fuse satellite should probe such more distant environments and hopefully answer some of these questions . the value of d / h for is larger than that usually considered typical for the milky way . this robust result establishes a new lower limit to the primordial d / h ratio . within the framework of standard big bang nucleosynthesis ( walker et al . 1991 ) , the large value of d / h found toward is equivalent to a cosmic baryon density of @xmath3 . this error simply reflects the uncertainty in the d / h ratio toward reported in this paper . we regard this value of @xmath144 as an upper limit since no correction has been applied for the destruction of deuterium in stars . this upper limit on @xmath145 is consistent with the preferred values of @xmath145 derived from recent analyses of the boomerang and maxima cosmic microwave background measurements ( e.g. , lange et al . 2000 ; tegmark & zaldarriaga 2000 ; hu et al . however , any lowering of this upper limit on @xmath145 to correct for astration will lead to a marginal disagreement with simple inflation models ( see figure 4 in tegmark & zaldarriaga 2000 ) and requires adjustments of other cosmological parameters . alternatively , the @xmath145 upper limit from d / h may be taken as a prior assumption for the constraint of other cosmological parameters using the cmb data ( e.g. , see hu et al . 2000 ) . we wish to thank the us and german space agencies , nasa and dara , for their joint support of the mission that made these observations possible . the successful execution of our observations was the product of efforts over many years by engineering teams at princeton university observatory , ball aerospace systems group , and daimler - benz aerospace . important contributions to the success of imaps also came from the efforts of d. a. content and r. a. keski - kuha and other members of the optics branch of the nasa goddard space flight center and from o. h. siegmund and s. r. jelinsky at the berkeley space sciences laboratory . we also thank bruce draine for insightful discussions on atomic interactions with dust grains and pahs . this work was supported in part by nasa grant nag5 - 616 to princeton university . the data were obtained from the national space science data center at nasa - goddard . anders , e. , & grevesse , n. 1989 , geochim . acta , 53 , 197 audouze , j. , & tinsley , b. 1974 , apj , 192 , 487 bauschlicher , c. w. 1998 , apj , 509 , l125 bevington , p. r. , & robinson , d. k. 1992 , data reduction and error analysis for the physical sciences , 2nd ed . ( new york : mcgraw hill ) bianchi , l. , & bohlin , r. c. 1984 , a&a , 134 , 31 bluhm , h. , marggraf , o. , de boer , k. s. , richter , p. , & heber , u. 1999 , a&a , 352 , 287 boesgaard , a. m. , & steigman , g. 1985 , araa , 23 , 319 bohlin , r. c. , savage , b. d. , & drake , j. f. 1978 , apj , 224 , 132 bruston , p. , audouze , j. , vidal - madjar , a. , & laurent , c. 1981 , apj , 243 , 161 dearborn , d. s. p. , steigman , g. , & tosi , m. 1996 , apj , 465 , 887 diplas , a. , & savage , b. d. 1994 , apjs , 93 , 211 edvardsson , b. , anderson , j. , gustafsson , b. , lambert , d. l. , nissen , p. e. , tomkin , j. 1993 , a&a , 275 , 101 epstein , r. l. , latimer , j. , & schramm , d. n. 1976 , nature , 263 , 198 ferlet , r. 1981 , a&a , 98 , l1 fitzpatrick , e. l. , & spitzer , l. 1994 , apj , 427 , 232 goldbach , c. , ldtke , t. , martin , m. , & nollez , g. 1992 , a&a , 266 , 605 glz , m. , et al . 1998 , in proc . iau colloq . 166 , the local bubble and beyond , eds . d. breitschwerdt , m. j. freyberg , j. trmper ( berlin : springer ) , 75 hoffleit , d. & jaschek , c. 1982 , the bright star catalogue , 4th ed . , ( new haven : yale u. obs . ) howarth , i. d. , prinja , r. k. , & massa , d. 1995 , apj , 452 , l65 howk , j. c. , & sembach , k. r. 2000 , aj , in press , astro - ph/9912388 hu . , w. , fukugita , m. , zaldarriaga , m. , & tegmark , m. 2000 , astro - ph/0006436 hurwitz , m. , et al . 1998 , apj , 500 , l1 jenkins , e. b. 1971 , apj , 169 , 25 jenkins , e. b. , & peimbert , a. 1997 , apj , 477 , 265 jenkins , e. b. , joseph , c. l. , long , d. , zucchino , p. m. , carruthers , g. r. , bottema , m. , & delamere , w. a. 1988 , in uv technology ii , ed . r. e. huffman ( bellingham : international society for optical engineering ) , 213 jenkins , e. b. , reale , m. a. , zucchino , p. m. , & sofia , u. j. 1996 , ap&ss , 239 , 315 jenkins , e. b. et al . , 2000 , apj letters , in press jenkins , e.b . , tripp , t. m. , woniak , p. r. , sofia , u. j. , & sonneborn , g. 1999a , apj , 520 , 182 ( paper i ) jenkins , e. b. , et al . 1999b , baas , 31 , 942 jura , m. , 1982 , in advances in ultraviolet astronomy ( nasa cp-2238 ) ed . y. kondo ( greenbelt : nasa ) , 54 lampton , m. , margon , b. , & bowyer , s. 1976 , apj , 208 , 177 lange , a. e. , et al . 2000 , astro - ph/0005004 lemoine , m. , et al . 1999 , , 4 , 231 linsky , j. l. et al . 1993 , apj , 402 , 694 linsky , j. l. , diplas , a. , wood , b. e. , brown , a. , ayres , t. r. , & savage , b. d. 1995 , apj , 451 , 335 massa , d. et al . 1995 , apj , 452 , l53 massa , d. , van steenberg , m. e. , oliversen , n. , & lawton , p. 1998 , in ultraviolet astrophysics beyond the final archive ( esa sp-413 ) , ed . w. wamsteker & r. gonzalez riestra ( noordwijk : esa ) , 723 meyer , d. m. , cardelli , j. a. , & sofia , u. j. 1997 , apj , 490 , l103 meyer , d. m. , jura , m. , hawkins , i. , & cardelli , j. a. 1994 , apj , 437 , l59 moos , h. w. , et al . 2000 , apj letters , in press morton , d. c. 1978 , apj , 222 , 863 morton , d. c. , & bhavsar , s. p. 1979 , apj , 228 , 147 morton , d. c. & dinerstein , h. l. 1976 , apj , 204 , 1 morton , d. c , & underhill , a. b. 1977 , apjs , 33 , 83 mullan , d. j. , & linsky , j. l. 1999 , apj , 511 , 502 peimbert , m. 1999 , in chemical evolution from zero to high redshift , eds . j. r. walsh & m. r. rosa ( berlin : springer ) , 30 press , w. h. , et al . 1992 , numerical recipes in fortran , 2nd ed . ( cambridge : cambridge univ . press ) prinja , r. k. et al . 1992 , apj , 390 , 266 puldrach , a. w. a. , kudritzki , r. p. , puls , j. , butler , k. , & hunsinger , j. 1994 , a&a , 283 , 525 reid , a. h. n. , & howarth , i. d. 1996 , a&a 311 , 616 rogerson , j. b. , spitzer , l. , drake , j. f. , dressler , k. , jenkins , e. b. , morton , d. c. , & york , d. g. 1973 , apj , l97 rogerson , j. b. , & york , d. g. 1973 , apj , 186 , l95 sahu , m. s. 1992 , ph.d . thesis , univ . groningen sahu , m. s. et al . 1999 , apj , 523 , l159 savage , b. d. & sembach , k. r. 1996 , araa , 34 , 279 schaerer , d. , schmutz , w. , & grenon , m. 1997 , apj , 484 , l153 schmutz , w. et al . 1997 , a&a , 328 , 219 scully , s. t. , cass , m. , olive , k. a. , schramm , d. n. , truran , j. , & vangoini - flam , e. 1996 , apj , 462 , 960 shull , j. m. , & van steenberg , m. e. 1985 , apj , 294 , 599 sofia , u. j. , & jenkins , e. b. 1998 , apj , 499 , 951 strickland , d. j. , & lloyd , c. 1990 , observatory , 110 , 1 tegmark , m. , & zaldarriaga , m. 2000 , astro - ph/0004393 v3 tenorio - tagle , g. 1996 , aj , 111 , 1641 tosi , m. 1988a , a&a , 197 , 33 tosi , m. 1988b , a&a , 197 , 47 tosi , m. , steigman , g. , matteucci , f. , & chiappini , c. 1998 , apj , 498 , 226 van der hucht , k. a. , conti , p. s. , lundstrm , i. , & stenholm , b. 1981 , space sci 28 , 307 vidal - madjar , a. 2000 , in the light elements and their evolution , eds . l. da silva , m. spite , & j. r. de medieros , a.s.p . , in press vidal - madjar , a. , & gry , c. 1984 , a&a , 138 , 285 vidal - madjar , a. , laurent , c. , bonnet , r. m. , & york , d. g. 1977 , apj , 211 , 91 vidal - madjar , a. , laurent , c. , bruston , p. , & audouze , j. 1978 , apj , 223 , 589 vidal - madjar , a. , et al . 1998 , a&a , 338 , 694 walker , t. p. , steigman , g. , schramm , d. n. , olive , k. a. , & kang , h .- s . 1991 , apj , 376 , 51 watson , w. d. , 1973 , apj , 182 , l73 welty . d. e. , morton , d. c. , & hobbs , l. m. 1996 , apjs , 106 , 533 york , d. g. , & rogerson , j. b. 1976 , apj , 203 , 378 york , d. g. , spitzer , l. , bohlin , r. c. , hill , j. , jenkins , e. b. , savage , b. d. , & snow , t. p. 1983 , apj , 266 , l55

coherent state theory is shown to reproduce three categories of representations of the spectrum generating algebra for an algebraic model : ( i ) classical realizations which are the starting point for geometric quantization ; ( ii ) induced unitary representations corresponding to prequantization ; and ( iii ) irreducible unitary representations obtained in geometric quantization by choice of a polarization . these representations establish an intimate relation between coherent state theory and geometric quantization in the context of induced representations .

quant-ph0201129

the process of quantizing a classical system has been of interest since the early days of quantum mechanics and remains an active field of research . the most sophisticated quantization procedure is provided by _ geometric quantization _ this procedure , founded on kirillov s so - called _ orbit method _ @xcite , was inititated by kostant @xcite , and souriau @xcite , who extended the orbit method by capitalizing on the physical insights that come from its applications to quantum mechanics . thus geometric quantization incorporates ideas that physicists have used for many years , but provides new avenues to address quantization of more difficult systems . our primary concern in this paper is the significance of gq for the representation theory of lie groups and lie algebras . its relevance to representation theory is based on the observation that quantizing a model with an sga ( spectrum generating algebra ) , defined in the following section , is equivalent to constructing an appropriate irreducible unitary representation of that algebra @xcite . conversely , as brought to light by gq , the construction of a unitary irrep of a lie group or algebra is often equivalent to quantizing some classical hamiltonian system . thus , the theory of induced representations plays a central role in the quantization of a model and in quantum mechanics in general , as emphasized by mackey @xcite . establishing an explicit correspondence between induced representation theory and gq sheds new light on both theories . the theory of induced representations is viewed in this paper from the perspective of coherent state theory @xcite , which incorporates other inducing constructions in terms of structures and concepts that relate naturally to quantum mechanics and gq . it is shown that coherent state theory reproduces three categories of representations of the sga for an algebraic model , and that these categories are related to structures in geometric quantization . first , there are classical realizations of the sga as functions on a phase space ; gq begins with such a classical realization . the coherent state construction also yields the induced unitary ( reducible ) representations that correspond to prequantization . finally , the unitary irreducible representations , corresponding to full quantization , are obtained through means related to the choice of a polarization in gq . these techniques are illustrated by a variety of examples . a relationship between cohererent state theory and geometric quantization follows naturally from the problem that gave rise to coherent state representations ; namely , `` construct an irrreducible unitary representation of a group from the properties of a set of coherent states '' @xcite . as will be shown in the following , the expectation values of the lie algebra over a set of coherent states give a classical representation of the lie algebra . thus , regaining a unitary irrep is equivalent to quantization of this classical representation . it is found that reconstructing the irrep is possible in coherent state theory only for particular orbits which give rise to the classical representations that are described as _ quantizable _ in geometric quantization . it is the principal aim of this paper to show that coherent state methods provide an intuitive and practical framework for implementing the procedures of induced representations and geometric quantization . we thereby attempt to make these fundamental theories accessible to a wider community . the illustrative examples presented here are intentionally simple . however , it is noted that all three theories have been deployed in non - trivial ways . the theory of induced representations has made seminal contributions in physics , e.g. , to the representation theory of the poincar group @xcite and the space groups of crystals @xcite ; indeed , it is central to representation theory and quantum mechanics @xcite . geometric quantization leads to a deep understanding of the route from classical mechanics to quantum mechanics . it has been used , for example , to develop theories of quantized vortices in hydrodynamics @xcite and nuclei @xcite and to reproduce the irreducible unitary representations of compact lie groups and the holomorphic discrete series of reductive lie groups @xcite . other applications can be found in the book of guillemin and sternberg @xcite . similarly , coherent state theory has provided a fundamental understanding of classical - like behaviour in quantum mechanics , e.g. , in the field of quantum optics @xcite . it has also been used extensively @xcite , particularly in its vector coherent state extension @xcite ( outlined in the following paper ) , in the quantization of numerous physical models by explicit construction of the irreducible representations of their spectrum generating algebras ( cf . , @xcite for a long list of references ) . our hope is that the complementary aspects of these powerful methods will result in their successful application to even more challenging problems . it has long been known @xcite that there is a close relationship between the theories of coherent states , induced representations and geometric quantization . indeed , the kirillov method @xcite , from which geometric quantization emerged , was expressed as an inducing construction . this relationship was used , for example , by streater @xcite to construct the irreducible unitary representations of the semidirect product oscillator group , following both mackey and kirillov methods , and by dunne @xcite to construct the irreps of @xmath0 . it is particularly well known that the coadjoints orbits of a lie group play central roles in all three theories ( cf . , for example , chapter 15 of kirillov s book @xcite ) . the contribution of this paper is to explore the detailed relationships between the theories so that their complementary strengths may be better understood and more readily applied to problems in physics . moreover , by establishing detailed relationships , we hope to make it easier to move between the different expressions of quantization and induced representations in the solution of complex problems . with a relationship between scalar coherent state theory and gq in place , the generalization to vector coherent state theory @xcite , in which irreps of an algebra are induced from multidimensional irreps of a subalgebra , indicates new ways of applying geometric quantization to models with intrinsic degrees of freedom . conversely , the different perspective of geometric quantization suggests possibilities for generalization of the theory of induced representations . these issues will be discussed in the sequel paper . an algebraic model is defined below as a model with an sga . the quantization of an algebraic model and the construction of the irreducible unitary representations of its sga are then related problems . however , whereas quantization starts with the classical hamiltonian dynamics on a phase space , the theory of induced representations starts with the abstract sga . these and other concepts invoked in the two constructions are reviewed in this section . in classical mechanics , observables are realized as smooth real - valued functions on a connected phase space @xmath1 , i.e. , elements of @xmath2 . they form an infinite - dimensional lie algebra with lie product given by a poisson bracket . in quantum mechanics , observables are interpreted as hermitian linear operators on a hilbert space @xmath3 ; they are elements of @xmath4 and form an infinite - dimensional lie algebra with lie product given by commutation . the algebras @xmath2 and @xmath4 for a given physical system are different @xcite . however , a simple relationship may be established between finite - dimensional subalgebras of @xmath2 and @xmath4 . thus , it is convenient to consider an abstract lie algebra of observables @xmath5 that is real and finite - dimensional and can be represented classically by a homomorphism @xmath6 and quantum mechanically by a unitary representation @xmath7 . let @xmath8 and @xmath9 denote classical and quantal representations , respectively , of an element @xmath10 . then , if elements @xmath11 , @xmath12 , and @xmath13 satisfy the commutation relations @xmath14 = \rmi \hbar \ , c \ , , \ ] ] the corresponding linear operators and functions satisfy @xmath15 = \rmi \hbar \ , \hat c \ , , \ ] ] and @xmath16 where @xmath17 denotes the classical poisson bracket . in the commutation relations . the poisson bracket needs no such factor . ] ( more precisely , the classical homomorphism is given by @xmath18 , so that @xmath19 . ) let @xmath20 denote the classical algebra @xmath21 . we shall describe the algebra @xmath22 ( and hence the algebra @xmath5 ) as a _ spectrum generating algebra _ for the classical system if the values of the observables in @xmath22 are sufficient to uniquely specify a point in @xmath1 and their gradients span the tangent space of @xmath1 at every point . ( other compatible definitions of a sga can be found in the literature @xcite . for example , a classical sga can be defined by requiring that the only functions in @xmath2 that poisson commute with all elements of @xmath22 are the constant functions @xcite . ) a finite - dimensional lie algebra @xmath5 is said to be a sga for a quantal system if the hilbert space for the system carries an irreducible representation of @xmath5 . however , to be useful , one may require also that the hamiltonian and other important observables of the system should be simply expressible in terms of @xmath5 , e.g. , by belonging to its universal enveloping algebra . thus , in quantizing a classical model , we shall seek a quantal system with the same sga as the classical model . a model dynamical system that has a finite - dimensional sga is said to be an _ algebraic model_. let @xmath23 be a group of canonical transformations ( i.e. , symplectomorphisms ) of a classical phase space @xmath1 for a model . then , if @xmath23 acts transitively on @xmath1 , it is said to be a _ dynamical group _ for the model . if an element @xmath24 sends a point @xmath25 to @xmath26 , then @xmath1 is the group orbit @xmath27 and diffeomorphic to the factor space @xmath28 with isotropy subgroup @xmath29 a remarkable fact @xcite that will be used extensively in the following is that a phase space with a dynamical group can be identified with a coadjoint orbit . conversely , every coadjoint orbit is a phase space . moreover , the lie algebra @xmath5 of @xmath23 is a sga for the model . note that in quantum mechanics , one is often interested in projective representations of a given dynamical group . thus , we consider projective as well as true representations ; in practice , it is often simpler to choose a dynamical group that is simply connected so that all its representations are true representations . recall @xcite that @xmath23 has a natural adjoint action on its lie algebra @xmath5 @xmath30 where , for a matrix group , @xmath31 . @xmath23 also has a coadjoint action on the space @xmath32 of real - valued linear functionals on @xmath5 ( the dual of @xmath5 ) . the action of an element @xmath33 on an element @xmath10 is given by the natural pairing of @xmath5 and @xmath32 , expressed as @xmath34 for matrices . the coadjoint action is then defined , for @xmath35 and @xmath24 , by @xmath36 , where @xmath37 thus , the coadjoint orbit @xmath38 is diffeomorphic to the factor space @xmath39 with isotropy subgroup @xmath40 . we shall refer to an element @xmath41 of @xmath32 as a _ density_. now if a density @xmath33 is chosen such that @xmath42 then there is a diffeomorphism @xmath43 in which @xmath44 and @xmath45 . this map is known as a _ moment map _ @xcite . such a moment map defines a classical representation @xmath46 of the lie algebra @xmath5 as functions over the classical phase space @xmath1 , defined by @xmath47 with poisson bracket given by @xmath48 where @xmath49 is the antisymmetric two - form as a set of invariant vector fields on @xmath1 , @xmath49 becomes a two - form on @xmath1 . ] with values at @xmath50 given by @xmath51 ) \,,\quad \forall\ , a , b\in \mathfrak{g}\,.\ ] ] thus , the moment map @xmath52 defines a symplectic form @xmath49 on @xmath53 . this form is known to be nondegenerate and the map @xmath54 is a symplectomorphism . geometric quantization formalizes the ideas of dirac @xcite concerning the quantization of a classical system . the idea is to replace the classical ( poisson bracket ) algebra of functions on a phase space by a unitary representation ; i.e. , map each classical observable ( function ) @xmath55 to a hermitian operator @xmath56 such that , if @xmath57 then @xmath58 = \rmi\hbar \ , \hat{f}_3 \ , .\ ] ] there are some additional requirements of the _ dirac map . _ first , all constant functions must map to multiples of the identity operator @xmath59 . second , the unitary representation should be irreducible . gq solves the first part of the dirac problem by a construction known as prequantization . however , it is now known that there are generally no irreducible unitary representations of the full algebra of observables @xcite . thus , a complete solution of the dirac problem is impossible . when there exists a finite - dimensional sga for the model then , under certain quantizability conditions , the restriction to the sga of the representation given by prequantization becomes fully reducible . moreover , gq gives a prescription for the reduction in terms of a _ polarization_. for details of the techniques of gq , see woodhouse @xcite . it will be shown in the following that the reducible representation of a sga given by prequantization of a classical phase space diffeomorphic to @xmath60 integrates to the reducible representation of @xmath23 induced from a one - dimensional representation of @xmath61 in the standard theory of induced representations . likewise , the irreducible representations obtained by introducing a polarization are obtainable by the coherent state inducing construction . in this section , we show how the coherent state construction reproduces the three categories of representations of the sga of an algebraic model : classical realizations , reducible unitary representations corresponding to prequantization , and irreducible unitary representations of a full quantization . let @xmath23 with lie algebra @xmath5 denote the dynamical group of an algebraic model and let @xmath62 denote an abstract ( possibly projective ) unitary representation of @xmath23 on a hilbert space @xmath3 . there is no need to make a precise specification of @xmath62 . for example , if @xmath23 has a right invariant measure @xmath63 , @xmath3 could be the space @xmath64 of square integrable functions with respect to this measure and @xmath62 the regular representation . or , for application to models of many - particle systems , @xmath3 might be the standard many - particle hilbert space @xmath65 of square - integrable functions of many - particle cartesian coordinates and @xmath62 a weil @xcite or schrdinger representation ( see section [ subsubsec : schrodinger ] ) . for notational convenience we denote the representation @xmath66 of an element @xmath67 by @xmath68 . in section [ sect : coadorbits ] , the classical phase space for an algebraic model is shown to be expressible as a coadjoint orbit . in this section , we show that these coadjoint orbits can be projected from group orbits in the hilbert space . for any state @xmath69 of unit norm ( i.e. , @xmath70 ) there is a system of coherent states @xcite @xmath71 and a corresponding set of dual states @xmath72 moreover , there is a natural identification of any state with a density , i.e. , an element of the dual algebra @xmath32 . thus , the coherent states define a system of densities @xmath73 where @xmath74 it follows that the coherent states determine a coadjoint orbit @xmath75 they also determine a map @xmath76 from the lie algebra @xmath5 ( cf . section [ sect : coadorbits ] ) to functions on @xmath53 in which an element @xmath67 is mapped to a function @xmath77 with values @xmath78 the map @xmath76 is a classical representation of @xmath5 . for , if @xmath79 = \rmi\hbar \ , \hat c$ ] , then the corresponding functions @xmath80 , @xmath81 , and @xmath82 satisfy a poisson bracket relationship defined by @xmath83|g\rangle = \mathcal{c}(g ) \ , .\ ] ] the poisson bracket can be expressed in terms of local coordinates for @xmath53 as follows . first observe that @xmath53 is diffeomorphic to the coset space @xmath60 , where @xmath84 then the lie algebra @xmath85 of the stability subgroup @xmath61 is the set @xmath86 ) = 0 \ , , \forall\ , a\in \mathfrak{g}\ } \ , .\ ] ] thus , if @xmath87 is a basis for @xmath85 and @xmath88 completes a basis for @xmath5 , coordinates are defined for elements of @xmath23 about a point @xmath89 by setting @xmath90 it follows that at @xmath91 , @xmath92|0\rangle = 0 \ , .\ ] ] however , the corresponding derivatives with respect to the @xmath93 coordinates do not , in general , vanish . thus , the set @xmath93 serve as local coordinates for @xmath94 . with respect to these coordinates , we then define @xmath95 where @xmath96 is a coefficient in the expansion @xmath97 and @xmath98|0\rangle \ , .\ ] ] the poisson bracket of equation ( [ eq : pb ] ) is then expressed in the familiar form @xmath99 where the matrix @xmath100 is defined such that @xmath101 in this form , the poisson bracket can be extended to all @xmath102 functions on @xmath53 . it will be noted that the above results are expressed in terms of a specific choice of the basis elements @xmath103 of the lie algebra . thus , it is important to ask if the results depend on the choice . two distinct kinds of transformation of the basis are possible : @xmath104 and @xmath105 combinations of the two kinds are also possible but it is instructive to consider them separately . transformations of the first kind leave the subspace of the lie algebra spanned by the elements @xmath103 invariant . they generate normal coordinate transformations on @xmath53 . however , all physical expressions , like the values of observables and their poisson brackets , are manifestly covariant relative to such coordinate transformation . transformations of the second kind , corresponding to changes of the linear span of the @xmath106 basis , are known as _ gauge transformations_. because of the definition of the intrinsic subalgebra , the symplectic form @xmath107 is seen to be invariant under a gauge transformation . moreover , from the definition ( [ eq:25 ] ) , it follows that @xmath108|0\rangle \,.\ ] ] thus , it follows that @xmath109 and hence the poisson bracket are also gauge invariant . note that different classical representations result from different choices of the state @xmath110 , even in the case when the representation @xmath62 is irreducible . given an abstract unitary representation @xmath62 of the dynamical group @xmath23 over a hilbert space @xmath3 , a coherent state representation @xmath111 is defined by specification of a functional @xmath112 . a state @xmath113 then has coherent state wave function @xmath114 defined over @xmath23 by @xmath115 a hilbert space @xmath116 for a coherent state representation @xmath111 is the completion of the space of such coherent state wave functions with respect to the inner product @xmath117 where the inner product on the right is that of @xmath3 . the coherent state representation @xmath111 is then defined by @xmath118(g ' ) = \psi ( g'g ) \,.\ ] ] clearly there are many coherent state representations depending on the choice of the functional @xmath119 and hilbert space @xmath3 . if the functional @xmath119 is chosen such that @xmath120 where @xmath121 is a one - dimensional representation of the subgroup @xmath122 , we say that the coherent state representation @xmath111 is induced from the representation @xmath121 of @xmath61 . if @xmath23 is compact , the space @xmath116 of coherent state wave functions is contained in @xmath64 and @xmath111 is isomorphic to a subrepresentation of the regular representation @xcite ( this can be seen from the inner product given in section [ sect : innerproducts ] ) . more generally , if @xmath123 is such that @xmath116 consists of all functions which satisfy equation ( [ eq : hconstraint ] ) and whose absolute values are in @xmath124 , then @xmath111 is said to be a _ standard _ ( mackey ) induced representation @xcite . in general , the representation @xmath111 is highly reducible . it will be shown in section [ subsec : irreps ] that , by imposing extra conditions on the choice of the functional @xmath119 , it is possible to proceed directly to the irreducible representations of quantization . the above coherent state wavefunctions are defined as functions over the group @xmath23 . however , in practical applications , it is generally more useful to represent them as functions over a suitable set of @xmath39 coset representatives @xcite . recall that a set of coset representatives @xmath125 defines a unique factorization @xmath126 , with @xmath127 , of every @xmath24 . hence , it follows from the identity ( [ eq : hconstraint ] ) that the restriction of @xmath128 to the subset @xmath129 is sufficient to uniquely define @xmath114 . often it is also convenient to consider factorizations of the type @xmath130 with @xmath131 and @xmath132 , where @xmath133 is a subset of @xmath134 coset representatives and @xmath135 and @xmath136 are the complex extensions of @xmath137 and @xmath23 , respectively . note that the identity ( [ eq : hconstraint ] ) does not require that the wavefunctions are functions on @xmath138 ; in general , @xmath114 need only be a section of a complex line bundle associated to the principal bundle @xmath139 . the hilbert space @xmath116 , then , can be viewed as a space of sections of a complex line bundle over @xmath138 @xcite . we now show that , if a representation @xmath111 is induced from a representation @xmath121 of @xmath61 having the property that @xmath140 then the corresponding representation of the lie algebra @xmath5 , defined by @xmath141(g ) = \langle \varphi | t(g ) \hat{a } |\psi\rangle = \psi(ga ) , \quad a \in \mathfrak{g } \ , , \ ] ] where @xmath142 is a subrepresentation of that given by prequantization . ( such a relationship was shown in more restricted contexts , for example , by dunne @xcite and rawnsley @xcite . ) note , however , that for this prequantization to be possible , the representation @xmath121 of @xmath85 defined by @xmath41 must be a subrepresentation of the restriction to @xmath143 of some unitary representation @xmath62 of @xmath5 . when this condition is satisfied , we say ( in the language of geometric quantization ) that the classical representation of @xmath5 defined by @xmath41 is _ quantizable_. first observe that @xmath144 where @xmath145 . substitution of the identities @xmath146 where @xmath147 into the expansion equation ( [ eq : expansionofx ] ) of @xmath148 , then gives the explicit expression @xmath149(g ) = \sum_i a^i(g ) \chi ( a_i ) \psi(g ) + \rmi \hbar \sum_\nu a^\nu(g ) ( \partial_\nu \psi)(g ) \ , \label{eq : csrep}\ ] ] for the action of @xmath150 on coherent state wave functions . the expression ( [ eq : csrep ] ) of @xmath111 depends on the expansion ( [ eq : expansionofx ] ) of @xmath148 . thus it is coordinate - dependent and gauge - dependent . however , it can be expressed in a covariant form by taking advantage of the symplectic structure of the classical phase space . by expanding the classical representation @xmath151 of @xmath67 , @xmath152 equation ( [ eq : csrep ] ) becomes @xmath153(g ) = \mathcal{a}(g)\psi(g ) + \rmi\hbar \sum_\nu a^{\nu}(g ) ( \nabla_\nu \psi)(g ) \ , , \ ] ] where @xmath154 the first term , @xmath155 , of equation ( [ eq:3.30 ] ) is manifestly covariant . moreover , from the definition , @xmath156(g ) = \psi(a(g)g)$ ] , it follows that @xmath157 thus , the second term is also covariant . we shall refer to the operator @xmath158 , defined such that @xmath159(g ) = \sum_\nu a^{\nu}(g ) ( \nabla_\nu \psi)(g)\,,\ ] ] as a covariant derivative , in accordance with standard terminology . the induced representation @xmath150 of an arbitrary element @xmath160 is then expressed in the covariant form @xmath161 a similar expression can also be derived within the framework of berezin quantization @xcite . it is shown in the appendix that , for a particular choice of gauge , @xmath162 is expressible as a sum @xmath163 where @xmath164 is a hamiltonian vector field defined by the classical function @xmath165 and @xmath166 is a one - form ( gauge potential ) . it is also shown that the symplectic two - form @xmath49 of the manifold @xmath53 is the exterior derivative @xmath167 thus , @xmath166 is a symplectic potential for @xmath53 . note that the values of the symplectic potential @xmath166 depend on the choice of @xmath103 ( cf . equation ( [ eq : coordinatechart ] ) ) . as a consequence , @xmath166 is only defined to within a gauge transformation @xmath168 . however , its exterior derivative is gauge independent . the coherent state representation @xmath169 is now observed to be of the standard form of prequantization in the theory of geometric quantization . thus , prequantization of an algebraic model is equivalent to a standard ( mackey ) representation induced from a one - dimensional irrep of a suitable subgroup . however , depending on the choice of functional @xmath170 and starting hilbert space @xmath3 , the hilbert space @xmath116 of a coherent state representation may be an invariant subspace of that of prequantization . indeed , as we show in the following section , it is often possible to choose the functional @xmath170 such that the coherent state representation is irreducible . geometric quantization shows that prequantization can be extended to the whole infinite dimensional classical algebra of all functions on the phase space . however , the extension is not fully reducible and , as presently formulated , does not apply generally to an arbitrary coherent state representation , i.e. , a coherent state representation that is not equivalent to a standard induced representation . the theory of induced representations also extends , albeit in different ways . for example , it is possible to induce coherent state representations from a representation of a subgroup @xmath171 for which @xmath172 is not symplectic . it is also possible to induce representations that are not unitary . some of the possibilities will be illustrated with examples in section [ sec : scalarcsexamples ] . the full quantization of an algebraic model corresponds to construction of an irreducible unitary representation of its sga . the usefulness of scalar coherent state induction , and indeed the full vcs theory , resides in its facility to construct such representations in a practical and computationally tractable manner . all that is needed is a functional @xmath170 that uniquely characterizes an irrep . such a functional can often be defined , for example , by extending the condition ( [ eq : hconstraint ] ) to a suitable subgroup @xmath173 in the chain @xmath174 , where @xmath136 is the complex extension of @xmath23 . in making the extension from a real subgroup @xmath122 to a complex subgroup @xmath175 , we recall that a similar extension @xcite , applied to kirillov s orbit method , resulted in a method for constructing certain irreducible representations for semisimple lie groups . in making this extension , a technical concern is that , whereas a representation of a sga @xmath5 extends linearly to the complex extension @xmath176 of @xmath5 , the corresponding extension of the generic unitary representation @xmath62 of the real group @xmath23 may not converge for all of @xmath136 . however , it is sufficient for the purpose of defining an irreducible coherent state representation if the action of @xmath62 on some dense subspace @xmath177 can be extended to a suitable subset @xmath178 of a subgroup @xmath175 which contains @xmath61 . let @xmath179 denote a one - dimensional irrep of @xmath175 which restricts to a unitary irrep @xmath121 of @xmath180 . now suppose a functional @xmath181 is chosen such that @xmath182 it will be shown by examples in the following sections that , for many categories of groups , there are natural choices of @xmath173 and its representation @xmath179 for which the corresponding coherent state representation is irreducible . subgroups satisfying these conditions are familiar in the holomorphic induction of irreducible representations @xcite . for example , if @xmath23 were semisimple and the isotropy subgroup @xmath183 for the coadjoint orbit @xmath39 , as defined above , were a cartan subgroup , then a suitable subgroup @xmath175 would be the borel subgroup generated by @xmath61 and the exponentials of a set of raising ( or lowering ) operators . a suitable one - dimensional representation of @xmath173 would then be defined by a dominant integral highest weight for a unitary irrep of @xmath23 . more generally , if @xmath61 were a levi subgroup , @xmath173 would be parabolic . non - unitary irreps can also be induced in this way . however , they are not normally described as quantizations . the construction outlined above is a generalization of holomorphic induction and , for convenience , in this situation we will speak of `` the representation of @xmath23 induced from a representation of a subgroup @xmath184 . '' apart from imposing the stronger condition ( [ eq : nconstraint ] ) , the coherent state construction is the same as in section [ sect : scalarreps ] . however , the stronger condition restricts the set of coherent state wave functions to a subset with the result that the coherent state representation becomes an irreducible subresentation of that given by prequantization . now if a unitary coherent state representation @xmath111 of a dynamical group @xmath23 induced from a representation @xmath185 of a subgroup @xmath175 defines an irreducible representation of the lie algebra @xmath5 and if the representation @xmath179 satisfies the equality @xmath186 then we say that @xmath111 is a quantization of the classical representation of @xmath5 defined by @xmath41 . note , however , that for this quantization to be possible the representation @xmath121 of @xmath187 must extend to a representation @xmath179 of a subalgebra @xmath188 which is contained in a unique irrep of @xmath176 which restricts to a unitary irrep of @xmath5 . in the theory of geometric quantization , one would say that the choice of subgroup @xmath175 defines an _ invariant polarization _ @xcite . recall that a basis for @xmath189 defines a basis for the complex extension of the tangent space at every point of the classical phase space @xmath39 . a polarization provides a separation of the tangent space at each point of this phase space into canonical space - like and momentum - like subspaces . let @xmath190 denote the lie algebra of @xmath173 . according to woodhouse @xcite , the subalgebra @xmath191 generates an invariant polarization if it satisfies the conditions : 1 . @xmath192)=0 $ ] for any @xmath193 2 . @xmath194 3 . @xmath190 is invariant under the adjoint action of @xmath61 . the first condition ensures that the polarization is isotropic , i.e. , contains no canonically conjugate observables . the second condition ensures that @xmath190 is a maximal subalgebra for which the first condition holds ; the polarization is then said to be lagrangian on @xmath195 this condition ensures that @xmath190 is sufficiently large that a representation of the group @xmath173 characterizes an irrep of @xmath196 the final condition ensures that the polarization is well - defined on @xmath195 these conditions extend the definition of a parabolic subalgebra for a semisimple lie algebra to the general situation . as we illustrate with several examples in section [ sec : scalarcsexamples ] , the choice of a suitable subgroup @xmath175 for a coherent state quantization also defines an invariant polarization according to the above criteria . for a coherent state irrep that belongs to the discrete series , an inner product is defined in the following standard way . for a given reference state @xmath197 , let @xmath198 denote the integral @xmath199 where @xmath200 is a right - invariant measure on @xmath23 . this integral converges if @xmath197 is a normalizable state vector in an irreducible subspace of @xmath62 that carries a discrete series representation . the integral then defines @xmath198 as a well - defined operator on the hilbert space . moreover , it commutes with the representation @xmath201 of any element @xmath202 . hence , by schur s lemma , @xmath198 is a multiple of the identity on the irreducible subspace containing the vector @xmath197 . it follows that the space of coherent state wave functions @xmath203 where @xmath3 is the hilbert space for the representation @xmath62 , has inner product given to within a convenient norm factor by @xmath204 if the representation @xmath121 of @xmath61 is unitary , the coherent state wave functions have the property @xmath205 then the integral over the group in equation ( [ eq : normofvector ] ) can be restricted to an integral over the coset space @xmath206 with a right invariant measure induced from that on @xmath23 . when @xmath123 is a functional on a dense subspace @xmath207 of @xmath3 , the integral @xmath198 may not converge . however , the corresponding integral over @xmath138 may converge and , if so , it is sufficient to define an inner product of coherent state wave functions in parallel with mackey s construction of inner products for induced representations of semi - direct product groups . inner products for more general coherent state representations are constructed by k - matrix methods @xcite and the related integral methods of rowe and repka @xcite . examples are given in the following to illustrate systematic procedures for carrying out the prescriptions of geometric quantization within the framework of coherent state representation theory . the first example , for the nilpotent heisenberg - weyl algebra , serves as a useful prototype for more general applications . the familiar quantizations of this algebra known as schrdinger quantization and the bargmann - segal representation @xcite are both illustrated . the second and third examples are prototypes of semisimple and semidirect sum lie algebras , respectively . the examples show that coherent state theory provides simple and natural routes through the ( sometimes subtle ) methods of geometric quantization . often there is more than one path . there may be a choice of polarization ( as illustrated by two representations , one real and one holomorphic , for the hw algebra ) and a choice of the functional form of the resulting hilbert space ( illustrated for @xmath208 ) . a generic unitary representation @xmath62 of the hw algebra is spanned by hermitian operators , @xmath209 on a hilbert @xmath3 with commutation relations @xmath210 = \rmi \hbar\ , \hat{i } \ , , \qquad [ \hat{q},\hat{i } ] = 0 \ , , \qquad [ \hat{p } , \hat{i } ] = 0 \ , .\ ] ] the representations of the so - called oscillator group , which contains the heisenberg - weyl group as a normal subgroup , were constructed by streater @xcite using both mackey and kirillov methods . let @xmath211 denote any normalized state for which @xmath212 if elements of the hw group are parameterized @xmath213 the group conjugates of @xmath214 are @xmath215 and a classical realization of the hw algebra is given by the functions @xmath216 of @xmath217 and @xmath218 with @xmath219 the poisson bracket of @xmath220 and @xmath221 , for example , is given by @xmath222|0\rangle = \mathcal{i}(p , q ) \ , .\ ] ] an induced representation of the hw algebra , equivalent to a prequantization , is now constructed by coherent state techniques as follows . ( note that we employ a different coordinate chart than used in section [ sect : scalarreps ] ; although equivalent results are obtained in any chart , the coordinates used here are standard for this example . ) choosing @xmath223 to be some normalized state satisfying equation ( [ eq : hw(3)normalizedstate ] ) , as above , and factoring out the phases generated by the identity @xmath224 , a state @xmath225 of a model with the hw algebra as its sga is assigned a coherent state wave function @xmath114 defined over the classical phase space ( the @xmath226 plane ) by @xmath227 the corresponding coherent state representation @xmath111 of an element @xmath228 of the heisenberg - weyl algebra , defined generally by @xmath229(p , q ) & = \langle 0|\rme^{-\frac{\rmi}{\hbar } p \hat{q } } \rme^{\frac{\rmi}{\hbar } q \hat{p}}\hat{a } | \psi \rangle \\ & = \langle 0| \rme^{-\frac{\rmi}{\hbar } p \hat{q } } \bigl ( \hat{a } + \frac{\rmi}{\hbar } q[\hat{p},\hat{a } ] \bigr ) \rme^{\frac{\rmi}{\hbar } q \hat{p } } |\psi\rangle \ , , } \ ] ] is then the induced representation @xmath230 this representation is obtained in geometric quantization starting with the poisson bracket of any two functions @xmath165 and @xmath231 in the hw algebra in the form @xmath232 corresponding vector fields are then defined by @xmath233 and a symplectic form , for which @xmath234 is given by @xmath235 this two - form is exact and expressible as the exterior derivative @xmath236 of a variety of one - forms . for the @xmath237 coordinates defined by setting @xmath238 , cf . equation ( [ eq : pqcoords ] ) , the identities @xmath239 imply that an appropriate one - form is @xmath240 prequantization of the @xmath241 functions then gives @xmath242 which is identical to the induced representation of equation ( [ eq : inducedhw ] ) . to obtain an irreducible representation , a functional @xmath123 on a suitably - defined dense subspace @xmath243 may be chosen such that @xmath244 this choice corresponds to choosing the real polarization @xmath190 spanned by the operators @xmath245 and @xmath224 . note that the state @xmath197 is not a normalizable state vector of the hilbert space of square - integrable functions on the hw group . nevertheless , the bra vector @xmath123 is a well - defined functional on @xmath207 . the coherent state wave functions , for states in this dense subspace , are then the @xmath217-independent functions , given by @xmath246 and the coherent state representation of the algebra reduces to the familiar irreducible schrdinger representation @xmath247 to obtain a classical bargmann - segal representation of the hw algebra , choose any normalized state @xmath110 in the hilbert space such that @xmath248 where @xmath249 with group elements parameterized by the factorization @xmath250 the group conjugates of @xmath251 are given by @xmath252 and the corrresponding conjugates of @xmath245 and @xmath253 are @xmath254 this parametrization leads to a classical realization of the hw algebra in which @xmath255 map to functions @xmath256 of @xmath257 and @xmath258 defined by @xmath259 and with poisson bracket given , for example , by @xmath260|0\rangle = \mathcal{i}(z , z^ * ) \ , .\ ] ] with @xmath223 , a state @xmath225 is now assigned a coherent state wave function @xmath114 defined over the complex @xmath257 plane by @xmath261 the corresponding coherent state representation @xmath111 of an element @xmath228 of the heisenberg - weyl algebra , defined generally by @xmath262(x ) = \langle\varphi| \rme^{-z^*\hat{a}^\dagger } \rme^{z\hat{a}}\hat{a } | \psi \rangle = \langle\varphi| \rme^{-z^*\hat{a}^\dagger } ( \hat{a } + z[\hat{a},\hat{a } ] ) \rme^{z\hat{a } } |\psi\rangle \ , , \ ] ] then gives the prequantization @xmath263 to obtain an irreducible representation , define @xmath197 to be the vacuum state for which @xmath264 this state satisfies the equation @xmath265 and defines a complex polarization @xmath191 spanned by the operators @xmath266 and @xmath267 . the coherent state wave functions are now the holomophic functions , given for @xmath225 in the dense subspace of @xmath3 generated by the action of finite powers of @xmath268 on the vacuum state by @xmath269 the corresponding coherent state representation of the hw algebra is now the well - known bargmann - segal representation @xmath270 which is known to be irreducible . the hilbert @xmath116 space for this irrep is inferred in coherent state theory from the requirement that @xmath271 should be the hermitian adjoint of @xmath257 for a unitary representation . thus , @xmath116 has an orthonormal basis @xmath272 with @xmath273 and inner product @xmath274 from the observation that @xmath275 it is also determined that @xmath116 is the space of holomorphic functions with norm @xmath276 this hilbert space @xmath116 is the well - known bargmann - segal space of entire analytic functions . representations of the groups @xmath278 and @xmath0 were constructed within the framework of kirillov s orbit method by dunne @xcite . we consider here only @xmath277 ( isomorphic to @xmath279 ) as an example of a semisimple lie algebra . suppose the regular representation @xmath62 of the @xmath277 algebra is spanned by three components of angular momentum @xmath280 with commutation relations @xmath281 = \rmi \hat{s_k}\ , , \quad i , j , k\ ; { \rm cyclic,}\ ] ] acting on @xmath282 . elements of the su(2 ) group can be parameterized in many ways . the standard parameterization , in terms of euler angles , @xmath283 leads to a classical realization of the @xmath277 lie algebra and a prequantization . however , with this parameterization , it is not so easy to identify a polarization and an irreducible subrepresentation . parameterizations that lead naturally to irreducible quantizations are defined as follows . because the lie algebras @xmath277 and @xmath284 are both real forms of @xmath285 , it follows that the irreps of @xmath277 are defined by corresponding finite - dimensional irreps of @xmath284 . thus , it is useful to regard the operators @xmath286 as spanning a finite - dimensional irrep of @xmath284 and use the iwasawa factorization to represent an element @xmath287 in the parameterized form @xmath288 where @xmath289 . let @xmath290 be a normalized state such that @xmath291 where @xmath292 is real . we then obtain the classical realization of the @xmath277 algebra , @xmath293 , as functions on a cylinder @xmath294 the poisson bracket of these functions @xmath295 \rme^{-\rmi\theta\hat{s}_2 } \rme^{-z\hat{s}_- } |0\rangle \,,\ ] ] gives @xmath296 it can also be expressed in the classical form @xmath297 the quantizability condition is that @xmath298 should be an integer . prequantization is then given by choosing @xmath197 to be an eigenstate of @xmath299 with eigenvalue @xmath292 ( a half integer ) so that @xmath300 coherent state wave functions for the induced representation are now defined by @xmath301 and the corresponding representation of the infinitesimal generators of @xmath208 is defined in the usual way by @xmath302 ( z,\theta ) = \langle \varphi| \rme^{z\hat s_- } \rme^{\rmi\theta \hat s_2 } \hat s_i|\psi\rangle \ , .\ ] ] this equation gives @xmath303 immediately as @xmath304 from @xmath305 ( z,\theta ) & = \langle\varphi| \rme^{z\hat s_-}\rme^{\rmi\theta \hat s_2 } \hat s_1 |\psi\rangle \nonumber \\ & = \langle\varphi| \rme^{z\hat s_- } [ \hat s_1\cos\theta + \hat s_3 \sin\theta ] \rme^{\rmi\theta \hat s_2 } |\psi\rangle \ , , \\ { [ } \gamma ( { s_3})\psi ] ( z,\theta ) & = \langle\varphi| \rme^{z\hat s_- } [ \hat s_3\cos\theta - \hat s_1 \sin\theta ] \rme^{\rmi\theta \hat s_2 } |\psi\rangle \ , , \end{aligned}\ ] ] and the observation that @xmath306 \rme^{z\hat s_- } \ , , } \ ] ] it follows that @xmath307 + \cos \theta\big[\frac{\partial}{\partial z } + \frac{\partial}{\partial \theta}\big ] \ , , \\ \gamma ( { s_3 } ) = \cos \theta \big [ m + z\frac{\partial}{\partial z}\big ] - \sin \theta\big[\frac{\partial}{\partial z } + \frac{\partial}{\partial \theta}\big ] \ , .}\ ] ] an irreducible subrepresentation results if @xmath197 is chosen to be a highest weight state , so that @xmath298 is a positive integer ( which we now call @xmath308 ) , and satisfies the equations @xmath309 where @xmath310 is the adjoint of @xmath311 . the coherent state wave functions then become independent of @xmath257 , @xmath312 and are seen to be functions on the circle . the coherent state representation reduces to @xmath313 which is that of an @xmath277 irrep of angular momentum @xmath314 . it corresponds to the quantization obtained by choosing the polarization @xmath190 to be the borel subalgebra of @xmath315 spanned by @xmath316 and @xmath317 . the inner product and hilbert space for this irrep are inferred in coherent state theory from the requirement that , for a unitary representation , @xmath318 should be the hermitian adjoint of @xmath319 and vice versa . thus , an orthonormal basis @xmath116 is constructed by the systematic methods of k - matrix theory @xcite . the inner product is also given in integral form by the methods of @xcite . equivalent holomorphic representations are obtained by choosing the same polarization but a different factorization of an @xmath320 group element @xmath321 then , for @xmath110 again such that equation ( [ eq : classicalsu2state ] ) is satisfied , we obtain a classical realization of the @xmath277 algebra , @xmath322 , with @xmath323 the poisson bracket of these functions @xmath324 \rme^{-z\hat{s}_+ } \rme^ { -y\hat{s}_- } |0\rangle \,,\ ] ] again gives @xmath325 and is now expressed in the classical form @xmath326 with @xmath197 an eigenstate of @xmath299 of eigenvalue @xmath292 ( with @xmath298 an integer ) , coherent state wave functions are given by @xmath327 and the corresponding representation of the infinitesimal generators of @xmath208 is defined in the usual way by @xmath302 ( y , z ) = \langle\varphi| \rme^{y \hat s_- } \rme^{z \hat s_+ } \hat s_i|\psi\rangle \ , .\ ] ] one finds that @xmath328 this representation is also obtained by prequantization of the above classical realization . an irreducible subrepresentation is again obtained by requiring @xmath197 to be a highest weight state satisfying equation ( [ eq : su2highestweight ] ) . the coherent state wave functions then become independent of @xmath329 , @xmath330 and holomorphic functions of the variable @xmath257 . the coherent state representation reduces to @xmath331 which is that of an @xmath277 irrep of angular momentum @xmath314 . it corresponds to the quantization obtained by choosing the polarization @xmath190 to be the borel subalgebra of @xmath315 spanned by @xmath316 and @xmath317 . note that the last two examples involve the same polarization , but give different realizations . the euclidean group in two dimensions @xmath332 so(2)$ ] can be realized as the group of translations and rotations in a real two - dimensional euclidean space . its infinitesimal generators are two components @xmath333 of a momentum vector and an angular momentum @xmath334 . alternatively , it can be realized as the dynamical group of a two - dimensional rotor , e.g. , a particle moving in a circle . a set of observables for such a rotor is given by a pair of @xmath335 coordinate functions and again an angular momentum . let @xmath62 be the regular e(2 ) representation , with observables in the algebra satisfing the commutation relations @xmath336=0 \ , , \qquad [ \hat{l},\hat{x}]= \rmi\hbar \hat{y } \ , , \qquad [ \hat{l},\hat{y } ] = -\rmi\hbar \hat{x } \ , .\ ] ] group elements in @xmath337 can be parameterized @xmath338 let @xmath110 be a state in the hilbert space of @xmath62 having expectation values @xmath339 this state defines a classical realization of the @xmath337 lie algebra in which @xmath340 map to @xmath341-independent functions @xmath342 over @xmath343 defined by @xmath344 and for which the poisson bracket is given by @xmath345 now let @xmath123 be a functional on a dense subspace @xmath207 of the hilbert space for the representation @xmath62 such that @xmath346 the space of coherent state wave functions , defined for each @xmath347 by @xmath348 is then isomorphic to the space of square integrable functions on a cylinder with respect to the standard @xmath349 measure . this space carries a reducible representation of @xmath337 for which @xmath350 this representation is the same as that obtained for @xmath337 by prequantization of the rotor . to obtain an irreducible representation , one must choose a subalgebra that in geometric quantization defines a polarization . a suitable subalgebra is the lie algebra of the normal subgroup @xmath351 . let @xmath123 be a functional on a dense subspace of the hilbert space of @xmath62 such that @xmath352 the space of coherent state wave functions , defined for each @xmath347 by @xmath353 is now @xmath354 , the space of square integrable functions on the circle with respect to the standard @xmath355 measure . the coherent state representation of the @xmath337 lie algebra is now irreducible on @xmath354 and given by @xmath356 the theory of geometric quantization provides a sophisticated perspective on the underlying principles for quantization of a classical model . on the other hand , the theory of induced representations is one of the most versatile procedures for constructing representations of lie groups and lie algebras . as emphasized by researchers in both fields , the two theories have much in common and both contribute substantially to the description of quantum systems . unfortunately , because of their formidable mathematical expressions , they are not readily accessible to most physicists . thus , it is useful to know that both theories can be expressed in the language of coherent state representations , a language that has been specifically developed to provide practical methods for performing algebraic calculations in physics . it has been shown in this paper that the coherent state construction yields the three types of representations of the sga of an algebraic model involved in a quantization scheme : classical realizations , prequantizations , and full quantizations ( unitary irreps ) . examples have also been given to illustrate how the coherent state approach provides an intuitive path through the techniques of geometric quantization for algebraic models . thus , we are optimistic that the coherent state methods presented here will serve to make the methods of induced representations and geometric quantization accessible to a wider community . by expressing the methods of geometric quantization in the language of coherent state theory , we are also optimistic that the many techniques developed for the practical application of induced representation theory to the solution of physical problems will be equally useful for practical applications of the methods of geometric quantization . it is interesting that the classical representations given by coherent state methods automatically take into account the inherent limitations , imposed by the uncertainty principle , on an experimentalist s ability to measure an observable precisely . in the conventional interpretation of quantum mechanics , the expectation @xmath357 of an observable is identified with the mean value of many ( precise ) measurements of the value of the observable when the system is in a state @xmath225 . thus , the distribution of experimental values , using an ideal measurement in which the only limitations on accuracy are quantum mechanical , is given by the variance @xmath358 it is remarkable then that the mean expectation values of the observables of a sga define functions on a coadjoint orbit which are precisely those of a classical representation . it is also noted that a given unitary irrep of a sga can give rise to many classical realizations . for example , if a freely rotating object had squared angular momentum given by @xmath359 in a unitary so(3 ) irrep , then the corresponding classical value given by @xmath360 with @xmath361 can have values ranging from zero to @xmath362 . the maximum value of @xmath362 would be obtained when @xmath225 is a minimum uncertainty ( e.g. , a highest weight ) state . if an experimentalist could put the object into such a minimal uncertainty state , then he / she would obtain an integer value for @xmath334 and have determined the quantal state of the rotor precisely . however , in a practical situation , the uncertainties in a given experimental situation inevitably exceed the minimal uncertainties permitted by quantum mechanics . it is now known that scalar coherent theory has a natural generalization to a vector coherent theory @xcite in which an irrep of a lie algebra is induced from a multidimensional irrep of a subalgebra . in a sequel to this paper we shall show that vcs theory can also be expressed in the language of geometric quantization and that it corresponds to the quantization of a model with intrinsic degrees of freedom . we thank c bahri for useful discussions . we are also indebted to g rosensteel and the referees for bringing important references to our attention . sdb acknowledges the support of a macquarie university research fellowship . this paper was supported in part by nserc of canada . if @xmath165 is a classical representation of an element @xmath160 as a function on @xmath363 then the corresponding hamiltonian vector field @xmath164 on @xmath53 is defined to operate on a function @xmath364 on @xmath53 such that @xmath365 is equal to the poisson bracket @xmath366 . this requirement means that @xmath164 must satisfy @xmath367(g ) = \sum_{\mu\nu } \partial_\mu \mathcal{a}(g)\ , \omega^{\mu\nu}\ , \partial_\nu f(g ) = \sum_{\nu } a^\nu ( g)\,\partial_\nu\mathcal{b}(g ) \ , , \ ] ] where @xmath96 is a coefficient in the expansion @xmath368 and @xmath369 is defined by @xmath370 * claim . * _ let @xmath165 be a classical representation of an element @xmath67 as a function on @xmath371 and @xmath164 the corresponding hamiltonian vector field on @xmath53 . then the covariant derivative @xmath162 ( cf equation ( 48 ) ) acts on a coherent state wave function by @xmath372(g ) = \sum_\nu a^\nu ( g ) \big ( \partial_\nu + \frac{\rmi}{\hbar } \rho ( a_\nu ) \big)\psi(g ) .\ ] ] it is expressible in the form @xmath373 where @xmath166 is a symplectic potential ( one - form ) for @xmath53 . _ before proving this claim , we consider first the expression of the vector field @xmath164 as a derivation in terms of local coordinates by means of the following observation . * observation . * _ if @xmath374 , then @xmath375 and @xmath376 where @xmath377 + \frac{1}{3 ! } [ x(x),[x(x),a_\nu]]+ \cdots\ ] ] _ * proof . * the first identity of equation ( [ eq : a.deriv ] ) follows from the observation that @xmath378 and hence that @xmath379 the second identity of equation ( [ eq : a.deriv ] ) is obtained directly from @xmath380 equation ( [ eq : b.deriv ] ) is obtained similarly . if @xmath381 then it follows from the observation that to leading order in @xmath382 , @xmath383 g(x ) .\ ] ] hence , with the expansion @xmath384 the observation leads to the expression for the derivatives of a function @xmath364 on @xmath23 @xmath385 however , when @xmath386 ( a classical observable in the sga ) , the second term on the right is zero , due to the fact that @xmath387 satisfies the equations @xmath388 , and hence @xmath389 thus , we obtain @xmath390 and @xmath391(g(x ) ) = \sum_{\mu\nu } a^\nu ( g(x ) ) \overline\lambda_\nu^\mu(x)\frac{\partial}{\partial x^\mu } \mathcal{b}(g(x ) ) \ , , \ ] ] where @xmath392 is the inverse of the matrix @xmath393 . * proof of the claim . * the action of @xmath394 , defined by equation ( [ eq : xhamlocal ] ) can be extended to coherent state wave functions . however , while the action ( [ eq : xhamlocal ] ) of @xmath395 on functions over @xmath396 is covariant , the action on coherent state wave functions is not covariant ; these wave functions are not defined on @xmath53 but have extra phase factors and , as a consequence , they are not gauge invariant . in particular , @xmath397 is not generally zero . thus , in evaluating the covariant derivative , defined by equation ( [ eq : covderiv ] ) , both terms on the right hand side of equation ( [ eq : xderiv ] ) must be included to give @xmath398 the covariant derivative of @xmath114 at @xmath399 is then @xmath400(g(x ) ) = \sum_\nu a^\nu ( g(x))\overline\lambda^\mu_\nu(x ) \big(\frac{\partial}{\partial x^\mu } + \frac{\rmi}{\hbar } \theta_\mu(x ) \big ) \psi(g(x ) ) \ , , \ ] ] where @xmath401 thus , with the interpretation of @xmath402 as a component of a one - form @xmath403 , so that @xmath404 the covariant derivative is expressed @xmath405 as claimed . it remains to be shown that the one - form @xmath166 is a symplectic potential , i.e. , that the symplectic form on @xmath53 is given by @xmath406 first observe , from equation ( [ eq : thetanu ] ) , that @xmath407 now , with @xmath408 written in the form @xmath409 it follows , from multiple use of the observation , that @xmath410 \,.\ ] ] thus , we obtain @xmath411)\,,\ ] ] which from the definition of @xmath408 ( equation ( [ eq : anuexpansion ] ) ) and @xmath412 ( equation ( [ eq : omega ] ) ) gives @xmath413 therefore , if @xmath49 is the two - form defined by equation ( [ eq : sympform ] ) , then for the vector fields defined by equation ( [ eq : xhamlocal ] ) , we obtain @xmath414 which is identical to the expression in equation ( [ eq : pbracket ] ) for the poisson bracket @xmath415 . this result confirms that the one - form @xmath166 is indeed a symplectic potential and completes the proof of the claim . 99 kostant b 1970 _ on certain unitary representations which arise from a quantization theory _ , in _ group representations in mathematics and physics , lecture notes in physics , . 6 _ ( berlin : springer ) ; _ quantization and unitary representations _ , in _ letures in modern analysis and applications iii , lecture notes in mathematics , vol . 170 _ ( berlin , springer ) mackey g w 1952 _ ann . of math . _ * 55 * 101 ; 1968 _ induced representation of groups and quantum mechanics _ ( new york : benjamin ) ; 1978 _ unitary group representations in physics , probability and number theory _ ( reading , ma : benjamin ) joseph a 1970 _ commun . _ * 17 * 210 ; gotay m j 1995 _ quantization , coherent states , and complex structures ( bialowieza 1994 ) _ ( new york : plenum ) ; gotay m j , grundling h b and hurst c a 1996 _ trans . _ * 348 * 1579 ; gotay m j and grundling h b 1997 _ rep . math . _ * 40 * 107

the stellar dark halo mass relation of galaxies at different redshifts , ( , ) , encloses relevant features concerning their physical processes and evolution . this sequence of relations , defined in the range , together with average cold dark matter ( ) halo mass aggregation histories ( mahs ) are used here for inferring _ average _ growth histories , the galaxian hybrid evolutionary tracks ( ghets ) , where hybrid remarks the combination of observational ( ) and theoretical ( ) ingredients . as a result of our approach , a unified picture of stellar and halo mass buildup , population migration , and downsizing of galaxies as a function of mass is presented . the inferred average growth histories ( ghets ) of highest and lowest mass galaxies are definitively quite different from the average mahs , ( ) , of the corresponding dark halos . depending on how a given ( ) compares with the mass at which the -to- ratio curve peaks at the epoch , ( ) , two evolutionary phases are evidenced : ( i ) galaxies in an active regime of growth when , and ( ii ) galaxies in a quiescent or passive regime when . the typical at which galaxies transit from the active ( star - forming ) to the quiescent regime , , increases with , log(/) , making evident a _ population downsizing _ phenomenon . this result agrees with independent observational determinations based on the evolution of the galaxy stellar mass function decomposition into blue and red galaxy populations . the specific star formation rate , ssfr , predicted from the derivative of the ghet is consistent with direct measures of the ssfr for galaxies at different redshifts , though both sets of observational inferences are independent . the average ghets of galaxies smaller than at ( , ) did not reach the quiescent regime , and for them , the lower the mass , the faster the later growth rate ( _ downsizing in ssfr _ ) . the ghets allow us to predict the transition rate in the number density of active to passive population ; the predicted values agree with direct estimates of the growth rate in the number density for the ( massive ) red population up to . we show that based models of disk galaxy evolution , including feedback - driven outflows , are able to reproduce the low - mass side of the relation at , but at highers strongly disagree with the ghets : models do not reproduce the strong downsizing in ssfr and the high ssfr of low mass galaxies .

1008.4796

according to the popular hierarchical clustering @xmath2 cold dark matter ( ) scenario , galaxies form and grow inside evolving dark matter haloes . a central question emerges then about what is the galaxy stellar mass , , associated on average with a given halo of mass , i.e. , the relation . the change with redshift of this relation , ( , @xmath0 ) , resumes the key astrophysical processes of galaxy stellar mass assembly in the context of the scenario . with the advent of large galaxy surveys in the last years , a big effort has been made in constraining the local relation ( 1 ) directly by estimating halo masses with galaxy - galaxy weak lensing , with kinematics of satellite galaxies or with x - ray studies ; and ( 2 ) indirectly by linking the observed statistical galaxy properties ( e.g. , the galaxy stellar mass function , the two point correlation function , galaxy group catalogs ) to the theoretical halo mass function ( hereafter @xmath12 ; for recent reviews on all of these methods see moster et al . 2010 ; behroozi et al . 2010 ; more et al . 2010 , and more references in these papers ) . the latter approach is less direct than the former one , but it allows to cover larger mass ranges , and it can be extended up to the redshifts where observed are reported ( for recent results see conroy & wechsler 2009 ; moster et al . 2010 ; wang & jing 2010 ; behroozi et al . 2010 ) . in moster et al . ( 2010 ) and behroozi et al . ( 2010 ; hereafter bcw10 ) , the ( , @xmath0 ) functionality has been constrained up to @xmath13 . in each case , different observational data sets for the and different methods for statistically assigning halo masses to the galaxies were used . the relation at @xmath10 is similar in both works ( see also baldry et al . 2008 ; drory et al . 2009 ; guo et al . 2010 ; wang & jing 2010 ) . however , the change with @xmath0 of this relation , ( , @xmath0 ) , particularly for @xmath14 , is different in both cases . although the uncertainties in the inferences of ( , @xmath0 ) are still significant , the current determinations can be used for preliminary explorations of the galaxy mass assembly process . an original approach has been introduced by conroy & wechsler ( 2009 ; hereafter cw09 ) , who proposed a parametric form for as function of @xmath0 constrained by both some observational reports of this function and the star formation rate ( sfr) relations to @xmath9 . then , the cumulative at each @xmath0 were matched to the cumulative @xmath12s in order to infer the relations at different @xmath11s ( abundance matching formalism , the simplest of the indirect methods ; e.g. , marinoni & hudson 2002 , kravtsov et al . 2004 , vale & ostriker 2004 ; see 2 ) . finally , the obtained relations at different @xmath11s were connected by using simple parameterizations of the average halo mass aggregation histories ( mahs ) in order to infer the average stellar mass buildup of galaxies as a function of mass . the previous approach has the advantage ( 1 ) of being flexible enough as to allow for explorations covering large mass and redshift ranges , and ( 2 ) of providing a bridge between observational results and theoretical work . one of the bases of this approach is the statistical abundance matching formalism , which has shown to give robust results in agreement with direct inferences of the local relation or with other indirect inferences based on observations of the two - point correlation functions or galaxy group catalogs ( see moster et al . 2010 ; bcw10 ; more et al . 2010 ; dutton et al . 2010a ) . from the purely empirical point of view , recent studies are posing a new conception in our understanding of galaxy stellar mass assembly . for example , drory & alvarez ( 2008 ) used empirical up to @xmath15 ( drory et al . 2005 ) to infer the assembly of large galaxies , and in combination with available observations of the sfr relation at different @xmath11s , they constrained the contribution of star formation ( sf ) and merging to stellar mass build up in galaxies . from this and other studies ( e.g. , bundy et al . 2006,2009 ; hopkins et al . 2007 ; pozzetti et al . 2010 ) , the trend of downsizing ( cowie et al . 1996 ) is clearly shown in the sense that the mass at which the sfr starts to drop strongly decreases with time ( from @xmath16 at @xmath17 to @xmath18 at @xmath19 according to drory & alvarez 2008 ) . this result combined with observationally inferred merger rates , led the latter authors to conclude that the ( massive ) red sequence is built up from top to down . this phenomenon was originally dubbed as _ archaeological downsizing _ ( thomas et al . 2005 ; see also for related results e.g. , daddi et . al 2004,2007 ; drory et al . 2004 ; bundy et al . 2005,2006 ; conselice et al . 2007 ; marchesini et al . 2009 ; perez - gonzalez et al . 2008 and more references therein ) , and it seems that its main driver are internal galaxy processes that efficiently quench sf in massive galaxies ( e.g. , bundy et al . 2006 ; peng et al . 2010 ) . the latter result has been confirmed by recent decompositions of the by galaxy type ( based on color , sfr , morphology , etc . ) up to @xmath9 ( e.g. , borch et al . 2006 ; bell et al . 2007 ; drory et al . 2009 ; ilbert et al . 2010 ; pozzetti et al . 2010 ; for a compilation of early observations , see hopkins et al . a systematic result obtained in these works is that blue / late - type ( active ) galaxies migrate to the red / early - type ( passive ) population involving a sequence of masses that decreases with time , a result in line with the archaeological downsizing phenomenon . the environment plays also an important role . recent observational studies , where the has been divided not only by galaxy types but also by environment , show that the population migration happens more efficiently and earlier in denser environments ( e.g. , peng et al . 2010 ) . at lower masses the inferences of the assembly are difficulted by the incompleteness limit of the samples as higher is @xmath0 . however , at least for @xmath20 , current studies show that the specific sfr ( ssfr ) of low - mass ( blue ) galaxies is surprisingly high , and higher than the ssfrs of more massive galaxies ( e.g. , bauer et al . 2005 ; noeske et al . 2007 ; zheng et al . 2007 ; bell et al 2007 ; elbaz et al . 2007 ; chen et al . 2009 ; damen et al . 2009a , b ; santini et al . 2009 ; oliver et al . 2010 ; rodighiero et al . downsizing in ssfr of low - mass galaxies _ seems to imply a delayed buildup as lower is the mass , something difficult to explain by the moment in -based models ( for models , discussions , and more references see noeske et al . 2007 ; fontanot et al . 2009 ; firmani , avila - reese & rodrguez - puebla 2010 ) . based on the scenario and the new observational inferences discussed above , a unified picture of stellar and halo mass buildup of galaxies of all sizes can be developed . by using relations constrained by bcw10 from @xmath21 to @xmath22 , we generate an analytical ( , @xmath0 ) relationship , which is continuous and differentiable at any @xmath0 . this sequence of relations in @xmath0 is combined with average halo mahs in order to : 1 . calculate the _ average _ stellar mass buildup tracks of galaxies , ( @xmath0 ) , from @xmath23 up to now ( hereafter , galaxian hybrid evolutionary tracks , ghets ) along with their halo mahs , ( @xmath0 ) ; 2 . calculate the average ssfr histories of galaxies from the corresponding specific stellar mass growth rate histories , ; 3 . determine the typical mass at each @xmath0 that marks the transition from active to passive galaxy population ( in terms of growth ) 4 . infer the transition rate in the number density of active to passive galaxies at each @xmath0 . our approach differs from cw09 in several aspects . these authors used observational inferences of the ssfr vs at different @xmath11s for constraining the evolution of their proposed ( up to @xmath9 ) , which is a partial input of their approach for calculating ( , @xmath0 ) . instead , we use more recent and updated direct constraints for this relationship ( bcw10 ) , which extends up to @xmath13 , and then predict the ssfr histories . in cw09 a simple parameterization for the average mahs was used , while we generate them by using an extended press - schechter formalism that gives results similar to those of large n - body cosmological simulations . we differ from cw09 also in several aims and results . the inferences of the quantities listed above constrain in general the multiple astrophysical processes participating in the formation and evolution of galaxies . in particular , we will compare the inferred evolutionary tracks ( ghets ) to predictions of standard @xmath24-based semi - numerical models of disk galaxy evolution and explore whether the so - called downsizing in ssfr is really an issue for models . in 2 , the method to calculate the ghets is explained . the inferred average ghets are presented in different diagrams in 3 . in 3.1 and 3.2 the evolution of the ssfr as a function of mass , the evolution of the transition mass , and the flow of active to passive galaxies at each epoch , are shown and compared with direct observational estimates . the reliability of our approach and the results are discussed in 3.3 . in 4 , -based models of disk galaxy evolution are used to calculate evolutionary tracks for low - mass galaxies and see whether they are able to reproduce or not the ghets , in particular the ssfr downsizing phenomenon . a summary and our conclusions are given in 5 . our inference of the galaxy stellar mass buildup as a function of mass ( called here ghets ) consists of two main steps : ( 1 ) we find an analytical functionality for ( , @xmath0 ) , and ( 2 ) after generating the halo mahs , ( @xmath0 ) , we combine them with ( , @xmath0 ) to infer the ghets , ( @xmath0 ) . our starting point is the determination of ( , @xmath0 ) up to the highest redshift as possible . as mentioned in section 1 , such a relationship has been recently constrained up to @xmath25 by using the statistical formalism of matching one - to - one the cumulative theoretical @xmath12 to the observed cumulative at different @xmath11s : @xmath26 ( e.g. , moster et al . 2010 ; bcw10 ) . here we will use the results from bcw10 , who started from an assumed parametric expression for ( , @xmath0 ) and explored the parameter space , including uncertainties and sample variance , with a monte carlo markov chain technique to fit jointly the observed at different @xmath11s ( they actually carried out this analysis separately for two redshift ranges : @xmath27 and @xmath28 ) . the authors have estimated and carefully taken into account all kind of uncertainties , which introduce significant scatter in the relations and affect even the shape of these relations ( see 3.3 for a discussion ) . the observational input used in bcw10 is the total local obtained in li & white ( 2009 ) from sloan digital sky survey ( sdss ) data , and the total at higher redshifts ( up to @xmath29 ) homogeneously obtained in prez - gonzlez et al . ( 2008 ) . for the halo mass function , bcw10 used the results from n - body cosmological simulations by tinker et al . since these @xmath12s refer only to distinct halos , a correction by the sub - halo abundance was introduced ( see details in bcw10 ) . the mass of the subhalo is fixed as that one at the time of accretion into the halo , @xmath30 , but the satellite galaxy s is associated with the current , i.e. , it is implicitly assumed that the satellite s will continue to evolve in the same way as for centrals of halo mass @xmath30 ( the effect of fixing instead the satellite mass at the redshift of accretion has virtually no effect on the overall relation as checked in bcw10 ) . thus , the ( , @xmath0 ) relationship used here refers to the _ overall _ galaxy population at each epoch , and this _ includes _ satellite galaxies , though under the assumptions mentioned above for this sub - population . on the one hand , there are some hints that the relation of only satellite galaxies should not differ significantly from the overall relation ( wang et al . 2006 ) . on the other hand , satellites are not a dominant population at any epoch , in particular at high redshifts . therefore , even if its relation is quite peculiar , its effect over the average relation is expected to be negligible . for more details on the assumptions and role of uncertainties in the inference of the ( , @xmath0 ) relationship we refer the reader to the original work by bcw10 ( see also it is worth mentioning that bcw10 refer to their relations as corresponding to `` central '' galaxies . this should be understood in the sense that for each ( sub)halo is considered only the of one galaxy , the central one , but not in the sense that the satellite galaxy population has been excluded from the analysis ( see above ) . a ( sub)halo may contain more galaxies besides the central one ( satellites ) . therefore , the total stellar mass within the ( sub)halo will be larger if we account for satellites . in bcw10 it was shown that the inclusion of satellites results in a `` total '' / ratio sligthly larger than the `` central '' one at small masses but significantly larger at high ( group and cluster ) masses ( bcw10 ; see also yang , mo & van den bosch 2009 ) . some of this `` excess '' stellar mass is expected to end actually in the central galaxy due to satellite infall and/or tidal stripping of stars ( other fractions of this stellar mass may end in the stellar halo or remain in the surviving satellites ) ; this could modify the obtained overall relation h@xmath31 at @xmath10 , the mass fraction of stars accreted by the central galaxy is negligible , indicating that the latter can not have grown substantially due to the accretion of satellite galaxy s stars ; rather their growth is dominated by individual evolution ( in situ sf ) . for halos larger than @xmath32 h@xmath31 , the central galaxy may increase lately in a significant fraction by satellite accretion ( dry mergers ) ; another fraction of the accreted stars end in the stellar halo ( intracluster stars ) . ] . in general , we do not expect that the overall relations inferred by bcw10 will be significantly affected by the ( unknown ) physics of satellite galaxies , at least for @xmath33 at any epoch , and for any mass at @xmath34 . figure [ galefficiency - b ] shows the -to- ratio as a function of at @xmath35 0.1 , 1.0 , 2.0 , 3.0 , and 4.0 reported in bcw10 ( solid , dashed , dot - dashed , dot - dot - dashed , and dot - dashed - dashed lines ) with the respective 1@xmath36 uncertainties ( error bars , showed only for @xmath37 and 4 ) . several features regarding galaxy mass assembly can be anticipated from this figure : \1 ) in the @xmath38 range the peak of the curves shifts from log(/)@xmath39 to @xmath40 , while log(/)@xmath41 first declines from @xmath42 at @xmath21 to @xmath43 at @xmath44 , and then recovers the value @xmath42 at @xmath13 . taking into account the uncertainty , the shape of the relations is conserved . around log(/)@xmath45 the curves cross each other . \2 ) on the left side of the diagram , below the crossing masses ( in particular from @xmath46 to @xmath47 ) , for a given , the -to- ratio increases as @xmath0 decreases , which reveals a strong late growth of for individual galaxies . hence , this side is expected to be populated mainly by _ actively _ growing ( blue ) galaxies . \3 ) just the opposite happens on the right side , above the crossing masses , which means that for individual galaxies , brakes its growth while continues growing . to illustrate this point , we plot in fig . [ galefficiency - b ] the evolution ( brown arrow ) of a completely passive galaxy of constant mass @xmath48 that starts in the isochrone @xmath23 with @xmath49 and evolves until grows to @xmath50 at @xmath6 according to a typical mah ( see 2.2 ) for this mass . if the galaxies of the sample that allow us to find the isochrones would be growing in , then the arrow would diverge from the @xmath6 isochrone . therefore , the fact that the arrow reaches the @xmath6 isochrone exactly reveals that this side of the diagram is dominated by galaxies with _ passive _ evolution . the previous considerations confirm the idea that ( , @xmath0 ) resumes the key astrophysical processes of galaxy stellar mass assembly and therefore it contains information about the evolution of individual galaxies . our aim now is to explore the implications of the ( , @xmath0 ) relations shown in fig . [ galefficiency - b ] . as we are interested in the properties of ( @xmath0 ) up to its second derivative ( ssfr and the rate of change of ssfr ) , we can not use the double parameterization in @xmath0 given by bcw10 in the ranges @xmath51 and @xmath52 , respectively , because discontinuities arise around @xmath46 . then , starting from the data of fig . [ galefficiency - b ] , we introduce a modified parameterization that is continuous in @xmath0 . the following equations , adapted from bcw10 , summarize the model to be used here for @xmath53 : @xmath54 @xmath55 the dependence on @xmath0 is introduced in the parameters of equation ( [ mh_ms ] ) as : @xmath56 where @xmath57 is the scale factor . the function @xmath58 controls the change with @xmath0 of the peak value ( ordinate ) of the curves in fig . [ galefficiency - b ] . if @xmath59 , then the curve peak ordinate is independent of @xmath0 . we chose @xmath58 in order to reproduce roughly the evolution of the peak found in bcw10 . the first two eqs . ( 2 ) control the position of the / peak at each @xmath0 ( the value of ) , while the last three ones control the shape of the / we fix the parameters in eqs . ( 1 ) and ( 2 ) expecting ( i ) to reproduce well the bcw10 ( /) relations at the redshift extremes ( @xmath21 and @xmath13 , see fig . [ galefficiency - b ] ) , ( ii ) to keep the generated relations at all @xmath11s within the 1@xmath36 uncertainty given in bcw10 , and ( iii ) to have the values of most of the parameters reasonably inside the uncertainties of bcw10 ( see their table 2 ) . the best set of parameters that we have found is reported in table 1 ( for comparison , we reproduce there also the best fit values from bcw10 for their @xmath60 case with free systematic parameters @xmath61 and @xmath62 ) . for the function @xmath63(@xmath0 ) , we use : @xmath64 the agreement of our model with the bcw10 results can be appreciated in fig . [ galefficiency - b ] ( dotted curves with the same color code as the results from bcw10 at @xmath65 ) . .best parameters for the ( , @xmath0 ) functionality [ cols= " < , > , > " , ] the ( , @xmath0 ) relationship proposed in the previous section allows us to transform an average dark halo mah , ( @xmath0 ) , into an average stellar mass growth history , ( ( @xmath0),@xmath0 ) ( the ghet ) . instead of using a parameterization for the average mahs as was done in cw09 , here we calculate them by using the special extended press - schechter formalism developed in avila - reese , firmani & hernndez ( 1998 , see also firmani & avila - reese 2000 ) . under the assumption that the primordial density field is gaussian , we calculate the overall mass distribution of progenitor halos at a given epoch @xmath66 that will be contained in a halo of mass @xmath67 at a later epoch @xmath68 . thus , after defining the cosmology and the mass power spectrum of fluctuations ( as in bcw10 , parameters very close to wmap5 are used here , komatsu et al . 2009 ) , we start from a given mass at a given time ( e.g. , @xmath6 ) and calculate its mass distribution of progenitors at a previous time step through monte carlo trials . the mass of the most massive halo ( main progenitor ) is used as the conditional for the next time step and so on . this way is obtained one realization of the mah . for each given at @xmath6 , we calculate 20000 random tracks ( mahs ) and then we average among them to get the average mah for this mass . in spite of the stochasticity , the mahs present a systematical ( hierarchical ) trend with mass : the smaller is ( @xmath6 ) , the earlier is its mass assembly on average ( avila - reese et al . 1998 ; van den bosch 2002 ; wechsler et al . 2002 ) . our average mahs are not easy to describe by a simple parametrization , but they are in good agreement with those measured in the outcome of cosmological n - body simulations ( e.g. , fakhouri , ma , & boylan - kolchin 2010 ) . the ghets in the vs. diagram , calculated as explained in 2 , are plotted with red thin solid lines in figure [ msmh ] . the blue dashed curves show the relations for @xmath65 from top to bottom ( bottom to top ) on the left ( right ) side . the ghets in figure [ msmh ] start from the @xmath23 relation and end at the @xmath6 one . two features of the ghets in such a diagram are remarkable . ( 1 ) below log(/)@xmath69 the average slopes of the ghets are @xmath70log/@xmath70log@xmath71 , greater than the slopes @xmath70log/@xmath70log@xmath72 of the relations in their low - mass side at any @xmath0 . ( 2 ) the ghets of massive halos ( log(/)@xmath73 at @xmath6 ) stop to grow , happening this earlier , the more massive are the galaxies . such a behavior is due to the folding seen in the -@xmath0 surface that inverts the spatial distribution of the ( , @xmath0 ) curves going from the left to the right side of the diagram . the same properties may be derived from figure [ galefficiency ] , where the ghets are shown as tracks that connect the / relations . here the folding is seen in the blue dashed lines as a maximum that shifts to the higher mass side as @xmath0 increases , making the lines cross each other . both figures help to understand why the ghets of massive halos attain a maximum . the ghets ending at @xmath6 with masses @xmath74log(/ ) @xmath75 are plotted with red solid lines in figure [ ghets ] . the corresponding halo average mahs are also plotted but , for comparative reasons , each one has been shifted vertically in such a way that each mah coincides with its related ghet at @xmath6 ( blue dashed lines ) . this figure shows in detail the difference between the average stellar and the halo evolutionary track shapes . for ( @xmath0=0)@xmath76 , each ghet grows faster than the corresponding mah since @xmath23 . for ( @xmath0=0)@xmath77 , as the system is more massive , the stellar assembly of the galaxy occurs earlier in time with respect to the corresponding halo . our analysis shows conclusively that the stellar mass growth of galaxies deviates from the halo mass growth , specially for low- and high - mass systems . figures [ msmh ] and [ galefficiency ] show that the more massive the galaxies , the earlier their ghets start to slow down their growth and at each epoch there is a typical halo or galaxy mass at which this growth stagnates ; this typical mass decreases with time . such a situation can be interpreted as a transition at each epoch from active to passive galaxy populations , a transition at early epochs happens only for the largest galaxies but at later epochs involves gradually smaller galaxies ( see 3.2 for a more quantitative discussion ) . this transition from active to passive regimes is observed in the evolution of the blue and red components of the . in fact , observations show directly that , despite the stellar mass growth , the blue ( late - type ) galaxy component of the is not enriched with massive galaxies with time ; on the contrary , within the uncertainty , it seems to become poorer in massive galaxies ( e.g. , bundy et al . 2006 ; bell et al . 2007 ; pozzetti et al . 2010 ) . such a phenomenon is explained by a migration of galaxies from the blue to the red component of the ; the red ( early - type ) galaxy component of the is continuously fed with the migrated galaxies . our result , in the light of such an interpretation , is remarkable because in our analysis the galaxy color bi - modality never has been taken into account . studies where the were decomposed into blue and red populations support our result in an independent way ( e.g. , bell et al . 2007 ; drory et al . 2009 ; pozzetti et al . 2010 ) . note that the prediction of a population transition and the typical mass that separates the two populations at each @xmath0 depends on the position of the folding in the @xmath0 surface . this feature reveals a fundamental aspect of the evolution . therefore , the accurate study of this feature deserves maximum efforts from the observational point of view . the fact that the more massive are the systems , the earlier they finish their assembling , transiting from the active to the passive population , will be called here _ population downsizing _ ( in the literature the term archaeological downsizing has been used to describe a related behavior of galaxies ; see e.g. , fontanot et al . 2009 , and the references therein ) . for systems less massive than @xmath78 at @xmath6 ( @xmath79 ) , the active ( growing ) phase continues still at @xmath6 on average ( the population downsizing has not happened ) , and the smaller is the system , the growth is more delayed and concentrated towards the present . the fact that the less massive is the system , the later happens its active phase of assembling , is called _ downsizing in ssfr _ ( see e.g. , fontanot et al . 2009 , firmani et al . 2010 and the references therein ) . in the different diagrams plotted in figures [ msmh ] , [ galefficiency ] , and [ ghets ] , we have introduced the corresponding evolutionary tracks ( crosses at @xmath80 and 4 connected with dotted lines ) for two ( low mass ) galaxies calculated by means of a based model of galaxy evolution which will be explained in detail in 4 . their final stellar masses are log(/ ) = 9.4 and 10.1 , respectively . we anticipate the rather different behavior of the model galaxian evolutionary tracks ( hereafter gets ) with respect to the ghets . for example , the average slope of the former in the diagram is @xmath70log/@xmath70log=1.4 , much less than the slopes of any ghet , and also less than the lower limit estimated above for the low mass side of the relations at each @xmath0 . this means that the model isochrones ( defined by connecting crosses of a given @xmath0 ) have a distribution and evolution opposite to that of the empirical relations . the stellar mass buildup of galaxies may happen due to in situ sf or accretion of stars , mainly in dry merger events . several theoretical ( e.g. , maller et al . 2006 ; guo & white 2008 ) , semi - empirical ( zheng , coil zehavi 2007 ; cw09 ; yang , mo & van den bosch 2009 ; wang & jing 2010 ) , and observational ( e.g. , bundy et al . 2006,2009 ; bell et al . 2007 ; drory & alvarez 2008 ) pieces of evidence show that the former channel completely dominates in low- and intermediate - mass galaxies at all epochs , while the latter may play a moderate role for massive ( @xmath81 , mainly red ) galaxies at later epochs ( @xmath20 ) . as a working hypothesis , we will assume that the stellar mass buildup implied by our derived average ghets is only due to in situ sf . therefore , we calculate the ssfr as the time derivative of ( @xmath0 ) divided by the current mass , , and divided by ( @xmath82 ) , where @xmath83 is the gas recycling factor due to stellar mass loss . the obtained ssfrs are then compared with directly observed ssfrs as a function of and @xmath0 . in figure [ dotm ] , the evolution of the ghet - based ssfr is plotted ( red solid line ) for the same cases shown in figure [ ghets ] . from top to bottom , the mass of the ghets increases , except for the minor inversion ( track crossing ) present in the top right side of the diagram . for all the masses , ssfr decreases as @xmath0 decreases . the overall behavior of decreasing ssfr with time agrees , by construction , with the known direct observational inferences of the cosmic stellar mass density history and its time derivative ( e.g. , drory et al . 2005 ; prez - gonzlez et al . 2008 ) , providing a comparison term to the observed cosmic sfr density history ( madau et al . 1996 ; see hopkins & beacom 2006 for an extensive compilation of observational inferences ) . we test now if the ghet - based ssfrs histories as a function of mass are in agreement with the direct observational measures . the observations of course do not refer to individual evolutionary tracks . what observers do is to determine the ssfrs of galaxy samples in different redshift bins . a commonly reported result is the average of the ssfr in different @xmath0 bins corresponding to a given range of ( the same at all @xmath0 s ) . in figures [ dotm ] and [ dotmobs ] , from top to down , the thick ( cyan , blue , green , and red ) lines correspond to the ssfr inferred from the ghets that in each @xmath0 have log(/)=9.5 , 10.0 , 10.5 , and 11.0 , respectively . in figure [ dotmobs ] , we reproduce averages of ssfr observational estimates at different @xmath0 s for galaxies in four bins ( centered in the same masses plotted for the ghets ) . these estimates were taken for @xmath10 from salim et al . ( 2007 ) excluding galaxies with agn , and for higher @xmath0 s ( until @xmath9 ) , from bell et al . ( 2007 ; only their chandra deep field south estimates are plotted here ) . the data used to get these estimates refer to the population of blue ( star - forming ) galaxies ( see firmani et al . 2010 for more details on the choice of the observational data ) . it should be stressed that the intrinsic spread around the plotted curves is rather large . for the largest mass bin ( about @xmath84 , lower red curves ) , the ghets imply clearly lower ssfrs than the directly measured ssfrs of star forming ( blue ) galaxies at @xmath20 . this is expected since , as discussed above , the most massive galaxies show evidence to have early transited to the passive population . for the galaxy mass @xmath85 , the transit to the passive , red sequence happens at @xmath86 ( see fig . [ mtran ] below ) as indicated by a vertical mark over the curve . bell et al . ( 2007 ) reported the estimated ssfrs also for the red sequence galaxies . their estimates for @xmath87 ( red circles ) and @xmath88 ( green circles ) are plotted in figure [ dotmobs ] . the directly measured ssfrs of massive ( red ) galaxies are very low , decreasing their values as @xmath0 decreases in the same manner as our predictions show . while speculative by the moment due to the large uncertainties , the fact that the values of our inferred ssfrs are systematically higher than the direct measures of red massive galaxies could be interpreted as an evidence of contribution of stellar accretion ( dry mergers ) to the stellar mass growth of these galaxies at late epochs ( see section 2.1 for a discussion on this possibility ) . recently , damen et al . ( 2009a ) presented estimates of the ssfr up to @xmath89 for massive galaxies ( @xmath90 ) in the same chandra deep field south used in bell et al . as mentioned above , for redshifts @xmath91 galaxies of @xmath92 or smaller are on average in their active ( sf - driven ) growth regime . the results by damen et al . are in rough agreement at these redshifts with our predictions . if any , the ghet - based ssfrs are slightly lower than the averages of the directly measured ssfrs . for the intermediate mass bin ( @xmath93 , green curves ) , the agreement between the average ssfrs inferred from the ghets and those measured directly for the star - forming ( blue ) galaxies ( bell et al . 2007 ) is satisfactory , given the spread of data in ssfr . if any , the ssfr from the ghets is slightly lower , in particular at low @xmath11s . again , due to the population downsizing , active ( blue ) galaxies of masses @xmath93 transit to the passive ( red ) population at @xmath94 ( see fig . [ mtran ] ) as indicated by the mark over the green curve , so that their ssfrs at lower @xmath0 s should be in between those measured for the blue and red sequences . the agreement in the mass range around @xmath88 with the damen et al . ( 2009a ) observational inferences up to @xmath89 is also good , though the samples of these authors do not separate galaxies in early- and late - type ones . for @xmath95 the slope of the ssfr(@xmath93)(@xmath96 ) relation for our prediction and for direct observations is about 3.3 . for the lowest mass bins ( @xmath97 and @xmath98 , blue and cyan curves ) , the agreement between ghet - predicted and directly - measured ssfrs is reasonable if takes into account ( i ) the large intrinsic spread of the observational inferences ( 1@xmath36 width of @xmath99 dex around the mean in the relation log(ssfr)log at @xmath10 , salim et al . 2007 ; for higher redshifts , the typical reported intrinsic 1@xmath36 widths are of @xmath100 dex , noeske et al.2007 ; zheng et al . 2007 ; damen et al . 2009a , b ) ; and ( ii ) the sample incompleteness , which increases as @xmath0 is higher ( for such small masses , this may bias the observational average estimates toward larger values for @xmath101 ) . if any , the predicted ssfr(=const)(@xmath96 ) curves for these low masses decrease slower ( are shallower ) than the curves inferred from direct observations for @xmath102 ( see also damen et al . better observational estimates for both the low - mass and ssfrs at high redshifts are needed to attain more conclusive results . the dot - dashed line in figure [ dotmobs ] shows the curve @xmath103(@xmath0 ) 1 gyr](@xmath82 ) corresponding to a constant sfr . galaxies above ( below ) this curve are currently forming stars at a rate higher ( lower ) than the past average . in figure [ ssfr - ms ] , the evolution of ssfr vs. for the same ghets of previous figures is plotted ( red solid lines ) . this plot remarks the fact that small galaxies keep high ssfrs and grow fast at late epochs , while very large galaxies had high ssfr and rapid growth in the remote past , being passive at later epochs . in figure [ ssfr - ms ] , from bottom to top the thick ( cyan , blue , green , orange , red ) lines connect the individual tracks at the redshifts @xmath104 and 4 ( isochrones ) , respectively . each one of these isochrones corresponds to the average ssfr relations at a given @xmath0 and can be compared with direct observational inferences of ssfr vs at different @xmath0 bins . these inferences for samples at different epochs ( from @xmath10 up to @xmath105 ) show that galaxies ( mainly star - forming ones ) are ordered in the logssfr log diagram along sequences of intrinsic widths 1@xmath106 dex ( e.g. , bauer et al . 2005 ; salim et al . 2007 ; schiminovich et al . 2007 ; noeske et al . 2007 ; bell et al . 2007 ; drory & alvarez 2008 ; damen et al . 2009a , b ; santini et al . 2009 ; oliver et al . 2010 ; rodighiero et al . 2010 ) . the averages of these sequences are such that the smaller is , the higher the ssfr . however , due to the sample incompleteness limit , which increases from a few @xmath107 to several @xmath108 from @xmath10 to 1 , respectively , these averages at low masses might be overestimated . the isochrones showed in figure [ ssfr - ms ] are in general within the 1 @xmath36 width of intrinsic spread of the observational ssfr- relations . at low masses ( where observational samples with measured sfrs are below the completeness limit ) , the isochrones tend to be below the direct observational estimates of ssfr at the given epoch but at intermediate masses the agreement is rather good , i.e. , at low masses , the ghet - based ssfr- relations tend to be shallower than those constructed from direct observations . at the high mass end , our isochrones fall faster than the averages of the observed ssfr sequences , which typically are reported for star - forming ( blue ) galaxies . have in mind that our results refer to _ average _ trends and in this sense the average high mass galaxies in our case already transited to the passive ( red ) population . in fact , the ssfr relation at a given @xmath0 should be constructed separately for active ( blue ) and passive ( red ) galaxies . therefore , in the ssfr diagram two sequences would appear : at low masses a sequence dominated by blue galaxies and at high masses the sequence dominated by red galaxies . from the point of view of the ghet analysis , the characteristic mass that divides both sequences is related to the transition mass inferred from the ghets at each @xmath0 ( see 3.2 below ) . as seen in fig . 7 , the average slope of the passive sequence ( high - mass side ) at any @xmath0 is steeper than that one of the active sequence ( low - mass end ) . interestingly enough , this is the trend actually revealed in those observational works where the ssfr relations at different @xmath11s ( out to @xmath109 ) are plotted separately for red and blue galaxies ( bell et al . 2007 ; oliver et al . 2010 ) . in general , we conclude that the ssfrs at different @xmath0 s and as a function of predicted on the basis of our ghets are reasonably consistent with recent observational studies based on direct measures of the ssfr of galaxies at different @xmath0 s ( especially those that divide galaxies into blue and red ones ) . this consistency is particularly important _ because the observational information concerning sfr never enters in our scheme_. therefore , on the one hand , this result is an independent test for the whole approach presented here ( it predicts the mean ssfrs of galaxies at different epochs consistent with direct observational estimates ) . on the other hand , it supports our hypothesis that the stellar mass buildup given by the average ghets is driven by in situ sf . neither does this mean that the growth of by accretion of stellar systems ( dry mergers ) may not happen , but it suggests that it is less dominant than the sf channel ( our approach and the accuracy of current estimates of ssfr at different redshifts do not allow for quantitative inferences of the contributions of one or another channel of galaxy mass growth ) . a natural question arises about the information the ghets provide on the stellar initial mass function ( imf ) . we have outlined the agreement between our ssfr derived from the stellar mass growth and the directly inferred ssfr . the latter inference is highly sensitive to the assumption of a universal imf . in this sense , we can say that such an agreement is congruent with a universal imf . nevertheless , the uncertainties are yet large , making it premature to draw such a conclusion as sufficiently robust . finally , the results mentioned above confirm and integrate into a unified picture two key facts of galaxy stellar mass assembly : ( 1 ) the downsizing in ssfr of low mass galaxies and ( 2 ) the population downsizing of massive galaxies . the former will be discussed on the light of -based models in 4 . the latter is quantified in the next section . our method reveals the existence of a characteristic stellar mass at each @xmath0 , ( @xmath0 ) , at which the average ghet sharply declines its growth rate . we interpret this as a transition from the active to the quiescent regime of sfr , the onset of the migration from the star - forming blue to the passive red population ( population downsizing ) . we may estimate quantitatively ( @xmath0 ) by identifying the mass at a given @xmath0 corresponding to the ghet that started to strongly decrease its ssfr , for example , when the slope of the track in ssfr vs. ( see crosses in figure [ ssfr - ms ] ) becomes steeper than @xmath110 ( the measured ssfr at this point increases from log(ssfr / gyr@xmath31)@xmath111 at @xmath10 to @xmath112 at @xmath113 ) . the stellar calculated this way is plotted vs @xmath0 in figure [ mtran ] ( diagonal crosses ) . the function : @xmath114 offers an approximate description of the results up to @xmath105 ( solid line ) . the focus now is to compare our result with independent observational pieces of evidence for a transition mass from active to quiescent / passive population as a function of @xmath0 . nowadays , as mentioned in the introduction , it is possible to follow the evolution of the early ( red ) and late type ( blue ) separately . in one of the most recent and complete works , based on the @xmath0cosmos survey , pozzetti et al . ( 2010 ) determined the mass where the early and late type cross ( different estimators for these two populations are used , see the figure caption ) from @xmath115 to @xmath116 . such a crossing mass , @xmath117 , is interpreted namely as the typical mass of late - type galaxies migrating to early - type ones ( see also bell et al . 2007 ) . the agreement between the pozzetti et al . ( 2010 ) results for @xmath117 ( circles with error bars connected by different dashed lines ) with our ( @xmath0 ) is satisfactory as can be seen in figure [ mtran ] . local estimates of @xmath117 by bell et al . ( 2003 ; filled triangle ) and baldry et al . ( 2004 ; filled square ) are also plotted , as well as the law inferred by drory & alvarez ( 2008 ) for the mass above which the sfr as a function of begins to drop exponentially , /= our result also agrees qualitatively with other observational studies , with different definitions of the characteristic mass above which the passive ( red ) population of galaxies dominates in number density ( e.g. , bundy et al . 2006 ; hopkins et al . 2007 ; vergani et al . 2008 ) . we may predict also the flow of galaxies transiting from active to passive populations at each @xmath0 per unit of co - moving volume . at each redshift interval @xmath120 ( time interval @xmath121 ) there are ghets transiting from the active to the passive regime , whose associated halo masses are in the interval @xmath122 . the transition rate in number density is then given by the abundance of halos in such a mass interval divided by @xmath123 , @xmath124log/@xmath123 . for the abundance of halos at each @xmath0 we use the @xmath12s given in tinker et al . ( 2008 ) adapted to our cosmology and corrected to include sub - halos according to the functions given in bcw10 . the result is shown in fig . [ trans - rate ] . the transition rate in comoving number density of active to passive galaxies up to @xmath46 scatters around @xmath125 gal gyr@xmath31 mpc@xmath126 without any clear trend with @xmath0 ( the scatter is mainly due to the discretization of the mahs in mass and @xmath0 ) . at @xmath127 , the rate is @xmath128 gal gyr@xmath31 mpc@xmath126 . the obtained passive population growth rates can be compared with estimates reported in the same work by pozzetti et al . they found an average growth rate in number density of the red population integrated above log(/ ) = 9.8 of @xmath129 gal gyr@xmath31 mpc@xmath126 for a redshift interval centered in @xmath130 ( solid triangle with error bar in fig . [ trans - rate ] ) ; this value is in reasonable agreement with our results . note that the flow of galaxies in our case is related to a small mass range around the transition mass given by the _ average _ ghets , while the direct observational determination takes into account a large range of masses . the fact that the latter is only slightly larger than the former implies that indeed most of the galaxies becoming red are those of masses close to the average transition mass . the results presented above are based mainly on the evolution of the relation . an important question is then how general and reliable is the ( , @xmath0 ) relationship used here . a comprehensive analysis of the statistical and systematical uncertainties in ( , @xmath0 ) and how they affect the shape of the - relations at different @xmath11s has been carried out by bcw10 . these authors separate the uncertainty sources into three classes : uncertainties ( i ) in the observational inference of , ( ii ) in the dark matter @xmath12 , and ( iii ) in the matching process arising primarily from the intrinsic scatter between and . they found that by far the largest contributor to the error budget of the local relation comes from assumptions in converting galaxy luminosity into , which amounts to uncertainties of @xmath131 dex in the normalization of the relation ( see the error bars in fig.[galefficiency - b ] ) . the contribution from all other sources of error , including uncertainties in the cosmological model , is much smaller , ranging from 0.02 to 0.12 dex at @xmath6 and from 0.07 to 0.16 dex at @xmath46 . at high redshifts ( @xmath132 ) , statistical uncertainties grow , becoming as significant as the systematic ones . on the other hand , the intrinsic scatter in given and the random statistical error in inferences have a significant effect on the shape of the relation at the massive end ( taken into account in bcw10 ) . at @xmath10 , the results from most direct and indirect studies to infer ( , @xmath10 ) , in spite of the different methods and different local used , agree well among them , especially if the uncertainties are taken into account ( see bcw10 for a comparison ) . the systematic shift of the ( /) curves to higher as @xmath0 increases ( see figs . [ galefficiency ] and [ galefficiency - b ] ) is also a generic result obtained by all the authors ( cw09 ; wang & jing 2010 ; moster et al . 2010 ; bcw10 ) . this implies that the trend reported here of downsizing in ssfr for galaxies less massive than @xmath133 is robust ; the downsizing can be respectively stronger or weaker depending on whether the low - mass end of the curves shifts more or less with @xmath0 . as the completeness of the samples at low masses improves , the estimate of the at their low - mass end will be better , and hence more accurately the evolution of the relation at lower masses will be constrained . the results of different authors may differ regarding the exact location of the (or / ) curves at different @xmath0 s , especially for @xmath134 . for example , in moster et al . ( 2010 ) , the peaks in ( /) significantly decrease with @xmath0 in such a way that the curves intersect only at very high masses . this implies that the ghets keep growing until @xmath6 even for large galaxies , the transition to passive population happening only for the most massive ones . in a similar way , cw09 concluded from their analysis that out to @xmath9 and for @xmath135 there is not a characteristic mass at which the growth rate of galaxies is truncated . however , this is in conflict with independent observations that show instead strong population downsizing ( see 3.2 and fig . [ mtran ] ) . in the case of bcw10 , the peaks in ( /) slightly decrease out to @xmath9 and for higher @xmath11s , increase attaining at @xmath13 the same level as at @xmath21 ( figs . [ galefficiency - b ] and [ galefficiency ] ) . however , the uncertainties in the determination of the ( /) curves are large and other behaviors with @xmath0 are allowed . the relationship used here closely resembles the bcw10 results as discussed in 2 . in order to probe the robustness of the conclusions presented here , we have performed our analysis by using ( , @xmath0 ) relationships different to the one assumed in 2 but yet within the uncertainties of the bcw10 inferences . for example , we explored an ( , @xmath0 ) relationship such that the peaks of the ( /) curves remain at the same level for all redshifts ( i.e. , @xmath63(@xmath0)@xmath136 in eq . the obtained ghets are qualitatively similar to the ones presented here and eq.([mtranfit ] ) slightly changes to @xmath137 . it is noteworthy that the high mass ghets in fig . [ ghets ] , presenting a stagnation in their growth , show a shape a little sensitive to the parameter @xmath138 . by assuming @xmath139 in eq . ( 2 ) , the most massive ghets , after reaching a maximum stellar mass at approximately the same @xmath0 as presented in fig . [ ghets ] , decline in mass but not by more than 0.1 dex . we conclude that the main results reported here are robust at least for the range of variations in the ( , @xmath0 ) relationship that remains within the current uncertainties in the inference of this relationship . any model of galaxy formation and evolution within the hierarchical scenario should predict how does the -to- ratio ( the `` galaxy formation efficiency '' ) of individual galaxies evolve . several complex astrophysical processes intervene in shaping the evolution of this ratio . most of the semi - analytical and semi - numerical models as well as the numerical simulations in that regards subgrid physics introduce physical schemes for modeling these processes and for reproducing the _ present day _ relation ( or the luminosity function ) and other properties / correlations of galaxies . the question now is whether such models are able to reproduce the whole evolution outlined here . different approaches to explain the population downsizing have been recently proposed , while the ssfr downsizing is just starting to be recognized as a problem ( e.g. , fontanot et al . 2009 ; firmani et al . 2010 ; coln et al . 2010 ) . in each one of the figures where the ghets were presented in this paper , we have plotted the gets corresponding to two low mass disk galaxy models that end today with log(/ ) = 9.40 and 10.15 ( log(/ ) = 11.25 and 11.73 ; dotted lines ) . the models were calculated with a semi - numerical method that self - consistently solves dynamical and hydrodynamical equations of : halo mass virialization given the mah , disk formation in centrifugal equilibrium inside the growing halo , gravitational drag of the disk over the halo inner regions , sf triggered by gravitational instability and self - regulated by a vertical energy balance between energy input due to sne and turbulent energy dissipation , and sn driven mass outflows ( firmani & avila - reese 2000 ; firmani et al . the main assumptions implicit in the models are those of : spherical and cylindrical symmetries for the halo and disk , respectively , gas infalling on dynamical time scales ( cold mode ) , adiabatic invariance during halo contraction , and detailed angular momentum conservation for the infalling gas . the baryon fraction is assumed initially as the universal one , but the feedback driven outflow reduces significantly this fraction . the parameters of the outflow model are fixed in such a way that namely the ( low - mass ) relation at @xmath10 is reproduced . in firmani ( 2010 ) we have shown that this happens only in the case of the so called energy driven outflow and for a high sn energy transfer efficiency relation . ] . because the exploration in this paper is only at the level of average trends , the halo mahs used in the models are the averages among 20000 realizations for each given mass , and the used halo spin parameter @xmath140 was fixed to 0.03 at all @xmath0 s , a value slightly smaller than the mean of @xmath6 relaxed halos in n - body cosmological simulations ( e.g. , bett et al . 2007 ) . on the one hand , the n - body simulations show that @xmath140 does not significantly changes with time for halos growing in the accretion mode ( e.g. , peirani , mohayaee & de freitas pacheco 2004 ; donghia & navarro 2007 ) . on the other hand , our models show that the radius evolution of disk galaxies agrees with observational inferences when the gas @xmath140 parameter is roughly constant in time ( firmani & avila - reese 2009 ) . the gets in figure [ ghets ] show that the stellar mass growth is roughly proportional to the halo mah as @xmath0 decreases . the model gets and the ghets in the diagram ( fig . [ msmh ] ) are very different ; the average slopes of the gets are @xmath70log/@[email protected] , much less than the slopes of the corresponding ghets . the model -to- ratio in figure [ galefficiency ] decreases toward the past due mainly to the feedback driven outflow effect , but slightly , while the ghets of the same masses evidence a very fast decreasing . the and ( /) relations ( figs . [ msmh ] and [ galefficiency ] , respectively ) predicted by the model gets at low masses change with @xmath0 ( slightly increases with @xmath0 for a fixed ) in an opposite direction with respect to the empirically inferred observations . this discrepancy represents a serious problem for all the theory behind the models . the dramatic differences between the gets and ghets are also seen in the ssfr evolution plots ( figs . [ dotm ] and [ ssfr - ms ] ) . we remark here the downsizing aspect . the get - based ssfrs of the two galaxies evolve almost parallel ( with a law ssfr@xmath142 ( 1+@xmath0)@xmath143 ) , being the ssfr of the larger galaxy slightly higher since high @xmath11s ( weak upsizing ) . instead , the ghet - based ssfrs cross each other at high @xmath0 ( see the upper side of figure [ dotm ] ) , in such a way that larger galaxies have higher ssfr at very high @xmath11s , but at lower @xmath11s , the situation is inverted in such a way that as @xmath144 , the smaller is the galaxy , the higher the ssfr ( downsizing ) . the comparison of theoretical models with the ghet - based ssfrs suggests that some astrophysical mechanism not considered up to now should systematically delay the onset of sf activity the smaller the halo is . the low ssfrs ( early stellar mass assembly ) of small galaxies predicted by the models , which include energy - driven outflows , are actually generic to all -based models . this issue has been reported in different ways by the semi - analytic models . most of these models show that the stellar population of small galaxies ( @xmath145 ) is assembled too early , becoming these galaxies older , redder , and with lower ssfrs at later epochs than the observed galaxies in the same mass range ( e.g. , somerville et al . 2008 ; fontanot et al . 2009 ; santini et al . the problem is both at the level of satellite and central galaxies . for example , it was shown that models typically overpredict the observed stellar population ages ( pasquali et al . 2009 ) and stellar masses ( liu et al . 2010 ) of central galaxies in low mass halos . by means of a galaxy evolutionary model similar to ours , dutton , van den bosch & dekel ( 2010b ) also find that the ssfr of model low - mass galaxies is below the average of observations , specially at redshifts @xmath146 , though these authors conclude that their models are able to reproduce roughly the main features of the observed sfr sequence . finally , it should be said that n body + hydrodynamics simulations of individual low mass galaxies in a cosmological context also face the issue of too low ssfrs ( as well as too high stellar and baryonic mass fractions ; e.g. , coln et al . 2010 ) . in firmani et al . ( 2010 ) we explored the inclusion in our semi - numeric models of re - accretion of the ejected gas for a broad range of possibilities . as expected , the ssfr of galaxies increases in general due to re - accretion but this increasing goes in the opposite direction of downsizing : only moderately for low - mass galaxies and too much for the larger galaxies . empirically inferred relations from @xmath10 to @xmath25 were connected with average halo mahs in order to infer the corresponding individual _ average _ stellar mass growth histories , called here ghets . we have adopted an ( , @xmath0 ) relationship continuous in the @xmath147 redshift range and in agreement with the inferences by bcw10 , who used the technique of jointly matching the abundances of observed to the theoretical @xmath12s at different @xmath11s . the main results obtained here , which allow to establish a unified description of average galaxy evolutionary tracks for a large range of masses , are as follows . @xmath148 the folding in the @xmath0 surface , reflected mainly as peaks in the ( /) diagram that shift to higher masses as @xmath0 increases , introduces an important feature in the behavior of the ghets . for masses much smaller than the peak mass at a given @xmath0 , ( @xmath0 ) , the ghets are in an active growth phase , whereas masses close to or greater than ( @xmath0 ) are in a quiescent or completely passive ( stagnated ) phase ( figs . [ msmh ] and [ galefficiency ] ) . therefore , at each @xmath0 there is a characteristic stellar mass at which on average galaxies slow down their growth and transit from an active ( star forming blue ) population to a passive ( red ) population . this transition mass decreases with time , giving rise to a phenomenon called here _ population downsizing_. @xmath148 galaxies less massive than at @xmath6 ( @xmath149 , @xmath150 ) have still growing ghets . besides , the lower the mass , the faster the later growth , due likely to a delayed and lately active sf phase , a phenomenon called _ downsizing in ssfr_. @xmath148 the shapes of the average stellar and halo mass assembling histories are quite different ( figure [ ghets ] ) . for galaxies that at @xmath6 have @xmath151 , their dark mahs at later epochs grow slightly slower as smaller is the mass , while their stellar ghets grow much faster . for @xmath152 , the larger the galaxy , the earlier its ghets attain a stellar mass stagnation ; until this epoch , the more massive the galaxy , the faster the growth with respect to the corresponding growth . @xmath148 by neglecting any stellar mass growth by accretion ( dry mergers ) , a ghet allows to find the corresponding ssfr history , ssfr(@xmath0 ) , by calculating and correcting by the stellar mas - loss recycling factor @xmath153 . the inferred ssfrs at different @xmath0 s and as a function of ( figs . [ dotm ] , [ dotmobs ] , and [ ssfr - ms ] ) are reasonably consistent with recent observational studies based on direct measures of the ssfr of star - forming galaxies at different @xmath0 s . the ghet - based ssfrs corresponding to masses that at a given @xmath0 are larger than the transition mass ( @xmath0 ) , are much lower than the measured ssfrs of ( rare ) luminous blue galaxies , but agree or are even slightly larger than the measured ssfrs of red galaxies . the overall consistency between the ghet - based ssfr- relations at different @xmath11s with direct inferences of these relations suggests that the accretion of stellar systems ( mergers ) plays a minor role in the assembling of galaxies , excepting perhaps those that transited to the passive sequence ( the most massive ones ) . @xmath148 by using the ssfr vs. evolutionary tracks , we calculated the characteristic transition mass at each @xmath0 , ( @xmath0 ) , above which the average ghet starts to significantly decrease its growth rate , transiting from the active to the passive galaxy population . the result is roughly described by the relation log(/ ) = @xmath154 ( at least up to @xmath105 ) . this result agrees with recent observational determinations of the evolution of the mass , at which the early- and late - type components cross each other , from @xmath155 to the present ( fig . [ mtran ] ) . @xmath148 we determined also the transition rate in number density of active ( blue ) to quiescent ( red ) galaxy population . at @xmath21 such a rate is @xmath156 gal gyr@xmath31 mpc@xmath126 and up to @xmath46 the rates are within @xmath157 gal gyr@xmath31 mpc@xmath126 , in good agreement with direct observational inferences of growth rate in number density of red galaxies . @xmath148 we further explored whether -based models of galaxy evolution are able to predict galaxy evolutionary tracks ( called here gets ) in agreement with the ghets , in particular at low masses where the downsizing in ssfr is evidenced by the ghets . we have shown that while the models with sn energy driven outflows are able to reproduce the local relation ( see also firmani et al . 2010 ) , they fail in reproducing this relation at higher @xmath0 s . the difference between gets and ghets at low masses is rather large : the former ones show a fast decrease in ssfr with time almost independent of mass , while for the latter ones , the lower the mass , the slower the late decrease of ssfr ( downsizing in ssfr ) . from these results , we conclude that the general description of stellar mass buildup of galaxies provided by our average ghets appears rather successful : the predicted ssfr histories as a function of and the predicted transition mass as a function of @xmath0 , as well as the transition flux at different @xmath11s , are roughly consistent with direct ( but yet limited ) observational estimates of all these quantities . furthermore , our analysis reveals nicely the existence of a galaxy bi - modality related to the growth activity : as @xmath0 decreases , smaller and smaller masses transit on average from the active ( blue star - forming ) to the passive ( red ) population . an important ingredient of our approach is the connection of halos to the observed galaxies . therefore , the underlying hierarchical scenario is successful in the sense that the predictions of the model are in agreement with independent observations . the average stellar and dark mass buildup of galaxies were found to be significantly different , specially for low and high masses . this result is in qualitative agreement with cw09 in spite of the differences in the data , method , and redshift range between their and our work . however , the results from both works are different at a quantitative level , which is seen , for example , in the different conclusions regarding the existence of a characteristic mass at each epoch at which the ssfr of galaxies is truncated . note that the ssfr and the determination of this characteristic mass ( ) are related to the first and second derivatives of the growth , respectively . our results put in a unified picture both the population downsizing ( related to galaxies with @xmath158 at @xmath6 ) and the downsizing in ssfr ( related to smaller galaxies ) . the challenges are now to explain the processes that produce : ( 1 ) the transition from active to passive regimes in growth associated with a sharp cessation of the sfr in massive galaxies , ( 2 ) the fact that the typical mass of ssfr truncation , , decreases with cosmic time ( population downsizing ) , and ( 3 ) the fact that the smaller is the halo , the more is delayed the galaxy s growth ( downsizing in ssfr ) . items ( 1 ) and ( 2 ) seem to find partial explanations in aspects related to the same dark halo clustering ( environment ) evolution ( e.g. , neistein et al . 2006 ) as well as to the introduction of feedback processes due to the agns of massive galaxies ( e.g. , silk & rees 1998 ; kauffmann & haehnelt 2000 ; granato et al . 2004 ; cattaneo et al . 2005 ; de lucia et al . 2005 ; bower et al . 2006 ; croton et al . 2006 ) . item ( 3 ) , as shown here , is a sharp and less understood problem . later re - accretion of the gas ejected by galaxies due to sn - driven outflows does not account for a solution to this issue ( firmani et al . . a better understanding of the astrophysical processes intervening in galaxy formation and evolution as well as the role of environmental processes is necessary . we are grateful to dr . l. pozzetti for kindly making available to us in electronic form the results reproduced in our fig . [ mtran ] . we thank the anonymous referee for his / her careful review of our paper . v. a. acknowledges papiit - unam grant in114509 and conacyt grant 60354 for partial funding .

we present detailed comparisons between high quality observational colour - magnitude diagrams ( cmds ) of open star clusters and synthetic cmds based on montecarlo numerical simulations . the comparisons account for all of the main parameters which determine the shape of the cmd for a stellar population . for the four clusters studied , ngc 6819 , ngc 2099 ( m37 ) , ngc 2168 ( m35 ) and ngc 2323 ( m50 ) , we derive reddening , distance , age , binary fraction , star formation rate and indicative metallicity by comparing the locations and density of points in the observed cmds to the simulated cmds . we estimate the uncertainties related to stellar evolution theories by adopting various sets of stellar models for all of the synthetic cmds and discuss which stellar models provide the theoretical cmds that best reproduce the observations . [ firstpage ] colour - magnitude diagrams methods : n - body simulations open clusters and associations : general open cluster and associations : individual : ngc 2099 , ngc 2168 , ngc 2323 , ngc 6819

astro-ph0403420

theoretical isochrones are commonly fit to the major observational sequences of star clusters in order to both better understand the underlying physics of stellar evolution and to determine properties of the clusters , i.e. , the age . if the metallicity , reddening and distance to the cluster are well constrained from independent techniques , the comparisons typically involve matching the morphology of the turn - off and location of the red giant stars to predictions . recently , the newer method of using synthetic colour - magnitude diagrams to compare with observational data has proven to be much more informative and rewarding @xcite . these montecarlo simulations allow modelling of several additional parameters which dictate the distribution of points in the cmd , such as stochastic star formation ( sf ) processes , binary fraction , photometric spread , main - sequence thickness , data incompleteness and small number statistics . consequently , the results not only provide a measure of the properties of the cluster , but can also constrain the star formation history ( sfh ) and the initial mass function ( imf ) . furthermore , by comparing the simulations based on several different sets of evolutionary tracks , we can constrain which models use the best prescription of parameters ( such as treatment of overshooting , mixing length , etc ... ) . confronting the simulations with observations requires a large data set with accurate photometry . for this , we use the deep @xmath0 photometry presented in the cfht open star cluster survey ( @xcite , hereafter jski ) . jski observed 19 open star clusters in our galaxy and have yet published results on the four richest clusters , ngc 6819 ( @xcite , hereafter jskii ) , ngc 2099 ( @xcite , hereafter jskiii ) , and ngc 2168 and ngc 2323 ( @xcite , hereafter jskiv ) . these data were reduced and calibrated in a homogenous manner as described in jski . the resulting cmds exhibit very tight main sequences showing several ` kinks ' and slope changes which are predicted by theory . more importantly , the combination of very short and deep exposures , and the large aerial coverage of the detector ( 42@xmath1 28@xmath2 ) has allowed the measurement of stars from the brightest asymptotic giant branch ( agb ) and red giant branch ( rgb ) phases down to very low - mass main - sequence phases ( @xmath30.2 m@xmath4 ) . this allows our comparisons to yield evolutionary information over a wide mass range . the reduced data set in the cfht open star cluster survey has been requested by , and made available to , several investigators for additional science rewards outside our goals ( e.g. , astrometric studies , proper motions , variable stars , radial velocities , brown dwarfs , blue stragglers and galactic disk star distributions ) . the present study complements these efforts and analyses the four published clusters in a way that allows us to include them in a large homogeneous sample of open clusters aimed at studying the formation and evolution of the galactic disk ( bragaglia 2003 , and references therein ) . galactic open clusters are indeed particularly well suited to this purpose , since they span a range of ages from a few million to several billion years and can be observed in various regions of the galactic disk characterised by different star formation histories . they can be used to study both the present day disk structure and its temporal evolution ( janes & phelps 1994 , friel 1995 , tosi 2000 , bellazzini et al . old open clusters offer a unique opportunity to trace the whole kinematical and chemical history of our disk , if collected in populous and representative samples and accurately and homogeneously analysed ( see e.g. , twarog , ashman , & anthony - twarog 1997 ; carraro , ng , & portinari 1998 ) . here we apply the synthetic cmd method to ngc 6819 , ngc 2099 , ngc 2168 and ngc 2323 to derive their age , reddening , distance modulus and ( approximate ) metallicity homogeneously to the bragaglia ( 2003 ) cluster sample . the method also allows us to determine other features of these clusters , such as the existence ( or lack thereof ) of a significant fraction of unresolved binary systems , the original total mass of formed stars and the possible evaporation of some of the lower mass stars . the organisation of the paper is as follows , [ observations ] briefly summarises the data and the reduction procedures . further details are given in jski . in [ firstresults ] we present a summary of our main results which relate to this work from the published papers in the cfht open star cluster survey . [ synthetic ] sets up the numerical simulations and presents details on how the synthetic cmds were created . next , we compare the cmds and the corresponding luminosity functions from the observations with the simulations on a cluster - by - cluster basis ( [ vs ] ) . finally , we discuss the results in [ discussion ] and conclude the study in [ conclusions ] . an absolute requirement in using different data sets to make detailed comparisons between observations and models is that of homogeneity . to truly understand the variations in the different models parameters , we must minimize systematic errors . therefore , we desire a rich data set of open clusters , all reduced using the same approach , calibrated consistently , and with well determined incompleteness factors . the richest clusters from the cfht open star cluster survey , ngc 6819 , ngc 2099 ( m37 ) , ngc 2168 ( m35 ) and ngc 2323 ( m50 ) perfectly meet these requirements . the data were all imaged during an excellent three night observing run with the canada - france - hawaii telescope ( cfht ) on 1999 october , using the cfh12k camera . the optical detector is a high resolution ( 1@xmath510@xmath6 pixels ) ccd mosaic camera which projects to an aerial coverage of 42@xmath2@xmath528@xmath2 on the sky . [ cols="^,^,^,^,^,^",options="header " , ] \1 . @xmath7 these very short exposures were obtained at a later date . + most clusters were observed with exposure times sufficient to reach the magnitude of the coolest white dwarfs and of the faintest m - type stars on the main sequence . we also complemented the deep photometry with shorter exposures to cover the brighter turn - off and evolved stars . the individual images were registered and co - added ( where necessary ) using the fits large images processing software ( flips ) ( j .- c . cuillandre , 2000 , private communication ) . details of this procedure , as well as the pre - processing of the data are provided in 3 of jski . the actual photometry of the sources in the images was measured using point - spread - function extractor ( psfex ) , a highly automated program which will be integrated in the code of a future sextractor version ( bertin & arnouts 1995 ; e. bertin 2000 , private communication ) . we used a separate psf on each ccd of the mosaic . included in the psf are polynomial basis functions which can map the variations of the psf across the ccd . psfex is discussed further in 4 of jski . data for each of the clusters were taken in the @xmath8 and @xmath9 filters , and calibrated in the johnson photometric system using multiple images of the sa-92 and sa-95 standard star fields @xcite . the final averaged uncertainties in the zero - point terms were measured to be @xmath30.018 magnitudes , uncertainties in the airmass coefficients were @xmath30.015 magnitudes in @xmath9 and @xmath30.007 magnitudes in @xmath8 . colour coefficients , as well as airmass terms , were similar to nominal cfht expectations . details of the calibration of instrumental magnitudes to real magnitudes are provided in 5 and tables 2 and 3 of jski . the resulting , calibrated cmds for the open clusters exhibit unprecedented long , clear main sequences ranging from the brightest cluster members to the limiting magnitude in each case ( 23 @xmath10 @xmath11 25 ) . a summary of the observational log for each cluster ( exposure time , seeing , airmass , etc ... ) is presented in table 1 . figure 1 presents the final calibrated cmds for each of the four open clusters , with the best fit theoretical isochrones as determined in our previous published analysis of these clusters ( see jskii , jskiii , jskiv for details ) . these rich clusters are clearly delineated from the background galactic disk stars . the main sequences are very tight and show several features which are predicted by stellar evolutionary theory ( see 5 ) . for this , we used a new grid of model isochrones calculated by the group at the rome observatory according to the prescription described in @xcite . these models adopt convective core - overshooting by means of an exponential decay of the turbulent velocity out of the formal convective borders fixed by schwarzschild s criterion ; a free parameter which gives the e - folding distance of the exponential decay is set to @xmath12 = 0.03 @xcite . the convective flux has been evaluated according to the full spectrum of turbulence ( fst ) theory prescriptions @xcite . the theoretical isochrones were transformed into the observational plane by making use of the bessel , castelli & pletz ( 1998 ) conversions . the lower main sequence ( m @xmath13 0.7 m@xmath4 ) was calculated by adopting nextgen atmosphere models @xcite . for m @xmath13 0.47 m@xmath4 ( or t@xmath14 @xmath13 3500 k ) the transformations of @xcite in @xmath15 are not very reliable and so the faint end of the isochrones terminate at this mass . these models will be referred to as fst from here on . the tight , very rich , main sequence and turn - off consist of over 2900 cluster stars to our limiting magnitude . main - sequence fitting of the un - evolved cluster stars with the hyades star cluster yields a distance modulus of ( @xmath16)@xmath17 = 12.30 @xmath18 0.12 ( @xmath19 = 2500 pc ) , for a reddening of @xmath20(@xmath15 ) = 0.10 . these values are consistent with a theoretical isochrone of age 2.5 gyrs . detailed star counts in concentric annuli out to large angular radii set constraints on the cluster radius , @xmath21 = 9@xmath225 @xmath181@xmath220 , and clearly indicates mass segregation in the cluster . the global cluster mass function is found to be quite flat , @xmath23 = @xmath240.15 ( @xmath23 = 1.35 , salpeter ) . a large population of white dwarf stars is found in the cluster and is currently being investigated spectroscopically . the cluster cmd shows extremely well populated and very tightly constrained main sequence , turn - off , and red giant populations . the photometry for this cluster is faint enough ( @xmath25 24.5 ) to detect the end of the white dwarf cooling sequence . therefore , the white dwarfs are used as chronometers to age the cluster providing an independent age ( 566 @xmath26 myrs ) from the main - sequence turn - off plus red giant clump fit ( 520 myrs ) . we also derive the reddening ( e(@xmath15 ) = 0.21 @xmath18 0.03 ) and distance ( ( @xmath16)@xmath17 = 11.55 @xmath18 0.13 ) to ngc 2099 by matching main - sequence features in the cluster to a fiducial main - sequence for the hyades @xcite , after correcting for small metallicity differences . the cluster luminosity and mass functions indicate some evidence for mass segregation within the boundary of the cluster . ngc 2168 is a well studied nearby ( d = 912 @xmath27 pc ) young cluster ( 180 myrs ) showing a well defined main sequence . the reddening of the cluster , @xmath20(@xmath15 ) = 0.20 , and the metallicity , @xmath28 = 0.012 , were adopted from the literature @xcite . the cluster is found to contain @xmath31000 stars above our limiting magnitude with a global mass function slope very similar to a salpeter value ( @xmath23 = 1.29 ) . there is mild evidence for mass segregation in the cluster . white dwarf stars are found and can potentially set very important constraints on the high - mass end of the white dwarf initial - final mass relationship considering they must have evolved from quite massive progenitors . ngc 2323 is one of the youngest clusters in the cfht open cluster survey ( 130 myrs ) , and shows a main sequence extending over 14 magnitudes in the @xmath9 , @xmath15 plane . the cluster has had very few photographic / photoelectric studies , and prior to our efforts , no previously known ccd analysis . this is surprising considering the rich stellar population ( @xmath32100 stars above our magnitude limit ) and the relatively low distance modulus ( @xmath29)@xmath30 = 10.00 @xmath18 0.17 ) . the cluster shows clear evidence of mass segregation , an important result considering the dynamical age is only 1.3@xmath5 the cluster age . age , reddening and distances for the four clusters have been re - derived applying the synthetic cmd method ( tosi et al . 1991 ) to the empirical cmds described above . the best values of the parameters are found by selecting the values that provide synthetic cmds with morphology , number of stars in the various evolutionary phases and luminosity functions in better agreement with the empirical ones . the method has already been succesfully applied to several clusters ( bragaglia 2003 and references therein ) . the synthetic cmds are constructed via montecarlo extractions of ( mass , age ) pairs , according to an assumed imf , sf law , and time interval of the sf activity . each extracted synthetic star is placed in the cmd by suitable interpolations on the adopted stellar evolution tracks and adopting the bessel et al . ( 1998 ) tables for photometric conversion in the johnson - cousins photometric system . the absolute magnitude is converted to a _ provisional _ apparent magnitude by applying ( arbitrary ) reddening and distance modulus . the synthetic stars extracted for any magnitude and photometric band are assigned the photometric error derived for the actual stars of the same apparent magnitude . then , they are randomly retained or rejected on the basis of the incompleteness factors of the actual data , derived from extensive artificial star tests . once the number of objects populating the whole synthetic cmd ( or portions of it ) equals that of the observed one , the procedure is stopped , yielding the quantitative level of the sf rate consistent with the observational data , for the prescribed imf and sf law . to evaluate the goodness of the model predictions , we compare them with : the observational luminosity functions , the overall morphology of the cmd , the stellar magnitude and colour distributions , the number of objects in particular phases ( e.g. , on the red giant branch , in the clump , on the blue loops , etc . ) . a model can be considered satisfactory only if it reproduces all of the features of the empirical cmds and luminosity functions . given the uncertainties affecting both the photometry and the theoretical parameters ( stellar evolution tracks included ) , the method can not provide strictly unique results ; however , it allows us to significantly reduce the range of acceptable parameters . in this way we derive the age , reddening , distance of the cluster , and indicate the metallicity of the stellar evolution models in better agreement with the data . in principle , the latter should be indicative of the cluster metallicity , but this depends significantly on some of the stellar model assumptions , such as opacities . to estimate if , and how many , unresolved binary systems could be present in the cluster , the synthetic cmds have been computed both assuming that all the cluster stars are single objects and that an ( arbitrary ) fraction of them are members of binary systems with random mass ratio . again , the comparison of the resulting main - sequence morphology and thickness with the observed ones allows us to derive information on the most likely fraction of binaries . we also infer the astration mass of the cluster ( i.e. , the total mass that went in all of the stars formed in the cluster ) , according to the adopted imf . the method here also allows us to derive the best imf as long as the observed mass range of main - sequence stars is sufficiently large and the main sequence is sufficiently tight . however , the high background contamination affecting the systems examined here and the lack of sufficiently reliable external blank fields , make the derivation too unsafe to discuss it here . other factors ( see below ) , such as the actual size of the photometric errors also prevents us from comparing the imf in any significant detail . to test the effect of different input physics on the derived age , reddening and distance modulus , we have run the simulations with three different sets of stellar evolutionary tracks . the adopted sets were chosen because they assume different prescriptions for the treatment of convection and range from no overshooting to rather high overshooting from convective regions . despite these differences , they are all able to well reproduce the observed cmds of both star clusters and nearby galaxies . they are thus suitable to evaluate the intrinsic uncertainties still related to stellar evolution models . we use the fst tracks of @xcite with high ( @xmath12 = 0.03 ) and moderate ( @xmath12 = 0.02 ) overshooting ; the bbc94 tracks of the padova group @xcite , with overshooting ; and the franec tracks @xcite with no overshooting . for those cases where the metallicities are not well constrained , we allow different values in our comparisons . incompleteness factors and photometric errors are taken from the original papers in the cfht open star cluster survey series ( jskii , jskiii and jskiv ) and folded into the numerical simulations . all models assume that the star formation activity has lasted 5 myr ( i.e. , approximately an instantaneous burst ) and that the stars were formed following a single slope ( @xmath23 = 1.35 ) salpeter s imf over the whole mass range covered by the adopted tracks ( 0.6 100 @xmath31 ) . figures 2 , 4 , 6 and 8 show for each cluster the synthetic models in better agreement with the data for each adopted set of stellar tracks , as well as the cluster observational cmd on the same scale . in all of these figures the cmds in the top row are the synthetic ones , those in the second row are the same synthetic ones but overlapped with the stars of the equal area fiducial blank field , and the bottom cmd is the empirical one . figures 3 , 5 , 7 and 9 compare the empirical luminosity functions with those predicted by the best synthetic models . ultimately , the best models have been selected by visually comparing features between the synthetic models and the observations , such as the location and density of points on the cmd . however , we have also attempted to confirm our results by using both the two - dimensional kolmogorov - smirnov test on the cmd and @xmath32 tests on the luminosity and colour functions . in the cases of well constrained field contamination , such as behind ngc 6819 and ngc 2099 , the tests confirm our choices of best synthetic models . for the other two clusters , the tests do not help discriminate between best choice of models and we strictly use conclusions based on the visual location of points . also , these statistical tests are particularly not well suited for these comparisons because of the insignificant weight given to the small number of stars in key evolutionary stages which must be fit by the synthetic models ( such as the red giant clump ) . ngc 6819 is the richest cluster in this work . the cluster field contains 10510 _ bona fide _ stars ( i.e. , with _ parameter @xmath33 0.5 ) while the blank field of same area contains 8504 _ bona fide _ stars . the synthetic cmds are therefore computed with 2000 objects ( a fraction of which are assumed to be members of binary systems ) . all of the tracks used to simulate the ngc 6819 cmd are solar metallicity models as the cluster metallicity has been constrained to this value through the high resolution analysis of @xcite . figure 2 shows that the three sets of models provide slightly different ages ( and hence , reddening and distances ) , which strictly depend on the assumed amount of overshooting : the higher the overshooting , the older the age . this is of course due to higher overshooting models having higher stellar luminosities , thus allowing lower mass stars to reach the desired luminosity . the variations in the ages seen in the models also stress the importance of deriving uncertainties on cluster parameters by using different sets of tracks . the fst models with @xmath12 = 0.03 ( left ) lead to parameters virtually identical to those found in jskii . a better estimate of the age is found to be 2.30 @xmath18 0.15 gyrs , distance modulus @xmath34 = 12.0 , and reddening @xmath35 = 0.10 . the franec models ( right ) suggest younger ages , due to their lack of overshooting , 1.6 - 1.8 gyrs , and therefore need a slightly higher reddening ( @xmath35 = 0.15 ) to reproduce the colours . the bbc94 models ( middle ) , with intermediate overshooting , favor an intermediate age of 2.0 gyrs , with @xmath35 = 0.12 and @xmath34 = 12.2 . none of these models are found to perfectly reproduce all of the observed features in the ngc 6819 cmd . for example , the fst models predict a main - sequence slope steeper than the observed one and the two diverge at the faint end . these models also tend to place the red giant clump excessively towards the red , probably due to the fact that the models do nt include mass loss . the franec models better reproduce the main - sequence distribution but present a turn - off with a hook that is too pronounced , a likely consequence of the young age . given the high field contamination , however , this feature is not particularly evident in the combined cmds of the second row . a smaller cluster core cmd shows the hook prominently . the franec models also overpopulate the lower red giant branch . this is visible even despite the fact that this region of the cmd is plenty of contaminating objects . the bbc94 models provide the best fit to the data cmd , despite a main - sequence slope which is slightly too steep . most of these discrepancies between the observed features and the synthetic cmds result from problems in the stellar models . as pointed out by andreuzzi et al . ( 2004 ) studying the open cluster ngc 6939 with the same method , assuming a reddening dependence on the stellar colour as adopted by twarog et al . ( 1997 ) makes the synthetic main sequences even steeper and , hence , more inconsistent with the data . another evident result that we find for all of the clusters studied here , is that the observed main - sequence spread can not be reproduced if one assumes photometric errors as small as those listed in our photometry catalogue . to test the actual size of the photometric errors we revert back to the artificial star tests presented in jskii . a plot of the @xmath36 - @xmath37 distribution tells us that , in fact , the errors quoted agree with the artificial star tests . the synthetic cmds however , require that either the errors are larger , at least 0.01 magnitude even in the brighter bins , or there is some physical cause for the observed spread ( i.e. , differential reddening or metallicity spread ) . a spread due to binaries is modelled and therefore ruled out as a cause for this discrepency . the apparent spread could be caused by small offsets in the calibration of the different ccds of the mosaic , but internal metallicity spread in the cluster or small differential reddening can not be excluded . we are currently investigating this possibility in another project ( r. s. french et al . 2003 , private communication ) . for now , we have increased the errors to 0.01 magnitude for stars brighter than @xmath9 = 19 and doubled them for fainter bins ( @xmath38 = 0.01 at @xmath9 = 21 , @xmath38 = 0.04 at @xmath9 = 23 ) . empirically , this is the minimum size required to obtain a main - sequence spread comparable with the observed one . all of the shown cases in figure 2 assume that 20% of the stars are in binary systems with random mass ratio . we also tested simulations with 30% binaries and found minor differences . given the high field contamination , it is difficult to understand which fraction and assumption of the mass ratio is the best case . it could entirely be true that the fraction is actually smaller if the mass ratio is more skewed towards equal masses . models without binaries are not shown as they amplify the problem of the main - sequence width ( the predicted one is too tight ) and do not correctly reproduce the stellar distribution on the red side of the main sequence . figure 3 shows the empirical luminosity function of the 10510 cluster field stars ( empty circles ) compared to the luminosity functions resulting from the sum of the 2000 synthetic stars of figure 2 with the 8504 stars of the blank field . the agreement is excellent , except for the slight underestimate of the clump stars of the fst models . a clear result of our study is that all of the synthetic cmds for ngc 6819 show overpopulated faint main sequences ( @xmath39 20 ) , both for single and binary stars . this is a clear signature of low mass stellar evaporation ( or at least segregation outwards of our cluster boundary , 9@xmath225 ) . as we showed in jskii , the mass function of this cluster steadily marches to flatter values , and becomes inverted in the outermost annuli . therefore , the observations provide the answer to this overabundance . however the effect is not seen in the luminosity function comparisons , which are in excellent agreement with the observed one . this suggests either that the faint magnitude luminosity function is dominated by field contamination or that the apparent over - concentration of the synthetic main sequences is actually an artifact of the residual underestimate of the photometric errors . assuming a single slope salpeter imf , the mass of gas that formed the stars of the surveyed area of ngc 6819 between 0.6 and 100 @xmath31 is @xmath40 with the bbc94 models , @xmath41 and @xmath42 with the fst and the franec models , respectively . hence , different stellar models do not provide the same answer for a global parameter such as the star formation either . the differences are again related to the adopted input physics and to the amount of assumed overshooting in the stellar models . less overshooting implies higher masses for the observed living stars , i.e. shorter lifetimes , which , convolved with the negative slope of the imf , leads to a higher total astrated mass . the cluster field contains 12194 _ bona fide _ stars , while the blank field extrapolated to the cluster area contains 10576 stars . hence , the synthetic cmds have been computed with 1618 objects . in jskiii , we assumed the cluster metallicity of ngc 2099 to be solar , based on the isochrone fitting results of @xcite as well as our own best guess from the fit . a more recent study , @xcite , prefer a subsolar z = 0.008 metallicity , again based on isochrone fitting . therefore , one goal is to constrain the metallicity of the cluster based on synthetic cmd comparisons . so , for this cluster we use the the fst tracks of @xcite with @xmath12 = 0.03 , @xmath28 = 0.02 and @xmath28 = 0.006 , the bbc94 tracks of the padova group with @xmath28 = 0.02 @xcite and @xmath28 = 0.008 @xcite , and the franec tracks @xcite with @xmath28 = 0.02 , @xmath28 = 0.01 and @xmath28 = 0.006 . for the fst , the synthetic sequences using solar metallicity models look impressively consistent with the observed cmd in all the minimal details . however , the turn - off and clump shape are not very good : the sequence after the turn - off is too long and curved and the clump is a bit short and faint . it is also very difficult to single out the best age , as anything between 520 and 700 myrs does nt significantly change the important cmd features ( clump morphology and luminosity , shape of turn - off , etc ... ) , all looking reasonable . the clump becomes too faint below 500 myrs , indicating that this is a lower age limit and too blue beyond 800 myrs indicating an upper limit . the best fit resulting parameters for these tracks are therefore age = 620 @xmath18 60 myrs , @xmath35 = 0.21 @xmath18 0.02 , and @xmath34 = 10.70 @xmath18 0.20 . the bbc94 @xmath28 = 0.02 tracks excellently reproduce all of the main sequence and clump features , but have long post turn - off sequences , as we saw in the fst models . for solar metallicity , these tracks are in better agreement with the data than the fst models . the age is also better constrained , thanks to a higher sensitivity of the various features : age = 590 @xmath18 10 myrs , @xmath35 = 0.18 @xmath18 0.01 , and @xmath34 = 10.50 @xmath18 0.10 . the franec @xmath28 = 0.02 tracks provide the best agreement with the observed upper main sequence and turn - off features . however , the resulting clump is too vertical and overpopulated . part of the latter problem is due to the fact that their main sequence is not as deep as the others , because the minimum mass available to us is 0.7 m@xmath4 , instead of 0.6 m@xmath4 as for the fst and padova models . the code therefore can not distribute the required 1618 faint stars in these bins and , inevitably , puts a larger number of brighter stars according to a salpeter imf . the resulting best fit parameters in the franec models are age = 590 @xmath18 10 myrs , @xmath35 = 0.16 @xmath18 0.02 , and @xmath34 = 10.31 @xmath18 0.10 . figure 4 presents the lower metallicity synthetic cmds for ngc 2099 . for the @xmath28 = 0.006 fst models ( left ) , the length of the sequence just after the turn - off shortens and becomes more vertical , thus better reproducing the observed morphology . the clump morphology also improves , thanks to a slightly larger extension . however , the agreement in the main - sequence morphology is no longer as excellent as the higher metallicity case , but still quite good . it may well be that intermediate metallicity tracks would work even better therefore , overall the synthetic cmds from the fst tracks prefer the lower metallicity , with parameters : age = 520 @xmath18 40 myrs , @xmath35 = 0.36 @xmath18 0.01 , @xmath34 = 10.40 @xmath18 0.10 . the bbc94 lower metallicity tracks ( middle ) , @xmath28 = 0.008 , are the best models for this cluster . they reproduce all of the main - sequence features and also have a better shaped turn - off than with the solar metallicity models . a minor inconsistency is seen in the location of the clump , which is slightly too vertical , but this is a minor detail . the age is less well defined , with final best parameters age = 430 @xmath18 30 myrs , @xmath35 = 0.36 @xmath18 0.03 , @xmath34 = 10.50 . the lower metallicity franec tracks ( right ) can be ruled out as both the @xmath28 = 0.01 and @xmath28 = 0.006 models show extremely extended clumps ( real blue loops ) which are inconsistent with the observed clump for any reasonable age . also , the lifetimes in that phase favor the stars to reside on the blue side of the clump ( vertical ) , adding further inconsistency to the colour and shape . we have also perfomed several tests on the fraction of binary stars in ngc 2099 . at first glance , the empirical cmd does nt show any evidence of binaries , but we have found a posteriori that , without including them in the synthetic cmds , both the turn - off morphology and the colour distribution on the red side of the main sequence are not well reproduced . our best guess is that binaries are around 15 - 20% of the cluster stars and all of the results that we have discussed above assume this fraction ( 20% ) with a random mass ratio . figure 5 shows the empirical luminosity function of the 12194 cluster field stars ( empty circles ) compared to the luminosity functions resulting from the sum of the 1618 synthetic stars of figure 4 ( subsolar metallicity ) with the 10576 stars of the extrapolated blank field . the predicted luminosity functions are consistent with the data , but the agreement at the brightest magnitudes is not exceptional . the very good match at the faintest magnitudes shows that the extrapolation of the number of stars observed in the blank field to the larger area covered by the cluster is correct . the general fits of the data to synthetic cmds favor the lower metallicity models for ngc 2099 , thus requiring a higher reddening . these are found to better reproduce the shape of the upper main sequence and turn - off . notice that age and modulus derived here do not coincide with those derived jskiii with the same stellar tracks . the difference in age is probably due to the fact that with the isochrone fitting method one simply fits a curve to the points , whereas with the synthetic cmds one has to reproduce the points themselves , including morphology and quantity . the difference in modulus is , presumably , simply the consequence of the age difference . assuming a single slope salpeter imf , the mass of gas that formed the stars of the surveyed area of ngc 2099 between 0.6 and 100 @xmath31 is @xmath43 with the bbc94 models , @xmath44 and @xmath45 with the fst and the franec models , respectively . the cluster field contains 9298 _ bona fide _ stars . our previous work on ngc 2168 ( jskiv ) suffered due to the large size of the cluster ( r @xmath46 20@xmath2 ) . a very small blank field was constructed from the outer edges of the cfh12k ccd mosaic . in fact , some cluster members surely resided within the blank field , and therefore the cluster luminosity function could have been underestimated . using synthetic cmds , we find that we can only reproduce the observed stellar distribution in the cmd and the luminosity function if we use 500 - 600 stars down to @xmath9 = 19 . using the blank field from the outer edges of the ccd subtracts off too many stars and forces a poor agreement between the theory and the observations . this confirms our initial suspicion that the cluster is larger than our aerial coverage . the synthetic cmds shown here contain 550 stars brighter than @xmath9 = 19 . for the synthetic cmd comparisons , we again use both solar and subsolar metallicity models from each of fst , bbc94 , and franec . we immediately find that the subsolar metallicity models provide a much nicer agreement to the observations than solar metallicity models . the solar metallicity models not only predict an insufficient number of main - sequence stars just below the turn - off , they also contain very vertical upper main sequences . besides , for ages as old as @xmath3180 myrs the turn - offs are slightly too faint , while for younger ages , @xmath47120 myrs , the turn - off has the proper luminosity but is severely underpopulated when compared to the observations . flattening the imf did not help the overall fit . figure 6 shows the best synthetic cmds for ngc 2168 . for the fst tracks ( left ) , the models with maximum overshooting ( @xmath12 = 0.03 ) and low metallicity ( @xmath28 = 0.006 ) reproduce the data very nicely . these require a best fit distance modulus slightly smaller ( @xmath34 = 9.6 ) and a reddening slightly larger ( @xmath35 = 0.22 ) than found in jskiv . the differences can be most likely attributed to the lower metallicity ( jskiv uses @xmath28 = 0.012 ) that we chose in these comparisons . the best age for this model is 200 myrs , slightly older than what jskiv found ( 180 myrs ) . the absence of a population of post main - sequence stars coupled with the insensitivity of the turn - off luminosity and morphology with age in these models allows for a nice agreement between the theory and data with ages down to 120 myrs . the agreement is found to be not as good if we use the same metallicity fst models with slightly lower overshooting , @xmath12 = 0.02 . at an age of 180 myrs the turn - off is too faint , however , with younger ages ( 150 myrs ) we overpopulate the post main - sequence stars ( five are seen ) . at 120 myrs , the turn - off is too bright , hence this age is too young . the bbc94 models for ngc 2168 are shown in the middle panel of figure 6 . the @xmath28 = 0.08 tracks are found to excellently reproduce both the cmd morphology and the number of stars in the different phases . the main - sequence curves are best fitted using @xmath35 = 0.20 and @xmath34 = 9.8 . the best age estimate is found to be 180 myrs . an age of 150 myrs does nt provide any post main - sequence stars whereas an age of 200 myrs is too old as it predicts too many clump stars . the franec models ( right ) predict too many post main - sequence stars for both adopted metallicities , @xmath28 = 0.01 and @xmath28 = 0.02 . as summarised in jskiv , many studies of ngc 2168 exist in the literature with none finding more than a few clump stars . the franec models would require a very young age ( 110 myrs ) to fit the post main - sequence phases and this would in turn force a poor agreement in the turn - off region . the best fit model is , arguably , the one with age = 110 myrs , @xmath35 = 0.22 , and @xmath34 = 9.8 . our best estimate for the ngc 2168 binary fraction is found to be @xmath330% . this fraction puts a consistent number of stars , and sparse regions , on the red side of the cluster main - sequence in the synthetic cmds . all cmds shown in figure 6 contain this binary fraction . comparing the resulting best synthetic cmds with the observed one , it is clear that the predicted main sequence is too tight and slightly underpopulated at faint magnitudes ( @xmath39 15 ) . as addressed earlier in [ ngc6819 ] , the tightness is a question of photometric errors , possible differential reddening , or metallicity spreads within the cluster . the underpopulation in this case is most likely related to the uncertain method we have used to infer the number of cluster members . many tests that we have done indicate that we can not attribute more stars to the cluster without getting too many bright stars , both on the main sequence and in subsequent evolutionary phases . we also note here that @xmath9 = 19 is the limiting magnitude of the numerical simulations for this cluster . this is related to the availability of photometric conversion tables @xcite for the adopted metallicities . with that said , it is remarkable how well the overall shape of the upper main sequence is reproduced . the kinks and slope changes all fall in the correct place as dictated by the observations . this allows us to accurately determine the best fit parameters for ngc 2168 . interestingly , the best fit synthetic cmd for this cluster , the bbc94 model , provides an identical reddening , distance , and age to what we found using the fst isochrone in jskiv . the fst synthetic cmd derived parameters are also in excellent agreement with our observations considering that the metallicities are slightly different . figure 7 shows the empirical luminosity function of the 9298 cluster field stars ( empty circles ) compared to the luminosity functions resulting from the sum of the synthetic stars of figure 6 with the 8748 stars resulting by randomly multiplying the 1150 objects of the supposed blank field for an area normalisation factor . clearly , a major side effect of the lack of an appropriate external field is that the comparison of the luminosity functions becomes only a matter of self - consistency and can not be used for any consideration on mass segregation / evaporation . as mentioned above , the luminosity functions agree with the data only with 500 - 600 cluster members . one may think to improve the situation by steepening significantly the imf ( thus allowing for more stars in the fainter main sequence , without adding bright ones ) , but we consider it highly dangerous to modeify the imf without a good field subtraction . we stress that observations of a deep @xmath0 field offset from the centre of ngc 2168 by at least one degree would be very useful . assuming a single slope salpeter imf , the mass of gas that formed the stars of the surveyed area of ngc 2168 between 0.6 and 100 @xmath31 is @xmath48 with the bbc94 models , @xmath49 and @xmath50 with the fst and the franec models , respectively . the cluster field contains 11115 _ bona fide _ stars . taking into account that 1 ) the blank field contains 5142 stars but covers an area 1.8 times smaller than the cluster field , 2 ) the mass of the available stellar models do nt allow synthetic stars fainter that @xmath51 , the synthetic cmds have been initially computed with 565 objects brighter than @xmath9 = 18 . a posteriori , we have however found that in this way we predict luminosity functions clearly overestimating the bright portion ( where the cluster dominates ) and underestimating the faint portion ( where field contamination dominates ) . we empirically find that good luminosity functions are obtained only if the cluster contains 285 members brighter than @xmath52 , suggesting that the previously used blank field underestimated the actual contamination . again observations of an external field of appropriate size would be crucial for a safer derivation of some of the cluster properties . for the youngest cluster in this work , ngc 2323 , the metallicity is known to be solar and in fact , we find that the solar metallicity tracks provide the best agreement to the data in this work . the most difficult parameter to fit in ngc 2323 is the cluster age . although ages can be constrained from the turn - off morphology , the luminosity of the red giant clump is an alternative tool . ngc 2323 only shows two bright , red objects which are potential post main - sequence candidates . however , since these two objects do not have any membership probability from the literature , we can not trust them as cluster members . furthermore , the turn - off itself is not heavily populated preventing an accurate determination of its luminosity . as we ll see below , we find that models with a range of young ages are equally consistent with the observed distribution . figure 8 presents our best synthetic cmds for ngc 2323 . all of the shown cases are for solar metallicity and assume a 30% binary fraction with random mass ratio . the fst models ( left ) are found to perfectly reproduce the majority of the observed features . the agreement is also found to be independent of the adopted overshooting ( @xmath12 = 0.02 or @xmath12 = 0.03 ) . although the age is slightly higher ( by @xmath320 myrs ) in the higher overshooting case , the reddening , distance modulus and goodness of fit are the same . our best parameters are therefore age = 130 myrs , @xmath35 = 0.22 , and @xmath34 = 10.0 for @xmath12 = 0.03 . however , ages between 120 and 180 myrs are equally likely . we have chosen 130 myrs as this age predicts a post main - sequence star where a real star ( member or not ) is indeed located on the cmd . therefore , the results from this work for the fst models are identical to the parameters found in jskiv . the bbc94 stellar tracks also perfectly reproduce features on the cmd of this cluster . for @xmath35 = 0.22 , the best fit distance modulus is found to be @xmath34 = 10.2 . in practice , all ages between 80 and 150 myrs fit the data equally well . we ve again chosen 120 myrs as the best age as it predicts a post main - sequence star in the correct location as dictated by the potential red giant in the data . the franec models ( right ) predict a larger number of post main - sequence stars than the fst and bbc models . this is therefore inconsistent with the observational data . for solar metallicity , the best compromise between the number of post main - sequence stars , the turn - off luminosity , and the shape of the upper main sequence is found for an age of 110 myrs . the turn - off is still too faint , however , younger ages produce very vertical and blue upper main sequences . the best fit reddening is @xmath35 = 0.22 and the distance is @xmath34 = 10.0 . since the results from these latter comparisons were not very satisfactory , we also tried a lower metallicity , @xmath28 = 0.01 . the results are very similar , with an overproduction of post main - sequence stars and a worse shape to the upper main sequence . figure 9 shows the empirical luminosity function of the 11115 cluster field stars ( empty circles ) compared to the luminosity functions resulting from the sum of the synthetic stars of figure 8 with the 10576 stars of the extrapolated blank field . as already discussed for ngc 2168 , the lack of an appropriate estimate of the back / foreground stars does nt allow for a critical analysis of the luminosity function . we can see , however , from figure 8 , that within the uncertainties , the selected models reproduce quite well the observed luminosity function . assuming a single slope salpeter imf , the mass of gas that formed the stars of the surveyed area of ngc 2323 between 0.6 and 100 @xmath31 is @xmath53 with the bbc94 models , @xmath54 and @xmath55 with the fst and the franec models , respectively . table 2 summarises the results described in the previous section and includes other information about the four clusters studied here . we recall that stellar models that include overshooting from convective regions imply older ages than models without overshooting . this explains why our ages resulting from the fst tracks with @xmath12=0.03 are systematically older than those obtained with the bbc94 models and much older than those obtained with the franec ones . it is interesting to notice that while the fst models with higher overshooting reproduce the data better than those with @xmath12=0.02 , the bbc94 models appear to be those in better overall agreement , in spite of having a formal overshooting somewhat lower than the fst @xmath12=0.03 models . this indicates that what counts in the predictions of stellar evolution models is the combination of the many parameters and input physics . we find that in general the bbc94 models better reproduce both the morphological features and the numbers of stars in the various evolutionary phases of the observed clusters . llllll parameter & explanation & ngc 6819 & ngc 2099 & ngc 2168 & ngc 2323 + position : & + @xmath56 & ra & @xmath57 & @xmath58 & @xmath58 & @xmath59 + @xmath60 & declination & @xmath61 & @xmath62 & @xmath62 & @xmath63 + @xmath64 & galactic longitude & @xmath65 & @xmath66 & @xmath67 & @xmath68 + @xmath69 & galactic latitude & @xmath70 & @xmath71 & @xmath72 & @xmath73 + + distances : & + @xmath74 - previous study & apparent distance modulus & 12.30 @xmath18 0.12 & 11.55 @xmath18 0.13 & 10.42 @xmath18 0.13 & 10.68 @xmath18 0.14 + @xmath74 - fst ( @xmath12 = 0.03 ) & apparent distance modulus & 12.31 & 11.52 & 10.22 & 10.68 + @xmath74 - bbc94 & apparent distance modulus & 12.57 & 11.62 & 10.42 & 10.88 + @xmath74 - franec & apparent distance modulus & 12.47 & 11.42 & 10.48 & 10.68 + + @xmath35 - previous study & reddening & 0.10 & 0.21 & 0.20 & 0.22 + @xmath35 - fst ( @xmath12 = 0.03 ) & reddening & 0.10 & 0.36 & 0.20 & 0.22 + @xmath35 - bbc94 & reddening & 0.12 & 0.36 & 0.20 & 0.22 + @xmath35 - franec & reddening & 0.15 & 0.36 & 0.22 & 0.22 + + @xmath34 - previous study & true distance modulus & 11.99 @xmath18 0.18 & 10.90 @xmath18 0.16 & 9.80 @xmath18 0.16 & 10.00 @xmath18 0.17 + @xmath34 - fst ( @xmath12 = 0.03 ) & true distance modulus & 12.0 & 10.40 & 9.60 & 10.00 + @xmath34 - bbc94 & true distance modulus & 12.2 & 10.50 & 9.80 & 10.20 + @xmath34 - franec & true distance modulus & 12.0 & 10.30 & 9.80 & 10.00 + + @xmath19 - previous study & distance from sun & 2500 @xmath75 pc & 1513 @xmath76 pc & 912 @xmath27 pc & 1000 @xmath77 pc + @xmath19 - fst ( @xmath12 = 0.03 ) & distance from sun & 2512 pc & 1202 pc & 832 & 1000 + @xmath19 - bbc94 & distance from sun & 2754 pc & 1259 pc & 912 & 1096 + @xmath19 - franec & distance from sun & 2512 pc & 1148 pc & 912 & 1000 + + age : & + @xmath78 - previous study & isochrone fit age & 2.5 gyrs & 520 myrs & 180 myrs & 130 myrs + @xmath78 - fst ( @xmath12 = 0.03 ) & synthetic cmd age & 2.3 gyrs & 520 myrs & 200 myrs & 130 myrs + @xmath78 - bbc94 & synthetic cmd age & 2.0 gyrs & 430 myrs & 180 myrs & 120 myrs + @xmath78 - franec & synthetic cmd age & 1.6 - 1.8 gyrs & 400 myrs & 110 myrs & 110 myrs + + metallicity : & + @xmath28 & heavy metal abundance@xmath79 & 0.02 & @xmath470.02 & 0.012 & 0.020 + + binary fraction & + binaries & binary percentage & 20% & 20% & 30% & 30% + + size : & + @xmath80 & angular diameter & 19@xmath2 & 27@xmath228 & @xmath4632@xmath2 & 30@xmath2 + @xmath81 & linear diameter@xmath82 & 15.2 pc & 10.2 pc & @xmath468.5 pc & 9.6 pc + @xmath83 & mass of cluster@xmath84 & 4000 @xmath85 & 2500 @xmath85 & 850 @xmath85 & 480 @xmath85 + + \1 . these values were not spectroscopically determined and reflect those used in the original isochrone fits . + 2 . computed using the distances from the bbc94 results . m is the mass of gas that formed stars between 0.6 and 100 @xmath85 and is therefore only a lower limit to the total astrated mass . + the excellent delineation of the various main - sequence curves and kinks allows us to put fairly stringent limits on both the reddening and the distance modulus of each cluster , independent of the age . table 2 shows in fact that the variation of the values of these two quantities resulting from different sets of stellar tracks is quite small . this clearly helps to better define the age as well . as discussed for the case of ngc 6819 , the synthetic main sequences of all the four clusters are tighter than the observed main sequences when the formal photometric errors of the survey catalogues are assumed . to reproduce the observed width , the synthetic cmds require either photometric errors of at least 0.01 magnitude even in the brighter bins or some other factors such as an internal metallicity spread or differential reddening . the synthetic cmd tests clearly indicate that ngc 2168 is larger than previously thought and that our outer ccd blank field is not adequate . a similar problem , although to a lesser degree , may also exist in the ngc 2323 data set . future deep photometry of these clusters should include separate blank field observations . indeed , in spite of the great depth and excellent accuracy of our photometry , which allows us to measure the faintest , least massive stars of each cluster , the lack of appropriate decontamination fields has prevented us from safely deriving their present mass functions and , consequently , their imf , mass segregation , and evaporation status . the synthetic cmds have allowed us to infer the astrated mass of each cluster . we find that although this does depend on the adopted set of stellar evolutionary tracks , it only varies within a factor 1.2 from the minimum to the maximum value obtained with the stellar models considered here . we have chosen to simply apply salpeter s single slope imf for this exercise . we note , however , that increasing evidence suggests that the slope is most likely different ( steeper for more massive stars , flatter for less massive ones , e.g. , kroupa 2002 ) beyond the range of masses , @xmath86 , originally analysed by salpeter ( 1955 ) . although this simplistic assumption introduces a systematic error in our estimates , it is not a large concern as we are not providing absolute values of star formation rates but rather just comparing relative rates for the four clusters . given the angular size of the observed field of view for each cluster ( see table 2 ) , the actual area surveyed obviously depends on the adopted distance to the cluster . we have assumed both the instrinsic distance modulus and the astrated mass resulting from the best synthetic cmd comparisons , the bbc94 models . this corresponds to a distance from the sun of 2754 pc for ngc 6819 , 1259 pc for ngc 2099 , 912 pc for ngc 2168 , and 1096 pc for ngc 2323 . the surveyed areas are then 182 , 81 , 56 and 72 @xmath87 , respectively . the corresponding astrated mass per unit area is then 23.1 , 32.1 , 15.2 and 6.7 @xmath88 , equivalent to star formation rates of @xmath89 , and @xmath90 , respectively . these values refer only to stars with masses included in the adopted sets of evolutionary tracks , namely from 0.6 to 100 @xmath31 . to derive total astrated masses and star formation rates , one should extrapolate down to 0.1 @xmath31 , following the preferred imf . in practice , extrapolating salpeter s imf down to 0.1 @xmath31 , we must multiply by a factor 2.05 the astrated masses and the rates . if , at the other extreme , we prefer to assume the positive slope + 0.44 below 0.6 @xmath31 suggested by gould , bahcall & flynn ( 1997 ) , then the above values must be multiplied only by 1.27 . there is a significant difference ( more than a factor of four ) in the star formation rates of our four clusters , the most active having been ngc 2099 and the least active ngc 2323 , with no apparent relation with galactic location . these rates are 23 orders of magnitude higher than the average star formation rate of field stars in the solar neighbourhood , ( @xmath91 ( e.g. timmes , woosley & weaver 1995 ) , and are comparable to those of the most active dwarf irregular and blue compact galaxies ( e.g. tosi 2003 and references therein ) . the survival factor of the stars formed in any cluster clearly depends both on age and on evaporation and ngc 6819 turns out to be the only one of the four clusters having suffered a reduction from the number of formed stars to that of the stars still alive and present in the studied area . we re - derive key parameters for the four very rich open star clusters , ngc 6819 , ngc 2099 , ngc 2168 , and ngc 2323 . the parameters are measured by comparing high quality , deep empirical colour - magnitude diagrams with montecarlo simulations . the combination of comparing the morphology and number of stars in various evolutionary phases and cluster luminosity functions allows us to provide tight constraints on the reddening , distance , age and binary fraction for each cluster . in cases where the cluster metallicity is not certain , simulations with different abundances are also compared . in all cases the data are better reproduced when a fraction of unresolved binary systems between 20 and 30% is assumed . a summary of the results is given in table 2 . the synthetic cmds and lfs are generally found to be in excellent agreement with the observational data . this circumstance , combined with the fact that different sets of stellar evolution tracks provide different values for the cluster parameters , confirms how important it is to use more than one set of models to estimate the theoretical uncertainties . it also shows that a homogeneous approach is crucial to derive reliable overall cluster properties , such as age - metallicity relations . in fact , cluster dating based on different stellar models may lead not only to different absolute ages , but also to different age ranking , with significant drawbacks on the interpretation of the cluster properties in terms of galactic evolution . some discrepancies , such as the thickness of the main sequences _ vis vis _ the size of the photometric errors , and the possibility that cluster stars may be contaminating our blank field ( particularly in the case of ngc 2168 ) are identified and discussed . for each cluster , we also measure the astration mass ( i.e. , the total mass that went in all of the stars formed in the cluster ) according to a single slope salpeter imf . from this , we calculate the star formation rate between 0.1100 @xmath31 to be @xmath92 , and @xmath93 for ngc 6819 , ngc 2099 , ngc 2168 , and ngc 2323 respectively . these rates would be a factor 1.6 lower if the imf below 0.6 @xmath31 has a slope of + 0.44 , as inferred by gould et al . ( 1997 ) from hst data , rather than salpeter . the true value may in fact lie somewhere in between these two imfs , as recently discussed e.g. , by chabrier ( 2003 ) . in principle our data are deep enough to allow for a direct derivation of the cluster imf ; however , the four clusters are heavily contaminated by fore / background stars and the lack of appropriate decontamination fields has prevented us from a safe analysis of the star counts at the fainter magnitudes . the need of appropriate photometry in nearby fields is emphasised , also for the purpose of adequate studies of the clusters mass segregation / evaporation . we thank paolo ventura and franca dantona for having provided their stellar models prior to publication . the bulk of the synthetic cmd code was originally written by laura greggio . j. s. k. received financial support during this work through an nserc pgs - b graduate student research grant . this work has been partially funded through the italian miur - cofin-2003029437 . andreuzzi , g. , bragaglia , a. , tosi , m. , marconi , g. 2004 , mnras , in press , astro - ph/0311249 aparicio , a. , bertelli , g. , chiosi , c. , garcia - pelayo , j. m. 1990 , a&a , 240 , 262 barrado y navascus , d. , deliyannis , c. p. , & stauffer , j. r. 2001 , apj , 549 , 452 bellazzini , m. , ibata , r. , monaco , l. , martin , n. , irwin , n.j . 2003 , mnras , in press , astro - ph/0311119 bertelli , g. , bressan , a. , chiosi , c. , fagotto , f. , & nasi , e. 1994 , a&as , 106 , 275 bessell , m. s. , castelli , f. , & plez , b. 1998 , a&a , 333 , 231 bragaglia , a. , 2003 , in _ stars in galaxies _ , m.bellazzini , a.buzzoni , s.cassisi eds , mem.s.a.it . in press bragaglia , a. , carretta , e. , gratton , r. g. , tosi , m. , bonanno , g. , bruno , p. , cali , a. , claudi , r. , cosentino , r. , desidera , s. , farisato , g. , rebeschini , m. , & scuderi , s. 2001 , aj , 121 , 327 bragaglia , a. , & tosi , m. 2003 , mnras , 343 , 306 bressan , a. , fagotto , f. , bertelli , g. , & chiosi , c. 1993 , a&as , 100 , 647 canuto , v. m. , & mazzitelli , i. 1992 , apj , 389 , 724 carraro , g. , ng , y.k . , & portinari , l. 1998 , mnras , 296 , 1045 chabrier , g. 2003 , apj , 586 , l133 dominguez , i. , chieffi , a. , limongi , m. , & straniero , o. 1999 , apj , 524 , 226 fagotto , f. , bressan , a. , bertelli , g. , & chiosi , c. 1994 , a&as , 105 , 29 french , r. s. 2003 , private communication freytag , b. , ludwig , h. g. , & steffen , m. 1996 , a&a , 313 , 497 friel , e.d . 1995 , ara&a , 33 , 38 gould , a. , bahcall , j. , & flynn , c. 1997 , apj , 482 , 913 hauschildt , p. h. , allard , f. , & baron , e. 1999 , apj , 512 , 377 janes , k.a . , & phelps , r.l . 1994 , aj , 108 , 1773 kalirai , j. s. , fahlman , g. g. , richer , h. b. , & ventura , p. 2003 , aj , accepted . ( astro - ph/0306241 ) kalirai , j.s . , richer , h.b . , fahlman , g.g . , cuillandre , j. , ventura , p. , dantona , f. , bertin , e. , marconi , g. , & durrell , p. 2001a , aj , 122 , 257 kroupa , p. 2002 , science , 295 , 82 landolt , a. u. 1992 , apj , 104 , 340 reid , i. n. 1999 , ara&a , 37 , 191 salpeter , e. e. 1955 , apj , 121 , 161 sarrazine , a. r. , steinhauer , a. j. b. , deliyannis , c. p. , sarajedini , a. , bailyn , c. d. , kozhurina - platais , v. , von hippel , t. , & platais , i. 2000 , a&as , 32 742 skillman , e.d . , & gallart , c. 2002 , pasp , 274 , 535 timmes , f.x . , woosley , s.e . , weaver , t.a . 1995 , apjs , 98 , 617 tolstoy , e. , & saha , a. 1996 , apj , 46 2 , 672 tosi , m. 2000 , in _ the chemical evolution of the milky way : stars versus clusters _ , f. matteucci , f. giovannelli eds , ( kluwer , dordrecht ) , 255 , 505 tosi , m. 2003 , in _ stars in galaxies _ , a. buzzoni , s. cassisi and m. bellazzini eds , mem.s.a.it in press ( astro - ph/0305416 ) tosi , m. , greggio , l. , marconi , g. , & focardi , p. 1991 , aj , 102 , 951 twarog , b.a . , ashman , k.m . , & anthony - twarog , b. 1997 , aj , 114 , 2556 ventura , p. , zeppieri , a. , mazzitelli , i. & dantona , f. 1998 , a&a , 334 , 953

the molecular gas , h , that fuels star formation in galaxies is difficult to observe directly . as such , the ratio of to is an observational estimation of the star formation rate compared with the amount of molecular gas available to form stars , which is related to the star formation efficiency and the inverse of the gas consumption timescale . we test what effect an ir luminous agn has on the ratio in a sample of 24 intermediate redshift galaxies from the 5 mjy unbiased _ spitzer _ extragalactic survey ( 5muses ) . we obtain new co(1 - 0 ) observations with the redshift search receiver on the large millimeter telescope . we diagnose the presence and strength of an agn using _ spitzer _ irs spectroscopy . we find that removing the agn contribution to results in a mean for our entire sample consistent with the mean derived for a large sample of star forming galaxies from . we also include in our comparison the relative amount of polycyclic aromatic hydrocarbon emission for our sample and a literature sample of local and high redshift ultra luminous infrared galaxies and find a consistent trend between and , such that small dust grain emission decreases with increasing for both local and high redshift dusty galaxies .

1411.1065

star formation is one of the main internal driving forces of galaxy evolution , resulting in the chemical enrichment of a galaxy , the heating of the interstellar medium ( ism ) , and indirectly , the production of dust through the winds of dying stars . star formation converts a galaxy s molecular gas into stars through multiple complicated processes including gas accretion and the collapse and cooling of molecular clouds . although the star formation process itself is intricate , the overall conversion of gas into stars can be expressed simply by the schmidt - kennicutt ( sk ) law which directly relates the molecular gas content to the star formation rate ( sfr ) through a power - law equation , @xmath8 , albeit with significant scatter @xcite . the bulk of the present day stellar mass was formed at a peak epoch of star formation , from @xmath9 ( * ? ? ? * and references therein ) . during this era , the buildup of stellar mass was dominated by dusty galaxies referred to as luminous infrared galaxies ( lirgs , @xmath10 ) and ultra luminous infrared galaxies ( ulirgs , @xmath11 ) ( e.g. , * ? ? ? in the past two decades , the spate of far - ir / submillimeter space - based and ground - based telescopes have enabled astronomers to simultaneously study the star formation , through infrared ( ir ) emission , and molecular gas , through co emission , of dusty galaxies out to redshifts of @xmath12 ( * ? ? ? * and references therein ) . the ir luminosity , @xmath1 , is an ideal measure of the sfr for dusty galaxies as it is the integrated emission from the dust , presumably heated by star formation . on galaxy - wide scales , co traces the molecular hydrogen which is difficult to observe directly ; a conversion factor , @xmath13 , is used to relate the co luminosity directly to the h@xmath0 mass ( * ? ? ? * and references therein ) . in the past few years , a `` galaxy main sequence '' has been empirically determined for local and high redshift galaxies ; a tight relationship holds between sfr and stellar mass , and this relationship evolves with redshift ( e.g. , * ? ? ? * ; * ? ? ? galaxies that lie above the main sequence , that is , galaxies that have an enhanced sfr for a given stellar mass , are designated `` starbursts '' in this parameter space , as they are thought to be undergoing a short - lived burst of star formation , likely triggered by a major merger . the rate at which a galaxy can form stars is limited by the amount of molecular gas present . @xcite proposed an extended sk law which relates the specific star formation rate ( @xmath14 ) to the stellar mass surface density , suggesting that the existing stellar population may play a role in regulating the amount of star formation . the authors apply their extended sk law to an analytical model of gas accretion and find that it accurately reproduces the galaxy main sequence . a dichotomy between starbursts and normal star forming galaxies may also be observed by comparing @xmath1 with @xmath2 . first , there may be a `` normal '' rate of star formation measured in undisturbed disk galaxies for a given amount of molecular gas . then , there is an enhanced starburst mode , where a galaxy has a higher @xmath1 than expected for a given @xmath2 , possibly triggered by a major merger funneling gas towards the inner regions of a galaxy ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? this dichotomy has a direct effect on the calculation of the h@xmath0 mass , since @xmath13 is proposed to be a factor of @xmath15 lower for starbursting galaxies , due to significant amounts of co residing in the inter - cloud medium . correctly identifying starbursts is critical for calculating accurate gas masses . it now appears that every massive galaxy hosts a supermassive black hole at the center , implying that all galaxies have gone through an active galactic nuclei ( agn ) phase , and some lirgs and ulirgs show signs of concurrent black hole growth and star formation ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in the classical picture of galaxy evolution , a starburst is triggered by a major merger , and this phase can be followed by an agn phase , implying that an obscured growing agn may be observable during a galaxy s starburst phase ( e.g. , * ? ? ? if present , an agn can dramatically influence the internal evolution of a galaxy by heating the dust , expelling the gas , and ultimately quenching the star formation through feedback mechanisms ( e.g. , * ? ? ? the effect of an obscured agn on the ism can be probed through ir observations . in the mid - ir , star forming galaxies have prominent pah features arising from photodissociation regions ( pdrs ) , but agn emission can dilute these features , leaving a warm dust power - law continuum @xcite . radiation from the agn can heat the dust in the ism to temperatures @xmath16k , producing a significant contribution ( @xmath17 ) to the far - ir emission ( e.g. , * ? ? ? * ; * ? ? ? the degree of scatter in the relationship between @xmath1 and @xmath2 , often attributed to the two different modes of star formation , is possibly affected by the contribution from an agn to @xmath1 . @xcite observe high @xmath3 in qsos and infer that these ratios might be boosted by an agn contribution to @xmath1 . @xcite build on this study by using hcn as a tracer of star formation in qsos . using iras bright galaxies , the authors determine the median hcn / co ratio for normal star forming galaxies and use this , combined with @xmath3 and @xmath18 , to statistically correct @xmath1 of qsos for an agn contribution . an alternate approach is to use @xmath19 ( 40 - 500@xmath20 m or 42.5 - 11.5@xmath20 m ) instead of @xmath1 ( 8 - 1000@xmath20 m ) , as dust emission in this regime should come primarily from heating by young stars . @xcite conclude that @xmath21 in a sample of 17 qsos is similar to the ratios in local ulirgs ( many of which are known to host obscured agn ) , and @xcite finds high @xmath21 ratios in qsos relative to the iras galaxies . in this paper , we build upon observations in the high redshift and local universe with a study of intermediate redshift galaxies ( @xmath22 ) from the 5 mjy unbiased _ spitzer _ extragalactic survey ( 5muses ; p.i . george helou ) . these galaxies have extensive ir data from the _ spitzer space telescope _ and _ herschel space observatory _ , allowing us to accurately measure @xmath1 and quantify the contribution from an agn . we complement the existing ir data with new co(1 - 0 ) observations from the redshift search receiver , which has a large bandwidth of 38ghz and a resolution of 100km / s , on the large millimeter telescope alfonso serrano , enabling us to explore how much of the scatter in the @xmath23 relation is due to an agn contribution to @xmath1 . we discuss the details of our sample and observations in section 2 , the effect of an agn on the relationships between @xmath1 , pahs , and co(1 - 0 ) emission in section 3 , and summarize our findings in section 4 . throughout this paper , we adopt a flat cosmology with @xmath24kms@xmath25mpc@xmath25 , @xmath26 , and @xmath27 . lccc cccc 5muses-200 & 16:12:50.9 & + 53:23:05.0 & 0.043 & 0.70 & & @xmath28 & @xmath29 + 5muses-179 & 16:08:03.7 & + 54:53:02.0 & 0.053 & 0.32 & 348 @xmath30 20 & 11.3 @xmath30 0.93 & 1.46 @xmath30 0.12 + 5muses-169 & 16:04:08.3 & + 54:58:13.1 & 0.064 & 0.72 & 306 @xmath30 67 & 4.47 @xmath30 1.08 & 0.84 @xmath30 0.20 + 5muses-105 & 10:44:32.9 & + 56:40:41.6 & 0.068 & 0.84 & 251 @xmath30 20 & 7.19 @xmath30 1.27 & 1.54 @xmath30 0.27 + 5muses-171 & 16:04:40.6 & + 55:34:09.3 & 0.078 & 0.87 & 345 @xmath30 48 & 7.60 @xmath30 1.45 & 2.14 @xmath30 0.41 + 5muses-229 & 16:18:19.3 & + 54:18:59.1 & 0.082 & 0.72 & 386 @xmath30 33 & 11.3 @xmath30 1.48 & 3.53 @xmath30 0.46 + 5muses-230 & 16:18:23.1 & + 55:27:21.4 & 0.084 & 0.87 & 305 @xmath30 50 & 6.38 @xmath30 1.25 & 2.09 @xmath30 0.41 + 5muses-234 & 16:19:29.6 & + 54:18:41.9 & 0.100 & 0.70 & 217 @xmath30 25 & 6.24 @xmath30 0.69 & 2.93 @xmath30 0.33 + 5muses-132 & 10:52:06.6 & + 58:09:47.1 & 0.117 & 1.30 & 181 @xmath30 44 & 4.28 @xmath30 0.92 & 2.77 @xmath30 0.60 + 5muses-227 & 16:17:59.2 & + 54:15:01.3 & 0.134 & 0.24 & 247 @xmath30 26 & 1.17 @xmath30 0.34 & 1.00 @xmath30 0.29 + 5muses-141 & 10:57:05.4 & + 58:04:37.4 & 0.140 & 0.48 & 203 @xmath30 29 & 3.06 @xmath30 0.62 & 2.86 @xmath30 0.58 + 5muses-158 & 16:00:38.8 & + 55:10:18.7 & 0.145 & 0.73 & 458 @xmath30 78 & 5.17 @xmath30 1.23 & 5.21 @xmath30 1.24 + 5muses-225 & 16:17:48.1 & + 55:18:31.1 & 0.145 & 0.46 & 400 @xmath30 146 & 4.52 @xmath30 0.57 & 4.55 @xmath30 0.57 + 5muses-273 & 16:37:31.4 & + 40:51:55.6 & 0.189 & 0.34 & 241 @xmath30 69 & 1.81 @xmath30 0.40 & 3.16 @xmath30 0.69 + 5muses-136 & 10:54:21.7 & + 58:23:44.7 & 0.204 & 0.36 & 477 @xmath30 78 & 2.30 @xmath30 0.55 & 4.69 @xmath30 1.11 + 5muses-294 & 17:12:32.4 & + 59:21:26.2 & 0.210 & 0.38 & & @xmath31 & @xmath32 + 5muses-216 & 16:15:51.5 & + 54:15:36.0 & 0.215 & 0.27 & 331 @xmath30 60 & 1.82 @xmath30 0.36 & 4.14 @xmath30 0.83 + 5muses-194 & 16:11:19.4 & + 55:33:55.4 & 0.224 & 0.54 & & @xmath33 & @xmath34 + 5muses-249 & 16:22:14.8 & + 55:06:14.2 & 0.237 & 0.41 & & @xmath35 & @xmath36 + 5muses-250 & 16:23:13.1 & + 55:11:11.6 & 0.237 & 0.48 & 599 @xmath30 40 & 5.56 @xmath30 0.68 & 15.5 @xmath30 1.89 + 5muses-275 & 16:37:51.4 & + 41:30:27.3 & 0.286 & 0.61 & 950 @xmath30 163 & 7.77 @xmath30 1.38 & 32.0 @xmath30 5.68 + 5muses-313 & 17:18:52.7 & + 59:14:32.1 & 0.322 & 0.47 & & @xmath37 & @xmath38 + 5muses-156 & 15:58:33.3 & + 54:59:37.2 & 0.340 & 0.47 & & @xmath39 & @xmath40 + 5muses-101 & 10:41:59.8 & + 58:58:56.4 & 0.360 & 0.32 & & @xmath41 & @xmath42 5muses is a _ spitzer _ irs mid - ir spectroscopic survey of 330 galaxies selected from the swire and _ spitzer _ extragalactic first look survey fields ( details in * ? ? ? it is a flux limited sample selected at 24@xmath20 m using mips observations from the _ spitzer space telescope _ , with @xmath43mjy . crucially , 5muses is a representative sample at intermediate redshift ( the median redshift of the sample is 0.14 , 1.8 gyr ago ) of galaxies with @xmath44 , bridging the gap between large samples of local star forming galaxies , lirgs , and ulirgs and high redshift observations of lirgs and ulirgs . complete details of the _ spitzer _ data reduction are found in @xcite . for the present study , we make use of the _ spitzer _ irs short - low ( sl ) spectroscopy , spanning the range @xmath45 m . a 24@xmath20 m selection criteria could potentially bias the sample towards warmer sources . @xcite reports the far - ir observations of the 5muses sources with the _ herschel space observatory_. of 188 5muses sources covered by the observations , 154 ( 82% ) are detected at 250@xmath20 m . the authors also determine that the 250@xmath20 m flux densities of the whole population of galaxies in the observed field with @xmath43mjy ( the 5muses detection limit ) and that of the 5muses sample is drawn from the same distribution , indicating that the dust temperatures of 5muses are not significantly warmer than other ir luminous galaxies . in section [ sec : pah ] , we compare the 5muses sources to high redshift ulirgs , also selected at 24@xmath20 m , from the goods-_herschel _ survey . for the high redshift ulirgs , @xcite find no difference between 70@xmath20 m selected sources ( @xmath46 m rest frame ) and submillimeter selected sources in terms of pah and co luminosities . the high redshift ulirgs in this work are drawn from a larger parent population of _ spitzer _ irs sources @xcite . @xcite find that 97% of goods-_herschel _ sources with a 100@xmath20 m detection also have a 24@xmath20 m detection , ensuring that for high redshift lirgs and ulirgs , a 24@xmath20 m selection criteria does not bias the selection towards warmer sources . from the 5muses sample , we selected 24 sources that span a range of pah equivalent widths ( ew ; calculated for the full sample in * ? ? ? * ) , as pah ew is correlated with the mid - ir strength of an agn @xcite . all 24@xmath20 m sources have spectroscopic redshifts determined from fitting the mid - ir spectral features . these sources were chosen for a pilot co(1 - 0 ) study specifically because they exhibit a range of pah strengths and because the spectroscopic redshifts are such that we can observe the co(1 - 0 ) transition with the large millimeter telescope ( lmt ) . these sources were also selected according to @xmath1 , such that every source is observable in a reasonable ( @xmath47 min ) amount of time during the lmt early science phase . our sample spans a redshift range of @xmath48 and @xmath49 . details are listed in table [ properties ] . the lmt is a 50 m millimeter - wave radio telescope on volcn sierra negra , mexico , at an altitude of 4600 m @xcite . the high elevation allows for a median opacity of @xmath50 at 225ghz in the winter months . for the early science phase , the inner 32.5 m of the primary reflector is fully operational . the primary reflector has an active surface and a sensitivity of 7.0jy / k at 3 mm . the pointing accuracy ( rms ) is @xmath51 over the whole sky and 1 - 2@xmath52 for small offsets ( @xmath53 ) from known sources . during march and april 2014 , we observed our 24 sources with the redshift search receiver ( rsr ; * ? ? ? * ; * ? ? ? the rsr is comprised of a dual beam , dual polarization system that simultaneously covers a wide frequency range of 73 - 111ghz in a single tuning with a spectral resolution of 100km / s . this large bandwidth combined with a high spectral resolution is ideal for measuring the co integrated line luminosities in high redshift sources . in our intermediate redshift sample , the large bandwidth allows us to observe the co(1 - 0 ) transition for sources spanning a wide range of redshifts . the beam size of the rsr is frequency dependent , such that @xmath54 . the beam size depends on the frequency of the observed co(1 - 0 ) line , and at the median redshift , @xmath55 , the rsr has a beam size of @xmath56 . typical system temperatures ranged from @xmath57k during our observations . weather conditions varied over the eight nights of observations , with @xmath58 . all observations were taken at elevations between 40 - 70@xmath59 where the gain curve is relatively flat . on source integration times ranged from @xmath60 minutes and were determined by estimating @xmath2 from @xmath1 using the average ratio from @xcite . the integration times were estimated to obtain a @xmath61 detection of the expected integrated line luminosity in each source according to the lmt integration time calculator . we obtained @xmath61 detections of 17 out of 24 targets . with the sole exception of 5muses-313 , where the requested on source integration time was not completed , the desired rms was reached for all sources . the rms has been calculated using the full spectrum from 73 - 111ghz . we list the rms in table [ properties ] for each source . there is no correlation between rms and s / n , so low co(1 - 0 ) emission appears to be an intrinsic property of our undetected sources . data were reduced and calibrated using dreampy ( data reduction and analysis methods in python ) . dreampy is written by g. narayanan and is used specifically to reduce and analyze lmt / rsr data . it is a complete data reduction package with interactive graphics . for each observation scan , four distinct spectra are produced from the rsr . after applying appropriate instrumental calibrations from the dreampy pipeline , certain frequencies where known instrument artifacts are sometimes present were removed . there are no known bandpass features in the spectral regions where the co(1 - 0 ) lines are expected . for each observation , linear baselines were calculated outside the region of the co(1 - 0 ) line , and the rms estimated from the baseline of the full spectrum is used when all data for a given source are averaged together to produce the final spectrum . the averaging is a weighted average where the weights are set to @xmath62 . we checked whether aperture corrections were necessary using the empirical relation derived in @xcite and the optical radii from nasa / ipac extragalactic database . we calculated at most a 10% aperture correction for our lowest redshift galaxies which have the largest angular extent . given the uncertainty inherent in applying a correction formula , we have opted not to apply any aperture corrections . from the final spectrum for each source , in units of antenna temperature , we determined the locations of the co(1 - 0 ) peak by fitting a simple gaussian to the spectrum close to the frequency @xmath63 , allowing for a generous 10% uncertainty in @xmath64 , where @xmath64 was derived by fitting the mid - ir spectral features ( described in section [ sec : mir ] ) . in table [ properties ] , we list @xmath65 , the redshift determined from the peak of the co(1 - 0 ) emission . in all cases , the redshifts derived from fitting the co(1 - 0 ) line agree with the mid - ir redshifts within @xmath66 . the rsr is designed for blind redshift searches , and we test the reliability of our co(1 - 0 ) detections by performing blind line searches over the full 38ghz spectra . for 94% ( 16/17 ) of the objects that we claim detections for , a blind line search identifies the correct redshift . for 5muses-225 , known noise artifacts elsewhere in the spectrum confuse the redshift solution . however , since redshifts are known for our targets , we can nevertheless reliably measure the co(1 - 0 ) emission at the correct redshift . spectra ( with @xmath67 in mk ) and the best fit gaussians are shown in figure [ fig : co ] . we integrate under the gaussian to determine the line intensities , @xmath68 , in k km / s . we convert the line intensity to @xmath69 using a conversion factor of 7jy / k , calibrated specifically for the lmt early science results , and we convert to @xmath70 following the equation in @xcite after correcting for cosmology differences . the co(1 - 0 ) luminosities are listed in table [ properties ] . our analysis centers around comparing the molecular gas emission , traced by @xmath2 , with the dust emission , traced primarily through @xmath1 . @xmath71 m ) values are calculated in @xcite . the authors create synthetic irac photometry using the _ spitzer _ irs spectra and then fit a library of templates to the synthetic irac photometry and mips 24 , 70 , and 160@xmath72 m photometry . the final spectral energy distribution ( sed ) for each galaxy is created by combining the mid - ir spectroscopy with the best - fit template . finally , the authors calculate @xmath1 by integrating under the sed from @xmath73 m . twelve of the sources in our study also have photometry from spire on the _ herschel space observatory _ @xcite . we test whether including this longer wavelength data changes @xmath1 for these 12 sources , and find excellent agreement , such that there is less than a 3% difference between @xmath1 calculated with and without spire data . for the purposes of this study , we wish to use the standard definition : @xmath74 m ) . we correct the @xmath71 m ) values from @xcite to @xmath74 m ) by scaling by 0.948 , a conversion factor which was determined using composite lirg and ulirg seds from @xcite . in addition , @xcite uses a slightly different cosmology with @xmath75kms@xmath25mpc@xmath25 . over our range of redshifts , this requires an average conversion factor of 0.970 to account for the difference in cosmologies . to summarize , we calculate : @xmath76 these values are listed in table [ tbl : dust ] . when an agn is present , @xmath4 can have a non - neglible contribution from dust heated by an agn , and hence @xmath1 does not directly translate to an sfr . we diagnose the presence and strength of an agn through mid - ir ( @xmath77 m rest frame ) spectral decomposition . we follow the technique outlined in detail in @xcite which we summarize here . we fit the individual spectra with a model comprised of three components : ( 1 ) the star formation component is represented by the mid - ir spectrum of the prototypical starburst m 82 @xcite . we have verified our choice of starburst template by comparing with the starburst composite sed from @xcite and find no difference in the decomposition results ; ( 2 ) the agn component is determined by fitting a pure power - law with the slope and normalization as free parameters ; ( 3 ) an extinction curve from the @xcite dust models is applied to the agn component . the extinction curve is not monotonic in wavelength and contains silicate absorption features , the most notable for our wavelength range being at 9.7@xmath78 m . we tested applying additional extinction to the star formation component beyond that inherent in the m 82 template and found this to be negligible for all sources . we fit all three components simultaneously . for each source , we quantify the strength of the mid - ir agn , @xmath79 , as the fraction of the total mid - ir luminosity coming from the power - law continuum component . the pah ew from @xcite , initially used to select our sample , directly relates to the mid - ir agn fraction , supporting the reliability of our diagnostic . our decomposition technique is illustrated in figure [ fig : decomp ] . we use the mid - ir agn strength , @xmath79 , to determine the total contribution of the agn to @xmath4 . to estimate the conversion between @xmath79 and @xmath80 , we use 22 composite seds created from @xmath81 lirgs and ulirgs spanning a redshift range of @xmath82 ( kirkpatrick et al.2015 , in preparation ) . we decompose the mid - ir portion of the composite seds following the technique outlined above . we then decompose the full sed , from 0.5 - 1000@xmath20 m , by fitting a pure agn template and a star formation template . we find that the total ir agn contribution is @xmath83 of the mid - ir agn strength ( @xmath84 ) . for the 5muses sample , where we lack enough data points to accurately decompose the full - ir sed , we scale the mid - ir agn strength by 0.53 to estimate the total agn contribution , @xmath80 , to @xmath4 , and we use this value to calculate @xmath85 , the ir luminosity attributed to star formation . in the analysis that follows , we separate our sources according to @xmath79 , resulting in three categories : ( 1 ) purely star forming galaxies have @xmath86 ( @xmath87 ) ; ( 2 ) agn have @xmath88 ( @xmath89 ) ; and ( 3 ) composites have @xmath90 ( @xmath91 ) . we are also interested in the strength of the pah emission , since this is expected to trace star formation and photodissociation regions within molecular clouds . by comparing the pah emission with @xmath1 and co(1 - 0 ) emission , we have three separate tracers of the dust and gas in the ism which are expected to correlate with star formation . we quantify the pah emission as @xmath92 , the luminosity of the isolated 6.2@xmath20 m line . we fit a continuum on either side of this feature at 5.9@xmath20 m and 6.5@xmath20 m and remove the continuum component . we then integrate under the continuum subtracted emission feature to obtain @xmath92 . none of our sources are extended at 8.0@xmath20 m using the _ spitzer _ irac images , so aperture corrections are not necessary when comparing the _ spitzer _ and lmt luminosities . we list @xmath92 , @xmath85 , @xmath79 , and @xmath80 in table [ tbl : dust ] . lc cccc cccc 5muses-200 & 10.36 & 0.23 & 0.12 & comp & 10.31 & 8.04 & 10.48 & 0.09 & ms + 5muses-179 & 10.22 & 0.24 & 0.13 & comp & 10.17 & 7.95 & 11.08 & 0.01 & ms + 5muses-169 & 10.79 & 0.06 & 0.03 & sfg & 10.79 & 8.45 & 10.82 & 0.11 & ms + 5muses-105 & 10.88 & 0.16 & 0.08 & sfg & 10.85 & 8.38 & 10.79 & 0.13 & ms + 5muses-171 & 11.06 & 0.08 & 0.04 & sfg & 11.05 & 8.54 & 10.93 & 0.16 & ms + 5muses-229 & 11.10 & 0.12 & 0.06 & sfg & 11.08 & 8.62 & 11.36 & 0.06 & ms + 5muses-230 & 11.09 & 0.00 & 0.00 & sfg & 11.09 & 8.98 & 10.46 & 0.50 & sb + 5muses-234 & 11.03 & 0.15 & 0.08 & sfg & 11.00 & 8.52 & 10.47 & 0.40 & sb + 5muses-132 & 11.30 & 0.00 & 0.00 & sfg & 11.30 & 9.16 & 10.67 & 0.51 & sb + 5muses-227 & 11.08 & 0.57 & 0.30 & comp & 10.91 & 8.63 & 10.73 & 0.18 & ms + 5muses-141 & 11.14 & 0.67 & 0.36 & comp & 10.93 & 8.71 & 11.15 & 0.07 & ms + 5muses-158 & 11.41 & 0.23 & 0.12 & comp & 11.36 & 8.87 & 10.99 & 0.28 & sb + 5muses-225 & 11.09 & 0.29 & 0.15 & comp & 11.02 & 8.61 & 10.99 & 0.13 & ms + 5muses-273 & 11.40 & 0.45 & 0.24 & comp & 11.28 & 8.78 & 10.97 & 0.23 & ms + 5muses-136 & 11.39 & 0.77 & 0.41 & agn & 11.13 & 9.03 & 11.41 & 0.06 & ms + 5muses-294 & 11.55 & 0.16 & 0.08 & sfg & 11.52 & 8.97 & 11.49 & 0.12 & ms + 5muses-216 & 11.39 & 0.14 & 0.07 & sfg & 11.37 & 9.06 & 10.71 & 0.53 & sb + 5muses-194 & 11.72 & 0.81 & 0.43 & agn & 11.44 & 8.34 & 11.19 & 0.21 & ms + 5muses-249 & 11.63 & 0.12 & 0.06 & sfg & 11.61 & 9.26 & 10.96 & 0.52 & sb + 5muses-250 & 11.63 & 0.05 & 0.03 & sfg & 11.63 & 9.44 & 11.29 & 0.26 & ms + 5muses-275 & 12.00 & 0.79 & 0.42 & agn & 11.73 & 9.46 & 11.13 & 0.47 & sb + 5muses-313 & 11.81 & 0.41 & 0.22 & comp & 11.70 & 9.12 & 10.64 & 1.36 & sb + 5muses-156 & 12.06 & 0.31 & 0.16 & comp & 11.99 & 9.24 & 10.17 & 7.70 & sb + 5muses-101 & 11.91 & 0.82 & 0.43 & agn & 11.63 & 9.03 & 10.86 & 0.68 & sb . the specific star formation rate ( ssfr ) is the ratio of sfr to stellar mass . @xcite determined stellar masses for the 5muses sample by fitting @xcite population synthesis models to optical and near - ir broadband photometry assuming a chabrier imf . we adopt these stellar masses , and we calculate the sfr from @xmath85 according to @xmath93 assuming a kroupa imf and a constant star formation rate over the past 100myr @xcite . we do not convert between a chabrier imf and a kroupa imf for this formula , since the conversion is very small @xcite . the galaxy main sequence can be used to classify galaxies as either normal star forming galaxies or starbursts based on whether they have an enhanced sfr for a given @xmath94 . this relationship is a slowly varying function of redshift . @xcite present a relationship between ssfr and the time since the big bang in gyr , @xmath95 , for the galaxy main sequence ( see equation 13 in * ? ? ? the relation in @xcite is derived assuming a salpeter imf , an @xmath1-sfr conversion from @xcite , and a cosmology with @xmath75kms@xmath25mpc@xmath25 , @xmath96 , and @xmath97 . we convert the relation from a salpeter imf to a kroupa imf using @xmath98 @xcite . we convert @xmath95 to the cosmology used in this paper by multiplying by 1.02 , appropriate for our redshift range . finally , the sfrs are related by sfr@xmath99 @xcite . applying all of these conversions gives the ms relation appropriate for the present work : @xmath100 if a galaxy has a ssfr a factor of two greater than ssfr@xmath101 , it is classified as a starburst . we use @xmath65 and @xmath85 to calculate ssfr@xmath101 and list the stellar masses along with the main sequence / starburst designations in table [ tbl : dust ] . we plot ssfr vs. @xmath102 for our sample in figure [ fig : ms ] . we show the ssfr calculated using @xmath4 ( grey open symbols ) and the ssfr calculated using @xmath85 ( filled symbols ) . we color the sources according to whether they are star forming galaxies ( blue ) , composites ( green ) , or agn ( red ) , according to @xmath80 ( see section [ sec : mir ] ) . removing the agn contribution to @xmath4 , which we have done for all sources , has the effect of moving two of our sources from the starburst region onto the main sequence . two sources , 5muses-136 and 5muses-179 , lie below the main sequence , possibly indicating that they are transitioning to a more quiescent evolutionary stage . in this small sample , we observe no obvious association of agn strength with distance from the main sequence , similar to what is seen in @xcite . in the local universe , the sk law is traditionally expressed in terms of surface densities , which necessarily require resolved measurements of star formation and molecular gas @xcite . at higher redshifts , resolved measurements are often not feasible , and global measurements of the sfr and molecular gas are used instead . the ratio of @xmath1 , directly related to a global sfr , and @xmath70 is commonly referred to as a star formation efficiency @xcite . it is essentially the integrated version of the sk law without the uncertain conversion from co to h@xmath0 mass . this quantity is also related to the inverse of the gas depletion timescale which describes how long a galaxy could continue to form stars at the current rate if the gas reservoir is not replenished ( e.g. , * ? ? ? in general , @xmath1 correlates with @xmath70 , although there is significant scatter . @xmath1 is primarily measuring the reradiated light from newly formed stars ( with some contribution from an older stellar population ) , while @xmath70 is measuring the reservoir of molecular gas available to form stars in the future ; hence the dust and gas do not necessarily trace star formation on the same timescales . co emission can also be present between molecular clouds , introducing more scatter in the relationship between @xmath1 and @xmath2 . to account for this scatter , many authors propose two relationships , one for starbursts , undergoing an enhanced @xmath3 , and one for normal star forming galaxies ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? if a galaxy undergoes a burst of star formation that also triggers agn growth , this could account for some of the scatter in @xmath3 . initially , as an embedded agn grows more luminous , it heats some of the surrounding dust , but is enshrouded enough that the host galaxy is still visible ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? the increase in the amount of warm dust heated by the agn will enhance @xmath1 but may not yet affect @xmath70 , implying an artificially high @xmath3 unless the agn contribution is accounted for ( e.g. , by considering @xmath85 ) . as the agn becomes less enshrouded , the pah and cold dust emission from the host become less prominent because the dust heated by the agn is outshining the dust in star forming regions , and/or because feedback from the agn is quenching the star formation . if the agn is quenching the star formation , this will produce lower @xmath85 values and hence lower @xmath5 , unless feedback from the agn expels co on the same timescales . as yet , no study has attempted to quantify the effect of the agn on @xmath3 individually in galaxies due to the difficulty in measuring the amount of @xmath1 attributable to an agn . the 5muses sample has mid - ir spectra that exhibit both pah features and a strong underlying continuum , allowing us to cleanly separate out the agn luminosity and calculate @xmath85 . figure [ fig : lir_lco ] shows the relationship between @xmath85 and @xmath70 , and the points are colored according to @xmath80 . we overplot the relations between @xmath85 and @xmath70 determined for star forming galaxies ( sfgs , dashed line ; grey region indicates one standard deviation ) and starbursts ( dotted line ) in @xcite . these relations were robustly determined using large samples of normal star forming galaxies and starbursts from @xmath6 , and we are interested in testing how removing the agn component in our galaxies changes their position relative to these relations . the presence of two separate relations is discussed in depth in @xcite , where the authors also present a single relation fitting all available @xmath1 and @xmath103 measurements in the literature . we compare to the separate starburst and sfg relations to investigate any difference in @xmath5 according to ir power source or starburstiness . @xcite determined the relations for sfgs and starbursts using the far - ir luminosity ( fir : @xmath104 m ) as opposed to @xmath1 , so we scale the relationships according to @xmath105 , a ratio determined using the lirg and ulirg templates from @xcite . we overplot the @xmath106 scatter from @xcite as the grey shaded region . we plot the main sequence galaxies ( determined by equation [ eq : ms ] ) as filled circles and starbursts as filled stars . the starbursts on average have a factor of two higher @xmath85 @xmath70 than the main sequence galaxies . we find that both the main sequence and starbursts are consistent with the sfg relationship from @xcite , and there is no strong separation according to ir power source . we can look at this more simply by considering the ratio @xmath107 . in figure [ fig : sfe ] , we plot @xmath107 as a function of @xmath80 . we also include @xmath108 as the unfilled circles and stars . the grey shaded regions illustrate the standard deviation around the average @xmath3 calculated for starbursts and sfgs in @xcite . the standard deviation for the sfg region is 0.33 dex , slightly larger than the standard deviation illustrated in figure [ fig : lir_lco ] . this is due to the fact that the relation between @xmath1 and @xmath2 is non - linear , so the standard deviation relative to the mean @xmath3 is larger . none of our galaxies lie in the starburst region . @xcite find a mean @xmath109 of 44 . we calculate that the mean @xmath108 is 59 , but when we remove the agn contribution to @xmath4 , we calculate a mean @xmath107 of 52 , closer to the mean measured by @xcite . removing the agn component only mildly reduces the scatter about the mean @xmath107 . two galaxies , 5muses-179 and 5muses-275 , lie below the sfg region . the galaxy with the lowest @xmath5 , 5muses-179 , also lies decidedly below the main sequence region in figure [ fig : ms ] , further indicating that its star formation is highly inefficient , and this galaxy could be transitioning to a quiescent phase . we find no relationship between @xmath5 and @xmath80 . for this small sample , the star forming galaxies exhibit less scatter in @xmath5 than the composite galaxies , although both groups have the same average @xmath5 . the agn have a lower average @xmath5 indicating that these galaxies are not converting gas to stars at the same rate as the composite or sf galaxies ; this hints that the star formation might be beginning to quench in these agn sources . it is interesting to note that two galaxies with very high @xmath5 lower limits ( 5muses-249 and 5muses-294 ) show no trace of an agn according to their mid - ir spectra . this could be illustrative of the different timescales that agn signatures , starburst signatures , and enhanced @xmath5 ratios are visible . mid - ir spectroscopy and chandra x - ray observations provide evidence that the majority of local ulirgs and high redshift submillimeter galaxies ( smgs ) , which have a high merger ( and hence , starburst ) fraction and enhanced @xmath3 , are predominately powered by star formation in the mid - ir ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . a non - negligible fraction of lirgs and ulirgs host a mid - ir luminous agn , although optical morphologies of local lirgs reveal that mid - ir agn signatures are predominately found either in isolated disk galaxies or coalesced nuclei at the end of a merger @xcite . we expect , then , that if the agn and starburst phase do not overlap completely , agn should have lower @xmath5 ratios , and figure [ fig : sfe ] shows that they do , for this small sample , once the agn contribution to @xmath4 has been accounted for . also of interest is that we do not observe any dichotomy in @xmath5 either as a function of mid - ir power source or along the main sequence / starburst classification . @xcite observe a clear relationship between the gas depletion timescale and the ssfr in a large , complete , sample of local galaxies as part of the cold gass survey . we compare our sources with the cold gass galaxies in figure [ fig : tdep ] . @xcite calculate the depletion timescale as @xmath110 , but to avoid any uncertainties due to converting @xmath2 to @xmath111 ( see section [ sec : gas ] ) , we simply use the ratio sfr/@xmath112 . we have calculated the sfrs for our galaxies using @xmath85 . figure [ fig : tdep ] also includes the @xmath113 galaxies from @xcite , where we have corrected the sfrs , calculated using the conversion in @xcite , to be on the same scale as ours . the sfrs from @xcite are calculated by fitting the sed from the uv out to 70@xmath20 m . we do not further correct for differing cosmologies as the intrinsic scatter in figure [ fig : tdep ] is larger than any shift introduced in this manner . when looking at just our 5muses sample in figure [ fig : tdep ] , there is no strong correlation between sfr/@xmath2 and ssfr , just as we observed no separation according to ssfr in figure [ fig : sfe ] . however , when we extend the dynamical range of the plot by considering the local galaxies from @xcite and the @xmath113 galaxies from @xcite , there is a strong correlation ( a spearman s rank test gives a correlation coefficient of @xmath114 with a two - sided significance equal to 0.0 ) . our sample lies in the range expected . this suggests that we are not observing any differences between our starburst and main sequence galaxies in figures [ fig : lir_lco ] and [ fig : sfe ] simply because we are not probing a large enough range of @xmath2 and @xmath1 . ir , co , and pah luminosities are all commonly used as tracers of star formation in dusty galaxies . pah emission arises from pdrs surrounding young stars and has been demonstrated locally to be largely cospatial with the molecular clouds traced by co emission @xcite . if star formation is continuously fueled for @xmath115gyr , these tracers should all correlate . for most star forming galaxies , the ratio of @xmath116 is fairly constant , but there is an observed deficit of pah emission relative to @xmath1 in local ulirgs , possibly due to an increase in the hardness of the radiation field caused either by a major merger / starburst or an agn @xcite . this same deficit does not hold for similarly luminous galaxies at high redshift , however , where the majority of ulirgs are observed to have strong pah emission @xcite . @xcite explored the evolution of @xmath117 with redshift for a sample of ulirgs from @xmath118 as well as a sample of local ulirgs . specifically , the authors compare @xmath117 with @xmath1 and find that the deficit in @xmath119 relative to @xmath1 occurs at a higher @xmath1 for high redshift galaxies than is seen in the local universe . galaxies from @xmath9 typically have higher gas fractions than local counterparts @xcite . this increase in molecular gas could be linked to the relative increase in pah emission , since both are largely cospatial . indeed , when @xcite compare @xmath117 with @xmath109 , they find a consistent relationship for both the local and high redshift ulirgs . we now build on the analysis presented in @xcite by extending the parameter space explored to the lower luminosity 5muses sample . the 5muses galaxies combined with the local ulirgs comprise a low redshift sample for comparison with the high redshift ulirgs . we plot @xmath120 vs. @xmath85 for our sample in the left panel of figure [ fig : pah_co ] . we also include the high redshift and local ulirgs from @xcite where we have calculated @xmath85 for all ulirgs by scaling the mid - ir agn strength , determined by decomposing the mid - ir spectra . there is a decreasing trend between the 5muses sample and the local ulirgs , but the high redshift ulirgs are shifted from this relation . in the right panel , we plot @xmath120 vs. @xmath107 . @xmath2 is calculated using the estimated co(1 - 0 ) luminosity for all galaxies ( see * ? ? ? * for conversion details ) . in this panel , most galaxies follow the same decreasing trend , with a few obvious outliers . we overplot the best fit relation for the 5muses sample and the local and high redshift ulirgs . the shaded region indicates one standard deviation above and below the fit . there is a decreasing correlation between @xmath120 and @xmath107 for most galaxies . figure [ fig : pah_co ] suggests that the relative amount of emission from small dust grains is related to @xmath107 for dusty galaxies out to @xmath121 . that is , weaker pah emission is associated with a higher star formation efficiency and faster gas depletion timescales . the decrease in @xmath92 with increasing @xmath1 could indicate that pah emission in general is suppressed for more luminous galaxies . we do not find significantly lower @xmath120 ratios for our agn or composite galaxies as compared to our star forming galaxies , indicating that in our sample , the growing agn is not affecting this ratio . as discussed in @xcite , the pah deficit could be similar to the observed deficit in [ cii ] emission at high @xmath1 in local galaxies ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? @xcite probe the [ cii ] deficit in local lirgs and find that galaxies with compact mid - ir emission have a [ cii ] deficit , regardless of the mid - ir power source . for our sample , follow - up observations are required to trace the compactness of the galaxies . if our galaxies have extended dust emission , this would explain the similar @xmath7 ratios for the 5muses galaxies and the high redshift ulirgs , since high redshift ulirgs are known to have extended dust emission ( e.g. , * ? ? ? based on the relative strengths of the dust emission and co(1 - 0 ) emission , the 5muses sources , primarily lirgs , seem to be more accurate counterparts for the high redshift ulirgs than the local ulirgs , evidencing the evolution of ism properties with redshift . a morphological comparison of these sources could provide more insight into structure and compactness of the dust and gas emission . the gas fraction is expressed as @xmath122 , where @xmath123 . @xmath124 is a commonly adopted value for normal star forming galaxies , while in starbursts , the conversion @xmath125 has been measured ( * ? ? ? * and references therein ) . we have two observational indicators of starburstiness : @xmath5 and ssfr . to calculate @xmath126 , we explore two scenarios . first , we apply @xmath127 to our entire sample ( top panel of figure [ fig : gas ] ) . none of our galaxies have @xmath5 indicative of a starburst ( figure [ fig : sfe ] ) , so applying the same @xmath13 to the entire sample is a reasonable assumption . second , we use @xmath125 to calculate the gas mass for those galaxies with an ssfr indicative of a starburst ( bottom panel of figure [ fig : gas ] ) . we plot the gas fractions as a function of redshift we also plot gas fractions of normal star forming galaxies from the literature , and we have corrected the individual @xmath13 values used to be 4.6 . the average molecular gas fraction evolves with redshift , and we plot the measured relation , @xmath128 , determined from a stellar mass limited sample with @xmath129 @xcite , similar to the masses of our 5muses sample . our gas fractions ( in both panels ) lie in the range expected when comparing with the best fit line and the points from @xcite and @xcite . our co(1 - 0 ) detection rate is high ( 17 out of 24 sources ) , producing a large range of measured gas fractions . in the top panel , our starburst galaxies are systematically higher than the main sequence galaxies , and the scatter about the @xmath130 line is larger , which suggests that the lower @xmath125 conversion factor might be more appropriate for these sources if we expect similar gas fractions for main sequence and starburst galaxies . theoretically , the conversion factor depends on the geometry of the co and h@xmath0 distribution . when the co emission is extended and not confined to molecular clouds , warm , and has a high surface density , as is the case in mergers , then the lower @xmath13 value is appropriate ( * ? ? ? * and references therein ) . @xcite measure @xmath13 for a sample of high redshift main sequence and starburst galaxies and find an anti - correlation between @xmath13 and ssfr , which they interpret as evidence that the mechanisms responsible for raising a galaxy off the main sequence must also affect the physical conditions within the star forming regions . we also find that the gas fractions reproduce a similar separation between sources as the main sequence criterion , linking @xmath13 with the ssfr . in contrast , neither @xmath5 nor @xmath120 shows any separation between the starburst and main sequence galaxies , likely due to the limited range being probed . these ratios , then , are relatively stable for galaxies of a limited mass and luminosity range . @xcite argue against a bimodal @xmath13 conversion factor , and instead develop a fitting formula for the conversion factor that depends on the metallicity and co line intensity . we currently lack metallicities for our sample , so we can not directly apply the prescribed variable conversion factor . given the continuous relationship between ssfr and sfr/@xmath2 evidenced in figure [ fig : tdep ] , a continuous , rather than bimodal , conversion factor based on galactic environment may be the most appropriate choice and would mean that all galaxies in our sample obey a similar relationship between the molecular gas and the stellar mass . we present new lmt / rsr co(1 - 0 ) detections for 24 intermediate redshift galaxies from the 5muses sample . we use _ spitzer _ mid - ir spectra , available for all sources , to diagnose the presence and strength of an agn . we removed the agn contribution to @xmath4 and probe the star formation , gas , and dust emission using @xmath85 , @xmath2 , and @xmath92 . we find 1 . removing the agn contribution to @xmath4 results in a mean @xmath5 for our entire sample consistent with the mean @xmath3 derived for a large sample of star forming galaxies from @xmath6 . for our four agn sources , removing the agn contribution produces a mean @xmath5 lower than the mean @xmath5 for our star forming galaxies or composites . we find that @xmath5 is not strongly correlated with either the ssfr or the mid - ir power source over the range of luminosities probed . the average ratio of @xmath120 in our sample is similar to what is observed in high redshift ulirgs rather than local ulirgs . when we plot @xmath7 as a function of @xmath5 , we find all galaxies ( local ulirgs , our intermediate redshift 5muses sources , and high redshift ulirgs ) are consistent with the same declining relationship . 3 . our starbursts have gas fractions that are clearly offset from the main sequence galaxies if we apply a constant @xmath13 to all galaxies which might indicate that the two populations require different @xmath13 values . however , no dichotomy between starbursts and main sequence galaxies is evident when comparing other quantities that probe the ism ( @xmath85 , @xmath2 , or @xmath119 ) . we plan to improve this study in the future by expanding the sample size with more co(1 - 0 ) observations from the lmt , in order to further test how the gas and dust depends on @xmath80 . we also plan to measure the morphologies of this sample to look for merging signatures and evidence of extended dust / gas emission . we thank the anonymous referee for the helpful comments which have improved the clarity of this work . this work would not have been possible without the long - term financial support from the mexican science and technology funding agency , conacyt ( consejo nacional de ciencia y tecnologa ) during the construction and early operational phase of the large millimeter telescope alfonso serrano , as well as support from the the us national science foundation via the university radio observatory program , the instituto nacional de astrofsica , ptica y electrnica ( inaoe ) and the university of massachusetts ( umass ) . the umass lmt group acknowledges support from nsf uro and ati grants ( ast-0096854 and ast-0704966 ) for the lmt project and the construction of the rsr . ak would like to acknowledge support from a william bannick student travel grant . we are grateful to all the lmt observers from mexico and umass who took data for this project . this work is based in part on observations made with the _ spitzer space telescope _ , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa , and the _ herschel space observatory _ , a european space agency cornerstone mission with significant participation by nasa . this research has made use of the nasa/ ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . alexander , d. m. , bauer , f. e. , chapman , s. c. , et al . 2005 , , 632 , 736 alexander , d. m. & hickox , r. c. 2012 , newar , 56 , 93 armus , l. , charmandaris , v. , bernard - salas , j. , et al . 2007 , , 656 , 148 bendo , g. j. , wilson , c. d. , warren , b. e. , et al . 2010 , , 402 , 1409 bolatto , a. d , wolfire , m. , & leroy , a. k. 2013 , , 51 , 207 brandl , b. r , bernard - salas , j. , spoon , h. w. w. , et al . 2006 , , 653 , 1129 bruzual , g. & charlot , s. 2003 , , 344 , 1000 carilli , c. l. & walter , f. 2013 , , 51 , 105 chapman , s. c. , smail , i. , windhorst , r. , et al . 2004 , , 611 , 732 chung , a. , narayanan , g. , yun , m. , et al . 2009 , , 138 , 858 coppin , k. , pope , a. , menndez - delmestre , k. , et al . 2010 , , 713 , 503 daddi , e. , bournaud , f. , walter , f. , et al . 2010a , , 713 , 686 daddi , e. , elbaz , d. , walter , f. , et al . 2010b , , 714 , l118 desai . v. , armus , l. , spoon , h. w. w. , et al . 2007 , , 669 , 810 daz - santos , charmandaris , v. , t. , armus , l. , et al . 2010 , , 723 , 993 daz - santos , t. , armus , l. , charmandaris , v. , et al . 2013 , , 774 , 68 downes , d. , & solomon , p. m. 1998 , , 507 , 615 draine , b. t. 2003 , , 41 , 241 elbaz , d. , dickinson , m. , hwang , h. s. , et al . 2011 , , 533 , 119 erickson , n. , narayanan , g. , goeller , r. , grosslein , r. 2007 , aspc , 375 , 71 evans , a. s. , frayer , d. t. , surace , j. a. , sanders , d. b. 2001 , , 121 , 1893 evans , a. s. , solomon , p. m. , tacconi , l. j. , et al . 2006 , 132 , 2398 frster schreiber , n. m. , et al . 2003 , , 399 , 833 geach , j. e. , smail , i. , moran , s. m. , et al . 2011 , , 730 , l19 genzel , r. , tacconi , l. j. , graci - carpio , j. , et al . 2010 , , 407 , 2091 graci - carpio , j. , et al . 2011 , , 728 , l7 hopkins , p. f. , hernquist , l. , cox , t. j. , et al . 2006 , , 163 , 1 hughes , d. , juregui correa , j .- c . , schloerb , f. p. , et al . 2010 , spie , 7733 , 31 kaufman , m. j. , wolfire , m. g. , hollenbach , d. j. , et al . 1999 , , 527 , 795 kennicutt , r. c. , jr . 1998 , , 498 , 541 kennicutt , r. , c. , jr . & evans , n. 2012 , , 40 , 531 kirkpatrick , a. , pope , a. , alexander , d. m. , et al . 2012 , , 759 , 139 leroy , a. k. , walter , f. , bigiel , f. , et al . 2009 , , 137 , 4670 magdis , g. e. , rigopoulou , d. , helou , g. , et al . 2013 , , 558 , 136 murphy , e. j. , chary , r .- r . , dickinson , m. , et al . 2011a , , 732 , 126 murphy , e. j. , condon , j. j. , schinnerer , e. , et al . 2011b , , 737 , 67 magnelli , g. , saintonge , a. , lutz , d. , et al . 2012 , , 548 , 22 madau , p. & dickinson , m. 2014 , accepted . arxiv:1403.0007 menndez - delmestre , k. , blain , a. w. , smail , i. , et al . 2009 , 699 , 667 mullaney , j. r. , alexander , d. m. , goulding , a. d. , hickox , r. c. 2011 , , 414 , 1082 narayanan , d. , krumholz , m. r. , ostriker , e. c. , hernquist , l. 2012 , , 421 , 3127 noeske , k. g. , weiner , b. j. , faber , s. m. , et al . 2007 , , 660 , l43 petric , a. o. , armus , l. , howell , j. , et al . 2011 , , 730 , 28 pope , a. , chary , r .- r . , alexander , d. m. , et al . 2008 , , 675 , 1171 pope , a. , wagg , j. , frayer , d. , et al . 2013 , , 772 , 92 saintonge , a. , kauffmann , g. , kramer , c. , et al . 2011a , , 415 , 32 saintonge , a. , kauffmann , g. , wang , j. , et al . 2011b , , 415 , 61 sajina , a. , yan , l. , armus , l. , et al . 2007 , , 664 , 713 sanders , d. b. & mirabel , i. f. 1996 , , 34 , 749 sanders , d. b. , soifer , b. t. , elias , j. h. , et al . 1988 , , 325 , 74 sanders , d. b. , scoville , n. z. , soifer , b. t. 1991 , , 370 , 158 schmidt m. 1959 , , 129 , 243 shi , y. , helou , g. , yan , l. , et al . 2011 , , 733 , 87 siebenmorgen , r. , krgel , e. , & spoon , h. w. w. 2004 , , 414 , 123 solomon , p. m. , & vanden bout , p. a. 2005 , , 43 , 677 stacey , g. j. , hailey - dunsheath , s. , ferkinhoff , c. , et al . 2010 , , 724 , 957 tacconi , l. j. , genzel , r. , neri , r. , et al . 2010 , , 463 , 781 tacconi , l. j. , neri , r. , genzel , r. , et al . 2013 , , 768 , 74 tran , q. d. , lutz , d. , genzel , r. , et al . 2001 , , 552 , 527 veilleux , s. , et al . 2009 , , 701 , 587 weedman , d. , charmandaris , v. , & zezas , a. 2004 , , 600 , 106 wu , y. , helou , g. , armus , l. , et al . 2010 , , 723 , 895 xia , x. y. , gao , y. , hao , c .- n . , et al . 2012 , , 750 , 92 young , j. s. & scoville , n. z. 1991 , , 29 , 581 zahid , h. j. , dima , g. i. , kewley , l. j. , et al . 2012 , , 757 , 54

we present 1998 hst observations of m87 which yield the first single - epoch optical and radio - optical spectral index image of the jet at 0.15 resolution . we find , comparable to previous measurements , and ( ) , slightly flatter than previous workers . reasons for this discrepancy are discussed . these observations reveal a large variety of spectral slopes . bright knots exhibit significantly flatter spectra than interknot regions . the flattest spectra ( ; comparable to or flatter than ) are found in the two inner jet knots ( d - east and hst-1 ) which contain the fastest superluminal components . the flux maximum regions of other knots have . the maps of and appear poorly correlated . in knots a , b and c , and are essentially anti - correlated with one another . near the flux maxima of two inner jet knots ( hst-1 and f ) , changes in appear to lag changes in , but in two other knots ( d and e ) , the opposite relationship is observed . this is further evidence that the radio and optical emissions of the m87 jet come from substantially different physical regions . the delays observed in the inner jet are consistent with localized particle acceleration in the knots , with for optically emitting electrons in knots hst-1 and f , and for optically emitting electrons in knots d and e. synchrotron models fit to the radio - optical data yield hz for knots d , a and b , and somewhat lower values , hz , in other regions of the jet . if the x - ray emissions from knots a , b and d are co - spatial with the optical and radio emission , we can strongly rule out the `` continuous injection '' model , which overpredicts the x - ray emissions by large factors . because of the short lifetimes of x - ray synchrotron emitting particles , the x - ray emission likely traces sites of particle acceleration and fills volumes much smaller than the optical emission regions .

astro-ph0012044

the giant elliptical galaxy m87 hosts the best - known extragalactic synchrotron jet . as a result of its proximity ( distance = 16 mpc , tonry 1991 ) , particularly high resolution studies of its structure are possible ( @xmath11 pc ) . the synchrotron nature of its emissions were first demonstrated by baade ( 1956 ) . many workers have observed the m87 jet in the radio , optical and x - rays ( perola & tarenghi 1980 ; stocke et al . 1981 ; schreier et al . 1982 ; smith et al . 1983 ; biretta , owen & hardee 1983 ; killeen et al . 1984 ; keel et al . 1988 ; perez - fournon et al . 1988 ; owen , hardee & cornwell 1989 ; biretta , stern & harris 1991 , hereafter bsh91 ; meisenheimer , rser & schltelburg 1996 , hereafter mrs96 ; sparks , biretta & macchetto 1996 , hereafter sbm96 ; zhou 1998 ) . those data have revealed complex structure , with several bright radio - optical knots , as well as significant information on the broadband spectrum of the jet . the first hst observations of m87 ( boksenberg et al . 1992 ) revealed considerable detail in the optical jet . its large scale radio and optical structures are remarkably similar ; differences only appear at @xmath12 scales ( sbm96 ) . more recent hst observations have revealed apparent superluminal motion in several components at speeds up to 6@xmath13 ( biretta et al . 1999 ; biretta , sparks & macchetto 1999 , hereafter bsm99 ) , linking the properties of m87 with those of bl lacertae objects . hst observations have also served to illuminate the jet s axial structure for the first time , as the radio and optical polarized structures on @xmath14 scales show strong and consistent differences which indicate that higher - energy particles inhabit regions closer to the jet axis ( perlman et al . 1999 , hereafter p99 ) . the first ground - based multiband , contemporaneous optical photometry of the jet ( bsh91 ) revealed that its spectrum is considerably steeper in the optical than in the radio . more recent ground - based work ( mrs96 ) confirmed this picture , but disagreed with bsh91 on the spectral index of various jet components . those workers also presented the first detailed fits of synchrotron emission models to broadband , data produced both in their own observations and in those of previous workers ( references above ) and predicted synchrotron break frequencies @xmath15 hz . pre - costar hst data ( sbm96 ) revealed considerable narrowing of the jet from the radio to the optical , and flatter radio - optical spectra in bright knots . those workers also extracted optical spectra for large regions of the jet and were the first to fit synchrotron emission models to radio - optical data for the jet at hst resolutions . the model fits indicated stronger magnetic fields , and shorter particle lifetimes in bright knot regions . despite this large volume of work , several important questions remain open . for example , the presence of optical synchrotron radiation requires that electrons must be accelerated to at least @xmath16 . yet the mechanism by which particles are accelerated is poorly known , as are the dynamics of the synchrotron spectrum with time . moreover , the origin of the x - ray emission from the jet is still not completely constrained ( bsh91 , harris et al . 1997 , neumann et al . 1997 ) , and pre - costar hst data were not sufficient to provide a high signal to noise map of the jet s spectral index structure ( sbm96 ) . here we present 1998 hst observations of the m87 jet , spanning seven bands between @xmath17 microns . these data represent the highest resolution spectral data ever gathered on any extragalactic jet . to analyze the broadband continuum shape , we also include observations taken at 15 ghz with the vla in 1994 . these data allow us to resolve several important issues left unanswered by previous works . in 2 , we will discuss the observational setup and data reduction procedures . our results will be presented in 3 , including both radio - optical and optical spectral index maps of the jet . in 4 we compute synchrotron spectrum fits using our radio - optical data plus x - ray data from the literature . in 5 we compare our results with those of previous authors . in 6 we discuss the physical implications of our results . we imaged the m87 jet with the wfpc2 and nicmos aboard hst on 25 and 26 february 1998 . data were obtained through six filters , spanning the wavelength range 0.3 - 1.6 microns . a seventh near - ir band ( 2.05 microns ) was also scheduled for 26 february , but had to be reobserved on 4 april 1998 , due to a loss of guide - star lock . in table 1 , we list important details of these observations . to maximize the observable portion of the jet , the february observations were done with the hst oriented such that the jet fell along a chip diagonal . unfortunately an equally good orientation was not available on 4 april ; as a result , the f205w data do not include the region within @xmath18 of the nucleus . in the f110w and f160w bands , the small size of the nic1 detector made it necessary to observe at two positions . the majority of the observing time for those bands was used on the inner jet due to its lower surface brightness . to maximize resolution and minimize the effect of known instrumental problems such as grot ( sosey & bergeron 1999 ) , bad columns and warm pixels , observations in the f205w , f160w , f110w and f814w bands were dithered . we used two - position dithers in f205w , f160w and f110w , while for the f814w observation , we obtained four images in a three - position pattern with sub - pixel dithering . in the f205w band , two dithered 192s observations per orbit were done at a position @xmath19 away to measure the thermal background . we reduced each dataset using the best recommended flatfields , biases , darks and illumination correction images . the iraf task crrej was used to combine cr - split wfpc2 images and eliminate cosmic rays . for the f814w data , we used crrej first on the 2 images taken at the central dither position , then shifted the pc images taken at the 2 other positions to be consistent with one another using imlintran , and then combined those with crrej . the product images were then combined using drizzle ( fruchter & hook 1998 , gonzaga et al . 1998 ) onto a grid with pixels aligned 45@xmath20 to the original orientation of the image and smaller by a factor of @xmath21 . drizzle was also used for the other wfpc images in order to correct for the geometric distortion in the pc chip ( holtzmann et al . 1995 ) , using solutions by casertano et al . ( in prep . ) unequal pedestal effects in the nicmos data were eliminated with unpedestal ( van der marel 1998 ) . no overall pedestal correction was computed , since the signal from the galaxy overwhelms the pedestal . this resulted in only small differences ( @xmath22 adu / pix / s ) near the join of the two point positions used in the f110w and f160w images . in order to minimize color - dependent terms in the nicmos f110w and f160w flat - fields , we also used color - dependent `` superflats '' made from long observations of io ( storrs , bergeron & holfeltz 1999 ) . the color dependent terms are reduced significantly in the longer - wavelength bands ; moreover , the same concerns are also addressed by the thermal background images taken for the f205w data . the nicmos images were combined using scripts from the drizzle handbook ( gonzaga et al . 1998 ) and offsets computed from the _ jitter _ files using the script bearing by e. bergeron ( mentioned in gonzaga et al . ) . to correct for geometric distortion and the slightly rectangular pixels , we used solutions by cox et al . ( 1997 ) . thermal background images for the f205w band were combined in the same way and then subtracted . galaxy subtraction was done using the tasks ellipse , bmodel and imcalc . to successfully model the galaxy , we masked out the jet and sw hotspot , as well as all globular clusters and stars . this is of necessity an iterative process , as the fainter globular clusters do not become apparent until galaxy subtraction is done . failure to mask a globular cluster would produce a circular `` ringing '' at the distance of the cluster . in the wfpc2 images , we only fit a galaxy model to the data on the pc chip , in order to minimize the effect of the chip join , which was found to produce a circular `` ringing '' near the radius of knot f in fits done with the entire 4-chip wfpc2 mosaic . due to the faintness of the galaxy in the uv , we were only able to subtract the galaxy from the f300w image after smoothing it with a gaussian . in the nicmos mosaics , it was necessary to manually mask out regions where no data exists . the smallness of the nic1 chip only made it possible to model the galaxy in the f110w and f160w images to a distance of @xmath23 from the nucleus . this also necessitated an iterative approach to finding the correct ellipticity and pa of the model galaxy in the f110w and f160w bands ( _ n.b . , _ it was not possible to assume a surface brightness model derived from the wfpc2 images and scaled to the nuclear flux because of significant large - scale color gradients within the galaxy itself ) . subsequent testing revealed that in all bands we were able to achieve residuals no larger than a few adu / pix , with no large - scale trends which might indicate incorrect ellipticity or `` ringing '' in most regions , although there are significant dust known lanes in the inner galaxy ( sparks et al . the dust lanes affect only one region of the jet , knot hst-1 . rms errors from the galaxy subtraction are included in our error estimates in each case by adding in quadrature . flux calibrated images were obtained by multiplying the reduced and galaxy - subtracted image by using the task calcphot in synphot . we used calcphot to do this task rather than the information in the headers because it allowed us to assume a spectral index value ( we used @xmath24 , which is close to the mean spectral index of the jet ) and thus refine the flux calibration and the values in the headers were in all cases @xmath25 of the photflam value from synphot . this also allowed us to apply a decontamination correction ( baggett et al . 1997 , baggett & gonzaga 1998 ) , which affected most bands by @xmath26 , but had a larger effect on the f300w data ( 8% ) . to simulate the size of the aperture correction due to the extreme wings of the psf , we used the knot a+b+c region and a tinytim psf for the f205w+nic2 combination . those simulations led us to adopt an aperture correction of 1.107 for regions which extend @xmath27 ( on each side ) beyond regions of the jet with _ visible _ flux ( such simulations are recommended for estimating the aperture correction for extended sources ; see nicmos team 1999 ) . conversion from @xmath28 to @xmath29 was done at each band s pivot wavelength ( @xmath30 in table 1 ) . considerable effort was expended to estimate the uncertainties in these data . several possible sources of error were accounted for , including zeropoint errors in the synphot calibration ( colina et al . 1998 for nicmos ; baggett & gonzaga 1998 , baggett et al . 1997 and whitmore & heyer 1995 for wfpc2 ; typically this error is @xmath26 ) , possible errors due to assuming the wrong slope for the jet in synphot ( at most 1 - 2% , even if standard synphot coefficients instead of ones calculated for @xmath24 are used ) , flatfielding errors ( @xmath31 for wfpc2 , whitmore & heyer 1997 , but @xmath32 for nicmos , calzetti & gilmore 1998 ) and poisson errors both in the data and in the subtraction of the galaxy ( see above ) . these errors have been propagated by adding in quadrature . once these steps were done , all images were rotated so that the jet was along the x axis ( the centerline of the jet was assumed to lie at pa 290.5@xmath20 ) , using the positional and orientation information in the headers . images were then convolved with gaussians so that their resolution ( fwhm ) was equal to that of the f205w image ( tinytim psfs were not used because of the large number of bands and difficulties associated with deconvolution tasks ) . we then resampled all images to the nominal nic2 pixel size of @xmath33/pix and placed them on identical coordinate grids , with the nucleus at ( 20,200 ) on 400 @xmath34 400 pixel images . testing has revealed that the alignment of these images is accurate to within 0.15 pixels . the resolution of the spectral index maps we show in 3 and 4 is therefore @xmath35 fwhm , translating to 12 pc at the distance of m87 . to analyze the jet s broadband spectrum , we include previously published radio data ( zhou 1998 , p99 ) , obtained in february 1994 at 15 ghz . the data reduction procedures for this dataset have been described in zhou ( 1998 ) and p99 . to facilitate direct comparison between the radio and optical images , the radio image was resampled and shifted to the same frame of reference and pixel scale as the hst images ( see above ) . while the radio data were obtained in february 1994 , the possible effect of variability is small . even for the largest radio variability found between 1993 and 1997 ( @xmath36 for hst-1 ; zhou 1998 ) , the effect on @xmath4 is only @xmath37 . the resolution of the radio and f205w images is very similar ( @xmath38 fwhm compared to @xmath39 fwhm ) so we did not have to convolve the radio image with a gaussian before comparisons were done , and we were able to use the same aperture correction . in figure 1 , we show the wfpc2 300w and f814w images , and the f160w nicmos image , all at full resolution . the structures seen in the three bands are similar , indicating that spectral curvature exists in various jet components tend to be gradual . we examined the profiles on each hst image of slices through resolved areas of the jet . these comparisons were done on convolved data only , to exclude resolution effects . we confirm the result of sbm96 that the jet is narrower in the optical than the radio . however , based on our data there is no evidence that significant narrowing takes place in the optical . we generated optical and radio - optical spectral index maps ( @xmath6 and @xmath4 ; @xmath40 ) , using simple weighted least - squares fitting algorithms in idl . in figure 2a , we show the optical spectral index map . both the @xmath6 and @xmath4 images were clipped in regions where the signal to noise in either the f814w or radio image was less than 4 . in figure 2b , we show the radio - optical spectral index ( @xmath4 ) map . the same figure also shows maps of synchrotron model fit parameters in panels 2c and 2d ( see 4 for description of those and discussion ) . the convolved f814w image is shown for comparison in figure 2e . the lack of `` ringing '' near compact features in the @xmath6 map indicates that the differing resolutions in the hst images were correctly accounted for , and also that the registration is quite good . there is a large steep - spectrum region at the southern end of the a+b+c complex , but it is clearly real since ( 1 ) there is no corresponding feature near the southern end of the inner jet , and ( 2 ) there is more flux and extent to the south of the bright regions in a+b+c than to the north ( see the images in sbm96 and p99 as well as figure 1 ) . in figure 3 , we show runs of @xmath41 , @xmath42 and flux with distance along the axis of the jet in the inner jet ( @xmath43 from the nucleus ) . figure 4 shows the same plots for the outer jet ( @xmath44 from the nucleus ) . the spectral indices given in those figures are flux - weighted on a row - by - row basis , with flux summed in a @xmath45 wide window for figure 3 and a @xmath46 wide window for figure 4 . we also extracted fluxes for each knot ( and some sub - knot regions ) in the jet . these points are listed in table 2 , and plotted along with the best - fit power laws in figure 5 . as can be seen , the single power law model is a good fit for most regions , although a few show signs of gradual curvature in the optical . table 2 also defines the regions used for each knot . we also applied an aperture correction of 1.107 ( 2.2 ) to both the radio and optical data . in the optical , all knots in the jet are well fit by single power laws . evidence for spectral curvature in the optical can , however , be seen in d - e , d - m , a shock , b-1 , b-2 and c ( fig . 5 ) . in all of these cases the curvature appears to be very gradual and in the sense of spectral steepening towards the blue , as expected in all standard models of synchrotron emission ( see 4 ) . the jet average is @xmath1 , slightly flatter than reported in mrs96 ( @xmath47 ) , but significantly flatter than reported in bsh91 and sbm96 ( @xmath48 ) . we will examine possible reasons for this in 5 . we derive @xmath0 , very similar to values observed by previous workers ( bsh91 , mrs96 , sbm96 ) . bright regions of the knots have somewhat flatter optical spectra than other regions of the jet , with @xmath49 near flux maxima , and substantial steepening to @xmath50 immediately downstream . two regions are particularly flat ( @xmath51 ) : the flux maxima of knots hst-1 and d - east . regions near the flux maxima in e , i and a have only slightly steeper spectra ( @xmath52 ) . by comparison , the inter - knot regions show significantly steeper optical spectra ( @xmath53 ) , although our signal to noise in those regions is somewhat low and our errors are correspondingly larger . much smaller variations are observed in @xmath4 than @xmath6 . slightly steeper values ( @xmath54 ) are seen in most regions of the inner jet ( interior to knot a ) , with slightly flatter values ( @xmath55 ) observed near the flux maxima of knots hst-1 , d - east , e and f. by contrast , in the outer jet , a very different pattern is observed : @xmath4 is remarkably stable , varying by at most @xmath56 0.03 , with no apparent correlation between jet flux and @xmath4 . in most regions of the jet , the optical spectrum is somewhat steeper than the radio - optical continuum . this can be seen on figures 3c and 4c , where runs of @xmath57 are plotted . values of @xmath57 average @xmath58 in bright regions in the inner jet , and @xmath59 in the outer jet . the flux maximum regions of hst-1 , d - east , e , i and a , show no evidence of radio - optical curvature ( in fact at their flux maxima hst-1 and d have @xmath6 somewhat flatter than @xmath4 , but at only the @xmath60 level ) . significantly larger values of @xmath61 are observed downstream of the flux maxima in each knot , indicating increased curvature . interestingly , we observe a nearly constant value of @xmath42 in knot regions outside of flux maxima , with no trend for increasing curvature as the distance from a local maximum increases . perhaps the most glaring feature seen in figures 2 and 4 is the oscillation of spectral indices , and the overall anti - correlation of @xmath6 and @xmath4 , observed in the outer jet . as can be seen , the @xmath6 and @xmath62 maps track each other quite well but appear almost anticorrelated with @xmath4 , in the sense that the flattest @xmath6 points also represent the brightest points , but those same points have the steepest values of @xmath4 . the above discussion indicates that the spectral morphology of the m87 jet is highly complex . it is worthwhile to discuss the jet s spectral morphology and the relationship of the optical and radio spectral indices in greater detail , isolating each jet region in turn . the implications of various features of the jet s spectral morphology will be discussed in 6 . we turn first to the inner jet , which has a knotty morphology , dominated by five discrete regions . proceeding outwards , the first knot is hst-1 , @xmath18 from the nucleus . at @xmath39 resolution , hst-1 consists of a bright peak with fainter regions downstream . at higher resolution it breaks up into several components , which appear to move at speeds up to @xmath63 ( biretta et al . 1999 , bsm99 ) . near the optical flux peak are local minima in both @xmath6 and @xmath4 , but the beginning of the flat @xmath6 region appears to be @xmath39 upstream from the beginning of the flat @xmath4 region . given the proximity to the central nuclear source and the location of a nearby dust lane , this result should be confirmed in future studies . the brightest knot in the inner jet is knot d , located @xmath64 from the nucleus . knot d can be separated into three regions : the flux maximum , called d - east , the southern ` twist ' , called d - middle , and the western edge , called d - west . within this knot are superluminal components moving at speeds ranging from @xmath65 seen in both the optical ( biretta et al . 1999a , bsm99 ) and radio ( biretta et al . 1995 ) . the @xmath6 map tracks closely the distribution of optical flux all throughout knot d , with the flattest spectral indices observed in a compact ( @xmath66 ) region at the flux maximum . less prominent local minima , are seen near the flux maxima in d - middle and d - west . the distribution of @xmath4 is somewhat different : the minimum in @xmath4 occurs @xmath39 upstream of the extrema in @xmath6 and @xmath62 ; the increase in @xmath6 which follows the flux maximum also appears to lag the increase in @xmath4 by this same amount . these two features show up most dramatically in the run of @xmath57 ( fig . the @xmath4 image does not show significant local extrema near the locations of either d - middle or d - west , but some slight flattening trend is seen over this region . regions closer to the jet axis appear to have flatter values of @xmath6 than the northern or southern edges , but this is not at all clear for @xmath4 . at about @xmath46 from the nucleus we observe knot e , which is barely resolved along the jet axis at @xmath39 resolution ( at higher resolution two condensations are seen ; sbm96 ) . speeds of @xmath67 have been detected in our hst monitoring program ( bsm99 ) . knot e does not exhibit the sharp spectral changes seen in knot d , but nevertheless a flatter-@xmath6 region is seen centered roughly on the flux maximum . there is also a local flattening in @xmath4 which appears more extended along the jet axis than the flat-@xmath6 region ; however @xmath4 reaches minimum @xmath68 upstream of @xmath6 . unlike in knot d , however , the downstream ends of the flat @xmath6 and @xmath4 regions appear spatially coincident . knot f , located @xmath69 from the nucleus , appears rather amorphous at @xmath39 resolution but breaks up into two diffuse blobs at higher resolution . radio monitoring has detected speeds near @xmath13 in this region ( biretta et al . 1995 ) , but our optical monitoring has not yet detected proper motion ( bsm99 ) . knot f s spectral morphology appears rather subtle ( figures 2a - b , 3 ) . both the @xmath6 and @xmath4 images show flatter spectral indices closer to the jet axis than along the northern or southern axes . neither spectral index map shows strong local extrema , but instead both show a flat spectrum region roughly coincident with the rather broad flux maximum region . at its upstream end , @xmath4 appears to lag @xmath6 by @xmath70 , but the situation is reversed at the downstream end , where the flat @xmath6 region terminates @xmath14 before the termination of the flat @xmath4 region . knot i , located @xmath71 from the nucleus , appears to have motions near @xmath13 from hst monitoring ( bsm99 ) . examination of figures 2a and 2e reveals a fairly compact flat-@xmath6 region centered on the flux maximum . however , the changes in @xmath4 are not well localized to the flux maximum region . as has been noted by many authors , the morphology of the outer jet is quite different from that of the inner jet . the outer jet features four knot complexes , traditionally called a ( @xmath72 from the nucleus ) , b ( @xmath73 from the nucleus ; _ n.b . _ neither the downstream end of knot a nor the upstream end of knot b has a sharp `` edge '' ) , c ( @xmath74 from the nucleus ) and g ( @xmath75 from the nucleus ) . looking at the outer jet in figures 2a - b and 4 , one is immediately struck by the strong anti - correlation between @xmath6 and @xmath4 . this anti - correlation begins in the inter - knot region between i and a , @xmath76 from the nucleus , where we observe a flattening in @xmath4 with a possibly double structure , corresponding to a region with steep @xmath6 . further downstream , the flux maxima in a , b and c all correspond to regions with flat @xmath6 but steep @xmath4 values , as do the ` filamentary ' structures in this region ( although the departures are smaller than for the flux maxima ) . also agreeing with this trend are the steeper than average values of @xmath6 and flatter than average values of @xmath4 seen in the b - c interknot region . a close look at figures 2a - b and 4 , in fact , reveals , that local minima in @xmath6 coincide very closely with local maxima in @xmath4 and vice versa ; moreover the locations of increases in @xmath6 also coincide very closely with decreases in @xmath4 and vice versa . we fitted synchrotron spectrum models to the integrated radio through x - ray fluxes for knots a , b and d as well as the larger a+b+c complex , and on a pixel to pixel basis to radio through optical data for the entire jet , using programs written by c. carilli and j. p. leahy ( carilli et al . 1991 , leahy 1991 ) . in these fits we include x - ray data in our analysis for comparison with the total fluxes of knots a , b and d , where x - ray emissions have been detected ( schreier et al . 1982 , bsh91 ) . for knot a , we use x - ray fluxes from a january 1998 rosat hri observation ( harris et al . 1998 ) . for knots b and d , fluxes have not yet been extracted from rosat data , so we used 1 kev fluxes from _ einstein _ ( bsh91 ) . the conversion of x - ray counts to flux for all regions was done assuming @xmath77 and log @xmath78 ( bsh91 ) . even though the x - ray data for b and d are not contemporaneous , this will have only a small effect on @xmath79 . a 50% change in flux ( 3@xmath34 as large as the variations in a documented by harris et al . 1997 , 1998 ) would only produce a change of @xmath80 . the models we fit were : \(1 ) a jaffe & perola ( 1973 ) model ( hereafter jp ) , which allows for continuous isotropization of the pitch - angle distribution of the electron population with time , but no further particle injection . \(2 ) a kardashev - pacholczyk model ( kardashev 1962 , pacholczyk 1970 ; hereafter kp ) , which does not allow pitch - angle scattering , and therefore allows the development of a high - energy `` tail '' of particles at oblique pitch angles . \(3 ) a `` continuous injection '' ( hereafter ci ) model ( heavens & meisenheimer 1987 , meisenheimer et al . 1989 ) under which a power - law distribution of relativistic electrons is continuously injected . this model was originally developed for the case of an unresolved source . meisenheimer et al . ( 1989 ) and heavens & meisenheimer ( 1987 ) have shown that this model applies when there exists within the telescope beam an unresolved zone of continuous high - energy particle injection ( _ e.g. _ , a shock front ) . each model can be characterized by an injection index @xmath81 and a break frequency @xmath82 . the differences between the jp , kp and ci spectrum show up most strongly blueward of the break . the jp spectrum has the fastest cutoff , the ci spectrum has the least steep break , and the kp model s cutoff is intermediate between the two . importantly , the overall trends fit by each model are identical ( _ e.g. , _ if a region has high @xmath82 , this will be true under all three models ) . therefore , the pixel - by - pixel fits were done assuming only the kp model . in fitting integrated fluxes , we tried only @xmath83 , as well as @xmath84 ( with the mean derived from the @xmath81 map ) . we did not attempt a @xmath85 minimization for these fits because all of those models have relatively fixed general characteristics ; moreover , as we will emphasize , because of the large gaps which exist in the mid to far ir as well as uv , there are significant limits to what this procedure can do . the results of these fits are shown in figures 6a - e , and the derived @xmath82 , @xmath81 and @xmath85 values are summarized in table 3 . inspection of figure 6 reveals a striking result : independent of whether the x - ray emissions are synchrotron in nature , the ci model is decisively ruled out by our data because it overpredicts by large factors ( 5 - 50 ) the x - ray fluxes . the ci spectrum could be made to work for knot a if the x - ray emitting electrons occupy a region about 1/5 the volume of knot a in the optical , but greater reductions are necessary for knots d ( factor 8) and b ( factor 50 ) . this topic will be discussed further in 6 . by comparison , both the jp and kp models fit the data reasonably well ; our data can not discriminate between these models because there are no data in the break region ( the euv ) . an interesting point in favor of a synchrotron nature for the x - ray emissions is that in knot d , jp and kp models fit to only the radio through optical data , predict the x - ray emissions to within a factor @xmath86 . pixel by pixel fits were done only to the radio through optical data . in figures 2c and 2d , we show ( respectively ) maps of @xmath82 and @xmath81 for the kp model only . note that good quality fits were only achieved in knot regions and so we have clipped these images so that they do not include interknot regions . for figure 2c , the color scale runs from @xmath87 to @xmath88 hz , while the color scale for figure 2d is identical to that used for figure 2b . we derive break frequencies @xmath89 hz in most inner jet regions . note that values @xmath89 hz should be treated as guidelines and/or lower limits . due to the gradual curvature of synchrotron spectra ( _ e.g. _ , figure 6 ) , a lack of significant radio - optical curvature makes fitting values of @xmath81 and @xmath82 difficult , and in such cases the code used outputs @xmath82 values 10 times higher than the highest frequency for which data are present ( here @xmath90 hz ) . somewhat lower break frequencies ( @xmath91 hz ) were found in knots a , b and c , with higher break frequencies in flat - spectrum regions near knot maxima . note that we do not quote break frequencies in inter - knot regions because the fits in those regions were poor due to their low signal to noise . these values agree well with those we find for the integrated fluxes of knots d , a and b , but with the addition of x - ray data . the @xmath82 ( figure 2c ) and @xmath6 ( figure 2a ) maps are extremely well correlated , with high break frequencies in the flattest spectrum regions , and low break frequencies in the fainter , steep spectrum regions . also , the @xmath81 map ( figure 2d ) largely correlates with the @xmath4 map ( figure 2b ) . both of these are to be expected given the rather large gap ( 4 decades in frequency ) between the radio and optical points . note that also from a physical point of view one expects @xmath4 and @xmath81 to be correlated unless there is significant low - frequency radio structure which is self - absorbed by 15 ghz . we do not have the data to test for this possibility but it is somewhat beyond the scope of this paper . we report average values of @xmath4 which are comparable to those reported by other authors . the average value of @xmath6 we find ( 0.9 ) is , however , flatter than values reported by previous authors , which range from @xmath92 ( mrs96 ) to @xmath93 ( bsh91 and sbm96 ) . it is worthwhile to look at the likely reasons behind these discrepancies . since the regions used to obtain fluxes ( as well as the calibration process ) varied in each case , the comparison is somewhat difficult . we therefore will speak in rather broad terms about the global picture , and concentrate on the a+b+c region , for which all previous workers except for sbm96 have compiled fluxes , summarized in mrs96 . as already mentioned , figures 6d and 6e contain our fluxes for the knot a+b+c region as well as the synchrotron spectrum fits for this region . we have overplotted on figures 6d - e the other fluxes quoted by mrs96 ; particularly relevant to this discussion is figure 6e which zooms in on the optical . as can be seen , our flux values are well in accord with all but two teams fluxes . the discrepant values are from bsh91 and killeen et al . we can not throw any additional light on why those values are discrepant ; we refer the reader to the discussion in mrs96 on this point . moreover , the overall value of spectral index one derives from our values is easily seen to be identical within the errors to what one would derive from all previous points except the iue values ( perola & tarenghi 1980 ) , which may be discrepant because of poor galaxy subtraction . those values bear some mention as they account for most of the curvature in the models fit by mrs96 to the data for a+b+c , and also have by far the largest error bars . also , those points , plus those of bsh91 , are the only ones consistent with a steep value for @xmath6 in the a+b+c region . as can be seen by comparing our figures 3a and 4a with figure 3 of mrs96 , our values and those of mrs96 for @xmath6 are completely consistent if a constant offset of @xmath94 is applied , and fine structure on scales smaller than an arcsecond is ignored . it is likely that the main reason for the constant offset is contamination from the galaxy , which would tend to redden the spectrum , since the mean color of starlight from m87 is significantly redder than the jet ( @xmath95 color of the galaxy s light averages 0.78 mag , boroson & thompson 1991 ; as compared to @xmath96 0.16 mag for knot a of the jet , bsh91 ) . this is perhaps not entirely unexpected given the much higher resolution of hst compared to ground - based telescopes . it is hard to imagine other significant color - based or systematic effects given that the overall flux scales agree quite well ( figure 6e ) . a second factor in this respect , however , could be the larger number of bands in our data set ( 7 compared to 3 ) which allows us to better fit spectral indices as well as see curvature . we fit models which are somewhat different than those fit by mrs96 . in particular we do not make assumptions about the nature of a spectral cutoff at high frequencies . therefore it is somewhat confusing to compare the values of _ break _ frequency we derive to the values of _ cutoff _ frequency they derive . it is also important to note that the comment made in mrs96 that there is evidence for a steeper cutoff than predicted by fermi acceleration was based only on the iue points and so this needs to be verified ( we have since obtained shorter wavelength data with stis ; those data will be the subject of a later paper ) . given all of this plus the somewhat flatter values of @xmath6 we find , it is therefore not surprising that our values for @xmath82 are higher than their @xmath97 values . in comparing the two tracks however ( cf . figure 8b in mrs96 and our figure 2c ) it is important to note that , particularly in the outer jet , the overall trends are exactly the same except for the resolution difference . the values reported by sbm96 were based on pre - costar hst data . a significant number of those datasets are low s / n and the spectral index fits in sbm96 were not weighted by signal to noise . moreover , the regions used by sbm96 contain large non - jet portions because they were rectangular regions defined on an image where the jet was inclined by roughly 21@xmath20 to the x axis . the data used here are considerably higher signal to noise in every band , and the jet is on the x axis ; in addition , we do not have to deal with the residuals caused by deconvolving the pre - costar psf in each band . we have refit spectral indices to the quoted flux values for various jet regions , using only the f140w , f220w and f372 m datasets ( the only ones with reasonably good s / n over large parts of the jet ) . when this is done we obtain optical spectral indices considerably flatter than those obtained by sbm96 ( probably due to our exclusion of the lower signal to noise data , which also would have had poorer galaxy subtraction ) , and indeed in bright regions such as knot a we obtain results comparable to within the errors . we have also resampled those images to @xmath33/pix , smoothed them to @xmath39 resolution , and fit a spectral index image using the same program described in 3 . the result is extremely noisy ( with error bars much larger than in our data ) , but spectral indices considerably flatter than 1 are found in bright regions ; in fact near the flux maxima of a and d - east spectral indices @xmath98 are found , with internal error bars @xmath99 . this tends to support the reliability of our results . finally , there is also a large discrepancy between our spectral index values and those quoted by bsh91 . as we have already commented ( and as previously noted by mrs96 ) , however , the flux scale used in bsh91 appears to be rather at variance with that of other workers . we do not have significant new light to throw on the flux scale issue but given the overall good agreement of our flux scale with that of previous workers we are suspicious of the spectral index values quoted by bsh91 . we have presented data which give the first high - resolution picture of the near - infrared to ultraviolet spectrum of the m87 jet , and indeed represent the first dataset of this type for any extragalactic jet . we have given a detailed discussion of the spectral morphology of the jet and the @xmath42 relationship in 3.3 and 3.4 , and discussed synchrotron spectrum model fits in 4 . these allow us to place significant constraints on the jet structure , the nature of the x - ray emission , and also the extent , need and nature of particle acceleration in the m87 jet . we discuss each of these topics in turn . our data are fully consistent with a synchrotron origin for the x - ray emission in the jet ( as proposed by bsh91 ) . the observation of variability in the x - ray emission from knot a and the nucleus ( harris et al . 1997 , 1998 ) , rules out a thermal bremsstrahlung nature for the x - ray emission . but neither our data nor the observed variability are completely sufficient to rule out an inverse - compton nature for the x - ray emission . if one assumes that the x - ray , optical and radio synchrotron emitting electrons in knots a , b and d are co - spatial , the ci model of meisenheimer et al . ( 1989 ) is ruled out because it overpredicts by large factors the observed x - ray emission . thus particle injection or acceleration can occur only in discrete sites occupying a small fraction of the jet volume . this agrees with the compactness required by the detection of x - ray variability in knot a on timescales of months ( harris et al . 1997 , 1998 ) . the ci model can be made to work for knot a if the x - ray emitting electrons in that knot occupy a region about 1/5 the projected area of knot a in the optical . similar arguments can be made for knots d ( factor 8) and b ( factor 50 ) , although their x - ray detections are less secure . measurement of the x - ray spectral index @xmath100 would allow us to discriminate between the synchrotron and inverse - compton models . inverse - compton x - ray emission would have a much flatter x - ray slope ( @xmath101 , comparable to the jet s @xmath4 and similar to the average hard x - ray slopes of low - energy peaked bl lacs ( lbl ) found by asca , see kubo et al . 1997 ) , than both the jp and kp synchrotron models , which predict @xmath102 and @xmath103 , as seen in the x - ray spectra of most high - energy peaked bl lacs ( hbl ; perlman et al . 1996 , sambruna et al . 1996 ) . unfortunately , because of the multi - component nature of the x - ray emission from the m87 region ( including cluster , galaxy and jet components ) , current x - ray data do not allow us to place good limits on the shape of the x - ray slope of m87 s jet . this could not be done from the _ einstein _ hri or rosat hri datasets due to their poor ( at most two - channel ) energy resolution . higher energy telescopes ( asca , sax and rxte ) have tried to determine the relative jet+nuclear x - ray emissions but with at best variable success due to their much poorer ( @xmath104 ) angular resolution . matsumoto et al . ( 1996 ) and allen et al . ( 2000 ) both find some evidence of a hard x - ray tail in asca data but with large error bars on the spectral shape ( @xmath105 between 2 - 10 kev ) . similar constraints ( @xmath106 ) , have recently been obtained from sax data ( guainazzi & molendi 1999 ) . however , a negative detection was obtained by rxte ( reynolds et al . 1999 ) , a non - imaging detector , which places a lower limit of @xmath107 on the spectral index of knot a , and @xmath108 on the core . comparison of the fluxes obtained in the two detections plus the upper limits , however , does reveal evidence of variability ( guainazzi & molendi 1999 , reynolds et al . thus current x - ray data can not constrain the spectral shape of any component in the m87 jet , but they do constrain the nature of the emission to nonthermal ( synchrotron or inverse - compton ) mechanisms because of the observed variability . as shown in figures 2 - 4 and 3 , the relationship between @xmath6 and @xmath4 in the m87 jet is rather complex . in the inner jet , the two appear correlated in most regions if a lag is applied to one or the other , while they are essentially anti - correlated in the outer jet . these trends are further evidence that it is impossible to model the m87 jet as an axially homogeneous jet where radio and optical synchrotron emitting electrons are completely co - spatial . instead , they support a model for the structure of the jet presented by p99 , whereby the energy spectrum of electrons in the jet , and magnetic field direction , vary significantly with distance from the jet axis . in the p99 model , higher - energy electrons are concentrated closer to the jet axis , where the magnetic field is most strongly affected by shocks and other disturbances , while lower - energy electrons occupy mainly the jet `` sheath '' at greater distances from the jet axis . at any one point , the observed spectral indices include both core and sheath contributions , with the core dominating in the optical and the sheath dominating in the radio . this model predicts that @xmath6 and @xmath4 should not vary together . it also predicts that some regions might have flatter _ observed _ values of @xmath6 than @xmath4 , since different regions can predominate in different bands . this is in fact observed at the flux maxima of hst-1 and d. the poor correlation between @xmath6 and @xmath4 in the outer jet is somewhat more surprising given the large degree of similarity between the radio and optical polarized structure in this region ( p99 ) . however , in the absence of strong shocks it is still entirely possible for energy stratification to occur without observing significant differences in polarimetry . indeed , we should point out that the mixing of previously stratified layers would not be expected to occur in a short , linear shock , and would likely produce turbulence which would be visible in the form of eddies ( not observed ) . moreover , the optical polarized structure of the outer jet , while much more similar to that seen in the radio , still does show evidence that shocks begin in the jet interior , particularly in the pre - shock regions of a and c ( refer to figures 2 , 5 and 6 of p99 ) . the details of the spectral index maps we find indicate that regions of the jet closer to the axis probably have somewhat steeper values of @xmath4 and hence @xmath81 , than the sheath . the finding that the x - ray emissions must fill a volume significantly smaller than the lower - energy emission also agrees with the stratification proposed by p99 . it should be noted that because of the observed variability , this result is independent of whether the x - ray emission mechanism is synchrotron or inverse - compton . the issue of whether _ in situ _ particle acceleration is required for the m87 jet , is complex and not well explored . as explained by bsh91 , particle acceleration is required if the magnetic field is near the equipartition value of @xmath109 g . this is underlined by our data , which yield `` synchrotron ages '' for knots d , a and b of 60 - 100 years under the jp model , and somewhat higher ( 110 - 280 years ) under the kp model , similar to values calculated by bsh91 under different assumptions . however , the radiative lifetimes of synchrotron emitting particles vary strongly with magnetic field ( @xmath110 ) , so that if the magnetic field is significantly weaker than the equipartition value ( heinz & begelman 1997 ) , it is not unfeasible to reconcile the radiative lifetimes of synchrotron emitting electrons with the length of the jet ( @xmath86 kpc ) . the local variations in optical spectral index and break frequency we find , as well as the optical , radio and x - ray variability observed in hst-1 , d and a ( biretta et al . 1999 , zhou 1998 , harris et al . 1997 , 1998 ) , represent strong evidence for particle acceleration near the flux maxima of those regions . such particle acceleration is in fact expected if knots d and hst-1 , where several variable superluminal components have been found in both hst and vla monitoring , are analogous to flaring regions in blazar jets . moreover , the details of the optical polarimetry in those and other regions ( p99 ) , where we observe magnetic field vectors to be strongly compressed transverse to the jet direction , creates all the conditions necessary for particle acceleration . if indeed the observed changes in spectral index are a product of particle acceleration , then figures 3 - 4 represent a time history of particle acceleration and synchrotron aging in the jet in regions where the implied lag @xmath111 light - crossing time of the jet . this interpretation allows us to comment on the nature of particle acceleration in the m87 jet , as well as the acceleration and cooling timescales , because it transforms the observed differences in the runs of @xmath6 and @xmath4 versus distance into time lag information . in two knot regions it appears that variations in @xmath6 lag variations in @xmath4 . in d - east , a lag of @xmath39 is observed , which translates to about 7.5 years using the observed speeds @xmath112 in this region ( bsm99 ) . the somewhat larger lag of @xmath68 seen in knot e translates to about 14 years given the speeds in bsm99 . there are also two knot regions ( hst-1 and f ) where we observe the opposite trend : @xmath6 leading @xmath4 . in the case of knot hst-1 the observed lag translates into a period of @xmath113 years given the speeds in bsm99 ( greater precision is not possible given the nearness of the point source ) , while for knot f we can not translate the lags into timescales directly since the velocity fields are not well known in those knots . if particle acceleration is present in the m87 jet , the mechanism is not well constrained . several mechanisms have been proposed , including magnetic reconnection ( lesch & birk 1999 ) and fermi acceleration in knots ( meisenheimer , yates & rser 1997 ) . the structure of the jet , which contains many shock - like , flat - spectrum knots ( see also p99 ) , argues against a major role for magnetic reconnection . in contrast , the observation of sharp changes in optical magnetic field position angle in these same flat - spectrum knots ( p99 ) , suggests compression of the magnetic field in the jet interior at these positions , more consistent with shocks . the observed spectral variations near knot maxima in the inner jet are consistent with models for flares produced by shock - induced particle acceleration . as kirk , rieger & mastichiadis ( 1998 ) have shown , the evolution of spectral index in a given band during a flare depends on the relationship of the acceleration and cooling timescales for electrons which emit in that band . where the acceleration timescale is much shorter than the cooling timescale , higher - energy emissions should lead lower - energy emissions ( georganopolous & marscher 1998 , takahashi et al . 1999 ) , so that a flare will propagate from the highest energies to lower energies . but if the cooling and acceleration timescales are close to equal near where the peak in @xmath114 is , the relationship between spectral changes is complex . in that case , at low energies , where the acceleration timescale is much faster than the cooling timescale ( note that @xmath115 for synchrotron radiation ) , their model predicts a spectral flattening in advance of the flux maximum , peaking slightly before flux maximum , followed by a decrease to values somewhat below the jet s nominal value after maximum flux , and then a return to the nominal value after the disturbance has passed . but at high energies , where @xmath116 , a spectral _ steepening _ is predicted in advance of the flux maximum , followed by a hardening as the number of newly accelerated electrons build up . under this model ( assuming no injection of fresh particles ) , flares represent the reacceleration of old particles within the jet , so they will propagate from low to high energies , _ i.e. _ observed spectral changes at higher energies will lag those at lower energies . we appear to be observing _ both _ situations at different locations in the optical emission from the m87 jet , perhaps not dissimilar from what has been seen ( on shorter timescales ) in blazars ; in particular recent observations of pks 2155@xmath117304 several examples of both types of flares were seen within a week of observation ( sambruna et al . 1999 ) . with the above data and analysis in mind we can compare our findings to those of other teams on various optical jets . unfortunately high - quality data ( either ground - based or hst ) are few and far between , and only exist on a few sources . the only other radio / optical jet to have been imaged polarimetrically with hst is 3c 273 . comparing the hst and radio polarimetry for its jet ( thomson , mackay & wright 1993 , conway et al . 1993 ) , 3c 273 seems to show more similarity between radio and optical polarized structure than does the m87 jet ( p99 ) , despite having a large number of knots . however , a deep comparison of its radio and optical polarized structure is currently impossible because of the low signal to noise of the existing , pre - costar foc polarimetry data . further , a deep analysis of its optical spectrum has not yet been done , so we need to rely on ground - based data to compare the polarized and spectral index maps . neumann , meisenheimer & rser ( 1997 ) and rser & meisenheimer ( 1991 ) have carried out a deep analysis of the near - ir to optical spectrum of the 3c273 jet ( note that the latter paper also contains ground - based polarimetry , which is also useful ) , which fortunately is largely unaffected by galaxy subtraction issues ( unlike ground - based work for m87 ) . they find considerably more variation in @xmath6 than @xmath4 for 3c 273 , as we do for m87 , and also find that 3c 273 s jet is significantly narrower in the optical than in the radio ( as we do for m87 ; see sbm96 ) . a more recent , preliminary comparison of the deep hst / wfpc2 and radio images ( rser et al . 1997 , jester et al . in prep . ) show some other differences between the jet s optical and radio structure : the optical jet appears to stop @xmath18 short of the terminal radio hotspot , and the optical and radio flux maxima of knot b are not spatially coincident . interestingly , they also find somewhat of an anti - correlation between optical polarization and spectral index structure , with low polarization and fairly flat spectral indices @xmath72 from 3c 273 s nucleus , but higher polarization and much steeper spectral indices further out . more recently , _ chandra _ has imaged the 3c273 jet , finding a rather different morphology in x - rays than either the optical or radio : one knot s flux maximum appears displaced , and the optical emission declines drastically in the knot d - h region , where the strongest radio emission is seen , analysis of the knots spectra reveal inconsistencies between the spectra of downstream knots and the predictions of ssc models ( marshall et al . further analysis and deeper comparisons of the x - ray , optical and radio structure of the 3c 273 jet will no doubt give us considerable insight into its physics . lara et al . ( 1999 ) find a steep optical spectrum for 3c264 s jet , and a broadband shape consistent with either the ci model or a significantly sub - equipartition magnetic field ( heinz & begelman 1997 ) , and that it is unlikely that the observed optical jet emissions are inverse - compton in origin . they invoke the lack of well - defined knots in the 3c264 jet as evidence against fermi acceleration in shocks . this , plus the much earlier deceleration found for the 3c264 jet ( lara et al . 1999 , baum et al . 1997 ) , are key distinctions between it and m87 , which may make 3c264 more amenable to the ci scenario than m87 , where several bright shocks are observed . high signal to noise x - ray or uv observations , as well as deeper hst imaging data in several bands ( the hst data analyzed by lara et al . were 300s snapshots ) , as well as hst polarimetry , would go a long way towards resolving the mechanism which dominates in the 3c264 jet . schwartz et al . ( 2000 ) and chartas et al . ( 2000 ) have discussed the discovery with _ chandra _ of an x - ray jet in the quasar pks @xmath118 , and the subsequent identification of optical emission from three knots in its jet in deep hst images of the field . this object is at much larger distances than 3c 273 , let alone m87 or 3c 264 , making detailed comparisons of structure difficult . however , comparisons of the structure in the radio and optical reveal large - scale correspondence but tantalizing hints of differences on scales of @xmath119 which require deeper investigation . the broadband spectra of the knots are also inconsistent with the predictions of simple ssc models , although comptonized emission from cmb photons remains a possibility ( tavecchio et al . 2000 ) one final object for which such analysis can in future be done is 3c 66b . sparks et al . ( in prep . ) are analyzing radio - optical spectral data for this object , which also shows interesting radio - optical spectral differences . by discussing the kinetic energy budget of optical jets , scarpa & urry ( 1999 ) find that their observed properties require either fairly high bulk doppler factors @xmath120 and viewing angles @xmath121 ( if equipartition is maintained ) , or magnetic fields significantly less strong than equipartition ( whereupon lower values of @xmath122 or larger viewing angles are possible ) . this is in agreement with the findings of heinz & begelman ( 1997 ) for m87 , and consistent with our results . a similar result has been found through comparison of the properties of dust disks in the 3cr sample ( sparks et al . the above sections underline the importance of future observations . since our data indicate break frequencies significantly in excess of @xmath90 hz , high - resolution imaging observations in the far uv and x - rays are now required to localize the peak of the synchrotron emission from bright knots in the jet . such data , if gathered contemporaneously with optical and radio observations , would allow much stronger constraints on synchrotron emission and particle acceleration models . the _ chandra _ observations are particularly critical in this regard . those observations will yield spatially resolved ( resolution @xmath123 ) spectra in the @xmath124 kev region ( _ i.e. , _ @xmath125 hz ) , allowing direct comparisons of optical and radio morphologies . such comparisons will directly test the findings of neumann et al . ( 1997 ) that the x - ray emission from knot a is not coincident with the radio / optical peak , and will allow us to localize sites of particle acceleration . the _ chandra _ observations will also directly test the nature of the jet s x - ray emission by measuring the x - ray slope ( 5.1 ) . our data indicate that observations in the far - uv are required to localize the synchrotron peak of much of the jet . no imaging mission is currently proposed at these wavelengths ; however , the prospects for detecting m87 in the far - uv are bright , since the @xmath126 along its line of sight is relatively low ( although euve did not detect m87 ; lampton et al . 1997 ) . future monitoring experiments will have unparalleled importance for models of jet flares in blazars ( for a review of the spectral evolution observed in hbl flares , see takahashi et al . once the timescale of x - ray variability is established , the combination of contemporaneous vla , hst and _ chandra _ observations will make it possible to for the first time map out the broadband spectrum and polarization of a flaring component , and watch its spectrum decay once the flare has passed . such a study would allow us to fully model the spectrum and magnetic field configuration in the flaring region . we wish to thank eddie bergeron and megan sosey , as well as shireen gonzaga , for considerable help with the hst data reductions . we also wish to thank our referee , hermann - josef rser , for many suggestions which significantly improved this paper . jpl thanks the stsci for hospitality during his sabbatical visit . esp thanks chris reynolds , sebastian heinz , mitch begelman , joel carvalho , tim heckman , julian krolik , eric agol , greg madejski and jean eilek for interesting discussions . research on m87 at stsci is supported by hst grants go-5941 , go-7274 , go-7866 , go-8048 and go-8140 . esp acknowledges support at johns hopkins university and the university of maryland , fromfromfromfrom nasa ltsa grant nag5 - 9534/nag5 - 9997 . * figure 1 . * the jet of m87 , as seen by hst , in the f300w ( 0.3 microns , top ) , f814w ( 0.8 microns , middle ) and f160w ( 1.6 microns , bottom ) bands . flux levels displayed are from -0.3 to 1 @xmath127jy / pix for the f300w image , -0.5 to 2 @xmath127jy / pix for the f814w image , and -1 to 4 @xmath127jy / pix for the f160w image , and within these limits the identical color scale ( shown at bottom ) was used for all three images . the images have been rotated so that the jet is along the x axis , and have identical pixel sizes , origins and orientations . the traditional names are used for bright knots . * figure 2 . * maps of fitted quantities for the m87 jet , with the convolved f814w image of the jet shown for comparison at the bottom . at top ( figure 2a ) , we show the optical spectral index image ; second from top ( figure 2b ) , we show the radio - optical spectral index image . we also show maps of the synchrotron break frequency ( third from top , figure 2c ) and injection index ( fourth from top , figure 2d ) in the kp model . identical color tables were used for all panels . the color scale in figure 2a runs from 1.2 to 0.4 , while that in figures 2b and 2d run from 0.85 to 0.6 , and that for the break frequency map ( figure 2c ) runs from @xmath90 to @xmath88 . * figure 3 . * runs of @xmath6 ( top , fig . 3a ) , @xmath4 ( second from top , fig . 3b ) , @xmath42 ( third from top , fig . 3c ) and f814w flux ( bottom , fig . 3d ) , plotted as a function of distance along the jet axis . in this figure we show the inner regions of the jet , out to @xmath128 from the nucleus . these values are _ flux - weighted _ , i.e. , flux is summed across the jet at each pixel in the x direction before spectral indices were computed . see 3 for discussion . * figure 4 . * runs of of @xmath6 ( top , fig . 4a ) , @xmath4 ( second from top , fig . 4b ) , @xmath42 ( third from top , fig . 4c ) and f814w flux ( bottom , fig . 4d ) , plotted as a function of distance along the jet axis . in this figure we show the outer regions of the jet , @xmath44 from the nucleus . these values are _ flux - weighted _ , i.e. , flux is summed across the jet at each pixel in the x direction before spectral indices were computed . see 3 for discussion . * figure 5 . * fluxes of knot regions . the regions are defined in table 2 . the error bars include flatfielding , zeropoint and poisson errors . best - fit power law spectra are shown as lines on these log - log plots , and the spectral indices are given in table 2 . * figure 6 . * synchrotron spectrum model fits to four regions of the jet : knot d ( figure 6a ) , knot a ( figure 6b ) , knot b ( figure 6c ) , and knots a+b+c ( figures . 6d - e ) . along with our data , we plot fits from the jp , kp and ci models , both including and not including the x - ray point . in figures 6d - e , we show also previous data as tabulated by mrs96 ( see references and discussion in that paper ) . in all plots , data from this work are plotted as diamonds ; in figs . 6d - e , data taken from mrs96 are plotted as squares . see 4 for discussion .

we consider design of the quantum stabilizer codes via a two - step , low - complexity approach based on the framework of codeword - stabilized ( cws ) codes . in this framework , each quantum cws code can be specified by a graph and a binary code . for codes that can be obtained from a given graph , we give several upper bounds on the distance of a generic ( additive or non - additive ) cws code , and the lower gilbert - varshamov bound for the existence of additive cws codes . we also consider additive cyclic cws codes and show that these codes correspond to a previously unexplored class of single - generator cyclic stabilizer codes . we present several families of simple stabilizer codes with relatively good parameters .

1108.5490

it was the invention of quantum error correcting codes @xcite ( qeccs ) that opened quantum computing ( qc ) as a theoretical possibility . however , high precision required for error correction @xcite combined with the large number of auxiliary qubits necessary to implement it , have so far inhibited any practical realization beyond proof - of - the - principle demonstrations@xcite . in any qecc , one needs to perform certain many - body quantum measurements in order to decide how to correct the encoded state . the practical difficulty is that a generic code requires measurements which are both complicated and frequent at the same time . it is therefore clear that a quantum computer can only be build via a thorough optimization at every step of the design . in particular , code optimization targets codes that combine good parameters with fairly simple measurements . it is also desirable to parallelize these measurements given a specific on - chip layout of a qc architecture . to date , the main focus of the qecc - research has been on finding good codes with the traditional code parameters , which are the block length @xmath0 , code dimension @xmath1 , and code distance @xmath2 ( or code rate @xmath3 and the relative distance @xmath4 ) . for stabilizer codes @xcite , we also consider the number of encoded qubits @xmath5 . a number of stabilizer codes@xcite have been designed that meet or nearly achieve the existing bounds on distance @xmath2 for the given @xmath6 and @xmath0 . code parameters can be further refined by going beyond the family of stabilizer codes . one example is a recently introduced framework of codeword - stabilized ( cws ) quantum codes@xcite . a qubit cws code @xmath7 ( in standard form ) is determined by a graph @xmath8 and a classical binary code @xmath9 . cws codes include all stabilizer codes as a subclass ( the corresponding binary code @xmath10 must be linear ) , but also the codes which have been proved to have parameters superior to those of any stabilizer code@xcite . unfortunately , typical gains in code dimension @xmath1 correspond to a fraction of a qubit . moreover , error - correcting algorithms known for general non - additive cws codes have exponential complexity@xcite , as opposed to polynomial complexity of the stabilizer codes . even for the relatively simple additive codes , their optimization is a very difficult problem that has exponential complexity . this is one of the main reasons as to why the two relatively simple code families are almost exclusively used among stabilizer codes to estimate the threshold accuracy required for scalable quantum computation : the concatenated codes@xcite and the surface codes@xcite which originated from the toric codes@xcite . both families have very low code rates that scale as inverse powers of code distance . in this work we explore how the framework of cws codes can be used to relegate the design of quantum stabilizer codes to classical binary linear codes in order to simplify the overall design . in particular , we formulate several theorems framing the parameters of an additive cws code which can be obtained from a given graph . we also suggest a simple decomposition of the @xmath11 generator matrix corresponding to the stabilizer in terms of the graph adjacency matrix and the parity check matrix of the binary code . finally , we design several graph families corresponding to regular lattices which result in some particularly good codes . these include graphs with circulant adjacency matrix which can be used to construct single - generator cyclic additive codes , a class of codes overlooked in previous publications . in particular , we prove the existence of single - generator cyclic additive codes with the parameters @xmath12 $ ] , @xmath13 and @xmath14 $ ] ( version of toric codes ) . note that these code families have distances that are not bounded , unlike any cws code families constructed previously@xcite . the paper is organized as follows . in sec . [ sec : notations ] , we introduce the notations and briefly review some known results for quantum and classical codes . in sec . [ sec : upper - bound ] , we establish several upper bounds on general cws codes . in sec . [ sec : additive - cws ] we give a cws decomposition of the @xmath15 matrix corresponding to the stabilizer generators . in sec . [ sec : gv ] , we formulate the gilbert - varshamov ( gv ) bounds for additive cws codes which can be obtained from a given graph . cyclic additive cws and more general single - generator additive cyclic codes are considered in sec . [ sec : cyclic - cws ] where we discuss their properties and give several examples . we give our conclusions in sec . [ sec : conclusions ] . a _ classical _ @xmath16-ary block error - correcting code @xmath17 is a set of @xmath1 length-@xmath0 strings over an alphabet with @xmath16 symbols . different strings represent @xmath1 distinct messages which can be transmitted . the ( hamming ) distance between two strings is the number of positions where they differ . distance @xmath2 of the code @xmath10 is the minimum distance between any two different strings from @xmath18 . in the case of _ linear _ codes , the elements of the alphabet must form a galois field @xmath19 ; all strings form @xmath0-dimensional vector space @xmath20 . a linear error - correcting code @xmath21_q$ ] is a @xmath6-dimensional subspace of @xmath20 . the distance of a linear code is just the minimum weight of a non - zero vector in the code , where weight @xmath22 of a vector @xmath23 is the number of non - zero elements . a basis of the code is formed by the rows of its _ generator matrix _ @xmath24 . all vectors that are orthogonal to the code form the corresponding @xmath25-dimensional dual code , its generator matrix is the parity - check matrix @xmath26 of the original code . for a _ binary _ code @xmath27 $ ] , the field is just @xmath28 . for a _ quaternary _ code @xmath29 , the field is @xmath30 , with @xmath31 for non - binary codes , there is also a distinct class of _ additive _ classical codes , defined as subsets of @xmath20 closed under addition ( in the binary case these are just linear codes ) . a code @xmath10 is cyclic if inclusion @xmath32 implies that @xmath33 . codes that are both linear and cyclic are particularly simple : by mapping vectors to polynomials in the natural way , @xmath34 , it is possible to show that any such code consists of polynomials which are multiples of a single generator polynomial @xmath35 , which must divide @xmath36 ( using the algebra corresponding to the field @xmath19 ) . the quotient defines the _ check polynomial _ @xmath37 , @xmath38 which is the generator polynomial of the dual code . the degree of the generator polynomial is @xmath39 . the corresponding generator matrix @xmath24 can be chosen as ( the first @xmath6 rows of ) the circulant matrix formed by subsequent shifts of the vector that corresponds to @xmath35 . _ qubit _ quantum error correcting codes are defined in the complex hilbert space @xmath40 , where @xmath41 is the hilbert space of a single two - level system . @xmath41 is formed by all vectors @xmath42 with @xmath43 , and the inner product such that the two states are orthonormal , @xmath44 , @xmath45 . any operator acting in @xmath40 can be represented as a linear combination of pauli operators which form the @xmath0-qubit pauli group @xmath46 of size @xmath47 , @xmath48 where @xmath49 , @xmath50 , and @xmath51 are the usual pauli matrices , and @xmath52 is the identity matrix . the weight @xmath53 of a pauli operator @xmath54 is the number of non - identity terms in the corresponding tensor product . all pauli operators are unitary ; they are also hermitian with eigenvalues @xmath55 when the phase factor @xmath56 in eq . ( [ pauligroup ] ) is real - valued , @xmath57 . a state @xmath58 is stabilized by a hermitian pauli operator @xmath59 if @xmath60 . a linear space @xmath61 is stabilized by a set of operators @xmath62 if each vector in @xmath61 is stabilized by every operator in @xmath62 . an @xmath63 _ quantum error - correcting code _ is a @xmath1-dimensional subspace of the hilbert space @xmath40 . such a subspace can be described by an orthonormal basis @xmath64 . let @xmath65 be some set of pauli errors . a qecc detects all errors @xmath66 if and only if @xcite @xmath67 where @xmath68 only depends on the error @xmath54 , but is independent of the basis vectors . a qecc has distance @xmath2 if it can detect all pauli errors of weight @xmath69 , but not all errors of weight @xmath2 . the errors in the set @xmath70 can be corrected if and only if all the nontrivial pairwise combinations of errors from @xmath70 are detectable @xcite . thus a distance-@xmath2 code corrects all errors of weight @xmath71 . the code @xmath72 is _ non - degenerate _ if linearly - independent errors from @xmath73 produce corrupted spaces @xmath74 that are linearly independent . otherwise , the code is _ degenerate _ , implying the existence of at least two _ mutually degenerate _ linearly independent operators @xmath75 which act identically on @xmath72 . the code is called _ pure _ if linearly independent errors from @xmath73 produce corrupted spaces that are not only linearly independent , but also mutually orthogonal . for all codes considered in this work , non - degenerate codes are also pure@xcite . two codes are considered equivalent if they differ just by qubit order , and/or discrete rotations leaving each of the single - qubit pauli groups invariant . the latter are called local clifford ( lc ) transformations . here we briefly review the well - known family of stabilizer codes @xcite . an @xmath76 $ ] _ stabilizer code _ @xmath61 is a @xmath77-dimensional subspace of the hilbert space @xmath40 stabilized by an abelian group @xmath78 with @xmath79 hermitian pauli generators , @xmath80 . explicitly , @xmath81 such a code exists only if @xmath82 . the group @xmath83 is called the stabilizer of the code . changing the sign(s ) of one or several of the generators @xmath84 results in replacing @xmath61 with one of @xmath85 equivalent codes whose direct sum ( together with @xmath61 ) is the entire space @xmath86 . the _ normalizer _ of @xmath83 is a set of pauli operators generating unitary transformations that leave @xmath83 invariant , @xmath87 elements of the normalizer form a group commuting with @xmath83 but not necessarily with each other . it is possible to construct @xmath88 logical operators @xmath89 , @xmath90 , @xmath91 belonging to @xmath46 with the usual commutation relations that generate the normalizer when the generators of @xmath83 are included @xcite . the abelian subgroup of @xmath92 , @xmath93 , becomes a maximal abelian subgroup of @xmath46 when the generator @xmath94 is also included . the group @xmath95 stabilizes a unique _ stabilizer _ state @xmath96 , an @xmath97 $ ] stabilizer code , while the operators @xmath89 generate the basis of the code , i.e. , @xmath98 by convention , the stabilizer state is considered non - degenerate , and its distance @xmath99 is the minimum weight of a non - trivial member of the group @xmath100 . for stabilizer codes , phases of ( hermitian ) pauli operators are only needed to choose one of the equivalent codes in eq . ( [ eq : stabilizer ] ) , as well as to introduce the commutation relations . it is convenient to drop the phases and map the pauli operators to two binary strings , @xmath101 @xcite , @xmath102 where @xmath103 , @xmath104 , and @xmath105 is generally different from that in eq . ( [ pauligroup ] ) . this map preserves the operator algebra , with a product of two pauli operators @xmath106 and @xmath107 corresponding to a sum of the corresponding binary vectors @xmath108 and @xmath109 . the map ( [ eq : umap ] ) can be taken one step further@xcite to quaternary codes , by introducing @xmath110 vectors @xmath111 [ see eq . ( [ eq : f4def ] ) ; note that this mapping differs slightly from that in ref . ] . we will denote this combined map as a function @xmath112 . note that up to a phase this association also allows us to define @xmath113 . to be specific , for the pauli operator @xmath113 we will set @xmath114 in eq . ( [ eq : umap ] ) , which corresponds to @xmath115 in eq . ( [ pauligroup ] ) . it is easy to check that two pauli operators commute if and only if the symplectic scalar product @xmath116 vanishes ( @xmath117 ) . in terms of the corresponding @xmath118 , this corresponds to the vanishing of the _ trace inner product _ @xmath119 where @xmath120 , @xmath121 . a dual code to an additive @xmath11 code @xmath29 ( equipped with trace inner product ) is defined as @xcite @xmath122 if @xmath123 , one says @xmath29 is self - orthogonal . a classical additive code @xmath29 corresponding to a set of operators @xmath124 is self - orthogonal if and only if @xmath124 is an abelian group . thus any quantum stabilizer code can be described as a self - orthogonal classical additive code over @xmath11 . the following theorem is applicable to additive @xmath11 codes ( variant of theorem 2 from ref . @xcite ) : [ th1]suppose @xmath29 is an additive self - orthogonal code in @xmath125 , containing @xmath126 vectors , such that there are no vectors of weight @xmath127 in @xmath128 . then @xmath129 defines a stabilizer of an additive qecc with parameters @xmath76 $ ] . [ ex:713 ] the well - known calderbank - shor - steane ( css ) @xmath130 $ ] code @xcite has the stabilizer with the generators@xcite @xmath131 and the logical operators @xmath132 as any css code , this code is linear . qubit permutations also give an equivalent cyclic linear code with the generator polynomial @xmath133 ; @xmath35 is a factor of @xmath134 . the corresponding check polynomial is @xmath135 . cws codes @xcite represent a general class of nonadditive qeccs . a general cws code is defined in terms of a stabilizer state @xmath136 and a set of @xmath1 mutually commuting _ codeword operators _ @xmath137 . explicitly [ cf . ( [ eq : code - basis ] ) ] , @xmath138 for non - trivial cws codes , this construction coincides with union - stabilizer ( ust ) codes@xcite , restricted to the zero - dimensional originating code . any stabilizer state is lc - equivalent to a _ graph state _ @xcite , a stabilizer state with the stabilizer group @xmath139 whose generators @xmath140 are determined by the adjacency matrix @xmath141 of a ( simple ) graph @xmath8 , @xmath142 where @xmath143 , @xmath144 denotes the @xmath145-th row of @xmath146 . in fact , such a graph is usually not unique , even after accounting for graph isomorphisms . the full set of lc - equivalent graph states can be generated by a sequence of _ local complementations _ , operations on a graph where the subgraph corresponding to a neighborhood of a particular vertex is inverted . such graphs are called _ _ locally equivalent__@xcite . any cws code @xmath63 is lc equivalent to a cws code in _ standard form _ , defined by an order-@xmath0 graph @xmath8 and a classical binary code @xmath10 containing @xmath1 binary words . the graph defines the graph state , while the vectors of the classical code @xmath147 are used to generate the code word operators , @xmath148 . thus , @xmath149 it is customary to use notation @xmath150 for cws codes in standard form . the key simplification of the cws construction comes from the fact that the basis states @xmath151 are eigenvectors of the graph stabilizer generators , @xmath152 thus , a pauli operator in the form ( [ eq : umap ] ) can be transformed to a @xmath51-only operator @xmath153 , where the _ graph image _ of the operator @xmath154 is the binary vector @xmath155 the error correcting properties of a quantum cws code @xmath150 and the classical code @xmath9 are related by the following [ th2 ] ( after theorem 3 from ref . @xcite ) consider a cws code @xmath156 . an error @xmath54 such that @xmath157 , is detectable in @xmath72 if and only if the binary vector @xmath158 is detectable within the code @xmath10 . an error @xmath54 such that @xmath159 is detectable in @xmath72 if and only if @xmath160 for all @xmath161 . the case @xmath162 corresponds to pure ( non - degenerate ) errors , while @xmath163 indicates that the error is in the graph stabilizer group @xmath164 ; the corresponding detectability condition is a requirement that the error must be degenerate . while in general cws codes are non - additive , they include all stabilizer codes as a subclass . a cws code @xmath165 is additive if @xmath10 is a linear code@xcite . the stabilizer @xmath83 of an additive cws code in standard form is a subgroup of the graph stabilizer @xmath164 ; it can be obtained from the graph - stabilizer generators by a symplectic gram - schmidt orthogonalization procedure@xcite . conversely , the representation ( [ eq : code - basis ] ) of an additive code corresponds to a general cws code ; an lc transformation may be needed to obtain the corresponding standard form , and one can always find a standard form where @xmath10 is linear@xcite . in the following we will always assume such a representation . -ring graph corresponding to the @xmath166 $ ] code in example ex . [ ex:513 ] . center : `` kite '' graph corresponding to the degenerate code @xmath167 $ ] from example [ ex:613 ] . right : the graph corresponding to the cyclic @xmath130 $ ] code from example [ ex:713cws].,title="fig : " ] -ring graph corresponding to the @xmath166 $ ] code in example ex . [ ex:513 ] . center : `` kite '' graph corresponding to the degenerate code @xmath167 $ ] from example [ ex:613 ] . right : the graph corresponding to the cyclic @xmath130 $ ] code from example [ ex:713cws].,title="fig : " ] -ring graph corresponding to the @xmath166 $ ] code in example ex . [ ex:513 ] . center : `` kite '' graph corresponding to the degenerate code @xmath167 $ ] from example [ ex:613 ] . right : the graph corresponding to the cyclic @xmath130 $ ] code from example [ ex:713cws].,title="fig : " ] [ ex:513 ] the smallest single - error - correcting code is the linear cyclic @xmath166 $ ] code@xcite with the generator polynomial @xmath168 which divides @xmath169 . this code is unique ; its stabilizer generators can be obtained as cyclic permutations of a single operator , @xmath170 , and the logical operators are @xmath171 the corresponding cws code@xcite can be generated from the @xmath172-ring graph in fig . [ fig : ex613 ] ( left ) , and the binary code has a single generator @xmath173 . note that both the graph and the binary code preserve the original cyclic symmetry . [ ex:613 ] there exist only two inequivalent single - correcting codes @xmath167 $ ] ; both are degenerate@xcite . one of the codes is obtained from the code in example [ ex:513 ] by adding a qubit ; the graph of the corresponding cws code can be chosen as a @xmath172-ring [ fig . [ fig : ex613 ] left ] and a disconnected vertex @xmath174 ; the binary code @xmath10 is generated by @xmath175 . the degeneracy group is generated by @xmath176 . the stabilizer generators for the second code are listed in ref . this code corresponds to the graph in fig . [ fig : ex613 ] ( center ) , while the binary code is generated by @xmath177 . while there are three bits which are not involved with the classical code , they can not be dropped as they are part of the entangled state . the degeneracy group is generated by @xmath178 ( the equivalence follows from the fact that the first two vertices of @xmath8 share all of their neighbors ) . [ ex:713cws ] the linear cyclic @xmath130 $ ] css code from the example [ ex:713 ] is lc equivalent to a cws code with the graph in fig . [ fig : ex613 ] ( right ) . the corresponding classical code is given by @xmath179 . note that neither the graph nor the binary code is explicitly symmetric with respect to cyclic permutations of the qubits . note also that an inequivalent cws cyclic @xmath130 $ ] code exists ; we constructed such a code among others in example [ ex : k - copies ] , see table [ tab : quantum - repetition ] . in this section we give upper bounds on general cws codes in terms of the properties of the corresponding graph @xmath8 and the binary code @xmath10 . [ lemma : upper - binary ] the distance @xmath2 of the cws code @xmath180 can not exceed that of @xmath181 . indeed , any `` classical '' error in the form @xmath182 is mapped by eq . ( [ clmapping ] ) to the binary vector @xmath183 . if @xmath54 is detectable by @xmath72 , @xmath183 should be detectable by @xmath10 . lemma [ lemma : upper - binary ] concerns with errors which are dealt with by the binary code . on the other hand , a cws code is an enlargement of the code formed by the graph state . the following observation has been made in ref . @xcite : [ lemma : upper - pure ] the distance @xmath2 of a nondegenerate cws code @xmath63 is limited by the distance @xmath184 of the graph stabilizer state , @xmath185 . it follows from the fact that any member of the graph stabilizer is either a degenerate error , or it is a non - detectable error . note , however , as illustrated by the example [ ex:613 ] , in general , the distance of a cws code can actually be bigger than that of the graph stabilizer state . for a binary code @xmath10 , we will say that the @xmath186-th bit is _ involved _ in the code if there are vectors in the code for which the value of @xmath186-th bit differ , @xmath187 . alternatively , if the all - zero vector @xmath188 is in the code ( which can always be arranged ) , the condition is that there is a vector @xmath189 where @xmath186-th bit is non - zero , @xmath190 . [ lemma : j - th ] for a cws code @xmath156 with @xmath191 , let us assume that @xmath186-th bit is involved in the code @xmath192 . then the graph - stabilizer generator @xmath193 violates the error detection condition in theorem [ th2 ] . since the generator @xmath193 is in the graph stabilizer , @xmath194 , one has to check the degenerate condition in theorem [ th2 ] . the commutativity of @xmath193 with a given @xmath195 is determined by the @xmath186-th bit of @xmath23 ; conditions of the lemma ensure that only one of the two vectors commute with @xmath193 . note that this means that the code distance can not exceed that of any @xmath193 corresponding to bits involved in the binary code . since at least @xmath2 bits must be involved in the binary code , lemma [ lemma : j - th ] guarantees the following bound [ th : max - weight ] the distance @xmath2 of a cws code @xmath156 can not exceed the @xmath2-th largest weight of @xmath196 , minimized over all graphs that are locally - equivalent to @xmath8 . we will also be using the following for a graph @xmath8 with all vertices of the same degree @xmath197 , the distance of a cws code @xmath156 can not exceed @xmath198 . in particular , for any ring graph @xmath199 , which gives @xmath200 , for any double - ring graph@xcite @xmath201 , and for a large enough square lattice wrapped into a torus , @xmath202 . obviously , to maximize the distance of a cws code , one may want to maximize the distance of the binary code @xmath10 . to this end , it is a good idea to make sure that every bit is involved in @xmath10 . for such codes , we have [ th : all - bit ] the distance of a cws code @xmath180 where the binary code @xmath10 involves all bits can not exceed the _ minimum _ weight of @xmath196 , minimized over all graphs that are locally - equivalent to @xmath8 . the stabilizer of an additive cws code @xmath156 is a subgroup of the abelian graph stabilizer @xmath164 , and its generators @xmath203 can be expressed as products of graph stabilizer generators @xmath140 @xcite . explicitly , @xmath204 where @xmath205 is the corresponding matrix of binary coefficients . with the help of eq . ( [ eq : graph - generators ] ) , we obtain the following decomposition for the generator matrix @xmath24 of the associated additive @xmath15 code @xmath29 , @xmath206 where @xmath141 is the symmetric graph adjacency matrix with zeros along the diagonal , and @xmath207 is the @xmath208 identity matrix . the relation between the binary code @xmath10 and the quaternary code @xmath29 is explicitly given by the following the additive @xmath15 code @xmath29 with the generator matrix ( [ gf4 ] ) is the map @xmath209 of the stabilizer @xmath83 of the additive cws code @xmath210 generated by the graph with the adjacency matrix @xmath146 and the linear binary code @xmath10 if and only if @xmath211 is the parity check matrix of @xmath10 . use the basis vectors in the form ( [ eq : cws - basis ] ) and the commuting operators ( [ eq : stabilizer - operator - cws ] ) corresponding to the rows of the matrix @xmath24 [ eq . ( [ gf4 ] ) ] . direct calculation gives @xmath212 were summation over repeated indices is assumed . the statement of the lemma ( both ways ) follows from the definition of the parity check matrix . we can now easily relate the error detection conditions for additive codes in theorems [ th1 ] ( codes over @xmath15 ) and [ th2 ] ( cws codes ) . the code @xmath29 in theorem [ th1 ] is given by additive combinations of the @xmath79 rows of the generator matrix ( [ gf4 ] ) . evaluating the outer trace product of the generator matrix ( [ gf4 ] ) with a vector @xmath213 , we obtain the condition for the vector to be in @xmath214 @xmath215 this uniform linear system of @xmath79 equations with @xmath216 variables has @xmath217 linearly - independent solutions . the corresponding basis can be chosen as a set of @xmath6 linearly - independent `` classical '' vectors @xmath218 , where @xmath219 and the corresponding @xmath220 , @xmath221 , plus @xmath0 linearly - independent vectors such that @xmath222 , @xmath223 . some linear combinations of the latter vectors are actually in @xmath29 . these can be found using the identity @xmath224 : the corresponding @xmath225 have to satisfy @xmath226 , which precisely corresponds to the degenerate case , @xmath227 , in theorem [ th2 ] . general theory of cws codes guarantees that generator matrix of a quantum code equivalent to any additive self - orthogonal code over @xmath15 can be decomposed in the form ( [ gf4 ] ) . conversely , any matrix in the form ( [ gf4 ] ) with binary matrices @xmath211 and @xmath146 generates a self - orthogonal code over @xmath15 as long as the matrix @xmath146 is symmetric . we use it in the following section to prove the lower gilbert - varshamov ( gv ) bound for the parameters of an additive cws code which can be obtained from a given graph . the gv bound is a counting argument which non - constructively proves the existence of codes with parameters exceeding certain threshold . the argument is based on the fact that the set of possible codes ( vector spaces ) vastly outnumbers the set of vectors . then , if we count all codes of a given length @xmath0 , and then subtract the number of codes that contain any vector of weight @xmath228 or less , the remaining codes ( if any ) will all have distance @xmath2 or more . this `` greedy '' argument ignores any possible double counting of codes that contain several small - weight vectors . note that the gv bound necessarily gives asymptotically _ good _ codes with relative distance @xmath4 and code rate @xmath229 bounded away from @xmath230 as @xmath231 . for the entire class of pure stabilizer codes , the asymptotic gv bound@xcite states that there exist codes of length @xmath231 such that @xmath232 where @xmath233 is the binary entropy function . we are going to prove that the same bound also holds for pure cws codes corresponding to a given graph @xmath8 , as long as @xmath234 [ see lemma [ lemma : upper - pure ] ] . we are using eq . ( [ gf4 ] ) to parameterize the stabilizer matrices ; the resulting @xmath11 codes are automatically self - orthogonal . let @xmath235 be the number of cws codes that have length @xmath0 , dimension at least @xmath6 , and correspond to a given graph @xmath236 . let also @xmath237 be the number of such codes which contain a given vector @xmath213 , @xmath238 , in @xmath239 [ see eq . ( [ gf4 ] ) ] . the corresponding condition ( [ eq : c - perp ] ) is given by the trace inner product ( [ orthogonality ] ) . for @xmath240 , the binary vector @xmath241 is always non - zero ( which also guarantees that @xmath242 ) . as a result , @xmath237 and @xmath235 represent the corresponding numbers for the _ binary _ codes . then the standard counting arguments @xcite show that @xmath243 here we use the fact that each of @xmath244 vectors @xmath23 belongs to the same number @xmath237 of binary codes ; also each of @xmath235 binary codes contains @xmath245 nonzero vectors @xmath23 . the number of quaternary vectors of weight @xmath246 is @xmath247 . thus , for any graph @xmath236 , there exists a distance-@xmath2 cws code as long as @xmath248 now we see that there exist @xmath76 $ ] cws codes for the graph @xmath236 with distance @xmath249 where @xmath250 is the distance of the graph state @xmath251 and @xmath252 note that thus obtained quantum codes are always pure , since the summation in the l.h.s . can only be extended up to @xmath253 . apart from this latter condition , eq . ( [ eq : sump ] ) is identical to the quantum gilbert - varshamov bound@xcite for pure stabilizer codes , and takes the asymptotic form ( [ eq : gv1 ] ) as @xmath254 . the exact gv bound @xmath255 for pure stabilizer codes ( without the restriction on the distance ) is recovered if we go over different graphs . indeed , the gv bound ( [ eq : gv1 ] ) also applies for the special case of @xmath256 , corresponding to stabilizer states or self - dual codes@xcite . the gv bound on the relative distance is monotonous in @xmath6 and reaches its maximum at @xmath257 then for given @xmath0 and @xmath258 one can always find a suitable graph such that the gv bound @xmath259 becomes more restrictive than the condition @xmath260 . in practice , graphs with large distance @xmath251 are complicated ( have too many edges ) . it is much easier to come up with graph families corresponding to a fixed graph - state distance @xmath251 . for @xmath231 , the corresponding code families approach the maximum rate @xmath261 and have asymptotic redundancy @xmath262 defined by eq . ( [ eq : sump ] ) : @xmath263 it is readily verified that the r.h.s of estimate ( [ eq : gv - redundancy ] ) has the order of @xmath264 if @xmath2=const and @xmath231 . ] circles represent the qubits . all @xmath265 translations of the empty - circle pattern form the classical codewords @xmath266 . the generators of the stabilizer are formed as the products of the graph - state stabilizer generators along the directions parallel to the dashed lines . ] [ ex : square - planar ] graphs in the form of sufficiently large finite square lattice fragments [ fig . [ fig3 ] ] have maximum distance @xmath267 , but this requires that the bits in the corners and around the perimeter be not involved in the classical code . somewhat better redundancy can be achieved by avoiding only the bits in the corners , which gives @xmath268 . consider the family of classical codes where the codewords are obtained by taking all translations of the pattern shown in fig . [ fig3 ] with open symbols . the weight of any linear combination of such codewords is at least 4 . the lattice shown corresponds to the code @xmath269 $ ] , while general @xmath270 lattice gives the code with the parameters @xmath271 $ ] . asymptotically , the redundancy for @xmath272 is @xmath273 , @xmath274 , which is bigger than the logarithm in eq . ( [ eq : gv - redundancy ] ) . however , the fraction of auxiliary qubits vanishes as @xmath275 for large @xmath0 . the code distance can be increased with higher - dimensional generalizations , e.g. , we can generalize this construction to @xmath276-dimensional hypercubic lattice with @xmath277 nearest neighbors so that the distance is @xmath278 in full analogy with the two dimensional case . the corresponding redundancy will scale with the area of the boundary . while the code in fig . [ fig3 ] serves as a good illustration of the concept of lattice codes with simple stabilizer structure , it is still far from optimal . on the @xmath279 square lattice we have constructed numerically a code @xmath280 $ ] with weight-@xmath281 codewords which can be mapped into each other by translations and rotations . this design is only one logical qubit short of the best - known generic code @xmath282 $ ] . consider graphs in the form of @xmath283 square lattices wrapped into tori due to periodic boundary conditions . for @xmath284 , these graphs have the distance @xmath285 . gv bound ( [ eq : greedy - full ] ) shows that the cws codes with the following parameters can be obtained for these graphs : @xmath286 $ ] , @xmath287 $ ] , @xmath288 $ ] , @xmath289 $ ] , @xmath290 $ ] , . consider graphs in the form of @xmath283 triangular lattices wrapped into tori due to periodic boundary conditions . these graphs have the distance @xmath291 . gv bound ( [ eq : greedy - full ] ) shows that the cws codes with the following parameters can be obtained for these graphs : @xmath292 $ ] , @xmath293 $ ] , @xmath294 $ ] , @xmath295 $ ] , . example [ ex:713cws ] shows that a cyclic additive code does not necessarily preserve its symmetry when converted to cws standard form . by a _ cyclic _ additive cws code we just mean a code which is cyclic in standard form , with a circulant graph . for such a code , eq . ( [ gf4 ] ) can be rewritten as the generator polynomial , @xmath296,\ ] ] where the polynomials @xmath297 and @xmath298 are binary , @xmath297 is the parity - check polynomial of a binary cyclic code ( and therefore must divide @xmath299 ) , while @xmath298 corresponds to a symmetric circulant matrix , @xmath300 any such _ symmetric _ polynomial @xmath298 leads to a valid self - orthogonal additive code . the dimension of the quantum code corresponding to the generator polynomial ( [ eq : gf4-polynomial ] ) is @xmath301 . previously , the additive cyclic qeccs were introduced in theorem 14 of ref . @xcite , stating that any such code has two generators . single_-generator additive code described by eq . ( [ eq : gf4-polynomial ] ) represents a new setting , in which the second generator is equal to zero . this condition gives a self - orthogonal additive code @xmath29 [ see sec . [ sec : intro - stab - codes ] ] with no binary codewords ( any @xmath302 , @xmath303 , has @xmath304 ) . a somewhat wider class of _ single - generator _ cyclic additive codes can be also defined via eq . ( [ eq : gf4-polynomial ] ) , without requiring the symmetry ( [ eq : symmetric - polynomial ] ) of @xmath298 . then two codes @xmath72 and @xmath305 that have a generator polynomial in the form ( [ eq : gf4-polynomial ] ) with the same @xmath306 are equivalent if and only if @xmath307 where @xmath308 is the generator polynomial of the binary code@xcite . such a polynomial ( [ eq : gf4-polynomial ] ) generates a self - orthogonal @xmath15 code if and only if @xcite @xmath309 this guarantees self - orthogonality for any @xmath298 as long as @xmath310 an alternative formulation of this sufficient condition is that the corresponding generator polynomial @xmath311 must contain no more than one root from each pair @xmath312 of mutually conjugate @xmath0th roots of unity , @xmath313 . in particular , a self - reciprocal ( _ palindromic _ ) polynomial ) which correspond to symmetric circulant matrices . palindromic polynomials have reflection symmetry with respect to their `` centers '' , while eq . ( [ eq : symmetric - polynomial ] ) corresponds to a symmetry with respect to the free term , with an implicit circulant symmetry . ] , @xmath314 always contains roots in pairs @xmath315 and @xmath316 . for such polynomials eq . ( [ eq : cyclic - orthogonality - guarantee ] ) always fails ( including the special case of @xmath317 which has only one root @xmath318 ) . the algebraic condition ( [ eq : linear - cyclic - condition ] ) on check polynomials for linear cyclic codes makes them simpler to implement but also dramatically restricts their number . in particular , the general counting approach to finding cws codes [ see sec . [ sec : gv ] ] , where one first chooses a graph , and then searches for a suitable binary code can hardly be applied to cyclic cws codes . even for classical binary cyclic codes , there are no counting arguments known to date that yield asymptotically good codes , let alone the stronger gv bound ( see research problem 9.2 in ref . ) . also , long bch codes one of the major subclasses of cyclic codes are asymptotically bad and have a slowly declining relative distance @xmath319 for any code rate @xmath146 . on the other hand , binary cyclic codes often achieve the best known parameters ( exceeding the gv bound ) on short lengths @xmath320 . thus , using simple cyclic codes in quantum design can yield both good parameters and feasible implementation on the short blocks . to better evaluate code distance of single - generator quantum cyclic codes ( [ eq : gf4-polynomial ] ) , we will modify our counting approach of sec . [ sec : gv ] and begin with a binary cyclic code . namely , we will fix some parity - check polynomial @xmath297 with a desired degree @xmath6 among the binary factors of @xmath299 . then we will search for a polynomial @xmath298 , either corresponding to a cyclic graph [ see eq . ( [ eq : symmetric - polynomial ] ) ] , or satisfying the more general orthogonality condition ( [ eq : cyclic - orthogonality ] ) . however , this transition will show that the parameters of quantum codes generated this way strongly depend on the chosen binary code . we will concentrate exclusively on the binary codes with irreducible generator polynomial @xmath311 . we will show that the distance of such a cyclic cws code is limited from below by the gv bound ( or the variants thereof ) and from above by the distance of the classical cyclic code . since gv bound always produces asymptotically good codes , the parameters of our quantum codes will be mostly limited ( at least , for long blocks ) by their binary counterparts . we begin with analyzing the condition ( [ orthogonality ] ) for a cyclic cws code . for a vector @xmath321 to be in @xmath322 this condition can be rewritten in terms of the corresponding polynomials , @xmath323=0\:\mod x^{n}-1\:,\ ] ] where the coefficients of the ( reversed for notational convenience ) polynomial @xmath324 are given by the components of the vector @xmath321 . since binary @xmath297 divides @xmath299 , we can rewrite this in terms of the corresponding generator polynomial @xmath308 for the binary code @xmath10 , @xmath325 now , if @xmath326 is mutually prime with @xmath311 , eq . ( [ eq : poly - orthogonality - two ] ) can be just solved for @xmath298 . in this case the answer is unique @xmath327 $ ] . on the other hand , multiple solutions for @xmath298 are possible when @xmath328\neq1 $ ] . in this work , we avoid the complications caused in the latter case with multiple factors , distance estimates of quantum codes lead to the estimates of weight spectra of classical cyclic codes which contain the code generated by @xmath311 , which is beyond the scope of this work . ] and only consider irreducible polynomials @xmath311 . overall , for any irreducible @xmath311 and any @xmath329 with @xmath330 and @xmath331 , eq . ( [ eq : poly - orthogonality - two ] ) has a unique solution for @xmath298 such that @xmath332 . respectively , there is no more than one additive quantum code such that @xmath333 . generally , only some of thus obtained @xmath298 correspond to self - orthogonal codes , see eq . ( [ eq : cyclic - orthogonality ] ) . below we complete the greedy argument by counting the polynomials @xmath298 corresponding to self - orthogonal codes , eq . ( [ eq : cyclic - orthogonality ] ) . we consider separately the case when the irreducible polynomial @xmath311 is palindromic [ see eq . ( [ eq : palindromic ] ) ] in lemma [ lemma : gv - cyclic ] below , and when it is not in [ lemma : cyclic - one ] consider a cyclic binary code @xmath334 $ ] with the generator polynomial @xmath311 which is both irreducible and non - palindromic , @xmath335 . then there exists a single - generator additive cyclic code @xmath76 $ ] with distance @xmath336 restricted by both the distance @xmath337 of the binary code and the following variant of gv bound @xmath338 the non - palindromic generator polynomial @xmath311 is one of the factors in @xmath299 which also contains its reciprocal , @xmath339 . this implies that the corresponding parity - check polynomial @xmath297 satisfies eq . ( [ eq : cyclic - orthogonality ] ) . further , since @xmath311 is irreducible , the solution @xmath298 of eq . ( [ eq : poly - orthogonality - two ] ) is unique assuming @xmath340 and @xmath341 , which gives the exponential term in the r.h.s . of eq . ( [ eq : gv - cyclic - one ] ) . ( [ eq : gv - cyclic - one ] ) improves on the standard gv inequality ( [ eq : sump ] ) by discarding a few sets of vectors . the first set are vectors with @xmath342 , which implies @xmath343 . the second set are vectors with @xmath344 , which all give @xmath345 . the third set are non - zero vectors with @xmath346 , which can never be in @xmath214 corresponding to the generator ( [ eq : gf4-polynomial ] ) . finally , note that any error vector of weight @xmath246 produces at least @xmath347 different cyclic shifts . all of these cyclic shifts give the same polynomial @xmath298 and can be discounted . the condition @xmath348 comes from lemma [ lemma : upper - binary ] . the lhs of bound ( [ eq : gv - cyclic - one ] ) limits the number of cyclic classes for @xmath349-ary vectors @xmath329 of weight @xmath350 most of these vectors have the maximum period @xmath351 therefore , it can be proven that the term @xmath352 in ( [ eq : gv - cyclic - one ] ) can be replaced with the smaller term that rapidly tends to 1/n for large n. in turn , bound ( [ eq : gv - cyclic - one ] ) adds about @xmath353 information qubits to bound ( [ eq : sump ] ) but tends to the standard quantum gv bound ( [ eq : gv1 ] ) as @xmath254 . in the following example , this bound coincides with inequality @xmath354 , which uniquely sets code distance @xmath2 . the family of the binary codes with the parameters @xmath355 $ ] , @xmath356 , constructed in ref . , has irreducible non - palindromic generator polynomials as required in lemma [ lemma : cyclic - one ] . for @xmath357 the sum in eq . ( [ eq : gv - cyclic - one ] ) has just one term at @xmath358 ; for @xmath0 odd @xmath359 . explicit calculation confirms that the parameters of these codes satisfy inequality ( [ eq : gv - cyclic - one ] ) , which proves the existence of single - generator cyclic quantum codes with exactly the same parameters , @xmath360 $ ] , but not necessarily cyclic cws codes . the smallest of these codes , @xmath361 $ ] , corresponds to a polynomial @xmath362 ( unique up to a reversal ) and can be obtained from an order-@xmath363 circulant graph corresponding to @xmath364 . this particular combination of parameters gives the best existing code@xcite . according to the bch bound @xcite , a cyclic code has distance @xmath365 ( @xmath198 is the `` _ designed _ '' distance ) if the corresponding generator polynomial @xmath311 has @xmath197 consecutive roots , e.g. , @xmath366 , where @xmath315 is the primitive @xmath367th root of unity . a polynomial @xmath368 which has root @xmath315 , necessarily has @xmath246 distinct roots @xmath369 for all @xmath370 if @xmath246 is the smallest number such that @xmath371 . we say that the code has zeros @xmath372 , where exponents @xmath145 form the set @xmath373 the code generated by @xmath368 has designed distance @xmath172 if @xmath374 or , equivalently , if @xmath375 for some @xmath246 . the polynomial @xmath368 is non - palindromic if @xmath376 . we can further obtain codes with irreducible non - palindromic generators and designed distances @xmath281 , @xmath377 , etc . , by imposing additional conditions , e.g. , @xmath378 , @xmath379 , etc . the values of @xmath380 for which this is possible form an infinite set , @xmath381 , where the subscripts indicate the designed distances . in fact , the first three codes represent the well known quadratic - residue codes with the higher distances ( exceeding the bch bound ) equal to 7 , 11 , and 11 , respectively . gv bound proves the existence of additive quantum cws codes with the parameters @xmath382 $ ] , @xmath383 $ ] , @xmath384 $ ] , @xmath385 $ ] , @xmath386 $ ] , @xmath387 $ ] , etc . the first three codes have the parameters as good as any known codes with such @xmath0 and @xmath6 . now let us consider the case of a palindromic polynomial @xmath311 . first , we prove [ lemma : symmetric - q - cws ] consider a binary code @xmath10 generated by a palindromic polynomial @xmath311 such that @xmath388 . then any quantum code ( [ eq : gf4-polynomial ] ) which satisfies self - orthogonality condition ( [ eq : cyclic - orthogonality ] ) is equivalent to a cyclic cws code with a symmetric polynomial @xmath298 , see eq . ( [ eq : symmetric - polynomial ] ) . the corresponding check polynomial @xmath297 is symmetric , thus the condition ( [ eq : cyclic - orthogonality ] ) can be rewritten as @xmath389 the condition @xmath388 guarantees that the palindromic polynomial @xmath311 has odd weight and even degree @xmath390 , in which case the `` central '' monomial @xmath391 has non - zero coefficient @xmath392 . given the block length @xmath0 , let us choose an equivalent code [ see eq . ( [ eq : circulant - r - equivalence ] ) ] with @xmath298 such that the coefficients @xmath393 the coefficients of the polynomial in the l.h.s . of eq . ( [ eq : cyclic - orthogonality - symmetric ] ) satisfy the same condition , except for the term @xmath394 , which has coefficient @xmath395 . the coefficients are arranged in such a way that the l.h.s . of eq . ( [ eq : cyclic - orthogonality - symmetric ] ) can only equal zero or @xmath396 . however , the latter possibility can be excluded by comparing the corresponding coefficients of the free term @xmath397 . the only remaining case corresponds to a symmetric @xmath298 . it is now clear that for a palindromic irreducible generator polynomial @xmath398 , one should reduce the count in the r.h.s.of eq . ( [ eq : gv - cyclic - one ] ) by replacing @xmath126 with @xmath399 , the number of symmetric polynomials that satisfy eq . ( [ eq : zero - set ] ) . this gives the following version of gv bound , @xmath400 while the resulting estimate is much weaker than the gv bound ( [ eq : gv - cyclic - one ] ) , it still gives asymptotically good codes . a better ( especially , for small @xmath2 ) bound is given in the following consider a cyclic binary code @xmath334 $ ] with @xmath401 and the generator polynomial @xmath311 which is both palindromic and irreducible . then there exists a cyclic cws code @xmath76 $ ] with the distance @xmath402 , where @xmath403 : @xmath404 [ lemma : gv - cyclic ] the restriction on the distance guarantees that @xmath398 , and therefore @xmath311 satisfies the conditions of lemma [ lemma : symmetric - q - cws ] ; in particular , @xmath79 is even . the inequality just corresponds to symmetric polynomials @xmath298 and the errors that are also symmetric , @xmath405 . the statement of the lemma follows from the fact that for any general error @xmath406 , there is a symmetric error @xmath407 whose weight is even and is limited by @xmath408 . among classical codes , the largest distance is obtained for the repetition codes , with the parameters @xmath409 $ ] . the parity - check polynomial is @xmath410 ; the generator polynomial @xmath411 is irreducible and palindromic for @xmath412 , and for all @xmath413 that satisfy the condition @xmath414 , where @xmath415 is the multiplicative order of @xmath16 modulo @xmath416 . this includes the following @xmath417 : @xmath418 lemma [ lemma : gv - cyclic ] shows that for @xmath0 from the set ( [ eq : prime - set ] ) , additive cyclic cws codes with parameters @xmath419 $ ] exist , where @xmath420 is obtained from eq . ( [ eq : gv - cyclic - two ] ) with @xmath421 . asymptotically , at large @xmath0 , this corresponds to cyclic codes with the relative distance given by half of that given by eq . ( [ eq : gv - zero - rate ] ) . ( cyclic analogs of the toric code ) in the setting of the previous example , cyclic cws codes @xmath166 $ ] , @xmath422 $ ] , @xmath423 $ ] , @xmath424 $ ] with @xmath425 were obtained numerically . the corresponding graph - state generators are @xmath426 for @xmath427 , @xmath428 for @xmath429 , etc . we obtain a family of cyclic codes with the weight-@xmath349 generators , @xmath430 , @xmath431 , etc . codes with generators @xmath432 , @xmath433 , @xmath434 , and @xmath435 have the same parameters ( the corresponding graphs are somewhat more complicated ) . the latter family can be generalized to codes with @xmath436 , @xmath421 , @xmath437 , @xmath438 ; the corresponding stabilizer generators @xmath439 having @xmath440 identity operators separating @xmath441 and @xmath442 . these cyclic codes are related to generalized toric codes@xcite ; the square - lattice qubit layout preserving the circulant symmetry is illustrated in fig . [ fig : skewed ] for @xmath443 . [ ex : toric - cws ] ] $ ] with generators @xmath444 , and a generalized toric code on square lattice . only qubits within the dashed square are participating in the code ; periodic boundary conditions are assumed . the vertical dashed line indicates a topologically nontrivial chain of errors which limits the distance of the code : @xmath445 is equivalent to @xmath446 . right : same for the code @xmath422 $ ] with the generator @xmath447.,title="fig : " ] ] $ ] with generators @xmath444 , and a generalized toric code on square lattice . only qubits within the dashed square are participating in the code ; periodic boundary conditions are assumed . the vertical dashed line indicates a topologically nontrivial chain of errors which limits the distance of the code : @xmath445 is equivalent to @xmath446 . right : same for the code @xmath422 $ ] with the generator @xmath447.,title="fig : " ] ( cws codes from @xmath6 copies of the repetition code ) let us take the binary code @xmath10 to be formed by @xmath6 copies of the repetition code with the distance @xmath448 . then the block size is @xmath449 , and the check polynomial is @xmath450 . the generator polynomial @xmath451 is always palindromic ; it is also irreducible if @xmath416 belongs to the set ( [ eq : prime - set ] ) , while @xmath452 , @xmath453 . for sufficiently large @xmath0 , lemma [ lemma : gv - cyclic ] gives asymptotically good codes with @xmath454 . since these parameters can not exceed those of the binary code @xmath10 which correspond to @xmath455 ( thus @xmath456 ) , for these values of @xmath416 and @xmath457 , there exist cyclic cws codes with the parameters of the corresponding binary code , @xmath458 $ ] . this prediction is readily verified empirically , see table [ tab : quantum - repetition ] . note that , as in the example [ ex : toric - cws ] , many of these codes have stabilizer generators with small weight . [ ex : k - copies ] .families of the cyclic codes obtained numerically from @xmath6 copies of the classical repetition code , @xmath459 , corresponding to @xmath460 . the expected distance saturation , @xmath461 , is reached already at @xmath462 , even for @xmath6 and @xmath416 different from those in example [ ex : k - copies ] . the operator strings in the last column are representative graph - state generators ( the remaining generators are obtained by cyclic shifts ) . [ cols="^,^,^,^,^,^,^",options="header " , ] in this paper , we analyze how the general cws framework can facilitate the search of the additive quantum codes with reasonably good parameters . unlike complete optimization of cws codes@xcite , which involves all non - isomorphic lc - inequivalent order-@xmath0 graphs and all binary codes of length @xmath0 , here one can independently pick a suitable graph @xmath8 , and then search for a linear binary code @xmath10 that can correct the error patterns induced by @xmath8 , see eq . ( [ orthogonality ] ) . the choice of the graph is discussed in sec . [ sec : upper - bound ] . in the simplest case of pure codes , one has to pick a graph with a sufficiently large graph - state distance @xmath184 . assuming that a regular graph is being sought in this design , we consider graphs with minimal vertex degrees @xmath463 or more . after the graph is chosen , the second step involves the search of an appropriate binary code . here in sec . [ sec : gv ] we prove the existence of the binary codes that give good quantum cws codes . the corresponding lower bound on the distance is given by the quantum gilbert - varshamov bound ( [ eq : sump ] ) . note that while this bound is proved for a given graph , it is the same bound that holds for a generic stabilizer code . our results show that by restricting the graph @xmath8 of a cws code to regular lattices , one can lower the complexity of the code search and still obtain codes with relatively good parameters . on the other hand , the graph structure could be mapped directly to a physical qubit layout . therefore , such codes can simplify both the hardware design and the error - correcting procedures , which will easily admit the property of translational invariance . an unexpected byproduct of this work is the discovery of a previously unexplored family of single - generator quantum cyclic codes ( sec . [ sec : one - gen ] ) . these codes are relatively easy to construct , and they are plentiful . we construct ( or prove the existence ) of several simple families of such codes that have unbounded distances . these include cyclic cws codes with weight-@xmath349 stabilizer generators , which turned out to be toric codes in disguise ( example [ ex : toric - cws ] ) , as well as a code family with the parameters of generalized repetition codes , @xmath464 $ ] ( example [ ex : k - copies ] ) . the main advantage of these families is a simple structure of their stabilizers . we are grateful to m. grassl for helpful comments . this work was supported in part by the army research office through the grant no . w911nf-11 - 1 - 0027 ( lp & i d ) , and by nsf through the grant no . 1018935 ( lp ) . 57ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty http://link.aps.org/abstract/pra/v52/pr2493 [ * * , ( ) ] http://dx.doi.org/10.1103/physreva.55.900 [ * * , ( ) ] http://dx.doi.org/10.1103/physreva.54.3824 [ * * , ( ) ] http://www.sciencemag.org/cgi/content/abstract/279/5349/342 [ * * , ( ) ] http://dx.doi.org/10.1103/physreva.66.032304 [ * * , ( ) ] http://dx.doi.org/10.1063/1.1499754 [ * * , ( ) ] http://dx.doi.org/10.1103/physreva.68.042322 [ * * , ( ) ] http://link.aps.org/abstract/pra/v69/e042314 [ * * , ( ) ] http://arxiv.org/abs/quant-ph/0402196 [ * * , ( ) ] , http://arxiv.org/abs/quant-ph/0506126 [ `` , '' ] ( ) , http://dx.doi.org/10.1038/nature03350 [ * * , ( ) ] http://dx.doi.org/10.1103/physreva.71.042322 [ * * , ( ) ] http://link.aps.org/abstract/prl/v98/e190504 [ * * , ( ) ] http://link.aps.org/abstract/prl/v85/p5452 [ * * , ( ) ] http://dx.doi.org/10.1038/414883a [ * * , ( ) ] link:\doibase 10.1038/nature01336 [ * * , ( ) ] http://dx.doi.org/10.1038/nature03074 [ * * , ( ) ] link:\doibase 10.1038/nphys1032 [ ( ) , 10.1038/nphys1032 ] http://dx.doi.org/10.1007/s11128-009-0105-1 [ * * , ( ) ] link:\doibase 10.1038/nature09071 [ * * , ( ) ] _ _ , http://arxiv.org/abs/quant-ph/9705052[ph.d . thesis ] , ( ) http://dx.doi.org/10.1109/18.681315 [ * * , ( ) ] @noop `` , '' ( ) , http://link.aps.org/abstract/prl/v99/e130505 [ * * , ( ) ] http://dx.doi.org/10.1109/tit.2008.2008136 [ * * , ( ) ] http://dx.doi.org/10.1063/1.3086833 [ * * , ( ) ] http://link.aps.org/abstract/pra/v78/e062315 [ * * , ( ) ] http://link.aps.org/abstract/prl/v101/e090501 [ * * , ( ) ] http://arxiv.org/abs/0709.1780 [ `` , '' ] ( ) , in http://dx.doi.org/10.1109/isit.2008.4594996[__ ] ( , ) pp . in http://dx.doi.org/10.1109/itw.2008.4578694[__ ] ( , ) pp . http://link.aps.org/abstract/pra/v79/e050306 [ * * , ( ) ] http://dx.doi.org/10.1103/physrevlett.104.190501 [ * * , ( ) ] link:\doibase 10.1103/physreva.81.052337 [ * * , ( ) ] http://arxiv.org/abs/quant-ph/9707021 [ * * , ( ) ] http://link.aps.org/abstract/pra/v78/e042303 [ * * , ( ) ] @noop _ _ ( , , ) http://dx.doi.org/10.1103/physreva.79.062322 [ * * , ( ) ] link:\doibase 10.1103/physreva.54.1098 [ * * , ( ) ] http://dx.doi.org/10.1103/physreva.54.4741 [ * * , ( ) ] http://arxiv.org/abs/quant-ph/9703016 [ `` , '' ] ( ) , in @noop _ _ ( , ) p. , @noop * * , ( ) , http://dx.doi.org/10.1103/physreva.69.022316 [ * * , ( ) ] in http://arxiv.org/abs/quant-ph/0602096[__ ] , vol . ( , , ) pp . http://dx.doi.org/10.1016/0012-365x(93)90357-y [ * * , ( ) ] http://dx.doi.org/10.1103/physrevlett.78.405 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.77.198 [ * * , ( ) ] link:\doibase 10.1109/tit.2004.838088 [ * * , ( ) ] @noop _ _ ( , , ) in http://www2.research.att.com/~njas/doc/self.pdf[__ ] , vol . , ( , , ) pp . @noop * * , ( ) http://arxiv.org/abs/quant-ph/9811052 [ `` , '' ] http://arxiv.org/abs/quant-ph/9810055 [ `` , '' ] link:\doibase 10.1103/physreva.76.012305 [ * * , ( ) ]

using self - consistent , physically motivated models , we investigate the x - ray obscuration in 19 type 2 [ oiii ] 5007 selected agn , 9 of which are local seyfert 2 galaxies and 10 of which are type 2 quasar candidates . we derive reliable line - of - sight and global column densities for these objects , which is the first time this has been reported for an agn sample ; 4 agn have significantly different global and line - of - sight column densities . five sources are heavily obscured to compton - thick . we comment on interesting sources revealed by our spectral modeling , including a candidate `` naked '' sy2 . after correcting for absorption , we find that the ratio of the rest - frame , 2 - 10 kev luminosity ( l ) to l}$ ] is 1.54 0.49 dex which is essentially identical to the mean type 1 agn value . the fe k luminosity is significantly correlated with l}$ ] , but with substantial scatter . finally , we do not find a trend between l and global or line - of - sight column density , between column density and redshift , between column density and scattering fraction or between scattering fraction and redshift .

1404.0012

supermassive black holes in the centers of galaxies grow by accretion in which they are observed as active galactic nuclei ( agn ) . x - ray emission is thought to originate in a corona surrounding the accretion disk where optical and ultraviolet photons from the disk are inverse - compton scattered to higher energies . according to conventional unified models @xcite , this central engine is enshrouded by a circumnuclear obscuring medium of dust and gas which has a toroidal geometry . the inclination of this system to the observer s line of sight therefore determines the observable properties . in this classification scheme , type 1 agn represent a face - on system , allowing an unobscured view of the central engine where broad and narrow optical emission lines are apparent , as well as the ultraviolet continuum , mid - infrared emission from the circumnuclear obscuring matter , and x - ray emission from the corona . conversely , the type 2 systems are oriented so that the accretion disk and corona are hidden from the observer s sight line , blocking the optical and ultraviolet continuum and optical broad lines , and possibly significantly impacting the observed x - ray emission . the physical processes that x - ray photons undergo are imprinted in the observed spectrum . an x - ray photon that is injected from the corona into the obscuring medium can pass unimpeded directly into our line of sight , be absorbed , or undergo one or more compton - scatterings , with multiple scatterings becoming more likely if the column density is compton - thick ( i.e. , n@xmath4 @xmath5 , where @xmath6 is the thompson cross - section for electron scattering ) . the probabilities of these outcomes depend on the column density of the reprocessor and the incident energy of the x - ray photon . fluorescent line emission from continuum photons ejecting inner shell electrons from atoms or ions in the obscuring medium can potentially be a dominant component of the observed spectrum . agn continuum photons can also not interact with the absorber by escaping through the opening of the torus and/or passing through gaps in a clumpy medium if the absorber is non - uniform and patchy , and then scatter off a more distant optically - thin medium , entering our line of sight . simple _ ad hoc _ x - ray models that parametrize the emission with an absorbed power law model or partial covering model will fail to capture the detailed physics inherent in these systems , especially when the column density exceeds @xmath7 @xmath5 , i.e. , when compton scattering becomes an important factor . even models that treat relativistic reflection off an accretion disk are unsuitable for these systems as they assume a column density that is infinite and a disk geometry does not adequately describe the absorber that reprocesses the x - ray emission . additionally , such modeling is one - dimensional , which does not allow a finite column density out of the line of sight to be measured , even though the fe k fluorescence line can provide information about the global matter distribution in which it is produced . with recent advances in spectral modeling that self - consistently treat compton scattering , fluorescent line emission , and scattering off a distant medium @xcite , we can now delve into the details of the obscuration in agn and obtain physically meaningful constraints on inclination angle , line - of - sight and global column density , and the percentage of scattered agn light . here we use physically motivated x - ray models to study the circumnuclear obscuration of a sample of seyfert 2 galaxies ( sy2s ) and type 2 quasar ( qso ) candidates selected based on their [ oiii ] 5007 line emission , where the sy2s are local ( @xmath8 ) , lower luminosity obscured agn compared with quasars , which have bolometric luminosities exceeding @xmath9 erg s@xmath10 . the [ oiii ] 5007 line forms in the agn narrow line region , 100s of parsecs above and below the circumnuclear obscuring medium , and it is primarily ionized by the agn continuum , making it a reliable indicator of intrinsic agn power @xcite . these samples are thus unbiased with respect to the circumnuclear obscuration while x - ray selected samples are inherently biased against the most obscured sources as the combined effects of photoelectric absorption and compton scattering modify the observed x - ray emission . the sy2 and qso samples have previously been analyzed and published in @xcite and @xcite , respectively , using simple power law or double power law models to fit the spectrum . the amount of implied obscuration was determined by normalizing the observed x - ray flux by the intrinsic luminosity ( [ oiii ] for the sy2 and qso sample , and by the mid - infrared [ oiv ] 26 @xmath11 m line and mid - infrared continuum for the sy2 sample ) and analyzing the equivalent width of the fe k@xmath3 line , which can exceed 1 kev in compton - thick sources as the fluorescent line in the reflection spectrum is superimposed on a depressed continuum ( e.g. , * ? ? ? * ; * ? ? ? @xcite also estimate the absorber column density by comparing the observed ratio of x - ray to [ oiii ] flux to a simulated unabsorbed ratio , using the unabsorbed ( type 1 agn ) values from @xcite as the basis for simulating 1000 random draws per source , and determining the column density that can account for the attenuation between the absorbed and simulated unabsorbed x - ray fluxes . in neither work is the fitted column density used as representative of the true obscuration as this is poorly determined with the simple models used . instead , the observed 2 - 10 kev fluxes and fe k@xmath3 ews , which are largely model - independent , were used as diagnostics of the column density . in both works , these obscuration proxies suggested that a majority of the sy2s and type 2 qso candidates are heavily obscured , and possibly compton - thick . in this analysis , we use the models from @xcite and ( * ? ? ? * mytorus ) to unveil the physical properties of the obscuration responsible for reprocessing the emission from the central engine within a spherical and toroidal geometry , respectively . these models treat the transmitted continuum , compton scattered emission and fluorescent line emission self - consistently , allowing for physically meaningful measurements of the global and line - of - sight column densities which are poorly determined with _ ad hoc _ models . we apply these model to the subset of [ oiii]-selected sy2s from @xcite and qso2 candidates from @xcite that have adequate signal - to - noise x - ray spectra to constrain the model parameters . we comment on what we can learn about the circumnuclear obscuration with the present data , including reliable measurements of the global as well as line - of - sight n@xmath12 , and highlight two interesting sources as revealed by our x - ray analysis . as the fe k@xmath3 emission is accommodated self - consistently within the global obscuring medium in these models , we overcome the shortcomings of previous analyses that used simple power - law modeling to measure the fe k@xmath3 emission either in front of or behind the line - of - sight absorbing screen . we then test whether the fe k@xmath3 luminosity indeed traces the intrinsic luminosity as reported in previous works ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? we also investigate whether there are differences between these qso2s and sy2s in terms of their column densities and scattering fractions and if there is a relationship between the column density and agn luminosity as posited by the `` receding torus '' model ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . throughout , we refer to sources as mildly obscured if the circumnuclear column density is below @xmath13 @xmath5 , moderately obscured for @xmath14 @xmath15 @xmath5 , heavily obscured for 10@xmath16 @xmath17 @xmath5 , and compton - thick if the column density exceeds @xmath18 @xmath5 . we adopt a cosmology of h@xmath19=70 km mpc @xmath10 s@xmath10 , @xmath20=0.27 , and @xmath21=0.73 . the sy2 sample and type 2 qso candidate sample are both selected from sdss based on their [ oiii ] 5007 flux and luminosity , with full details given in @xcite and @xcite . sy2s are identified using the bpt diagram @xcite , a line ratio diagnostic plot of [ nii]/h@xmath3 vs. [ oiii]/h@xmath22 that separates star - forming from active galaxies @xcite . the 20 objects lying in the sy2 locus of the bpt diagram with the highest [ oiii ] flux ( @xmath23 erg s@xmath10 @xmath5 ) with @xmath8 from sdss data release 4 @xcite comprise the parent sample . of these 20 , we were awarded _ xmm - newton _ observing time for 15 , and another 2 objects were available in the archive . the type 2 quasar candidates are selected following the procedures of @xcite who identified sources at redshifts @xmath24 lacking broad emission lines yet having strong narrow emission lines where the luminosity of the [ oiii ] line exceeded 10@xmath25 l@xmath26 . more recently , @xcite used the same technique on data release 6 of the sdss @xcite , which expanded the @xcite sample four - fold and included objects at @xmath27 , garnering 887 type 2 qso candidates . @xcite found 71 of these sources in the _ xmm - newton _ and _ chandra _ archive , with 54 having enough counts for at least a crude spectral fitting . one of these objects , sdss j123843.43 + 092736.6 , is part of the sy2 sample . though the parent sy2 and qso2 samples are unbiased with respect to their x - ray properties , the detailed modeling here requires relatively good signal - to - noise . we focus on objects that have enough detected photons to bin by at least 15 counts per bin . these sub - samples are therefore neither complete nor unbiased : we may be missing a significant fraction of the compton - thick sources from the parent samples that have weak observed x - ray emission . we therefore refrain from extrapolating our results to the general agn population . we note , however , that of the sy2s from @xcite not included in this analysis , all show evidence of heavy obscuration ( high fe k@xmath3 ew and/or low l@xmath28/l@xmath1}$ ] compared with sy1s ) . a little over half of the type 2 qsos not analyzed here are marked as likely compton - thick in @xcite since the 1@xmath29 confidence interval on the simulated @xmath30 values exceed 1.6@xmath31 @xmath5 . nine of the 17 [ oiii]-selected sy2s with x - ray coverage have an adequate number of counts for more detailed spectral fitting and are examined in this work ( table [ sy2_sample ] ) . twelve of the 54 type 2 qsos have enough counts to be useful for our analysis , though we exclude the sy2 ( sdss j123843.43 + 092736.6 ) and one source ( sdss j141120.52 + 521210.0 , a.k.a . 3c 295 ) which is at the center of a galaxy cluster and has prominent hot spots which complicates correctly modeling emission from the central engine . the 10 type 2 qso candidates used in this analysis are listed in table [ qso2_sample ] . we note that 3 of these sources have both _ xmm - newton _ and _ chandra _ observations , but as noted in @xcite , there is extended x - ray emission due to radio jets ( j083454.89 + 553411.1 ) or star formation ( sdss j090037.09 + 205340.2 and sdss j091345.48 + 405628.2 ) , so we utilize the _ chandra _ data to isolate the quasar - only emission . lrlrrrl mrk 0609 & 03 25 25.4 & -06 08 37 & 0.034 & 0402110201 + ic 0486 & 08 00 21.0 & 26 36 49 & 0.027 & 0504101201 + sdss j082443.28 + 295923.5 & 08 24 43.3 & 29 59 24 & 0.025 & 0504102001 + cgcg 064 - 017 & 09 59 14.8 & 12 59 16 & 0.034 & 0504100201 + sbs 1133 + 572 & 11 35 49.1 & 56 57 08 & 0.051 & 0504101001 + mrk 1457 & 11 47 21.6 & 52 26 58 & 0.049 & 0504101401 + sdss j115704.84 + 524903.6 & 11 57 04.8 & 52 49 04 & 0.036 & 0504100901 + sdss j123843.43 + 092736.6 & 12 38 43.4 & 09 27 37 & 0.083 & 0504100601 + cgcg 218 - 007 & 13 23 48.5 & 43 18 04 & 0.027 & 0504101601 + lrlrrrll sdss j083454.89 + 553411.1 & 0.241 & _ chandra _ & 04940 + sdss j083945.98 + 384319.0 & 0.425 & _ xmm - newton _ & 0502060201 + sdss j090037.09 + 205340.2 & 0.236 & _ chandra _ & 10463 + sdss j091345.48 + 405628.2 & 0.441 & _ chandra _ & 10445 + sdss j093952.74 + 355358.0 & 0.137 & _ xmm - newton _ & 0021740101 + sdss j103408.59 + 600152.2 & 0.051 & _ xmm - newton _ & 0306050701 + sdss j103456.40 + 393940.0 & 0.151 & _ xmm - newton _ & 0506440101 + sdss j122656.40 + 013124.3 & 0.732 & _ xmm - newton _ & 0110990201 + sdss j134733.36 + 121724.3 & 0.120 & _ chandra _ & 00836 + sdss j164131.73 + 385840.9 & 0.596 & _ xmm - newton _ & 0204340101 + the x - ray data are reduced with _ xassist _ version 0.9993 @xcite . this task processes the raw events files , calling the relevant routines from xmmsas and ciao for the _ xmm - newton _ and _ chandra _ observations , respectively . the data are filtered to omit time periods of background flaring . spectra and response files are extracted for user - inputted sources ( i.e. , at the coordinates of the [ oiii]-selected sy2s and type 2 qso candidates ) , with background spectra extracted from an annulus around the object of interest , or in the case of crowded fields , from a circular aperture within a source free region nearby . for the sources with _ xmm - newton _ observations , the pn , mos1 , and mos2 spectra are fit simultaneously . the spectra are initially fit in xspec with the models from @xcite which self - consistently describes compton scattering and fluorescent line emission within a spherical obscuring medium . we include a model component for scattering of the agn continuum off a distant optically thin medium if the data require it . although the @xcite model assumes full global covering , leakage of the intrinsic continuum of a few percent or less does not significantly break the self - consistency of the model . we also fit the mytorus model to the spectra , which we run in both the coupled and decoupled modes as described in detail in @xcite ; in the latter mode we are able to derive separate line - of - sight column densities ( n@xmath32 from the global average ( n@xmath33 ) . with both models , we test for the presence of thermal emission , which steepens the spectrum , by adding an _ apec _ model component ( see * ? ? ? * for a more thorough discussion of the interplay between thermal and non - thermal emission for sy2s in general ) . an initial best fit photon index ( @xmath34 ) exceeding 2 likely indicates that thermal emission impacts the observed soft x - ray spectrum , where adding an _ apec _ component better constrains our modeling of the agn continuum . @xcite performed a series of monte carlo simulations to emulate compton scattering of photons injected from a central x - ray point source into a spherical distribution of obscuring matter , with photoelectric cross sections from @xcite and cosmic abundances from @xcite . the obscuring medium is assumed to be neutral . this model self - consistently treats fluorescent line emission , using the fluorescent yields of @xcite . the total elemental abundances and iron abundances are free parameters , but we leave these frozen at solar . we also include model components for scattering of the intrinsic continuum off of distant material , as well as a gaussian convolution model for velocity broadening of the fe k complex ( @xmath35(@xmath36/6 kev)@xmath37 , with @xmath3 fixed at 1 , i.e. , the velocity width is independent of energy ) , if required by the spectrum . the model is then @xmath38 where the first constant accounts for the cross - calibration among the _ xmm - newton _ pn , mos1 , and mos2 detectors , n@xmath39 is the ( fixed ) galactic absorption , and @xmath40 is the fraction of the agn continuum that is scattered into our line of sight from a distant optically thin medium . in xspec `` speak , '' this translates into @xmath41 if scattering off a distant medium is required by the data ( i.e. , the _ const * zpow _ components are needed to adequately model the spectra ) , the photon index ( @xmath42 ) and powerlaw normalization are tied to the values from the spherical model . in this set - up , it is the same agn continuum that is reprocessed by the absorber that leaks through the obscuration to be scattered off distant matter that ultimately enters our line of sight . the normalization of the thermal model is an independent parameter . the mytorus model @xcite assumes that an isotropic emitting x - ray source is surrounded by a neutral , uniform absorbing toroidal medium with a half opening angle of 60@xmath43 ( @xmath44/(4@xmath45 ) = 0.5 ) . this medium has the cosmic abundances of @xcite and photoelectric cross sections from @xcite and @xcite . the inclination angle of this system ( @xmath46 ) can range from 0@xmath43 ( face - on ) to 90@xmath43 ( edge - on ) ; angles below 60@xmath43 do not intercept the torus . when a photon is injected from the corona into this medium , it can pass through unimpeded , be absorbed , or undergo one or more compton - scatterings , where the probabilities of these outcomes depend principally on the energy of the photon and the column density of the absorber . within the circumnuclear material , absorption of photons can incite fluorescent line emission , which can escape the medium without scattering , forming the `` core '' of the observed emission line , or be scattered , resulting in a `` compton - shoulder . '' mytorus models the `` zeroth - order continuum '' ( i.e. , transmitted component ) , compton scattered component and fluorescent line emission self - consistently by using pre - defined tables for a power law input spectrum realized from a suite of detailed monte - carlo simulations . these simulations include the k@xmath3 transitions of fe and ni and the k@xmath22 transition of fe , but the ni k@xmath3 line is not included in the spectral - fitting model . the three model components are fit simultaneously in xspec , where the power law index ( @xmath47 ) and normalization , line of sight column density ( n@xmath48 ) , inclination angle ( @xmath46 ) and redshift ( @xmath49 ) are tied together . the relative normalizations among the zeroth - order continuum , compton scattered component and fluorescent line emission may vary due to deviations from model assumptions ( different elemental abundances and/or torus half opening angle , an ionized rather than neutral medium ) and time delays between the transmitted and scattered component and the scattered component and line emission . such differences may not be apparent in the spectrum since it may be observed on scales where the deviations are averaged over , especially between the scattered and fluorescent line emission . these unknowns are embodied in the relative normalizations @xmath50 and @xmath51 , for the compton scattered and line emission respectively , which are parametrized by a multiplicative constant in xspec . we keep @xmath52 and @xmath53 tied together in order to preserve the self - consistency of the compton - scattered continuum and the fe k@xmath3 line emission , within the context of the assumptions of the model ( such as solar fe abundance ) . initially , @xmath50 ( = @xmath51 ) is set to 1 , i.e. , the scattered and line emission have the same normalizations as the zeroth - order continuum . if the data require it ( @xmath54-test probability @xmath55 ) , we untie @xmath50 (= @xmath51 ) from the zeroth - order continuum , but only accept the fit if @xmath50 is constrained . as @xmath50 and @xmath51 parametrize our ignorance of different possible physical scenarios , and disentangling one possibility from another requires additional information that we lack , we only list the best - fit values for reference and refrain from drawing any physical interpretation from these normalizations . in addition to the mytorus zeroth - continuum , compton scattered , and fluorescent line emission models , we include model components for galactic absorption ( n@xmath56 ) , a multiplicative factor for the _ xmm - newton _ pn , mos1 , and mos2 detector cross - normalizations , a gaussian convolution to the fluorescent line spectrum for velocity broadening , and possibly scattering of the continuum from an optically thin medium and/or a thermal component ( if the data require these components ) , similar to the spherical absorption model . our model in coupled mode , i.e. where the line - of - sight and global average column densities are equal , is then @xmath57 where the first constant , n@xmath56 , @xmath58 , and @xmath40 are the same as defined for the spherical geometry obscuration model , and @xmath50 and @xmath51 are defined above . again , the photon indices , power law normalizations , inclination angles , and circumnuclear column densities are tied among all model components , while the thermal component normalization , if present , is independent . in xspec , the relevant commands are @xmath59 here , _ mytorus_ezero_v00.fits _ , _ mytorus_scatteredh200_v00.fits _ , and _ mytl_v000010nep000h200_v00.fits _ are tables for the line - of - sight extinction , compton scattered component , and fluorescent line emission , respectively , with an exponential cutoff energy of 200 kev for the latter two components ; since we work in an energy range far below the termination energy , the choice of cutoff energy has a negligible effect on our results . we can model the compton - scattered continuum from the global matter distribution , which may in general be unrelated to the line - of - sight column density , by decoupling the inclination angle and column density of the mytorus compton - scattered continuum component . such a scenario ( which is no longer restricted to a toroidal geometry ) can be realized if a portion of the observed x - ray emission is viewed after reflection off the back side of the obscuring medium without further interaction with this matter , mimicking a patchy medium where this reflected emission is seen through a hole . the spectrum can in fact be dominated by this unobscured reflection component . in this decoupled case , the inclination angle for the zeroth - order continuum is fixed at 90@xmath43 so that its associated column density represents the line of sight column density only ( n@xmath48 ) . there is a compton scattered component for the far - side reflection , with the inclination angle fixed at 0@xmath43 as it mimics face - on reflection , and its associated column density ( n@xmath33 ) represents the global average . like the coupled case , the normalizations and photon indices among the zeroth order continuum , the compton scattered component , and scattering off a distant medium are tied together . the absorption measured with the spherical model , n@xmath60 refers to the radial distribution of the obscuring material and so the line - of - sight column density from the spherical model is directly comparable to the line - of - sight column density , n@xmath48 , measured by mytorus . when mytorus is run in the default ( i.e. , coupled ) mode , the global column density , n@xmath33 , is the same as the line - of - sight column density . as noted above , mytorus can disentangle the global average column density from the line - of - sight n@xmath12 when run in decoupled mode . for a toroidal distribution , the global column density is the angle average of the line - of - sight column density from the top to the bottom edge of the torus . when assuming an isotropic distribution of input photons , the equatorial column of the torus , n@xmath33 is then equivalent to ( @xmath61)@xmath62n@xmath60 ( see * ? ? ? if we are able to achieve a good fit to the spectra using the spherical absorption model of @xcite , then we accept this model as the preferred model since the added complexity of a toroidal geometry will not allow meaningful constraints on the inclination angle . in some cases where a decent fit ( reduced @xmath63 ) is obtained with the spherical absorption model , mytorus provides a better fit of the fe k complex , in which case we accept the mytorus model fits as a better representation of the circumnuclear obscuration ; indeed , these fits are a statistically significantly improvement ( i.e. , @xmath54-test probability exceeds 3@xmath29 ) . though the spherical model may be preferred , this does not necessarily rule out the possibility that the absorber has a more complex geometry , but rather indicates that this latter possibility is not required to accomodate the spectrum . for the objects where the mytorus fit is preferred , we inspect the results of fitting in coupled and decoupled mode . we accept the decoupled fit only if we obtain constraints on the fit parameters ( namely @xmath52 , n@xmath48 , and n@xmath33 which can be poorly determined in decoupled mode if there is a weak or absent iron line ) or if the fe k complex is better modeled in decoupled mode . we also note that even with the addition of the _ apec _ model component , the upper limit on the best fit @xmath34 is beyond the allowable model range ( 3 for the spherical model , 2.6 for mytorus ) for a handful of objects ; in these cases , we quote the 90% confidence lower limit . for several objects ( mrk 0609 , sdss j115704.84 + 524903.6 , sdss j123843.43 + 092736.6 , sdss j083454.89 + 553411.1 , sdss j103408.59 + 600152.2 ) , the reduced @xmath64 gives a high null hypothesis probability value ( @xmath65 ) which may indicate a poor fit . however , with the exception of mrk 0609 , inspection of the fit residuals shows no systematic offsets which suggests that the preferred model is adequate and that a significant model component is not missing . the nature of the residuals do not compromise our primary objective of measuring reliable column densities for the circumnuclear matter . as we discuss below , mrk 0609 has a typical type 1 x - ray spectrum , lacking x - ray evidence of circumnuclear obscuration . as the spherical absorption and mytorus models require the presence of an x - ray reprocessor with a minimum column density of 10@xmath66 @xmath5 and 10@xmath67 @xmath5 , respectively , they are not appropriate for this object ; a power - law model is sufficient for fitting this source @xcite . the spectral fitting results are reported in tables [ sy2_sph_fits]-[sy2_myt_decoup_fits ] and [ qso2_sph_fits]-[qso2_myt_decoup_fits ] , while the [ oiii ] , fe k@xmath3 and intrinsic , agn - only rest - frame 2 - 10 kev luminosities are listed in tables [ sy2_lums ] and [ qso2s_lums ] for the sy2 and qso2 sample , respectively . all [ oiii ] luminosities , unless otherwise noted , are observed and not corrected for dust extinction as h@xmath3 falls out of the optical bandpass at @xmath68 and we are therefore unable to calculate the balmer decrement ( h@xmath3/h@xmath22 ) for all objects in our sample . the estimation of the fe k@xmath3 luminosity depends on the best - fit model , since the fe k@xmath3 line can not be independently varied in the spherical model . for the mytorus model , the luminosity is integrated in xspec around the rest - frame fe k feature after turning off all model components other than @xmath51 , _ gsmooth _ ( if included ) , and _ mytorusl_. if a source is best - fit with the spherical model , the rest - frame luminosity of the continuum around the fe k feature is estimated and then subtracted from the total luminosity within the same energy range , giving the fe k@xmath3 luminosity . of the 9 [ oiii]-selected sy2s , 7 are best fit with the spherical absorption model , as summarized in table [ sy2_sph_fits ] and shown in figure [ sy2_sph_spec ] ; for none of these sources was the width of _ gsmooth _ constrained and it is therefore not listed in the table . the fe k line was also unresolved for these agn . three of these had 90% upper limits in @xmath34 beyond the maximum allowable limit of the model and are thus listed in table [ sy2_sph_fits ] with their 90% lower limits . one object , sbs 1133 + 572 , has a steep photon index that pegs at the maximum allowed value . inspection of its spectra indicates that the soft emission dominates the spectrum while the hard emission ( @xmath69 kev ) is heavily suppressed . however , the signal - to - noise of the spectrum precludes us from obtaining constraints on possible thermal emission by including an _ apec _ component . conversely , ic 0486 has a shallow fitted photon index . freezing @xmath34 to 1.8 leads to questionable best - fits for @xmath50 ( 11.1@xmath70 ) and f@xmath71 ( 16.7@xmath72 ) , and the line - of - sight and global column densities are nearly equivalent to those from the fit with @xmath34 free . only one source , cgcg 064 - 17 , shows no evidence of scattering of the agn continuum off a distant medium and a mildly obscured spectrum ( @xmath13 @xmath5 ) . this source also does not require a thermal component to better constrain the photon index , consistent with @xcite where we reported that adding a thermal model results in an unconstrained fit ; we note that in modeling cgcg 218 - 007 in that work , using a double power law to model the agn spectrum , a thermal model was unconstrained , yet with the spherical absorption model used here , thermal emission is detected . three sources are heavily obscured and 2 have best fit n@xmath48 ranges that are within the compton - thick regime , sbs 1133 + 572 and 2masx j11570483 + 5249036 . as discussed in detail below , one source , mrk 0609 , is unabsorbed . the remaining 2 objects are well fit with mytorus in decoupled mode ( see figure [ sy2_myt_spec_decoup ] ) , with line of sight column densities that differ from the global average by about an order of magnitude ( table [ sy2_myt_decoup_fits ] ) . we find the global obscuration in ic 0486 is high ( @xmath73 @xmath5 ) , though not compton - thick , while the global absorber in 2 masx j08244333 + 2959238 permits a compton - thick solution . llllllr mrk 0609 & 0.03 & 1.69@xmath74 & @xmath750.01 & ... & ... & 513.29(337 ) + cgcg 064 - 017 & 0.03 & 1.89@xmath76 & 0.76@xmath77 & ... & ... & 225.38(227 ) + sbs 1133 + 572 & 0.01 & @xmath782.84 & 68.7@xmath79 & 1.52@xmath80 & ... & 32.79(30 ) + mrk 1457 & 0.02 & @xmath781.56 & 26.3@xmath81 & 1.86@xmath82 & 0.25@xmath83 & 57.34(49 ) + j1157 + 5249 & 0.02 & @xmath782.11 & 106@xmath84 & 0.40@xmath85 & 0.24@xmath86 & 77.4(47 ) + j1238 + 0927 & 0.01 & 2.32@xmath87 & 36.1@xmath88 & 0.57@xmath89 & 0.86@xmath90 & 170.21(120 ) + cgcg 218 - 007 & 0.01 & 2.12@xmath91 & 53.8@xmath92 & 0.26@xmath93 & 0.60@xmath94 & 64.33(60 ) + llllllllll ic 0486 & 0.03 & @xmath751.43 & 1.21@xmath95 & 10.2@xmath96 & @xmath750.09 & 6.84@xmath97 & 7.09@xmath98 & ... & 388.69(370 ) + j0824 + 2959 & 0.03 & @xmath782.08 & 24.6@xmath99 & 90.2@xmath100 & 0.12@xmath101 & 3.65@xmath102 & 0.43@xmath103 & 0.74@xmath104 & 142.91(129 ) + llll mrk 0609 & 7.72 & 40.77 & 42.58@xmath105 + ic 0486 & 7.30 & 41.21 & 42.77@xmath74 + sdss j082443.28 + 295923.5 & 7.51 & 40.95 & 42.62@xmath106 + cgcg 064 - 017 & 7.77 & 40.22 & 42.48@xmath107 + sbs 1133 + 572 & 7.88 & 40.27 & 42.72@xmath108 + mrk 1457 & 7.80 & 40.56 & 42.61@xmath109 + sdss j115704.84 + 524903.6 & 7.59 & 40.49 & 42.99@xmath110 + sdss j123843.43 + 092736.6 & 8.39 & 41.84 & 43.97@xmath111 + cgcg 218 - 007 & 7.25 & 41.16 & 42.97@xmath112 + for one of the 10 qso2s , sdss j091345.48 + 405628.2 , both the spherical absorption and mytorus model in decoupled mode provided an acceptable fit to the data . though the fe k complex is slightly better accommodated by the mytorus model ( figure [ fek_j0913 ] ) , the improvement is marginal , so we present the results for both model fits . the predicted intrinsic 2 - 10 kev luminosity is comparable between both models ( table [ qso2s_lums ] ) and the line of sight column densities are consistent within the error bars to be heavily obscured to compton - thick . the decoupled mytorus fit does not constrain the upper limit on the global column density . the combination of heavy obscuration with a relatively high scattering fraction precludes us from determining the geometry of the absorber , though clearly leakage of the intrinsic continuum indicates that the obscuration is non - uniform . in the analysis below , we use the global and line - of - sight column densities measured by the decoupled mytorus model , noting that n@xmath48 is consistent with the results from the spherical model . as the possible scattering fraction between the two models shows a relatively wide range , we exclude this source when examining the scattering fraction distributions . of the remaining 9 qso2s , 7 are best fit by the spherical absorption model ( table [ qso2_sph_fits ] , figure [ qso2_sph_spec ] ) , 1 is well - accommodated by the coupled mytorus model ( table [ qso2_myt_coup_fits ] , figure [ qso2_myt_coup ] ) and 1 requires the mytorus model in decoupled mode ( table [ qso2_myt_decoup_fits ] , figure [ qso2_myt_decoup ] ) , as discussed in greater detail below . we measured scattering fractions for 5 of these objects and obtain lower limits on the line of sight column density for 2 sources . for all but 2 objects , we derive constraints on @xmath34 . one source is mildly obscured , 4 are moderately obscured , 2 are heavily obscured and 2 are heavily obscured to compton - thick along the line of sight . though many of the sy2s and qso2s are best fit by the spherical absorption model , this does not necessarily mean that the x - ray reprocessor is a uniform sphere . rather , limited statistics and model degeneracies preclude a more complex geometry from being required to explain the data . additionally , many of the objects best fit by the spherical model show evidence of agn continuum leakage , indicating that openings in the obscuring medium must exist for the photons to escape . lllllllr j0834 + 5534 & 0.04 & 2.08@xmath113 & 0.08@xmath114 & ... & ... & ... & 159.42(122 ) + j0839 + 3843 & 0.03 & @xmath751.60 & 1.67@xmath115 & ... & ... & ... & 24.97(29 ) + j0900 + 2053 & 0.03 & 1.77@xmath116 & 34.5@xmath117 & @xmath750.14 & 9.17@xmath118 & ... & 98.36(103 ) + j0913 + 4056 & 0.01 & 1.90@xmath119 & 62.0@xmath120 & ... & 12.7@xmath121 & ... & 106.57(88 ) + j1034 + 3939 & 0.01 & @xmath781.82 & 52.9@xmath122 & ... & 1.82@xmath123 & 0.77@xmath124 & 70.97(64 ) + j1226 + 0131 & 0.02 & 1.74@xmath125 & 2.33@xmath126 & ... & ... & ... & 84.12(82 ) + j1347 + 1217 & 0.02 & 1.41@xmath127 & 3.50@xmath128 & ... & 3.97@xmath129 & ... & 38.53(46 ) + j1641 + 3858 & 0.01 & 1.35@xmath130 & 2.17@xmath131 & ... & ... & ... & 99.75(81 ) + llllllllllr j1034 + 6001 & 0.01 & 1.87@xmath132 & @xmath7841 & 1 & @xmath7860.2 & ... & 3.05@xmath133 & 0.80@xmath134 & 54.94(32 ) + lllllllllr j0913 + 4056 & 0.01 & 2.03@xmath135 & 97.8@xmath136 & 34.5@xmath137 & ... & 1 & 4.74@xmath138 & ... & 101.61(87 ) + j0939 + 3553 & 0.01 & 1.71@xmath139 & @xmath78167 & 14.4@xmath140 & 0.15@xmath141 & 1 & 0.56@xmath142 & ... & 122.55(111 ) + llll sdss j083454.89 + 553411.1 & 8.69 & 41.61 & 43.52@xmath107 + sdss j083945.98 + 384319.0 & 8.61 & 42.28 & 44.36@xmath143 + sdss j090037.09 + 205340.2 & 8.98 & 42.92 & 44.67@xmath144 + sdss j091345.48 + 405628.2 ( spherical ) & 10.34 & 43.20 & 45.23@xmath145 + sdss j091345.48 + 405628.2 ( mytorus ) & 10.34 & 43.15 & 45.28@xmath146 + sdss j093952.74 + 355358.0 & 8.75 & 42.39 & 44.74@xmath147 + sdss j103408.59 + 600152.2 & 8.81 & 40.72 & 42.96@xmath148 + sdss j103456.40 + 393940.0 & 8.91 & 41.00 & 43.27@xmath149 + sdss j122656.40 + 013124.3 & 9.81 & 42.44 & 45.02@xmath150 + sdss j134733.36 + 121724.3 & 8.65 & 41.04 & 43.69@xmath105 + sdss j164131.73 + 385840.9 & 10.05 & 42.17 & 45.08@xmath151 + though the sdss spectrum from 2001 and the ctio spectrum from 2007 @xcite shows an absence of broad optical emission lines , there is no x - ray evidence to support that the view to the central engine is obscured . not only do we find no absorption above that from our galaxy , the fe k@xmath3 feature is not significant above the 2@xmath29 level @xcite , suggesting a lack of global obscuration . bepposax observations from 2000 also detected no signatures of absorption . additionally , the observed 2 - 10 kev x - ray luminosity of 4@xmath152 erg s@xmath10 is consistent with that of an unabsorbed nucleus , given the observed [ oiii ] 5007 luminosity of 2@xmath153 erg s@xmath10 and [ oiv ] 26@xmath11 m luminosity of 2.14 @xmath154 erg s@xmath10 and the relationships found between 2 - 10 kev x - ray and [ oiii ] and [ oiv ] luminosities for sy1s ( log(l@xmath155/l@xmath1}$ ] ) = 1.59 @xmath2 0.48 dex , heckman et al . 2005 ; log(l@xmath155/l@xmath156}$ ] ) = 1.92 @xmath2 0.60 dex , diamond - stanic et al . a proposed scenario where the source is heavily obscured while a small percentage of scattered light dominates the spectrum is not supported by our data as the observed luminosity would have to be over an order of magnitude lower for this situation to be applicable . furthermore , the _ spitzer _ observation of this source suggests that circumnuclear dust is not a dominant component of the mid - infrared spectrum . in agn with strong mir emission from the circumnuclear obscuring medium , the polycyclic aromatic hydrocarbon ( pah ) features are diluted by the mir continuum and they consequently have weak pah ews . in mrk 0609 , the pah ews at 11.3 @xmath11 m and 17 @xmath11 m are the highest among the 20 sy2s in the parent [ oiii ] sample @xcite . as no _ chandra _ data exist for this source , it is possible that a background unabsorbed agn contaminates the relatively larger _ xmm - newton _ and bepposax psfs . however , given that the extraction radius for _ xmm - newton _ is @xmath157 at the center of the host galaxy , a chance superposition of an unassociated source is improbable . mrk 0609 can possibly be a `` changing - look '' agn ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) where the optical observations in 2001 and 2007 caught the source in an obscured state while the 2000 and 2007 x - ray observations detected the object while unobscured . alternatively , mrk 0609 may belong to the class of `` true '' sy2s that lack circumnuclear obscuration : broad emission lines are not observed because they do not exist . such `` naked '' sy2s can be explained via an extension of the disk - wind model , where the broad line region forms in the wind launched by the accretion disk @xcite . theoretical work suggests that the outflow can not be maintained at low luminosities ( @xmath7510@xmath158 erg s@xmath10 ) or low eddington rates ( l@xmath159/l@xmath160 ) , causing the broad line region to disappear ( e.g. , * ? ? ? * ) . the critical eddington rate below which the outflow fails has been reported as 2.4@xmath161 @xcite and 1.3 @xmath162 @xcite , where @xmath163 is the black hole mass in units of 10@xmath164 m@xmath26 @xcite . based on the observed velocity dispersion , mrk 0609 has a black hole mass of 0.63 @xmath163 @xcite , constraining the upper limit on the eddington rate to be 0.0024 - 0.012 for this prediction to explain our observations . the reddening corrected [ oiii ] luminosity of mrk 0609 is @xmath165 erg s@xmath10 , which translates to a bolometric luminosity of @xmath166 @xcite and associated eddington rate between 0.06 - 0.09 . both the bolometric luminosity and eddington rate for mrk 0609 are inconsistent with predictions of the failed disk wind to explain naked sy2s . however , simultaneous optical and x - ray observations , which we are currently pursuing , are necessary to determine whether the obscuration in this system varies or if the broad line region is truly absent , and if by extension , the luminosity and accretion rate of mrk 0609 challenges current theories that explain these types of objects . we exclude mrk 0609 from the analysis below since we do not detect any intrinsic x - ray absorption in this source . the x - ray spectrum of sdss j093952.74 + 355358.0 reveals interesting characteristics in its circumnuclear absorber that suggests different global and line - of - sight column densities . the fe k@xmath3 feature is quite prominent ( [email protected] kev , * ? ? ? * ) , precluding the global and line - of - sight absorbers from being equal and compton - thin since the line of sight continuum would dilute the line compared with a pure reflection spectrum . if both column densities are compton - thick , then the continuum between 2 - 6 kev would be curved downwards rather than flat . this continuum curvature can be avoided if the global column density is compton - thin . to obtain the large fe k@xmath3 ew , however , the direct continuum has to be hidden behind a compton - thick absorber . the solid angle of this line of sight absorber must also be relatively small since a larger covering factor would produce more pronounced signatures of a reflection spectrum , such as a curved continuum between 2 - 6 kev . consequently , this absorber is likely in the form of a toroidal ring . we note that an alternate scenario where the fe abundance is super - solar is not supported by current data : freeing the fe abundance in the spherical absorption model does not adequately fit the fe k complex , while the mytorus decoupled model fit shows no significant residuals around the fe k@xmath3 line . as reported in table [ qso2_myt_decoup_fits ] , the global n@xmath33 is compton - thin while n@xmath48 is compton - thick . however , the exact value of the n@xmath48 is unknown , with possible values ranging from 1.67@xmath31 to above @xmath168 @xmath5 . as shown in figure [ j0935_highe ] , the different permissible values of n@xmath48 , from the low end to the best - fit value of @xmath169 @xmath5 to the high end of the range , predict very different spectral shapes at energies above 10 kev . hard x - ray observations , from , e.g. , nustar ( see , for instance , lansbury et al . , submitted ) or astro - h , are necessary to accurately constrain the line of sight column density in this source . no other type 2 qso is known to have a similar x - ray reprocessor geometry . for every source , we measure a global average ( n@xmath33 ) , as well as line - of - sight ( n@xmath48 ) , column density . as shown in figure [ nhz_v_nhs ] , n@xmath33 is largely consistent with n@xmath48 for most objects , but this is not always necessarily the case . four objects in particular are well fit mytorus in decoupled mode , where we find that the global distribution of matter is significantly different from that along the line - of - sight . as simpler models do not self - consistently treat the physics of the transmitted , compton - scattered , and fluorescent line emission , the fitted column densities are unreliable , preventing an accurate calculation of the inherent x - ray flux . we bypass this shortcoming in the present work and are able to derive realistic , absorption corrected 2 - 10 kev x - ray luminosities ( l@xmath0 , tables [ sy2_lums ] and [ qso2s_lums ] ) . the mean log ( l@xmath0/l@xmath1}$ ] ) for the sample is 1.54 @xmath2 0.49 dex ( [email protected] dex for sy2s and 1.51 @xmath20.59 dex for qso2s ) , which is remarkably consistent with the average value for sy1s of 1.59 @xmath2 0.48 dex @xcite . this similarity in normalized x - ray luminosities is to be expected if type 1 and type 2 agn have the same physical central engine , as posited by unification models , and affirms that the sophisticated models do accurately measure the circumnuclear column density . past x - ray studies have reported a correlation between the fe k@xmath3 luminosity and intrinsic agn luminosity ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? . however , such analyses relied on _ ad hoc _ x - ray modeling where the fe k@xmath3 feature was parametrized by a gaussian component that is placed either in front of or behind the line - of - sight absorbing screen . as the models we use here self - consistently treat the fluorescent emission associated with the transmitted continuum and compton - scattered emission within the global obscuring medium , our study better estimates the fe k@xmath3 luminosity . in figure [ compare_fek ] , we compare the fe k@xmath3 luminosity from modeling the feature with a gaussian component in @xcite and @xcite and with the values we derive using self - consistent models . there is general agreement between both approaches , but in several cases , disagreements on the order of one magnitude are apparent . we show the fe k@xmath3 luminosity from the fits presented here as a function of the [ oiii ] luminosity in figure [ fek_lum ] ( a ) . we highlight with filled symbols those type 2 agn that have the fe k@xmath3 feature visibly noticeable in their spectra to test whether the fluorescent emission has to be prominent for a trend to be apparent . the fe k@xmath3 and [ oiii ] luminosities are correlated significantly ( @xmath170=0.735 , which gives a correlation probability @xmath7899.9% ) , and we find a best - fit slope of [email protected] , which is largely consistent with the values of [email protected] from @xcite , [email protected] from @xcite and [email protected] from @xcite . however , there is substantial scatter ( dispersion of l@xmath171/l@xmath1}$ ] is 0.66 ) , where the disagreement between the observed fe k@xmath3 luminosity and the best - fit relation is over an order of magnitude for half the sample , regardless of whether the fe k@xmath3 is visible or not . for the subset of 15 objects that have measured h@xmath3 and h@xmath22 fluxes , we corrected [ oiii ] luminosity for redenning ( l@xmath1,corr}$ ] ) , assuming the standard r(v)=3.1 extinction curve @xcite and an intrinsic h@xmath3/h@xmath22 ratio of 3.1 , to test whether significant amounts of host galaxy dust affects the observed [ oiii ] emission , contributing to the scatter . with the remaining 4 objects , we used the observed [ oiii ] line luminosity as a lower limit and then employed survival analysis ( asurv rev 1.2 , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) to test the strength of the correlation and find the trend between the two quantities . as depicted in figure [ fek_lum ] ( b ) , the correlation does not become more significant ( spearman s @xmath170=0.713 , giving correlation probability of @xmath17299.75% ) , and the dispersion of l@xmath171/l@xmath1,corr}$ ] increases somewhat ( 0.72 ) . we can therefore conclude that dust attenuating the [ oiii ] line emission is not responsible for the dispersion in the fe k@xmath3-[oiii ] luminsosity relationship . such scatter reflects the complexities of radiative transfer effects when the obscuring medium is not optically - thin @xcite . the correlation in figure [ fek_lum ] ( c ) illustrates that the fe k@xmath3 line is excited by the x - ray continuum , where the slope is approximately proportional to the fe k@xmath3 ew ( i.e. , the difference in the fe k@xmath3 line luminosity divided by the difference in the x - ray luminosity provides an estimate of the line ew ) . the spread in this relation gives an indication of the range of physical properties , such as the global n@xmath173 distribution and covering factor . the relatively tight correlation ( dispersion of l@xmath171/l@xmath0 is 0.36 ) implies that the differences in the structure of the x - ray processor are not large among these sources . the ratio of the observed 2 - 10 kev luminosity to the intrinsic luminosity ( e.g. , l@xmath174/l@xmath1}$ ] ) and ew of the fe k@xmath3 line are often used as model - independent proxies of circumnuclear obscuration ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? here we test the relationship between these obscuration diagnostics and the reliable line - of - sight and global column densities we derive here . as our goal is to test the results of proxies calculated from the simple power - law modeling with the present analysis , we use the observed l@xmath174 and fe k@xmath3 ews reported in @xcite and @xcite , rather than from the current modeling . figure [ nh - v - diag ] ( a , c ) shows n@xmath48 and n@xmath33 as a function of l@xmath174/l@xmath1}$ ] , where the vertical dashed lines indicate the range l@xmath174/l@xmath1}$ ] values observed in sy1s @xcite . the sources with the lowest normalized x - ray flux are heavily obscured to compton - thick . however , there are a handful of objects with normalized x - ray luminosities consistent with sy1s yet are heavily obscured ( @xmath175 @xmath176 n@xmath177 @xmath5 ) while the rest are moderately obscured ( @xmath14 @xmath15 @xmath5 ) to mildly obscured ( n@xmath178 @xmath5 ) . these results suggest that a low l@xmath174/l@xmath179}$ ] ratio is indicative of heavy obscuration , but the central engine can also be highly obscured when the normalized 2 - 10 kev luminosities are consistent with those of unabsorbed seyferts . we note that 2 of these 4 anomalous sources ( sdss j082443.28 + 295923.5 and cgcg 218 - 007 ) have rather large fe k@xmath3 ews ( @xmath1720.75 kev ) , making it likely that attenuation of the observed [ oiii ] line by dust in the host galaxy or even the outer parts of the torus inflates the normalized x - ray luminosity . comparison of n@xmath48 and n@xmath33 with the previously reported fe k@xmath3 ews from @xcite and @xcite , figure [ nh - v - diag ] ( b , d ) , also shows mixed results . we note that the reliability of ews calculated with previous modeling depends on how accurately the continuum around the line is measured . with the exception of sdss j083454.89 + 553411.1 which has an ew of @xmath172600 ev , agn with an ew above 400 ev are heavily obscured or compton - thick . at an ew of @xmath172200 ev , however , about half of the sources are consistent with being heavily obscured while the other half are moderately to mildly obscured . similar to the comparison with l@xmath174/l@xmath1}$ ] , a high fe k@xmath3 ew value ( @xmath78400 - 500 ev ) is a useful diagnostic of heavy absorption , but an agn can be heavily obscured at lower ews . we note that it is the same 2 sources with high l@xmath174/l@xmath1}$ ] values and low fe k@xmath3 ews that have fitted column densities exceeding 10@xmath16 @xmath5 ( sdss j123843.43 + 092736.6 and sdss j090037.09 + 205340.2 ) . the parent sy2 and type 2 qso candidate samples from @xcite and @xcite , respectively , are selected based on their optical emission , and are thus unbiased with respect to their x - ray properties . however , the detailed modeling used in this work requires relatively good signal - to - noise in x - rays , so the sub - samples analyzed here are unavoidably biased to be x - ray bright . we caution that the trends we examine between sy2s and type 2 qsos pertain to the sources in this work and it would be inappropriate to extrapolate these results to the general agn population . in figure [ hist ] , we show the distributions of the line - of - sight ( a ) and global average ( b ) column densities for the sy2s ( solid blue ) and type 2 qsos ( dashed red ) . a larger percentage of sy2s are heavily obscured while a majority of the quasars are almost evenly split between moderately and heavily obscured . most of the seyferts and quasars that have evidence for scattering of the agn continuum have a leakage fraction under 2% , though this fraction reaches between @xmath180% for individual sources ( figure [ hist ] ( c ) ) . though figure [ hist ] ( a , b ) indicates that these sy2s tend to be more heavily obscured than the qso2s , we do not find a trend in general between n@xmath173 , line - of - sight or global , and the absorption corrected x - ray luminosity ( figure [ lum_v_obs ] ) , as may be expected by the `` receding torus model '' ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? this result is consistent with that of @xcite where they used the [ oiii ] luminosity as a proxy of intrinsic agn luminosity for the same parent sample from which this work represents a sub - set . however as the agn we are studying are not representative of the general population , and in particular exclude objects with low column densities and unabsorbed type 1 agn , our results should not be construed as a challenge to the receding torus model . we also note that we find no trends between n@xmath48 or n@xmath33 and redshift , n@xmath48 or n@xmath33 and scattering fraction , scattering fraction and redshift , and scattering fraction and l@xmath174/l@xmath179}$ ] . ( a ) line of sight column density and ( b ) global average column density as a function of intrinsic 2 - 10 kev luminosity . similar to @xcite , we do not find a dependence of obscuration as a function of agn luminosity , inconsistent with the expectations of the `` receding torus model '' @xcite , though the sources presented here represent neither an unbiased nor complete sample . symbol and color coding same as figure [ nhz_v_nhs].,title="fig : " ] we used the physically motivated , self - consistent spherical absorption model of @xcite and the mytorus model @xcite to unravel the complexities of the x - ray reprocessor in 19 type 2 [ oiii]-selected agn , 9 of which are seyfert 2 galaxies and 10 of which are obscured quasar candidates . we report , for the first time for an agn sample , reliable measurements of the global , as well as line - of - sight , column densities . * along the line - of - sight , 1 sy2 and 1 qso2 are mildly obscured ( @xmath13 @xmath5 ) , 1 sy2 and 4 qso2s are moderately obscured ( @xmath14 @xmath181 n@xmath182 @xmath5 ) , 4 sy2s and 2 qso2s are heavily obscured ( @xmath175 @xmath181 n@xmath183 @xmath5 ) , and 2 sy2s and 3 qso2s are heavily - obscured to compton - thick ( @xmath184 @xmath5 ) . one source is unabsorbed . * 13 objects have evidence of scattering of the intrinsic agn continuum off a distant medium . the scattering fraction is largely under 2% , but reaches between @xmath172 7 - 9% for individual sources . * in 4 agn , we measure global column densities that are significantly different from the line - of - sight column density . in one of these objects , sdss j082443.28 + 295923.5 , the line - of - sight column density is compton - thin while the global column density is compton - thick , while in 1 other object , sdss j093952.74 + 355358.0 , the line - of - sight column density is heavily - obscured to compton thick with a compton - thin global column density . * we identified a candidate changing - look agn or naked sy2 galaxy , mrk 0609 . if mrk 0609 is a true sy2 , both its bolometric luminosity and eddington rate exceed the predicted critical values of disk - wind models that explain such sources @xcite . we are pursuing simultaneous x - ray and optical observations to test this possibility . * the spectral features of sdss j093952.74 + 355358.0 indicate that the line - of - sight obscuration is in the form of compton - thick toroidal ring which is embedded in a compton - thin global matter distribution . this is the first type 2 qso where this geometry has been observed . * the ratio of the intrinsic 2 - 10 kev luminosity to the [ oiii ] luminosity , [email protected] dex , is basically equivalent to that for type 1 agn @xcite , affirming that these sophisticated models reliably measure the circumnuclear column density , allowing the inherent x - ray luminosity to be revealed . * we find a significant correlation between the fe k@xmath3 and [ oiii ] luminosities , though with wide scatter . estimating the fe k@xmath3 luminosity from the [ oiii ] luminosity can lead to errors of over an order of magnitude . * we compared n@xmath48 and n@xmath33 with obscuration diagnostics used often in the literature from _ ad hoc _ spectral modeling , i.e. , l@xmath174/l@xmath179}$ ] and the fe k@xmath3 ew ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , where lower values of the former ( i.e. , @xmath1851 ) and higher values of the latter ( i.e. , @xmath1721 kev ) imply heavy - to - compton - thick obscuration . sources that are flagged as heavily obscured via these proxies do indeed have measured column densities consistent with heavy or compton - thick obscuration , but a couple of objects with values similar to unabsorbed agn also have measured n@xmath48 indicative of heavy obscuration . * though a larger percentage of the sy2s are more heavily obscured than the qso2s , no trend exists between line - of - sight or global obscuration and agn luminosity , parametrized here as the intrinsic , rest - frame 2 - 10 kev luminosity . we also find no relationship between n@xmath48 or n@xmath33 and scattering fraction , n@xmath48 or n@xmath33 and redshift , scattering fraction and redshift , and scattering fraction and l@xmath174/l@xmath179}$ ] . we reiterate that these results hold for the objects we studied here which are not necessarily representative of type 2 agn in general . we thank the referee for a careful reading of this manuscript and for helpful suggestions . p.g . acknowledges support from stfc ( grant reference st / j003697/1 ) . this work has made use of data from the sdss . funding for the sdss and sdss - ii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , the u. s. department of energy , the national aeronautics and space administration , the japanese monbukagakusho , the max planck society , and the higher education funding council for england . the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium for the participating institutions . the participating institutions are the american museum of natural history , astrophysical institute potsdam , university of basel , university of cambridge , case western reserve university , university of chicago , drexel university , fermilab , the institute for advanced study , the japan participation group , johns hopkins university , the joint institute for nuclear astrophysics , the kavli institute for particle astrophysics and cosmology , the korean scientist group , the chinese academy of sciences ( lamost ) , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , ohio state university , university of pittsburgh , university of portsmouth , princeton university , the united states naval observatory , and the university of washington .

in supersymmetric extensions of the standard model , higgsino - like charginos and neutralinos are preferred to have masses of the order of the electroweak scale by naturalness arguments . such light , and states can be almost mass degenerate , and their decays are then difficult to observe at colliders . in addition to the generic naturalness argument , light higgsinos are well motivated from a top - down perspective . for instance , they arise naturally in certain models of hybrid gauge - gravity mediation . in the present analysis , we study two benchmark points which have been derived in the framework of such a model , which exhibit mass differences of in the higgsino sector . for chargino pair and neutralino associated production with initial - state photon radiation , we simulate the detector response and determine how accurately the small mass differences , the absolute masses and the cross sections can be measured at the international linear collider . assuming that 500 fb has been collected at each of two beam - polarisations , we find that the mass differences can be measured to 40300 mev , the cross sections to 25% , and the absolute masses to 1.53.3 gev , where the range of values correspond to the different scenarios and channels . based on these observables we perform a parameter fit in the mssm , from which we infer that the higgsino mass parameter can be measured to a precision of about gev . for the electroweak gaugino mass parameters , , which are chosen in the multi - tev range , a narrow region is compatible with the measurements . for both parameters independently , we can determine a lower bound .

1307.3566

the lhc experiments have discovered a standard model ( sm)-like higgs boson with @xmath11 gev @xcite . however , it is not clear whether this newly discovered particle is a pure sm higgs boson , or whether it is part of an extension of the sm , which might also include multiple or composite higgs - like states . precise measurements of the higgs couplings and branching ratios will be required to resolve this question . another , more direct way to gain insights about physics beyond the sm ( bsm ) would obviously be to discover more new particles directly . a well - motivated extension of the sm is the minimal supersymmetric sm ( mssm ) , a fully renormalisable theory of high predictive power . in this model , the u(1 ) , su(2 ) and su(3 ) gauge couplings can unify , and a dark matter candidate can be naturally accommodated . furthermore , supersymmetry stabilises the higgs potential with respect to quadratically divergent radiative corrections . in the mssm the higgs mass is no longer a free parameter but becomes strongly constrained . if nature is supersymmetric at the electroweak scale , a large number of new superpartner mass parameters is introduced whose values depend on the unknown mechanism of supersymmetry breaking at high energies . this poses a challenge for any phenomenological analysis . due to the large number of unknown parameters it is very difficult to make reliable predictions for susy background processes , and in many cases the worst background contributions for susy signals originate in fact from susy processes . the experiments at the lhc have not yet found any hints for supersymmetry in squark and gluino searches @xcite , which points towards a rather heavy coloured superpartner spectrum with soft masses above a . however , there are large regions in the mssm parameter space which lead to a sm - like higgs with @xmath12 gev , a light electroweak susy sector , and a rather heavy coloured susy sector beyond the current lhc discovery range . in general , if the electroweak sector is light enough , it should be possible to discover neutralinos and charginos via drell yan production at the lhc @xcite , and the lhc experiments have already put limits on the electroweak - ino parameter space under several simplified model assumptions @xcite . however , in cases where these light states are very close in mass and their decay products are therefore very soft , it will be almost impossible to observe such processes at the lhc . an example is given by scenarios in which the higgsino mass parameter @xmath6 is much smaller than the electroweak gaugino masses @xmath9 and @xmath10 ( current bounds require only @xmath13{gev}$ ] in order to avoid the limits from lep chargino searches @xcite ) . then , the new light states are a chargino and two neutralinos which are higgsino - like and almost mass degenerate . the chargino and the heavier neutralino therefore decay only into extremely soft sm particles and the invisible lightest neutralino . the most `` natural '' versions of the phenomenological mssm which are still not ruled out are indeed characterised by higgsino masses of a few 100 , together with sub- third - generation squarks and a moderately heavy gluino , while the rest of the spectrum can be in the multi - tev range @xcite . models with light higgsinos are also attractive for cosmology , either in combination with gravitino dark matter or with a higgsino - like neutralino as a non - thermally produced lightest supersymmetric particle ( lsp ) ; see e.g. @xcite . while the light higgsino scenario is extremely challenging for the lhc , it may well be observable in the clean environment at an @xmath14 linear collider , for instance , at the international linear collider ( ilc ) @xcite . at the ilc , the striking feature of a precisely known initial state allows to successfully apply the initial - state radiation ( isr ) method , i.e. to select only those signal events that are accompanied by a sufficiently hard photon . in this context , polarised beams help to significantly enhance the signal cross section , at least in the chargino case . the subject of the present work is the phenomenology of light quasi - degenerate higgsinos at the ilc . we study their production and decay , and investigate in how far one can derive the fundamental mssm parameters without assuming any specific susy breaking scheme if only the light higgsinos are kinematically accessible . we study the processes @xmath15 and @xmath16 at @xmath17 gev with polarised beams and initial - state radiation at the ilc . we simulate the detector response for two mssm benchmark points with different @xmath18 gev ) mass splittings between the chargino and neutralinos , leading to different final state sm particles . the uncertainties on masses , mass differences and cross sections as determined by our simulation study are then used as inputs for a fit in order to reconstruct the mssm parameters @xmath9 , @xmath10 , @xmath6 and @xmath19 . our overall goal is to determine both how well the chargino and neutralino masses can be measured at the ilc , and what could be inferred from such a measurement about the underlying susy model . in the following chapters we first give an overview of the theoretical arguments in favour of light higgsinos and the phenomenological aspects of such scenarios in the mssm . the expected sm background will also be discussed , including beam conditions and detector capabilities at the ilc . in section 3 we define two benchmark scenarios on which our study will be based and provide the details for the detector simulation . in section 4 we discuss the expected experimental results at the ilc . in section 5 we present our results on reconstructing the fundamental parameters and the precision we expect to achieve with our analysis strategy . there are two main reasons why higgsinos are distinguished among the mssm superpartners of the sm particles . first , the size of the higgsino mass parameter @xmath6 is not necessarily related to the scale of susy breaking , as it is allowed by unbroken supersymmetry ( it is indeed the only dimensionful mssm parameter with this property ) . it can therefore be generated independently of the soft supersymmetry breaking terms , and at leading order does not receive any soft - term - dependent radiative corrections in the evolution from the high scale . second , @xmath6 enters the tree - level higgs potential and is therefore directly connected to the electroweak scale . for instance , the tree - level @xmath20 boson mass is given in terms of the higgs soft masses @xmath21 , the ratio @xmath19 of higgs vacuum expectation values , and @xmath6 as @xmath22 to avoid finely tuned large cancellations among these parameters , it follows that @xmath6 should not be too far above the electroweak scale , even if some or all of the squark , slepton and gaugino masses are much larger . these two arguments are the main motivation for investigating models where the higgsinos are light , while the other mssm superpartners may be much heavier . from a top - down perspective , light higgsinos are equally well motivated . for instance , they are generic in certain models of hybrid gauge - gravity mediation @xcite , whose main features we will now briefly recapitulate . in gauge - mediated susy breaking , some of the hidden sector states carry sm gauge charges . these messenger fields acquire large supersymmetric masses @xmath23 , and messenger loops generate soft terms of the order @xmath24 where @xmath25 is the scale of susy breaking , and @xmath26 roughly counts the number of contributing messenger states . in addition to these , generically there are also gravity - mediated contributions to soft terms as well as a gravity - mediated effective @xmath6 parameter , generated by gravitationally suppressed interactions between the hidden sector and the mssm . they are of the order of the gravitino mass , @xmath27 in models where the mssm is embedded in a more fundamental theory , for instance in superstring compactifications , the details of supersymmetry breaking are often unknown , while the spectrum of sub - planckian states can still be calculated . in phenomenologically promising models , these typically consist of the mssm along with a multitude of additional exotic states , some of which are charged under the sm gauge groups ( see e.g. @xcite and references therein ) . these exotic fields can be given large masses and act as gauge mediation messengers . a natural scale for them to decouple is the scale of the grand unification , @xmath28{gev}\approx\frac{1}{16\pi^2}m_{\rm planck}\,.\ ] ] with some simple assumptions about the couplings of the messengers to the susy - breaking fields , and choosing parameters such that @xmath29 is of the order of the electroweak scale , one can obtain a realistic soft term spectrum . for large messenger multiplicities @xmath30 , the soft terms are predominantly generated by gauge mediation and can be in the multi - tev range , while @xmath6 is induced by gravity mediation and of the order of @xmath31100@xmath32200@xmath33{gev}$ ] . the benchmark points we are using for the present analysis are derived from such hybrid gauge - gravity mediation models with large messenger numbers . however , we stress that these are by no means the only possibility to obtain light higgsinos from high - scale models , and that our results on the light higgsino phenomenology are independent of a specific susy breaking model . in a more general context , scenarios with a rather heavy coloured sector and a mass - degenerate light chargino / neutralino sector can easily be constructed in the unconstrained mssm , where the soft susy - breaking parameters are taken as independent . one simply chooses sufficiently large values for the squark and gluino masses ( and possibly also for the slepton masses ) . at the tree level the chargino / higgsino sector @xcite is determined by the u(1 ) bino mass parameter @xmath9 , the su(2 ) wino mass parameter @xmath10 , the higgsino mass parameter @xmath6 and @xmath19 . the neutralino mass matrix in the @xmath34 basis is given by @xmath35 since @xmath36 is complex symmetric , it can be diagonalised by one unitary matrix @xmath37 via @xmath38 ( where the matrix @xmath39 contains the neutralino masses @xmath40 on the diagonal ) . the mixing matrix @xmath37 relates the neutralino mass eigenstates to the neutral gauge eigenstates . for full analytical solutions of the system see @xcite . the chargino mass matrix in the @xmath41 basis , @xmath42 is not symmetric and therefore has to be diagonalised via the bi - unitary transformation @xmath43 . @xmath44 and @xmath45 relate the chargino mass eigenstates to the charged gauge eigenstates . in general , @xmath9 , @xmath10 and @xmath6 are complex - valued parameters . by a suitable rotation of the gaugino and higgsino fields , @xmath10 can be made real and positive without loss of generality . in the following we assume that also @xmath9 and @xmath6 are real ( positive or negative ) so no new sources of cp violation are introduced . one can easily classify three limiting cases : in the case @xmath46 , one obtains @xmath47 with a bino - like @xmath48 , @xmath49 with wino - like @xmath50 and @xmath51 , and higgsino - like heaviest states with masses @xmath52 , @xmath53 , cf . ( [ eq : y ] ) and ( [ eq : x ] ) . by contrast , the hierarchy @xmath54 leads to the lightest two states being mass degenerate and wino - like , @xmath55 but a bino - like @xmath50 with mass @xmath56 , while @xmath57 and @xmath58 remain higgsino - like and heavy as before . choosing @xmath59 ( or @xmath60 ) leads to mass - degenerate lightest states with @xmath61 but @xmath62 ( or @xmath63 ) and @xmath64 ( or @xmath65 ) . the lighter neutralinos and the lightest chargino are then mostly higgsino - like and have unsuppressed couplings to the sm gauge bosons . this is the scenario which we are investigating . it is instructive to explicitly write down the tree - level masses of the three light states in the case that @xmath9 and @xmath10 are large . they are found by diagonalising the mass matrices of eqs . and , and given by @xmath66 up to terms suppressed by higher powers of @xmath67 , see e.g. @xcite . the factors @xmath68 denote the sign of the mass eigenvalue , that depends on the cp - quantum number of the corresponding neutralino @xcite . the wino and bino admixtures are small , of the order @xmath69 . for instance , for @xmath70 , the mass eigenstates are decomposed as @xmath71 where we have only written the respective leading - order terms of the weyl gaugino and higgsino components . for multi - tev gaugino masses , the gaugino admixtures are of the order of a few percent , and the mass splittings are of the order of @xmath72@xmath73 gev . explicitly , the tree - level mass difference between the lightest neutralino and the lightest chargino reads , again for @xmath70 , @xmath74 \,,\ ] ] which for @xmath75 further simplifies to @xmath76 for the neutralinos , the mass difference is @xmath77 remarkably , it does not depend on @xmath19 in the leading approximation . the corresponding expressions for @xmath78 are similar and the phenomenology of that parameter range has been discussed in @xcite . in our case the sign of @xmath6 has negligible impact on observables . equations hold for positive and negative values of @xmath9 . it is interesting to note that negative values of @xmath9 allow for cancellations of wino and bino contributions to neutralino masses . generally , they will result in different mass patterns depending on the sign of @xmath9 . in particular , for low values of @xmath19 the lightest neutralino can be heavier than the light chargino , i.e. @xmath79 . both possibilities , @xmath80 and @xmath81 , will be analysed in section [ sec : paradet ] . comparing with the analytical formulae of the neutralino ( chargino ) masses and couplings @xcite , it is obvious that one can express the neutralino and chargino sector as functions of @xmath82 and @xmath83 . in the large @xmath19 limit , @xmath84 and @xmath85 , therefore all neutralino and chargino masses as well as the cross sections are only weakly dependent on @xmath19 . this weak dependence is also reflected in our final results , cf . section [ sec : paradet ] . radiative corrections to the chargino and neutralino observables , see e.g. @xcite ( and references therein for a more complete list of studies ) , can be sizeable and increase the mass splittings by several hundreds of mev . these corrections will lift the degeneracy between higgsino states . if the tree - level mass difference is substantially less than a gev , a correction of 200 mev @xcite will have a profound impact on the lifetime and decay branching ratios . this may also be exploited to better determine the underlying mssm parameters , see @xcite . a detailed analysis of nlo corrections in this regard , is , however , beyond the scope of this study and in the following we will assume tree - level relations . at the ilc higgsino production proceeds via @xmath20 or @xmath86 exchange in the @xmath87-channel , see fig . [ fig : chaprod ] . in general , @xmath88- and @xmath89-channel exchange of selectrons or sneutrinos is also possible , but in the higgsino case they are highly suppressed due to the small yukawa couplings . hence , even if the selectrons and sneutrinos were relatively light , they would not affect any observables . higgsinos will be produced in @xmath90 or @xmath91 pairs ( the @xmath92 and @xmath93 couplings vanish in the pure higgsino limit ) . and @xmath94 have opposite weak isospin , and that both @xmath95 and @xmath50 contain equally large @xmath96 and @xmath94 admixtures as seen from eq . . ] both @xmath97 and @xmath98 will undergo three - body decays into @xmath99 via off - shell @xmath20 and @xmath100 bosons @xcite . the decay @xmath101 is also possible but kinematically highly suppressed . the detector signature comprises exclusively a few low - energetic leptons or hadrons , whose softness poses a major challenge to experiment . for small mass splittings , the radiative decay @xmath102 via a @xmath100-chargino loop can also become very important @xcite . more details on higgsino production and decay will be given below when we discuss the details of our simulation . - and @xmath89-channel selectron / sneutrino exchange graphs are negligible in our scenario due to a small yukawa coupling between sleptons and higgsinos . [ fig : chaprod],title="fig:",scaledwidth=60.0% ] - and @xmath89-channel selectron / sneutrino exchange graphs are negligible in our scenario due to a small yukawa coupling between sleptons and higgsinos . [ fig : chaprod],title="fig:",scaledwidth=60.0% ] for the discussion of the sm processes at the ilc it is convenient to classify them by the initial state and the number of final state fermions . due to the very strong fields between the colliding bunches , there is a sizeable photon component from back - scattered synchrotron radiation in the ilc beams @xcite . therefore , the initial states can be @xmath14 , but also @xmath103 , or @xmath104 . for @xmath14 and @xmath103 , the final state has an even number of fermions : 2f , 4f , 6f , ... , while the @xmath104 initial state gives rise to final states with an odd number of fermions : 1f , 3f , 5f , ... in this notation , @xmath86 can represent both real beam photons or virtual ones . for initial virtual photons , the final state is understood to also contain the electron and/or positron that emitted the photon ( the beam - remnant " ) , in addition to the fermion(s ) . the signal comprises events with a few low - energy charged particles and no other activity . possible sm processes giving this topology are of two types . either ( _ i _ ) events with large amounts of energy in invisible neutrinos , or ( _ ii _ ) events where a large fraction of the energy escapes with particles going in un - instrumented regions of the detector ; in an ilc detector , the only such regions are the holes in the low - angle calorimeter through which the beam pipes enter the detector . the first background type mainly contains events with two @xmath105s , possibly together with additional neutrinos , i.e. the subset of the classes @xmath106 2f and @xmath106 4f where one of the fermion pairs are @xmath105 leptons , and the other one ( if present ) is a neutrino pair . these channels do not have cross sections much larger than the signal . in addition , only a small fraction of the @xmath105s will decay to the very soft visible system which is typical for the signal . hence , while it is not negligible , this background type is expected to be manageable without too strong cuts on the event - topology . the potentially most severe background processes are of type ( _ ii _ ) , and are expected to be the high cross section multi - peripheral process @xmath107 , i.e. the @xmath108 2f class . in such events , the final beam - remnant electrons and positrons carry nearly the full beam energy in most cases , have low transverse momentum , and therefore leave the detector unseen through the outgoing beam pipe . the @xmath109-pair , on the other hand , can be emitted at any angle , and tends to have low momentum . in other words , these events have a large resemblance to the signal . however , if one of the incoming electrons emits an isr photon at an angle and energy high enough to be detected , this electron will recoil against the photon , changing its direction sufficiently to deflect it into the acceptance of the detector . in order to reduce this background to acceptable levels , we therefore require that the accepted signal candidate events should contain a detected isr photon . also the other component of the @xmath108 2f class , where the initial photons are real photons from the photon component of the beams , is suppressed by this requirement . this is simply due to the fact that , in this case , there are no initial charged particles that could emit an isr photon . instead , the most severe remaining background is expected to arise from the @xmath110 3f class , in particular from the @xmath88-channel diagram with incoming real photons . this process is similar to compton scattering , with the difference that the scattered photon is virtual , and gives rise to an @xmath109-pair . in this case , the hard isr photon does not recoil against the beam electron , but against the @xmath109-system , like in the signal events . however , in the event as a whole , a large amount of energy is carried away by the _ exchanged _ electron once it has been brought back on - shell by the beam - photon . this electron will receive most of the photon s energy , and closely follow its direction in the lab frame , i.e. along the beam pipe . as the second incoming particle is neutral in this case , no isr will be emitted from it , and thus it can not be deflected into the detector acceptance either . for our simulation we chose two benchmark points with the lightest higgs mass which is compatible with the lhc discovery ( within the theory uncertainty ) , namely @xmath111 and @xmath112 gev . we will refer to them as dm1600 and dm770 , according to their chargino lsp mass difference . in the following sections we will discuss their low - scale spectrum and the higgsino production and decay modes at the ilc as well as their implementation into our simulation . we especially comment on the decay modes and branching ratios in the nearly mass - degenerate @xmath113 system . beyond the two specific mass differences considered here , it has been shown recently that the isr technique and the selection strategy described in the following are also sensitive in case of larger mass differences of a few gev @xcite . as soon as the chargino lsp mass difference becomes even larger , analyses without isr requirement start to gain sensitivity and eventually take over . for mass differences smaller than the ones considered here , striking signatures like displaced vertices , kinks , anomalous ionisation etc . will make searches experimentally more easy . in this sense , the benchmarks chosen here fall into the experimentally most challenging regime . lccccccccc&n_1&n_2&n_3&m_&m_a&&^&m_1^&m_2^ + & 20&28&11&190gev&2160gev&43.81&160gev&1.72tev&4.33tev + & 46&46&20&250gev&4050gev&47.66&160gev&5.37tev&9.47tev + the general tree - level spectrum of our model was calculated using the spectrum generator softsusy @xcite . the model we are considering is characterised by the high - scale parameters specified in table [ tab : highspec ] . the parameters in the gaugino , squark , and slepton sector at the high scale are obtained from eqs . ( 3 ) and ( 4 ) in @xcite , while the higgs sector parameters are computed such that both the model predictions at the high scale are matched and electroweak symmetry breaking at the proper scale is obtained . with the input parameters in table [ tab : highspec ] , we obtain a spectrum which is mostly at the multi - tev scale , with only the two lightest neutralinos as well as the light chargino masses within current collider ranges , and the sm - like higgs boson with a mass within a @xmath114 theory uncertainty interval from the lhc central value of @xmath115 . the physical chargino and neutralino masses are calculated by softsusy using the running mass parameters @xmath116 , @xmath117 and @xmath118 as given in table [ tab : highspec ] , which includes one - loop corrections @xcite . the light chargino and neutralino spectrum , and the underlying on - shell @xmath6 and @xmath119 parameters , @xmath120 and @xmath121 . the most convenient choice should include `` representative '' states for each of the components , bino , wino and higgsino . this choice guarantees that the nlo corrections will be small @xcite . next , for fixed @xmath122 and using tree - level relations , see eqs . and , one calculates on - shell mssm parameters . the masses of the remaining charginos and/or neutralinos will be shifted by appropriately calculated radiative corrections . ] are given by @xmath123 note that the above masses include higher - order corrections , and therefore differ from the tree - level predictions in eq . at the sub - permil level . and @xmath19 ; see section [ sec : paradet ] for further details . ] as stated before , the two lightest neutralinos are nearly completely higgsino - like , with negligible gaugino admixtures . the mass splittings are very small with @xmath124 and @xmath125 in the dm770 ( dm1600 ) scenario . this has interesting consequences especially for the decay , which we will discuss in the following . + the production cross sections have been calculated with ` whizard 1.95 ` @xcite using the ilc set - up provided by the ilc generator group @xcite . it includes soft initial - state photon emission using an electron structure function @xcite as well as the beam energy - spectra corresponding to the ilc technical design report ( ilc tdr ) @xcite . due to the small mass difference , only very few and soft visible particles will result from the higgsino decay . for reasons mentioned above , a realistic analysis will therefore need to require an additional photon from initial - state radiation to be seen in the detector . the isr photon , in addition to its usefulness to suppress background , provides an elegant way to determine the @xmath126 and @xmath127 masses from their recoil against the photon . this has already been applied successfully for near - degenerate winos in amsb scenarios @xcite . we therefore determine the observable cross sections with one final state photon included in the hard matrix element . this photon is required to have a minimal invariant mass of @xmath128gev with the corresponding beam electron and an energy of more than @xmath129 , corresponding to the cut used later at the analysis stage . we have verified that the effect of double - counting with the soft initial - state photon emission is negligible . and @xmath130 as a function of the centre - of - mass energy.,title="fig:",scaledwidth=49.0% ] and @xmath130 as a function of the centre - of - mass energy.,title="fig:",scaledwidth=49.0% ] figure [ fig : xsec_roots ] shows the observable chargino and neutralino cross sections in the dm770 scenario as a function of the centre - of - mass energy , and for two different configurations of the polarisation of the beams , with the positron beam polarised to @xmath131 , the electron beam to @xmath132 . etc . will be used to denote the assumed beam polarisation . ] figure [ fig : xsec_pol ] shows the same cross sections for a fixed centre - of - mass energy of @xmath133 as a function of the positron polarisation . the chargino production can be strongly enhanced by choosing the appropriate beam helicities , whereas the differences are significantly smaller in the neutralino case . beyond the signal enhancement , beam polarisation can be used to prove that the observed process is mediated purely by @xmath87-channel exchange of a @xmath20-boson , i.e. by analysing the corresponding scaling factors between the polarised and unpolarised cross sections together with the order of magnitude of the cross section , see below and cf . discussion in @xcite . effects from beam polarisation at an @xmath134 collider on chargino / neutralino pair - production have already been studied in @xcite and have successfully been exploited for full parameter determination in @xcite . gev as a function of the positron polarisation for @xmath135 and @xmath136.,title="fig:",scaledwidth=49.0% ] gev as a function of the positron polarisation for @xmath135 and @xmath136.,title="fig:",scaledwidth=49.0% ] as mentioned , higgsinos are produced by @xmath20 @xmath137 exchange in the @xmath87-channel for neutralinos ( charginos ) respectively , cf . the diagrams shown in fig . [ fig : chaprod ] . it is then straightforward to explain the differences in production cross sections in figs . [ fig : xsec_roots ] and [ fig : xsec_pol ] by considering only these diagrams . for neutralino production , the ratio of @xmath138 for fully polarised beams can directly be obtained from the ratio of the respective fermion couplings to the @xmath20 boson , given by @xmath139 ^ 2\sim 1.3 $ ] , with @xmath140 , @xmath141 . for arbitrarily polarised beams the scaling factor between polarised and unpolarised cross section is given by @xmath142 where @xmath143 denotes the normalised effective number of interactions and @xmath144 the effective polarisation ( for details see @xcite ) . it is expected to achieve high @xmath145 and @xmath146 for centre - of - mass energies of @xmath147 gev at the ilc @xcite which leads to scaling factor of about 1.2 ( 1.7 ) with @xmath148 ( @xmath149 ) configuration of polarised @xmath150-beams . in case of chargino pair production , the large difference in the cross sections stems from positive ( in lr configuration ) or negative ( in rl configuration ) interference of the @xmath86 and @xmath20 @xmath87-channel contributions , which is again due to the structure of the fermion-@xmath20 couplings . these characteristics ( which are independent of additional photon radiation ) are well reflected in the total cross sections displayed in figures [ fig : xsec_roots ] and [ fig : xsec_pol ] . due to the quite pronounced mass degeneracy in the @xmath151 system , the @xmath152 decays require a specific treatment , which we will now briefly describe . in case of small mass splittings , a perturbative treatment on parton level , followed by standard hadronisation , leads to large uncertainties for slight variations of the masses of the partonic decay products . in the scenarios considered here , this especially plays a role in the decays of the chargino , where the mass difference can be in the sub - gev range . in fact , the same situation is encountered in the sm for the decay of @xmath105s into hadrons . in this case , a more appropriate description then follows an effective theory approach using hadronic currents . for the case of supersymmetry , such a description has been implemented in the event generator ` herwig++ ` for @xmath152 decays @xcite . we therefore used ` herwig++ ` to generate the according branching ratios and decays widths , they are given in table [ tab : chi1 ] . on the other hand , the branching ratios of the @xmath153 are calculated at parton - level followed by hadronisation , with the results listed in table [ tab : neu2].-lsp in dm1600 scenario and the @xmath153-lsp in dm770 scenario are of similar order . however , since for neutral current decays there is no analogue to the charged current @xmath105-decay in the the sm , no dedicated code is yet available which treats the decays of @xmath153 using hadronic currents . the effects of including such an improved description are in the line of future work . ] the decay widths of loop - induced processes have been cross checked with available analytic results @xcite . they agree within @xmath154 with the ` herwig++ ` values . .[tab : chi1 ] chargino @xmath155 decay modes according to ` herwig++ 2.6.0 ` . [ cols="<,^,^",options="header " , ] the cross section errors obtained in sec . [ sec : expprec ] did not take into account the uncertainty on the branching ratios . the branching ratios only depend on the available phase space for the decay of the virtual @xmath100 boson for charginos , given by the chargino lsp mass difference . with the previously obtained precision on the mass difference , the uncertainty on the br of the selected decay modes has to be taken into account . in case of the dm1600 scenario , due to the significant uncertainty on the chargino lsp mass difference , it amounts to @xmath156 for the chargino cross section measurement . for the dm770 scenario it is @xmath157 , comparable to the experimental uncertainty . these additional uncertainties are included in table [ tab : input ] . using ` minuit ` @xcite we minimise a @xmath158 function defined as @xmath159 where the sum runs over the input observables @xmath160 , with their corresponding experimental uncertainties @xmath161 . the theoretical values calculated using the fitted mssm parameters , eq . , are denoted by @xmath162 . note that throughout the simulation the one - loop corrected masses , given by eqs . and , were used . the difference between the tree - level values and the ones obtained from ` softsusy ` is accordingly taken into account in the fit in order to ensure consistency . the variation of the observables , with respect to the central values , on the susy parameters , eq . , is in the following calculated at the tree level . therefore , the results of the fit are to be understood as the tree - level fit . it turns out that using the experimental information one can not constrain @xmath19 . this is not very surprising , as with the values of parameters used here , there is very little mixing between higgsinos and gauginos which could have provided some sensitivity to @xmath19 . as can be seen from eq . , the chargino and neutralino masses depend only on @xmath163 and the sensitivity is rather weak , especially in the large @xmath19 limit . in order to have a converging fit we therefore fix @xmath19 to values in the range @xmath164 $ ] and fit the remaining three parameters of eq . . in principle , the cross sections depend also on the sneutrino and selectron masses but since the couplings to higgsinos are negligible , no dependence on these parameters enters our observables . the results of the fit are shown in tables [ tab : fit124 ] and [ tab : fit127 ] . the crucial observable for the determination of @xmath9 and @xmath10 is the mass difference between the lsp and the light chargino . on the other hand , the cross section measurements improve the determination of the @xmath6 parameter . in general , only lower limits can be obtained for @xmath165 and @xmath10 , which are given in tables [ tab : fit124 ] and [ tab : fit127 ] for positive and negative @xmath9 . however , it turns out that @xmath9 and @xmath10 are strongly correlated and the allowed regions are only narrow strips in the @xmath9@xmath10 parameter space , see figs . [ fig : deltamcal124 ] and [ fig : deltamcal127 ] . the @xmath6 parameter can be determined with an accuracy @xmath166 gev in dm770 scenario and @xmath167 gev in dm1600 scenario . m1.5 cm > m2 cm > m2 cm > m2 cm r2 cm & & & & & & input & lower & upper & lower & upper * @xmath168 & @xmath165 [ tev ] & @xmath169 & @xmath170 @xmath171 & no & @xmath170 @xmath171 & no & @xmath10 [ tev ] & @xmath172 & @xmath173 @xmath174 & no & @xmath173 @xmath174 & no & @xmath6 [ gev ] & @xmath175 & @xmath176 & @xmath177 & @xmath178 & @xmath179 * @xmath180 & @xmath165 [ tev ] & @xmath169 & @xmath181 @xmath171 & @xmath182 @xmath183 & @xmath181 @xmath171 & @xmath182 @xmath183 & @xmath10 [ tev ] & @xmath172 & @xmath166 @xmath184 & @xmath185 ( no ) & @xmath166 @xmath184 & @xmath185 ( no ) & @xmath6 [ gev ] & @xmath175 & @xmath186 & @xmath187 & @xmath188 & @xmath189 cr > m1.5 cm > m2 cm > m1.5 cm > m2 cm r1.5 cm & & & & & & input & lower & upper & lower & upper * @xmath168 & @xmath165 [ tev ] & @xmath190 & @xmath191 @xmath192 & no & @xmath191 @xmath192 & no & @xmath10 [ tev ] & @xmath193 & @xmath194 @xmath195 & no & @xmath194 @xmath195 & no & @xmath6 [ gev ] & @xmath196 & @xmath197 & @xmath198 & @xmath176 & @xmath199 * @xmath180 & @xmath165 [ tev ] & @xmath190 & @xmath194 & no & @xmath194 & no & @xmath10 [ tev ] & @xmath193 & @xmath200 & @xmath201 & @xmath200 & @xmath202 & @xmath6 [ gev ] & @xmath196 & @xmath176 & @xmath203 & @xmath204 & @xmath203 inclusion of the cross section measurements at @xmath205 gev only affects the determination of the @xmath6 parameter . the low cross section results in a large error , larger by a factor of @xmath206 than for @xmath207 gev . however , being close to the production threshold and on the quickly rising slope of the cross section , see fig . [ fig : xsec_roots ] , helps to improve the determination of the @xmath6 parameter in both scenarios . taking the cross section at a higher centre - of - mass energy , e.g. @xmath208 gev , does not yield a significant improvement . the increased statistics does not compensate the advantage of being close to the production threshold . the masses and mass differences require larger statistics , and therefore , measurements at higher centre - of - mass energies ( e.g. at @xmath209gev ) are indispensable . finally , we note that the main driver for the rather high @xmath6 error is the variation of @xmath19 , which shifts the central values of the @xmath6 parameter by approximately @xmath210 gev . this additional dependence will remain regardless of additional measurements and/or increased accuracy . in order to better understand the interdependence of the @xmath9 and @xmath10 determination , in the top panels of figs . [ fig : deltamcal124 ] and [ fig : deltamcal127 ] we show the 1-@xmath211 contours in the @xmath9@xmath10 plane for an integrated luminosity of @xmath212 . different contours correspond to different values of @xmath19 . for low values of either @xmath9 or @xmath10 the allowed region for the other parameter extends to arbitrary high values and can not be reliably constrained . significant variation with @xmath19 is also visible , especially in case of scenario dm770 . both cases of the sign of the @xmath9 parameter are considered . in general , for @xmath81 lower absolute values of gaugino masses are consistent with data . in the most extreme cases , when @xmath213 , a direct production of @xmath214 could be possible at the ilc with an increased centre - of - mass energy . we also note that for negative @xmath9 low values of @xmath19 are experimentally excluded since they lead to an inverted mass hierarchy between the light chargino and the lightest neutralino , @xmath215 . finally , we assess how much of an improvement can be expected after a high - luminosity run with a total of @xmath216 integrated luminosity at @xmath207 gev . we assume that with the increased luminosity the experimental errors would be reduced by a factor of 2 . we furthermore include the measurement of the mass difference @xmath217 , which was not studied here . for the @xmath6 parameter the increased luminosity narrows the allowed region by 23.5 gev , as can be seen in the last rows of tabs . [ tab : fit124 ] and [ tab : fit127 ] . in case of the @xmath9 and @xmath10 parameters the results are shown in the bottom panels of figs . [ fig : deltamcal124 ] and [ fig : deltamcal127 ] for scenarios dm1600 and dm770 , respectively . the neutralino mass difference , @xmath217 , helps to break the correlation between @xmath9 and @xmath10 from the chargino lsp mass difference , since it exhibits a different dependence on the fundamental parameters , in particular by introducing @xmath19 , see eqs . and . hence , the inclusion of the neutralino mass difference helps to constrain the parameters in the low @xmath19 regime , where the @xmath19 dependence is numerically most significant . this effect would not be achieved by solely increasing the luminosity . the absolute mass measurements , which are obviously improved with higher statistics , have only a very limited impact on determination of the multi - tev parameters , @xmath9 and @xmath10 . the neutralino mass difference , due to its different dependence on the fundamental parameters , provides important additional information to the fit . this underlines the importance of including this measurement in future analyses . contours for determination of @xmath9 and @xmath10 in scenario dm1600 . the input values of @xmath9 and @xmath10 are indicated by the star , and @xmath218 . * top : * the fit using input values listed in tab . [ tab : input ] for three different values of @xmath19 . * bottom : * projected fit results after a high - luminosity run , @xmath219 , with experimental errors improved by factor 2 and with the mass difference @xmath217 measurement included for a fixed @xmath220 and @xmath221 . for @xmath222 , other values of @xmath19 do not further extend allowed regions and the respective contours are inside the one for @xmath220 . for @xmath223 , solutions exist only for @xmath224 near @xmath225.,title="fig:",scaledwidth=100.0% ] contours for determination of @xmath9 and @xmath10 in scenario dm1600 . the input values of @xmath9 and @xmath10 are indicated by the star , and @xmath218 . * top : * the fit using input values listed in tab . [ tab : input ] for three different values of @xmath19 . * bottom : * projected fit results after a high - luminosity run , @xmath219 , with experimental errors improved by factor 2 and with the mass difference @xmath217 measurement included for a fixed @xmath220 and @xmath221 . for @xmath222 , other values of @xmath19 do not further extend allowed regions and the respective contours are inside the one for @xmath220 . for @xmath223 , solutions exist only for @xmath224 near @xmath225.,title="fig:",scaledwidth=100.0% ] contours for determination of @xmath9 and @xmath10 in scenario dm770 . the input values of @xmath9 and @xmath10 are indicated by the star , and @xmath226 . * top : * the fit using input values listed in tab . [ tab : input ] for three different values of @xmath19 . * bottom : * projected fit results after a high - luminosity run , @xmath219 , with experimental errors improved by factor 2 and with the mass difference @xmath217 measurement included for fixed @xmath220 . other values of @xmath19 do not further extend allowed regions and the respective contours are inside the shown one . , title="fig:",scaledwidth=100.0% ] contours for determination of @xmath9 and @xmath10 in scenario dm770 . the input values of @xmath9 and @xmath10 are indicated by the star , and @xmath226 . * top : * the fit using input values listed in tab . [ tab : input ] for three different values of @xmath19 . * bottom : * projected fit results after a high - luminosity run , @xmath219 , with experimental errors improved by factor 2 and with the mass difference @xmath217 measurement included for fixed @xmath220 . other values of @xmath19 do not further extend allowed regions and the respective contours are inside the shown one . , title="fig:",scaledwidth=100.0% ] for the scenario dm1600 , @xmath10 is now constrained in the @xmath227@xmath228 tev range , while @xmath80 in the 16 tev range . if @xmath229 its value is confined in a small range , @xmath230 for moderate values of @xmath19 , @xmath224 . again , such a scenario could in principle be probed with an upgraded centre - of - mass energy . also for dm770 one obtains a visible improvement , but still with rather broad ranges of allowed values for @xmath9 and @xmath10 ; however , we only found valid solutions for @xmath80 . in both cases , the contours for other values of @xmath19 are mostly covered by the high @xmath19 contour and are therefore not shown separately . this result is due to the weak dependence on @xmath19 of the complete neutralino / chargino sector , cf . section [ sec : pheno ] . additional measurements , involving e.g. third - generation sfermions , would be needed to obtain meaningful constraints on @xmath19 . the current lhc results provide strong bounds on constrained supersymmetric models . in particular , the measured higgs mass of @xmath231 gev results in a tension with the naturalness argument . however , in a specific class of scenarios , a rather light @xmath6-parameter in conjunction with a heavy coloured susy and gaugino spectrum can be well - motivated from a model - building perspective . such scenarios result in a strong mass degeneracy between the higgsino - like light neutralinos and chargino , @xmath232 and @xmath2 . due to the close mass degeneracies , three - body decays as well as radiative decays are dominant leading to extremely challenging experimental signatures . we therefore performed a simulation for the ild detector in two different benchmark scenarios at a centre - of - mass energy of @xmath209gev , assuming an integrated luminosity of @xmath209fb@xmath4 each for the two relevant polarisation configurations , which corresponds to about @xmath128 years of ilc operation with design accelerator parameters . in both scenarios , the @xmath6-parameter is around @xmath233gev , but the susy gaugino parameters @xmath9 and @xmath10 are in the multi - tev region . this leads to mass differences between the light chargino and lightest neutralino of about @xmath234gev and @xmath235mev in the two benchmarks , respectively . these small mass differences lead to a very soft transverse momentum spectrum of the visible decay products , which ends at @xmath236 and @xmath237gev for the two different scenarios . in this kinematic regime , background processes from collisions involving the photon component of the electron and positron beams are about @xmath236 orders of magnitude larger than the signal . however , they can be heavily reduced by requiring the signal to be accompanied by a hard isr photon . in conjunction with the known beam energy , the @xmath2 and @xmath238 masses can be determined with a precision between @xmath72@xmath157 from the reduced centre - of - mass energy of the system recoiling against the photon . the mass difference with the lsp can be reconstructed by boosting the visible decay products into the rest frame of the produced gaugino pair . this allows one to clearly resolve the lightest and next - to - lightest susy particles and to determine the corresponding mass difference in the two benchmarks with a relative accuracy of 17% and 5% , respectively . finally , the polarised cross sections can be measured with an accuracy of a few percent . the polarisation dependency , in particular the scaling factor between polarised and unpolarised cross sections , gives a clear indication of the higgsino nature . in the last step of this analysis , we investigated whether the relevant susy parameters , namely the gaugino mass parameters @xmath9 and @xmath10 , the @xmath6-parameter and @xmath122 can be extracted from the experimental observables . we find that the @xmath6-parameter can be determined to @xmath239 . in case of @xmath9 and @xmath10 , lower bounds in the multi - tev range can be set , which depend slightly on the value of @xmath122 , which can not be fixed from these measurements alone . additional solutions arise if one allows the bino mass parameter to be negative , @xmath81 . some of these , leading to a light @xmath214 , could be probed by direct production of the additional light neutralino . if the measurement uncertainties could be reduced by a factor of 2 , e.g. by increasing the luminosity , or by improving the analysis , and if a measurement of the mass difference between @xmath238 and @xmath240 could be performed with the same precision as for the chargino mass difference , the constraints on @xmath9 and @xmath10 would be significantly more restrictive and become less dependent on @xmath122 . our results demonstrate the strong physics potential of the ilc . high precision is mandatory in order to resolve such challenging scenarios that could be easily missed at the lhc . the high experimental precision of thr ilc of course has to be balanced by suitable higher order predictions implemented in a monte carlo generator on a fully differential level . first steps in this direction have already been presented in @xcite , but a full study including decays would be needed and is in the line of future work . we thank frank gde for providing the tracking efficiency histograms in presence of pair background . we thank the ilc generator group for providing the sm whizard samples . we thank aoife bharucha for discussions on loop corrections . we thankfully acknowledge the support by the dfg through the sfb 676 `` particles , strings and the early universe '' . this work has been partially supported by the micinn , spain , under contract fpa2010 - 17747 ; consolider - ingenio cpan csd2007 - 00042 . we thank as well the comunidad de madrid through proyecto hephacos s2009/esp-1473 and the european commission under contract pitn - ga-2009 - 237920 . s. chatrchyan _ et al . _ [ cms collaboration ] , phys . b * 716 * ( 2012 ) 30 [ arxiv:1207.7235 [ hep - ex ] ] . g. aad _ et al . _ [ atlas collaboration ] , eur . j. c * 73 * ( 2013 ) 2362 [ arxiv:1212.6149 [ hep - ex ] ] ; atlas - conf-2013 - 054 ; atlas - conf-2013 - 047 ; atlas - conf-2013 - 037 ; atlas - conf-2013 - 026 ; atlas - conf-2013 - 024 ; atlas - conf-2013 - 007 ; atlas - conf-2012 - 144 ; atlas - conf-2012 - 140 . s. chatrchyan _ et al . _ [ cms collaboration ] , phys . d * 87 * ( 2013 ) 072001 [ arxiv:1301.0916 [ hep - ex ] ] ; eur . j. c * 73 * ( 2013 ) 2404 [ arxiv:1212.6428 [ hep - ex ] ] ; jhep * 1303 * ( 2013 ) 037 [ arxiv:1212.6194 [ hep - ex ] ] , jhep * 1303 * ( 2013 ) 111 [ arxiv:1211.4784 [ hep - ex ] ] ; phys . b * 725 * ( 2013 ) 243 [ arxiv:1305.2390 [ hep - ex ] ] ; eur . j. c * 73 * ( 2013 ) 2568 [ arxiv:1303.2985 [ hep - ex ] ] ; eur . j. c * 73 * ( 2013 ) 2493 [ arxiv:1301.3792 [ hep - ex ] ] ; arxiv:1301.2175 [ hep - ex ] ; phys . * 111 * ( 2013 ) 081802 [ arxiv:1212.6961 [ hep - ex ] ] ; cms - pas - sus-13 - 011 ; cms - pas - sus-13 - 008 ; cms - pas - sus-13 - 007 . [ atlas collaboration ] atl - phys - pub-2013 - 007 . j. beringer _ [ particle data group collaboration ] , phys . d * 86 * ( 2012 ) 010001 ; a. heister _ et al . _ [ aleph collaboration ] , phys . b * 533 * ( 2002 ) 223 [ hep - ex/0203020 ] ; j. abdallah _ et al . _ [ delphi collaboration ] , eur . phys . j. c * 31 * ( 2003 ) 421 [ hep - ex/0311019 ] ; m. acciarri _ et al . _ [ l3 collaboration ] , phys . b * 482 * ( 2000 ) 31 [ hep - ex/0002043 ] ; g. abbiendi _ et al . _ [ opal collaboration ] , eur . j. c * 35 * ( 2004 ) 1 [ hep - ex/0401026 ] . m. papucci , j. t. ruderman and a. weiler , jhep * 1209 * ( 2012 ) 035 [ arxiv:1110.6926 [ hep - ph ] ] . r. allahverdi , b. dutta and k. sinha , phys . d * 86 * ( 2012 ) 095016 [ arxiv:1208.0115 [ hep - ph ] ] . c. h. chen , m. drees and j. f. gunion , phys . * 76 * ( 1996 ) 2002 [ hep - ph/9512230 ] , addendum / erratum : hep - ph/9902309 . h. baer , v. barger and p. huang , jhep * 1111 * ( 2011 ) 031 [ arxiv:1107.5581 [ hep - ph ] ] . h. baer , v. barger , p. huang and x. tata , jhep * 1205 * ( 2012 ) 109 [ arxiv:1203.5539 [ hep - ph ] ] . f. brmmer and w. buchmller , jhep * 1107 * ( 2011 ) 010 [ arxiv:1105.0802 [ hep - ph ] ] . f. brmmer and w. buchmller , jhep * 1205 * ( 2012 ) 006 [ arxiv:1201.4338 [ hep - ph ] ] . h. p. nilles , s. ramos - sanchez , m. ratz and p. k. s. vaudrevange , eur . j. c * 59 * ( 2009 ) 249 [ arxiv:0806.3905 [ hep - th ] ] . a. maharana and e. palti , int . j. mod . phys . a * 28 * ( 2013 ) 1330005 [ arxiv:1212.0555 [ hep - th ] ] . h. e. haber and g. l. kane , phys . * 117 * ( 1985 ) 75 . a. bartl , h. fraas , w. majerotto and n. oshimo , phys . d * 40 * ( 1989 ) 1594 . ; m. m. el kheishen , a. a. aboshousha and a. a. shafik , phys . rev . d * 45 * ( 1992 ) 4345 . ; m. guchait , z. phys . c * 57 * ( 1993 ) 157 [ erratum - ibid . c * 61 * ( 1994 ) 178 ] ; g. j. gounaris , c. le mouel and p. i. porfyriadis , phys . d * 65 * ( 2002 ) 035002 [ hep - ph/0107249 ] . s. y. choi , j. kalinowski , g. a. moortgat - pick and p. m. zerwas , eur . j. c * 22 * ( 2001 ) 563 [ addendum - ibid . c * 23 * ( 2002 ) 769 ] [ hep - ph/0108117 ] . j. f. gunion and h. e. haber , phys . d * 37 * ( 1988 ) 2515 . s. ambrosanio and b. mele , phys . rev . d * 55 * ( 1997 ) 1399 [ erratum - ibid . d * 56 * ( 1997 ) 3157 ] [ hep - ph/9609212 ] . s. y. choi , m. guchait , j. kalinowski and p. m. zerwas , phys . b * 479 * ( 2000 ) 235 [ hep - ph/0001175 ] . a. bharucha , a. fowler , g. moortgat - pick and g. weiglein , jhep * 1305 * ( 2013 ) 053 [ arxiv:1211.3134 [ hep - ph ] ] . a. bharucha , private communication . a. bharucha , j. kalinowski , g. moortgat - pick , k. rolbiecki and g. weiglein , eur . j. c * 73 * ( 2013 ) 2446 [ arxiv:1211.3745 [ hep - ph ] ] . g. a. moortgat - pick and h. fraas , phys . d * 59 * ( 1999 ) 015016 [ hep - ph/9708481 ] ; g. a. moortgat - pick , h. fraas , a. bartl and w. majerotto , eur . j. c * 7 * ( 1999 ) 113 [ hep - ph/9804306 ] ; g. a. moortgat - pick , h. fraas , a. bartl and w. majerotto , eur . j. c * 9 * ( 1999 ) 521 [ erratum - ibid . c * 9 * ( 1999 ) 549 ] [ hep - ph/9903220 ] . h. e. haber and d. wyler , nucl . b * 323 * ( 1989 ) 267 . m. berggren , t. han , j. list , s. padhi , s. su and t. tanabe , `` electroweakino searches : a comparative study for lhc and ilc ( a snowmass white paper ) , '' arxiv:1309.7342 [ hep - ph ] . k. yokoya and p. chen , lect . notes phys . * 400 * ( 1992 ) 415 . p. chen , phys . d * 46 * ( 1992 ) 1186 . b. c. allanach , comput . commun . * 143 * ( 2002 ) 305 [ hep - ph/0104145 ] . t. fritzsche and w. hollik , eur . j. c * 24 * ( 2002 ) 619 [ hep - ph/0203159 ] ; t. fritzsche , berechnung von observablen zur supersymmetrischen teilchenerzeugung an hochenergie - collidern unter einschluss hoeherer ordnungen " , phd thesis , cuvillier verlag , gttingen 2005 , isbn 3 - 86537 - 577 - 4 . w. kilian , t. ohl and j. reuter , eur . j. c * 71 * ( 2011 ) 1742 [ arxiv:0708.4233 [ hep - ph ] ] . t. behnke , j. e. brau , p. n. burrows , j. fuster , m. peskin , m. stanitzki , y. sugimoto and s. yamada _ et al . _ , `` the international linear collider technical design report - volume 4 : detectors , '' ( 2013 ) , pp . 3236 , arxiv:1306.6329 [ physics.ins-det ] . m. skrzypek and s. jadach , z. phys . c * 49 * ( 1991 ) 577 . c. adolphsen , m. barone , b. barish , k. bsser , p. burrows , j. carwardine , j. clark and h. m. durand _ et al . _ , `` the international linear collider technical design report - volume 3.ii : accelerator baseline design , '' ( 2013 ) , pp . 810 , arxiv:1306.6328 [ physics.acc-ph ] . c. hensel , `` search for nearly mass degenerate charginos and neutralinos in @xmath14 collisions , '' desy - thesis-2002 - 047 . ` http://www-library.desy.de/cgi-bin/showprep.pl?desy-thesis-02-047 ` t. fritzsche and w. hollik , nucl . suppl . * 135 * ( 2004 ) 102 [ hep - ph/0407095 ] . w. oller , h. eberl and w. majerotto , phys . d * 71 * ( 2005 ) 115002 [ arxiv : hep - ph/0504109 ] . w. kilian , j. reuter and t. robens , eur . j. c * 48 * ( 2006 ) 389 [ hep - ph/0607127 ] . t. robens , `` event generation for next to leading order chargino production at the international linear collider , '' hep - ph/0610401 . g. moortgat - pick , t. abe , g. alexander , b. ananthanarayan , a. a. babich , v. bharadwaj , d. barber and a. bartl _ et al . _ , phys . * 460 * ( 2008 ) 131 [ hep - ph/0507011 ] . k. desch , j. kalinowski , g. a. moortgat - pick , m. m. nojiri and g. polesello , jhep * 0402 * ( 2004 ) 035 [ hep - ph/0312069 ] . g. moortgat - pick , i. fleck , s. riemann , f. simon , o. s. adeyemi , g. alexander , m. s. amjad and v. v. andreev _ et al . _ , `` helmholtz aliance linear collider forum : proceedings of the workshops hamburg , munich , hamburg 2010 - 2012 , germany , '' desy 12 - 123h . d. grellscheid and p. richardson , arxiv:0710.1951 [ hep - ph ] . m. bahr , s. gieseke , m. a. gigg , d. grellscheid , k. hamilton , o. latunde - dada , s. platzer and p. richardson _ et al . _ , eur . j. c * 58 * ( 2008 ) 639 [ arxiv:0803.0883 [ hep - ph ] ] . t. sjostrand , s. mrenna and p. z. skands , jhep * 0605 * ( 2006 ) 026 [ hep - ph/0603175 ] . t. behnke , j. e. brau , p. n. burrows , j. fuster , m. peskin , m. stanitzki , y. sugimoto and s. yamada _ et al . _ , `` the international linear collider technical design report - volume 4 : detectors , '' ( 2013 ) , pp . 181314 , arxiv:1306.6329 [ physics.ins-det ] . c. bartels , m. berggren and j. list , eur . j. c * 72 * ( 2012 ) 2213 [ arxiv:1206.6639 [ hep - ex ] ] . s. agostinelli _ et al . _ [ geant4 collaboration ] , nucl . instrum . a * 506 * ( 2003 ) 250 . t. abe _ et al . _ [ ild concept group - linear collider collaboration ] , `` the international large detector : letter of intent , '' arxiv:1006.3396 [ hep - ex ] . m. berggren , `` sgv 3.0 - a fast detector simulation , '' arxiv:1203.0217 [ physics.ins-det ] . d. schulte , `` study of electromagnetic and hadronic background in the interaction region of the tesla collider , '' tesla-97 - 08 . mora de freitas and h. videau , `` detector simulation with mokka / geant4 : present and future , '' lc - tool-2003 - 010 . f. james and m. roos , comput . commun . * 10 * ( 1975 ) 343 .

using a statistical sample of dark matter haloes drawn from a suite of cosmological-body simulations of the cold dark matter ( cdm ) model , we quantify the impact of a simulated halo s mass accretion and merging history on two commonly used measures of its dynamical state , the virial ratio and the centre of mass offset . quantifying this relationship is important because the degree to which a halo is dynamically equilibrated will influence the reliability with which we can measure characteristic equilibrium properties of the structure and kinematics of a population of haloes . we begin by verifying that a halo s formation redshift correlates with its virial mass and we show that the fraction of its recently accreted mass and the likelihood of it having experienced a recent major merger increases with increasing and decreasing . we then show that both and increase with increasing and decreasing , which implies that massive recently formed haloes are more likely to be dynamically unrelaxed than their less massive and older counterparts . our analysis shows that both and are good indicators of a halo s dynamical state , showing strong positive correlations with recent mass accretion and merging activity , but we argue that provides a more robust and better defined measure of dynamical state for use in cosmological-body simulations at . we find that is sufficient to pick out dynamically relaxed haloes at=0 . finally , we assess our results in the context of previous studies , and consider their observational implications . [ firstpage ] methods :-body simulations galaxies : formation galaxies : haloes cosmology : theory dark matter large - scale structure of universe

1109.2671

one of the fundamental assumptions underpinning modern theories of galaxy formation is that galaxies form and evolve in massive virialised haloes of dark matter @xcite . characterising the properties of these haloes is an important problem , both theoretically and observationally , and its study has been one of the main objectives of cosmological @xmath0-body simulations over the last two decades . the majority of these simulations have modeled halo formation and evolution in a purely cold dark matter ( cdm ) universe ( cf . * ? ? ? * ) , with the focus primarily on their equilibrium structure ( cf . various studies have revealed that cdm haloes in dynamical equilibrium are triaxial structures ( e.g. * ? ? ? * ) supported by velocity dispersion rather than rotation ( e.g. * ? ? ? * ) , with mass profiles that are divergent down to the smallest resolvable radius ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and an abundance of substructure ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? the qualification that a halo is in dynamical equilibrium is a particularly important one when seeking to characterise the structure and kinematics of simulated haloes in cosmological simulations . previous studies have shown that dynamically unrelaxed haloes tend to have lower central densities ( see , for example , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and higher velocity dispersions ( see , for example , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) than their dynamically relaxed counterparts . this means that a dynamically unrelaxed halo is likely to have a measurably lower concentration @xmath8 and higher spin parameter @xmath9 than its dynamically relaxed counterpart ( see , for example , * ? ? ? * ; * ? ? ? * ) , and so care must be taken to avoid contaminating halo samples with dynamically unrelaxed systems when measuring , for example , spin distributions ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and the correlation of halo mass and concentration @xmath10 ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? yet haloes do not exist in isolation , and the degree to which they are dynamically relaxed or unrelaxed bears the imprint of both their environment and their recent mass assembly and merging history . as previous studies have shown , dynamically unrelaxed haloes tend to have suffered one or more recent significant mergers ( e.g. * ? ? ? * ; * ? ? ? for this reason , it is common practice to use dynamical state and recent merging history interchangeably , with the understanding implicit that unrelaxed haloes are ones that have suffered one or more recent major mergers . however , it is important to establish this practice on a more quantitative footing and to assess how well a halo s dynamical state and its recent mass assembly history correlate . this is because of the need to identify robustly haloes that are in dynamical equilibrium or indeed disequilibrium in cosmological simulations for further discussion . ] . the goal of this paper is to quantify this relationship using a statistical sample of haloes drawn from cosmological @xmath0-body simulations of the cdm model . the cdm model is the ideal testbed for this study because of the fundamental role merging plays in halo mass assembly ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and because we expect massive haloes , which on average form later than their less massive counterparts , to have more violent recent merging histories . such an undertaking has practical implications . for example , if we want to robustly characterise the predicted variation of , say , concentration @xmath8 with virial mass @xmath4 on galaxy group and cluster mass scales ( @xmath11 ) , then it is essential that we can identify relaxed systems in a robust fashion . should we use mass assembly histories directly and select only haloes that have quiescent recent merging histories , or are commonly used measures that estimate dynamical state based on material within the halo s virial radius @xmath12 adequate ? this is particularly important for comparison with observations that provide crucial tests of the theory , such as the analysis the @xmath13 relation for groups and clusters drawn from the sloan digital sky survey by @xcite . in this paper , we examine how a halo s mass assembly history and dynamical state varies with its virial mass @xmath4 and its formation redshift , and adopt simple measures to characterise a halo s recent mass assembly and merging history namely , the fraction of mass assembled ( @xmath14 ) ; the rate of change of mass with redshift @xmath15 ; and the most significant merger @xmath16 . we compare these with two measures of the halo s dynamical state the virial ratio @xmath17 where @xmath18 and @xmath19 are the kinetic and gravitational potential energies of halo material ( cf . * ? ? ? * ; * ? ? ? * ) , and the centre - of - mass offset @xmath20 where @xmath21 and @xmath22 are the centres of density and mass of halo material and @xmath12 is the halo s virial radius ( cf . * ? ? ? * ; * ? ? ? * ; * ? ? ? previous studies have shown that both @xmath1 and @xmath2 increase in the aftermath of a major merger ( e.g. * ? ? ? * ; * ? ? ? * ) , and we will clarify precisely how they relate to a halo s mass assembly and merging activity in general . we note that our work develops earlier ideas presented in @xcite , in which we investigated the relationship between halo mass @xmath4 and spin @xmath9 , and it complements that of @xcite , who address related but distinct issues in their critique of the application of the virial theorem ( cf . [ ssec : virial_theorem ] ) to simulated high redshift dark matter haloes . + the layout of the paper is as follows . in [ sec : sims ] , we describe our approach to making initial conditions ; finding and analysing dark matter haloes in evolved outputs ; constructing merger trees of our dark matter haloes ; and our criteria for defining our halo sample . in [ sec : hierarchical ] , we examine the relationship between a halo s virial mass @xmath4 , its formation time @xmath3 and measures of its mass accretion and merging history . in [ sec : equilibrium ] , we present commonly used measures for assessing the dynamical state of a dark matter halo the virial ratio @xmath23 ( cf . [ ssec : virial_theorem ] ) and the centre - of - mass offset @xmath24 ( cf . [ ssec : centre_of_mass_offset ] ) and investigate how these measures correlate with @xmath4 and @xmath3 . in [ sec : linking_mass_assembly_and_dynamical_state ] we combine the insights from the previous two sections and show how a halo s dynamical state depends on its recent mass accretion and merging history . finally , we summarise our results in [ sec : summary ] and comment on the implications of our findings for both observational studies and galaxy formation modeling . we have run a series of @xmath25 particle cosmological @xmath0-body simulations following the formation and evolution of structure in the cdm model . we use a sequence of boxes of side @xmath26 varying between @xmath27 and @xmath28 from @xmath29=100 to @xmath30=0 . in each case we assume a flat cosmology with a dark energy term , with cosmological parameters @xmath31 , @xmath32 , @xmath33 , and a normalisation @xmath34 at @xmath35 . various properties of these simulations are summarised in table [ tab : sim_props ] . initial conditions were generated using a standard procedure that can be summarised as follows ; 1 . generate the cdm transfer function for the appropriate cosmological parameters ( @xmath36 and @xmath37 ) using the boltzmann code cmbfast @xcite . this is convolved with the primordial power spectrum @xmath38 , @xmath39 , to obtain the unnormalised power spectrum , which is normalised by requiring that the linear mass @xmath40 equal @xmath41 on a scale of @xmath42 at @xmath7=0 . 2 . create a statistical realisation of a gaussian random field of density perturbations in fourier space , whose variance is given by @xmath43 , where @xmath44 and whose mean is zero . take the inverse transform of the density field and compute positions and velocities using the zeldovich approximation . 4 . impose these positions and velocities on an initial uniform particle distribution such as a grid or `` glass '' . note that throughout our we use a `` glass''-like configuration as our initial uniform particle distribution @xcite . + all simulations were run using the parallel treepm code gadget2 @xcite with constant comoving gravitational softening @xmath45 and individual and adaptive timesteps for each particle , @xmath46 , where @xmath47 is the magnitude of a particle s gravitational acceleration and @xmath48 determines the accuracy of the time integration . .*properties of the simulations*. each of the simulations contains @xmath25 particles . in addition , @xmath26 is the comoving box length in units of @xmath49 ; @xmath50 is the number of runs in the series ; @xmath51 is the particle mass in units of @xmath52 ; @xmath45 is the force softening in comoving units of @xmath53 ; and @xmath54 is the halo mass corresponding to @xmath55=600 particles , in units of @xmath56 . [ cols="<,^,^,^,^,^",options="header " , ] [ tab : sim_props ] [ [ halo - catalogues ] ] halo catalogues + + + + + + + + + + + + + + + groups were identified using the mpi - enabled version of ahf , otherwise known as s alo inder @xcite . ahf is a modification of mhf ( mlapm s halo finder ; see * ? ? ? * ) that locates groups as peaks in an adaptively smoothed density field using a hierarchy of grids and a refinement criterion that is comparable to the force resolution of the simulation . local potential minima are calculated for each of these peaks and the set of particles that are gravitationally bound to the peaks are identified as groups that form our halo catalogue . for each halo in the catalogue we determine its centre - of - density @xmath21 ( using the iterative `` shrinking spheres '' method described in @xcite ) and identify this as the halo centre . from this , we calculate the halo s virial radius @xmath12 , which we define as the radius at which the mean interior density is @xmath57 times the critical density of the universe at that redshift , @xmath58 , where @xmath59 and @xmath60 are the hubble parameter at @xmath7 and the gravitational constant respectively . the corresponding virial mass @xmath4 is @xmath61 we adopt a cosmology- and redshift - dependent overdensity criterion , which for a @xmath62cdm cosmology with @xmath63 and @xmath64 gives @xmath65 at @xmath7=0 ( c.f . [ [ merger - trees ] ] merger trees + + + + + + + + + + + + halo merger trees are constructed by linking halo particles at consecutive output times ; * for each pair of group catalogues constructed at consecutive output times @xmath66 and @xmath67 , the `` ancestors '' of `` descendent '' groups are identified . for each descendent identified in the catalogue at the later time @xmath68 , we sweep over its associated particles and locate every ancestor at the earlier time @xmath66 that contains a subset of these particles . a record of all ancestors at @xmath66 that contain particles associated with the descendent at @xmath68 is maintained . * the ancestor at time @xmath66 that contains in excess of @xmath69 of these particles and also contains the most bound particle of the descendent at @xmath68 is deemed the _ main progenitor_. typically @xmath70 , i.e. the main progenitor contains in excess of half the final mass . each group is then treated as a node in a tree structure , which can be traversed either forwards , allowing one to identify a halo at some early time and follow it forward through the merging hierarchy , or backwards , allowing one to identify a halo and all its progenitors at earlier times . a degree of care must be taken when choosing which haloes to include in our sample , to ensure that our results are not affected by the finite resolution of our simulations . one of the key calculations in this study is of a halo s virial ratio @xmath23 ( see [ ssec : virial_theorem ] ) , where @xmath18 and @xmath19 are the kinetic and gravitational potential energies of material within @xmath12 . the gravitational potential energy is particularly sensitive to resolution ; if a halo is resolved with too few particles , its internal structure will not be recovered sufficiently accurately and the magnitude of @xmath19 will be underestimated . we estimate how many particles are needed to recover @xmath19 robustly from a @xmath0-body simulation in figure [ fig : monte_carlo_energy ] . here we generate monte carlo @xmath0-body realisations of a halo whose spherically averaged mass profile is described by the @xcite profile , @xmath71 here @xmath72 is the radius @xmath73 normalised to @xmath12 , @xmath74 is the concentration parameter and @xmath75 is the characteristic density , @xmath76 the resulting gravitational potential energy is given by @xmath77.\ ] ] in a @xmath0-body simulation or realisation , we calculate @xmath19 by randomly sampling particles within @xmath12 and rescaling ; this gives @xmath78 where there are @xmath79 particles in the halo , each of mass @xmath80 . we sample @xmath81 particles from @xmath79 , @xmath82 is the magnitude of the separation between particles @xmath83 at @xmath84 and @xmath85 at @xmath86 , and the prefactor @xmath87 accounts for particle sampling . @xmath45 is the gravitational softening and @xmath88 corresponds to the softening kernel used in gadget2 . for the monte carlo realisations in figure [ fig : monte_carlo_energy ] we set @xmath45 to be vanishingly small , but for the simulations we use @xmath45 as it is listed in table [ tab : sim_props ] . figure [ fig : monte_carlo_energy ] shows @xmath89 measured for monte carlo realisations of a halo with @xmath9010 and @xmath12=200 kpc as a function of @xmath79 . for comparison the horizontal dotted lines indicate the value of @xmath89 ( @xmath91 ) we expect from equation ( [ eq : wnfw ] ) . if @xmath92 or fewer , the measured @xmath89 deviates from the expected @xmath89 by greater than @xmath93 ; therefore we might regard @xmath94 as the lower limit on @xmath79 for a halo to be included in our sample . however , we adopt a more conservative @xmath95 in the remainder of this paper ; this is because the structure of simulated haloes are affected by finite gravitational softening ( cf . * ? ? ? * ) , they are are seldom ( if ever ) smooth and spherically symmetric ( e.g. * ? ? ? * ) , as we assumed in our simple calculation , and there can be a range of concentrations at a given mass @xcite , which will affect any estimate of @xmath19 as inspection of equation ( [ eq : wnfw ] ) reveals . , the potential energy will be inaccurate.,width=302 ] in this section we establish quantitative measures for a halo s mass accretion and merging histories , and we examine how these measures relate to virial mass @xmath4 and formation redshift @xmath3 . [ [ quantifying - formation - redshift ] ] quantifying formation redshift + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + varies with virial mass @xmath4 at @xmath7=0 . we determine @xmath3 directly from a halo s merger tree for a halo of @xmath4 identified at @xmath7=0 , we identify the redshift @xmath3 at which the mass of its main progenitor first exceeds half its virial mass at @xmath7=0 . data are binned using equally spaced bins in log@xmath96 ; filled circle and bars correspond to medians , upper and lower quartiles . the solid , upper and lower dashed curves corresponds to the median @xmath3 and its upper and lower quartiles predicted by extended press - schechter theory ( cf . * ? ? ? * ) .,width=302 ] we begin our analysis by verifying the correlation between virial mass @xmath4 and formation redshift @xmath3 for our halo sample . we adopt the convention of @xcite and define @xmath3 as the redshift at which the mass of the main progenitor of a halo of mass @xmath97 identified at @xmath7 first exceeds @xmath98 . this is equivalent to @xmath99 in the survey of halo formation redshift definitions examined by @xcite . our expectation is that more massive cdm haloes will assemble more of their mass at later times than their lower mass counterparts and this is borne out by figure [ fig : mass_versus_zform ] . here we show the variation of @xmath3 with @xmath4 for our halo sample ; the filled circles and bars indicate the medians and upper and lower quartiles respectively , within logarithmic mass bins of width 0.5 dex . the relationship between the mean and median @xmath3 with @xmath4 can be well approximated by @xmath100 and @xmath101 where @xmath102 is @xmath4 in units of @xmath103 . this is in very good agreement with the mean variation reported for the `` overall '' sample of haloes drawn from the millennium and millennium ii simulations ( cf . @xcite and @xcite respectively ) in table 3 of @xcite , who found @xmath104 we show also the variation predicted by extended press - schechter ( eps ) theory for our choice of cdm power spectrum see the solid and dashed curves , indicating the median , upper and lower quartiles of the distributions ( cf . these curves were generated using realisations of @xmath105 monte - carlo merger trees for haloes with @xmath7=0 masses in the range @xmath106 . we note a slight but systematic offset between the medians evaluated from the simulated haloes and those predicted by eps theory , such that the simulated haloes tend to form earlier than predicted . this effect has been reported previously by both @xcite and @xcite . [ [ quantifying - recent - mass - accretion - history ] ] quantifying recent mass accretion history + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + because more massive systems tend to form later than their less massive counterparts , it follows that the rate at which a halo assembles its mass should increase with increasing @xmath4 and decreasing @xmath3 . the recent comprehensive study by @xcite provides a useful fitting formula that captures the complexity of a halo s mass accretion history and allows haloes to be categorised into different types i to iv , which depend on their growth rates . however , we adopt two simple well - defined measures of a halo s mass accretion rate that have a straightforward interpretation ; * @xmath107 , the fraction of mass that has been accreted by a halo during a time interval @xmath108 ; and * @xmath109 , the rate of fractional change in a halo s virial mass with respect to redshift over a redshift interval @xmath110 . note that @xmath111 is equivalent to the @xmath111 free parameter used in @xcite . we find that @xmath107 and @xmath111 are sufficient as simple measures of the mass accretion rate and we use them in the remainder of this paper . for the fiducial timescale @xmath108 , we use twice the dynamical timescale @xmath112 estimated at the virial radius , @xmath113 note that @xmath112 depends only on @xmath7 and is the same for all haloes . for our adopted cosmological parameters , @xmath114 at @xmath7=0 , and so @xmath115 gyr which corresponds to a redshift interval of @xmath116 at @xmath7=0 . merging proceeds on a timescale @xmath117 , with @xmath118 as the mass ratio of the merger decreases . our adopted timescale of @xmath119 for the response of a halo to a merger is reasonable when compared to typical values of @xmath120 expected for haloes in cosmological simulations , as estimated by @xcite and @xmath121 that are consistent with the results of cosmological simulations . here @xmath85 is the specific angular momentum of a merging subhalo , @xmath122 is the specific angular momentum of the circular orbit corresponding to the subhalo s orbital energy @xmath123 , and @xmath124 is the radius corresponding to this circular orbit . ] . we determine both @xmath125 and @xmath111 directly from each halo s merger tree by tracking @xmath97 of its main progenitor over the interval @xmath110 ; @xmath111 is obtained by taking the natural logarithm of the progenitor mass at each redshift and estimating its value by linear regression . haloes that have high mass accretion rates will have @xmath126 and @xmath127 . and formation redshift @xmath3 identified at @xmath7=0 , we follow its merger tree back for one dynamical time @xmath112 ( @xmath128 gyrs , @xmath129 ) and characterise its mass accretion history using two measures . the first is @xmath125 , the fraction of mass accreted over @xmath112 ( upper panels ) , and the second is @xmath111 , the average mass accretion rate of @xcite ( lower panels ) . data points and bars correspond to medians and upper and lower quartiles . note that we use equally - spaced logarithmic bins in @xmath4 and @xmath3.,width=302 ] in fig . [ fig : mass_versus_accretion_measures ] we show how a halo s mass accretion rate correlates with its virial mass and formation time . @xmath125 ( @xmath111 ) shows a steady monotonic increase ( decrease ) as @xmath4 increases over the range @xmath130 . for example , inspection of @xmath125 reveals that @xmath131 of the virial mass of a halo with @xmath132 has been accreted since @xmath133 , compared to @xmath134 for haloes with @xmath135 over the same period . @xmath125 ( @xmath111 ) shows a similar increase ( decrease ) with decreasing @xmath3 although it s interesting to note that the trend flattens off for haloes that form at @xmath136 . this analysis confirms our theoretical prejudice that more massive haloes and haloes that formed more recently tend to be the haloes with the measurably highest accretion rates . reassuringly , our results are in good agreement with the findings of recent studies . for example , @xcite examined the mass accretion and merging histories of a much larger sample of haloes drawn from the millennium and millennium - ii simulations and found that the mean instantaneous mass accretion rate varies with halo mass as @xmath137 ; this compares favourably with our equivalent measure , @xmath138 . @xcite looked at halo accretion rates , normalised to their maximum masses , over the redshift interval @xmath7=0.1 to 0 for haloes with masses @xmath139 and found only a weak dependence on halo mass , with higher mass haloes have higher rates . this is consistent with with our results for @xmath111 , whose median value changes by @xmath140 over the same range in halo mass . experienced by each halo since @xmath7=0.5 , where @xmath141 and @xmath142 correspond to the initial and final redshifts . we then compute the fraction of haloes @xmath143 at a given virial mass ( left hand panel ) and given formation redshift ( right hand panel ) that have experienced mergers with mass ratios @xmath16 in excess of @xmath144 ( filled circles ) , @xmath145 ( filled squares ) and @xmath146 ( filled triangles).,width=302 ] [ [ quantifying - recent - merger - activity ] ] quantifying recent merger activity + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + both @xmath125 and @xmath111 provide useful insights into a halo s total mass accretion rate , but they can not distinguish between smooth and clumpy accretion . in fig . [ fig : frequency_of_mergers ] we focus specifically on a halo s merger history by considering the likelihood that a halo of a given @xmath4 ( left hand panel ) or @xmath3 ( right hand panel ) has experienced _ at least one _ merger with a mass ratio @xmath16 since @xmath7=0.6 . each halo identified at @xmath7=0 has a unique merger history , which characterises not only how its @xmath4 grows as a function of time but also details of mergers it has experienced over time . using this merger history , we construct the distribution of mass ratios of mergers @xmath147 experienced by a halo of a given @xmath4 or @xmath3 between @xmath148 . we define @xmath149 where @xmath150 is the mass of the less massive halo prior to its merging with the more massive halo , @xmath151 is the virial mass of the more massive halo once the less massive halo has merged with it , and @xmath152 and @xmath142 are the redshifts of consecutive simulation snapshots . the maximum value of @xmath147 for a given halo gives us its @xmath16 and we use this to compute the fraction of haloes of a given @xmath4 or @xmath3 that have @xmath16 in excess of @xmath144 ( filled circles ) , @xmath145 ( filled squares ) and @xmath146 ( filled triangles ) . [ fig : frequency_of_mergers ] reveals that mergers with higher mass ratios ( i.e. minor mergers ) are more common than mergers with lower mass ratios ( i.e. major mergers ) , independent of @xmath4 and @xmath3 , and that more massive ( older ) haloes tend to experience more mergers than their lower mass ( younger ) counterparts . for example , the likelihood that a @xmath153 galaxy - mass halo experiences a merger with @xmath154 is @xmath155 , compared to @xmath156 for @xmath157 . in contrast , the likelihood that a @xmath158 cluster - mass halo experiences mergers with @xmath159 is @xmath160 . interestingly , we find that the fraction of haloes that have experienced a merger more significant than @xmath16 increases with @xmath4 approximately as @xmath161 . these results are broadly in agreement with the findings of @xcite . inspection of the leftmost panel of their figure 7 shows the mean number of mergers with mass ratios greater than 1:10 and 1:3 between @xmath7=0 and @xmath162 increases with increasing halo mass , such that a @xmath163 has a likelihood of @xmath164 ( @xmath165 ) to have experienced a merger with @xmath154 , and a likelihood of @xmath166 ( @xmath164 ) to have experienced a merger with @xmath167 . + as a function of halo mass and formation redshift*. we show how the fraction of haloes whose most significant merger s mass ratio is less than @xmath16 , as a function of virial mass ( left hand panel ) and formation redshift ( right hand panel ) . note that we select only haloes that have @xmath168 , and we consider only mergers between @xmath7=0.5 and @xmath7=0 . in the key , the numbers in brackets correspond to the lower and upper bounds in @xmath4 and @xmath3.,width=302 ] in fig . [ fig : cum_deltamax ] we show the full ( cumulative ) distributions of @xmath16 for haloes split into bins according to @xmath4 ( left hand panel ) and @xmath3 ( right hand panel ) ; note that we consider only haloes with @xmath169 . interestingly this figure reveals that the probability distribution of @xmath16 is insensitive to @xmath4 , but depends strongly on @xmath3 . for example , the median @xmath170 , independent of @xmath4 whereas it increases from @xmath171 for haloes with @xmath172 to @xmath173 for haloes with @xmath174 and @xmath175 for haloes with @xmath176 . + figs . [ fig : mass_versus_accretion_measures ] to [ fig : cum_deltamax ] demonstrate that there is a strong correlation at @xmath7=0 between a halo s virial mass @xmath4 , its formation redshift @xmath3 and the rate at which it has assembled its mass through accretion and merging over the last @xmath112 or equivalently @xmath177 . we use these results in [ sec : linking_mass_assembly_and_dynamical_state ] , where we investigate the degree to which a halo s @xmath4 , @xmath3 and mass accretion rate affect the degree to which it is in dynamical equilibrium . in this section we describe the two commonly used quantitative measures for a halo s dynamical state , the virial ratio @xmath1 and the centre of mass offset @xmath2 , and we examine their relationship with virial mass @xmath4 and formation redshift @xmath3 . the virial ratio @xmath1 is commonly used in cosmological @xmath0-body simulations as a measure of a halo s dynamical state ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? it derives from the virial theorem , @xmath178 where @xmath179 is the moment of inertia , @xmath18 is the kinetic energy , @xmath180 is the virial , and @xmath181 is the surface pressure integrated over the bounding surface of the volume within which @xmath179 , @xmath18 and @xmath19 are evaluated ( cf . * ? ? ? provided the system is isolated and bounded , the virial @xmath19 is equivalent to the gravitational potential energy . while not strictly true for haloes that form in cosmological @xmath0-body simulations , the convention has been to evaluate @xmath19 as the gravitational potential energy with this caveat in mind ( e.g. * ? ? ? we follow this convention and treat @xmath19 as the gravitational potential energy computed using equation ( [ eq : potential ] ) . if the system is in a steady state and in the absence of surface pressure , equation ( [ eq : virialtheorem ] ) reduces to @xmath182 , which can be written more compactly as @xmath183 ( e.g. * ? ? ? we refer to the ratio @xmath184 as the virial ratio and we expect @xmath185 for dynamically relaxed haloes . however , we might expect @xmath181 to be important for haloes that form in cosmological @xmath0-body simulations ; in this case @xcite have proposed modifying the virial ratio to obtain @xmath186 we calculate both @xmath18 and @xmath19 using all material within @xmath12 , while we follow @xcite by computing the surface pressure contribution from all particles that lie in a spherical shell with inner and outer radii of 0.8 and 1.0 @xmath12 , @xmath187 here @xmath188 corresponds to the volume of this shell and @xmath189 are the particle velocities relative to the centre of mass velocity of the halo . the energy associated with the surface pressure is therefore @xmath190 where @xmath191 is the median radius of the shell . + and @xmath192*. we bin all haloes in our sample at @xmath7=0 according to their @xmath1 and evaluate the median @xmath192 within each bin . the upper and lower quartiles of the distributions in @xmath1 and @xmath192 are indicated by bars.,width=302 ] figure [ fig : eta_vs_etadash_z0 ] shows how the median @xmath1 and @xmath192 for the haloes in our sample compare , with bars indicating the upper and lower quartiles of the distributions . we might expect that @xmath193 and insensitive to variation in @xmath1 ; however , this figure reveals that the relationship between @xmath1 and @xmath192 is not so straightforward . haloes that we would expect to be dynamically relaxed , with @xmath194 , have values of @xmath192<0 , suggesting that @xmath181 tends to over - correct . similar behaviour has been noted in both @xcite and @xcite for high redshift haloes ( @xmath195 ) . the relation between the median @xmath1 and @xmath192 is flat @xmath196 but rises sharply from @xmath197 to peak at @xmath198 before declining sharply for @xmath199 to a median of @xmath200 in the last plotted bin . interestingly , the width of the @xmath192 distribution increases with @xmath1 ; if @xmath1 tracks recent major merging activity as we expect , then this suggests that @xmath192 and consequently the surface pressure correction term @xmath181 is sensitive to mergers but in a non - trivial way . another commonly used measure of a halo s dynamical state is the centre - of - mass offset @xmath2 , @xmath201 which measures the separation between a halo s centre - of - density @xmath21 ( calculated as described in [ ssec : halo_properties ] ) and its centre - of - mass , calculated using all material within @xmath12 , normalised by @xmath12 ( cf . * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? @xmath2 is used as a substructure statistic , providing an estimate of a halo s deviations from smoothness and spherical symmetry . the expectation is that the smaller the @xmath2 , the more relaxed the halo ; for example , @xcite define dynamically relaxed haloes to be those with @xmath202 , while @xcite adopt @xmath203 . @xcite favoured a more conservative @xmath204 based on a thorough analysis . + and virial ratios @xmath1 and @xmath192*. we can clearly see the relation which is confirmed by measuring a spearman rank coefficient of 0.45 whereas we find an anti - correlation with spearman rank coefficient of -0.18 for @xmath192.,width=302 ] we can get a sense of how well @xmath2 measures the dynamical state of a halo by comparing it to @xmath1 and @xmath192 . in figure [ fig : deltar_vs_eta ] we plot the median @xmath1 and @xmath192 ( filled circles and squares respectively ) against the median @xmath2 ; as before , bars indicate the upper and lower quartiles of the distributions . this figure shows that both @xmath1 and @xmath192 correlate with @xmath2 but in different senses ; as @xmath2 increases , @xmath1 increases while @xmath192 decreases . the increase ( decrease ) is a gradual one ; for example , for @xmath6 , the median @xmath1 is flat with a value of @xmath205 , but for @xmath206 there is a sharp increase and @xmath207 , @xmath208 . although direct comparison is difficult , a similar trend can be gleaned from figure 2 of @xcite . we use the spearman rank correlation coefficient to assess the strength of the correlation between @xmath2 and @xmath209 ( cf . * ? ? ? * ) , and find strong positive and negative correlations for @xmath1 ( spearman rank coeffecient @xmath73=0.97 ) and @xmath192 ( @xmath73=-0.95 ) respectively . this is suggestive as we show below , @xmath2 correlates more strongly with merging activity than either of @xmath1 or @xmath192 ( cf . figure [ fig : deltar_vs_mah ] in [ sec : linking_mass_assembly_and_dynamical_state ] ) . both @xmath2 and @xmath1 increase with strength of merging activity , whereas @xmath192 appears to be over - corrected by @xmath181 ( as we have noted above ) . from this we conclude that @xmath181 ( as we evaluate it ) correlates with significant merger activity , which is confirmed by a spearman rank coefficient of 0.38 for the correlation between @xmath210 and @xmath16 . interestingly @xcite examined the correlation between @xmath2 and @xmath192 for high redshift haloes ( @xmath211 ) and noted a tendency for haloes with small values of @xmath192 to have larger values of @xmath2 . inspection of their figure 4 shows that this is true for haloes with @xmath212 ; for @xmath213 the relation with @xmath192 is flat . @xcite argue that , because there is no systematic shift in @xmath192 for @xmath214 , @xmath2 is not a useful measure of dynamical state at high redshifts . in figures [ fig : virial_ratio_measures ] and [ fig : deltar_measures ] we examine how @xmath1 , @xmath192 and @xmath2 vary with @xmath4 ( left hand panels ) and @xmath3 ( right hand panels ) for the halo population at @xmath7=0 . haloes are sorted in bins of equal width in mass ( @xmath215=0.5 dex ) and redshift ( @xmath110=0.25 ) , and we plot the median @xmath1/@xmath192/@xmath216 within each bin against the median @xmath4/@xmath3 ; bars indicate the upper and lower quartiles of the respective distributions . for reference , we also plot a horizontal dotted line in each panel of figure [ fig : virial_ratio_measures ] to indicate a virial ratio of unity . = 0 , we plot the median @xmath1 and @xmath192 versus @xmath4 ( left hand panel ) and @xmath3 ( right hand panel ) using equally spaced bins in @xmath217 and @xmath3 . data points and bars correspond to medians and upper and lower quartiles.,width=302 ] = 0 , we plot the median centre - of - mass offset @xmath2 versus @xmath4 ( left hand panel ) and @xmath3 ( right hand panel ) using equally spaced bins in @xmath217 and @xmath3 . data points and bars correspond to medians and upper and lower quartiles.,width=302 ] and @xmath192 at redshift @xmath35.,width=302 ] because more massive haloes tend to form at later times , and because these haloes tend to assemble a larger fraction of their mass more recently , we expect that @xmath1 and @xmath2 should increase with increasing @xmath4 and decreasing @xmath3 . this is borne out in figures [ fig : virial_ratio_measures ] and [ fig : deltar_measures ] . we find that the mean and median @xmath2 increases steadily with increasing @xmath4 as @xmath218 and @xmath219 where , as before , @xmath102 is @xmath4 in units of @xmath103 . this is consistent with the result of ( * ? ? * see their figure 9 ) , who found a similar trend for @xmath2 to increase with @xmath220 for a sample of cluster mass haloes ( @xmath221 ) in a @xmath222cdm model . their typical values of @xmath2 are offset to higher values than we find , but this can be understood as an effect of @xmath62 , the merging rate being suppressed in the @xmath223cdm model compared to the @xmath222cdm model . similarly , @xmath2 varies strongly with @xmath3 ; for @xmath224 we find that @xmath225 compared to @xmath226 for @xmath227 . the mean and median @xmath1 exhibit similar behaviour , increasing with increasing @xmath4 , albeit weakly , as , @xmath228 and @xmath229 this means that @xmath1 is systematically greater than unity for all @xmath4 that we consider @xmath230 for a typical @xmath231 halo , compared to @xmath232 for a typical @xmath233 halo . the same gradual increase in @xmath1 with decreasing @xmath3 is also apparent . as we might have anticipated from inspection of figures [ fig : eta_vs_etadash_z0 ] and [ fig : deltar_vs_eta ] , @xmath192 is systematically smaller than unity . its variation with @xmath4 is negligible ( @xmath234 ; a little surprising , when compared to @xmath235 at @xmath7=1 , as reported by @xcite ) but there is a trend for the median @xmath192 to decrease with decreasing @xmath3 . this makes sense because @xmath181 increases with the significance of recent mergers and haloes that have had recent major mergers tend to have smaller @xmath3 . this effect is also noticeable in the width of the @xmath192 distributions in each bin ( as measured by the bars ) , which are larger than than the corresponding widths of the @xmath1 distribution . we look at this effect in more detail by plotting the distributions of @xmath1 and @xmath192 shown in figure [ fig : eta_vs_mass_z0 ] . here it is readily apparent that there is a systematic shift towards larger @xmath1 as @xmath4 increases . interestingly the @xmath192 distribution remains centred on @xmath236 , but it spreads with increasing @xmath4 ; again , this suggests the sensitivity of @xmath192 to recent merging activity . [ fig : eta_vs_etadash_z0 ] to [ fig : eta_vs_mass_z0 ] demonstrate that there is a strong correlation at @xmath7=0 between a halo s virial mass @xmath4 , its formation redshift @xmath3 and its dynamical state , as measured by the virial ratio @xmath1 and the centre - of - mass offset @xmath2 . in contrast , the correlation with @xmath192 is more difficult to interpret , especially when @xmath1 is large . in these cases , we expect significant merging activity and as we note above , the correction by the surface pressure term @xmath181 increases the width of the original @xmath1 distribution by a factor of @xmath237 2 - 3 . it s also noteworthy that the median @xmath192 is systematically offset below unity . for this reason we argue that @xmath192 is not as useful a measure of a halo s dynamical state as @xmath1 . we have established quantitative measures of a halo s mass assembly and merging history and its dynamical state in the previous two sections , and we have investigated how these relate separately to a halo s virial mass @xmath4 and its formation redshift @xmath3 . in this final section we examine the relationship between a halo s mass assembly history and its dynamical state directly . , the fraction of mass accreted over @xmath112 ( left hand panel ) , and @xmath111 , the mean accretion rate over @xmath112 ( right hand panel ) , correlate with the standard ( @xmath1 , filled circles ) and corrected ( @xmath192 , filled squares ) virial ratio . data points correspond to medians and bars correspond to the upper and lower quartiles.,width=302 ] , which measures the mass ratio of the most significant recent merger since @xmath7=0.5 , correlates with the standard ( @xmath1 ) and corrected ( @xmath192 ) virial ratios respectively . filled circles ( squares ) correspond to medians of @xmath1 ( @xmath192 ) , while bars indicate the upper and lower quartiles.,width=302 ] in figures [ fig : virial_ratio_versus_accretion ] and [ fig : virial_ratio_versus_deltamax ] we show explicitly how a halo s recent mass accretion and merging history impacts on its virial ratio . as in section [ sec : hierarchical ] , we quantify a halo s mass accretion history by @xmath125 , the fractional increase in a halo s mass over the period @xmath112 ( equivalent to a redshift interval @xmath116 at @xmath7=0 ) , and @xmath111 , the mean accretion rate over the period @xmath238 ) . we use @xmath16 , the mass ratio of the most significant merger experienced by the halo over @xmath112 , to characterise a halo s recent merging history . and recent merging and accretion history*. here we examine whether @xmath2 correlates with the fraction of mass accreted over @xmath112 , @xmath239 , the mean accretion rate @xmath111 and the most significant merger @xmath16 . filled circles correspond to medians in the respective bins ; bars represent the upper and lower quartiles of the distribution.,width=302 ] we expect that the standard virial ratio @xmath1 should increase with increasing mass accretion rate and decreasing mass ratio of most significant merger , which is in good agreement with the behaviour that we observe . in particular , the median variation of @xmath1 with @xmath111 and @xmath16 can be well approximated by @xmath240 and @xmath241 ; the corresponding variation of @xmath2 can be well approximated by @xmath242 and @xmath243 . interestingly we note that the median corrected virial ratio @xmath192 declines with increasing mass accretion rate and mass ratio of most significant merger . both correlations indicate that merger events lead to a state that is less virialised , but , as we have noted already , the inclusion of the surface pressure term over - corrects the virial ratio . we see in figure [ fig : virial_ratio_versus_accretion ] that for @xmath244 , both the median @xmath1 and @xmath192 are flat ; @xmath245 whereas the median @xmath246 . above @xmath247 , the median @xmath1 increases sharply whereas it is the width of the @xmath192 distribution that shows the sharp increase . comparison with figure [ fig : virial_ratio_versus_deltamax ] provides further insight the median @xmath1 ( @xmath192 ) shows a gradual increase ( decrease ) with increasing @xmath16 , starting at @xmath248 ( @xmath236 ) for @xmath249 . for @xmath250=0.1 , @xmath251 ( @xmath246 ) . however , whereas the width of the @xmath1 distribution is largely insensitive to @xmath16 , the width of the @xmath192 distribution increases rapidly , bearing out our observations in the previous section . in figure [ fig : deltar_vs_mah ] we show how @xmath2 varies with @xmath125 , @xmath111 and @xmath16 . this reveals that @xmath2 increases with increasing mass accretion rate and mass ratio of most significant recent merger , as we would expect . although the scatter in the distribution is large , we can identify the remnants of recent major mergers ( @xmath252 ) as haloes with @xmath253 . haloes that have had relatively quiescent recent mass accretion histories ( @xmath254 , @xmath255 ) have @xmath6 . [ [ merging - timescale - dynamical - state ] ] merging timescale & dynamical state + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we conclude our analysis by investigating the timescale over which the effect of a merger can be observed in the virial ratio @xmath1 and the centre - of - mass offset @xmath2 . in figure [ fig : time_dep_dynstate ] we investigate how a typical halo s @xmath1 ( upper panel ) and @xmath2 ( lower panel ) , measured at @xmath35 , correlates with the redshift at which the halo suffered it s most significant merger , @xmath256 . for clarity , we focus on haloes for which @xmath257 , although we have verified that our results are not sensitive to the precise value of @xmath16 that we adopt ; filled circles correspond to medians and bars indicate upper and lower quartiles . the median @xmath1 increases with decreasing @xmath256 for @xmath258 before peaking at @xmath259 and declining at lower @xmath256 . the median @xmath2 shows a similar steady increase with decreasing @xmath256 below @xmath260 although there is evidence that it peaks at @xmath261 before declining at lower @xmath256 . the redshift interval corresponding to @xmath259 represents a time interval of @xmath262 or @xmath263 . , @xmath2 and @xmath256 , the redshift of the most significant recent major merger*. we identify all haloes in our sample at @xmath7=0 with @xmath264 and identify the redshift @xmath256 at which @xmath16 occurred . both @xmath1 and @xmath2 are evaluated at z=0 . filled circles and bars correspond to medians and upper and lower quartiles.,width=302 ] this is consistent with the finding of @xcite , who examined the velocity dispersion @xmath265 of material within @xmath12 of simulated galaxy cluster haloes ( see their figure 5 ) . they noted that merging leads to an increase in @xmath265 of the main ( host ) halo because the merging sub - halo acquires kinetic energy as it falls in the potential well of the more massive main halo . the peak in @xmath265 corresponds to the first pericentric passage of the subhalo , after which @xmath265 declines because subsequent passages are damped , and so the main halo relaxes . this will occur on a timescale of order @xmath266 , which is consistent with the peak in @xmath1 at @xmath259 . we would expect to see a peak in @xmath2 on roughly the merging timescale @xmath267 , which as we noted in [ sec : hierarchical ] is comparable to @xmath268 ( cf . * ? ? ? + we can take this analysis a little further by looking at the detailed evolution of @xmath1 and @xmath2 over time . in figure [ fig : dynstate_merge_tree ] we plot the redshift variation of @xmath4 ( normalised to its value at @xmath7=0 ; lower panel ) , @xmath1 ( middle panel ) and @xmath2 ( upper panel ) against the time since major merger , normalised by the dynamical time @xmath112 estimated at the redshift at which the merger occurred , @xmath256 . medians and upper and lower quartiles are indicated by filled circles and bars . for illustrative purposes , we show also the redshift variation of @xmath4 , @xmath1 and @xmath2 for a small subset of our halo sample ( red , blue , green , cyan and magenta curves ) . as in figure [ fig : time_dep_dynstate ] , we adopt @xmath269 . our naive expectation is that both @xmath1 and @xmath2 should increase in response to the merger , peak after @xmath270 and then return to their pre - merger values . if this behaviour is typical , then we expect pronounced peaks in the median values of @xmath1 and @xmath2 at @xmath271 . however , it is evident from figure [ fig : dynstate_merge_tree ] that there is no significant difference between the median @xmath1 and @xmath2 pre- and post - major merger , and so our naive expectation is not borne out by our results . this is not surprising if one inspects histories for @xmath1 and @xmath2 for individual haloes , in the spirit of @xcite ; @xmath1 and @xmath2 increase following a major merger , but the behaviour is noisy ( reflecting e.g. differences in orbital parameters of merging subhaloes , the redshift dependent virial radius , dependence on environment , etc ... ) and the timescale of the response varies from halo to halo simply averaging or taking the median washes any signal away . nevertheless it is worth looking at this in more detail , which we shall do in a forthcoming paper . , @xmath1 and @xmath2 to a major merger . * we include all haloes with @xmath264 at @xmath7=0 and plot the redshift variation of @xmath4 , @xmath1 and @xmath2 against the time since the major merger , normalised by the dynamical time of the halo at the redshift at which the merger occurred . filled circles and bars correspond to medians and upper and lower quartiles , while curves correspond to the histories of 5 individual haloes.,width=302 ] the aim of this paper has been to quantify the impact of a dark matter halo s mass accretion and merging history on two measures of dynamical state that are commonly used in cosmological @xmath0-body simulations , namely the virial ratio @xmath23 ( cf . * ? ? ? * ) and the centre - of - mass offset @xmath272 ( cf . * ? ? ? * ; * ? ? ? * ; * ? ? ? the virial ratio @xmath1 derives from the virial theorem and and the expectation is that @xmath185 for dynamically equilibrated haloes . the centre - of - mass offset @xmath2 can be regarded as a substructure statistic @xcite that provides a convenient measure of how strongly a halo deviates from smoothness and spherical symmetry . we expect a halo s dynamical state and its mass assembly history to correlate closely . understanding how is important because the degree to which a halo is dynamically equilibrated affects the reliability with which we can measure characteristic properties of its structure , such as the concentration parameter @xmath8 ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and kinematics , such as the spin parameter @xmath9 ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? for this reason , it is desirable to establish quantitatively how well they correlate and to assess how @xmath1 and @xmath2 can help us to characterise how quiescent or violent a halo s recent mass assembly history has been . our key results are that @xmath1 and @xmath2 show strong positive correlations with each other ( cf . figure [ fig : deltar_vs_eta ] ) as @xmath1 increases for a halo , so too does @xmath2 and that both are useful indicators of a halo s mass recent mass accretion and merging history . for example , @xmath1 and @xmath2 correlate strongly with @xmath16 , which measures the significance of a halo s recent merging activity ; haloes with @xmath273 ( cf . figure [ fig : virial_ratio_versus_deltamax ] ) and @xmath6 ( cf . figure [ fig : deltar_vs_mah ] ) will have quiescent recent mass assembly histories they are unlikely to have suffered mergers with mass ratios greater than 1:10 over the last few dynamical times . in contrast , interpreting the corrected virial ratio @xmath192=@xmath274 , where @xmath210 is the surface pressure energy , is problematic ( at least insofar as we have implemented it here , which follows the prescription of @xcite and has been applied in @xcite and @xcite ) . in principle , @xmath192 should account for the approximation that is made when we define a halo to be a spherical overdensity of @xmath57 times the critical density at a particular redshift . as we noted in [ sec : sims ] , haloes are more complex structures than this simple working definition gives them credit for , and by defining the halo s extent by the virial radius @xmath12 the likelihood is that material that belongs to the halo will be neglected . by correcting the virial ratio @xmath1 for what is effectively a truncation of the true halo , the corrected virial ratio @xmath192 takes account of the `` missing '' kinetic energy . however , our results imply that the correction itself ( the surface pressure energy @xmath181 ) is sensitive to a halo s merging history , and that it increases with increasing @xmath16 ( cf . figure [ fig : virial_ratio_versus_deltamax ] ) . for this reason we would caution against the use of @xmath192 to identify dynamically relaxed haloes , at least in the form that is currently used . interestingly , we find that systems with violent recent mass assembly histories ( most significant merger with a mass ratio @xmath275 between @xmath276 ) have values of @xmath1 and @xmath2 ( as measured at @xmath7=0 ) that peak at @xmath277 , which corresponds to a timescale of @xmath263 ( cf . figure [ fig : time_dep_dynstate ] ) . this is consistent with the earlier analysis of @xcite , who found that the velocity dispersion @xmath265 of material within the virial radius which is linked to the virial ratio @xmath1 peaks on first closest approach of the merging sub - halo with the centre of the more massive host halo . this should occur on a timescale of @xmath266 , after which @xmath265 and @xmath1 should dampen away . similar arguments can be made for @xmath2 . we note that these arguments can be made in a statistical sense , but if we look at the merging histories of individual haloes , the behaviour of @xmath1 and @xmath2 is much more complex , and as we demonstrate a simple timescale for their response to a major merger is difficult to define ( cf . figure [ fig : dynstate_merge_tree ] ) . we shall return to this topic in future work . + what is the significance of these results ? structure formation proceeds hierarchically in the cdm model and so we expect to find correlations between virial mass @xmath4 and formation redshift @xmath3 ( cf . figure [ fig : mass_versus_zform ] ) , which in turn result in positive correlations between @xmath4/@xmath3 and @xmath1/@xmath2 . ( cf . figures [ fig : virial_ratio_measures ] and [ fig : deltar_measures ] ) . this means that more massive haloes and those that formed more recently are also those that are least dynamically equilibrated , a fact that we should be mindful of when characterising the halo mass dependence of halo properties that are sensitive to dynamical state ( e.g. @xmath8 and @xmath9 ) . it s worth noting that the correlation between @xmath4 and @xmath1 is stronger than the correlation between @xmath4 and @xmath2 ; the median @xmath1 rises sharply with @xmath4 and there is no overlap between the width of the distributions of @xmath1 is the lowest and highest mass bins . in contrast , the median @xmath2 in the highest mass bin lies in the high-@xmath2 tail of the lowest mass bin . this is interesting because @xmath1 as it is usually calculated depends on @xmath19 , which is sensitive to the precise boundary of the halo . correcting for the surface pressure term does not appear to help , as we point out indeed , the surface pressure term itself correlates with merging activity . this points towards an ambiguity in the use of @xmath1 as we note , it rarely if ever satisfies @xmath1=1 . we discuss this point in a forthcoming paper , but we note that even in ideal situations , what one computes for @xmath1 depends on @xmath12 ( cf . * ; * ? ? ? * ) and so applying a flat cut based on a threshold in @xmath1 alone risks omitting massive haloes that might otherwise be considered dynamically equilibrated . for this reason we advocate the use of @xmath2 in cosmological @xmath0-body simulations as a more robust measure of a halo s dynamical state ; its calculation is computationally inexpensive , it is well defined as a quantity to measure , and its interpretation is both clear and straightforward . we find that @xmath278 , which corresponds to a @xmath279 , should be sufficient to pick out the most dynamically relaxed haloes in a simulation volume at @xmath7=0 . + although our focus has been fixed firmly on haloes in cosmological simulations , we note that our results have observational implications . whether or not an observed system for example , a galaxy cluster is in dynamical equilibrium will affect estimates of its dynamical mass if we assume a luminous tracer population that is in dynamical equilibrium ( e.g. * ? ? ? similarly , studies that seek to reconstruct a galaxy cluster s recent merging history tend to use signatures of disequilibrium ( e.g. * ? ? ? the most obvious measure of disequilibrium is the centre of mass offset @xmath2 , or its projected variant . although a more careful study in which we mock observe our haloes ( and a seeded galaxy population ) is needed , our results suggest that @xmath2 could be used to infer the redshift of the last major merger ( cf . figures [ fig : deltar_vs_mah ] and figure [ fig : deltar_measures ] , although care must be taken as figure [ fig : dynstate_merge_tree ] reveals ) . observationally , this would require measurement of , for example , projected displacements between gas and dark matter from gravitational lensing and x - ray studies . we note that @xcite have already tested this idea using idealised hydrodynamical simulations of mergers between galaxy clusters and found that the centroid offset between x - ray and projected mass maps captures the dynamical state of galaxy clusters well , but it is interesting to extend this idea using cosmological hydrodynamical simulations of galaxy groups and clusters . this will form the basis of future work . cp acknowledges the support of the stfc theoretical astrophysics rolling grant at the university of leicester . ak is supported by the _ spanish ministerio de ciencia e innovacin _ ( micinn ) in spain through the ramon y cajal programme as well as the grants aya 2009 - 13875-c03 - 02 , aya2009 - 12792-c03 - 03 , csd2009 - 00064 , and cam s2009/esp-1496 . he further thanks the aluminum group for chocolates . srk acknowledges financial support from swinburne university of technology s centre for astrophysics and supercomputing s visitor programme . he acknowledges support by the micinn under the consolider - ingenio , syec project csd-2007 - 00050 . the simulations presented in this paper were carried out on the swinburne supercomputer at the centre for astrophysics & supercomputing , the sanssouci cluster at the aip and the alice supercomputer at the university of leicester . v. , white s. d. m. , jenkins a. , frenk c. s. , yoshida n. , gao l. , navarro j. , thacker r. , croton d. , helly j. , peacock j. a. , cole s. , thomas p. , couchman h. , evrard a. , colberg j. , pearce f. , 2005 , , 435 , 629

repeated measurements as typically occurring in two- or multi - time correlators rely on von neumann s projection postulate , telling how to restart the system after an intermediate measurement . we invoke the principle of deferred measurement to describe an alternative procedure where co - evolving quantum memories extract system information through entanglement , combined with a final readout of the memories described by born s rule . the new approach to repeated quantum measurements respects the unitary evolution of quantum mechanics during intermediate times , unifies the treatment of strong and weak measurements , and reproduces the projected and ( anti- ) symmetrized correlators in the two limits . as an illustration , we apply our formalism to the calculation of the electron charge correlator in a mesoscopic physics setting , where single electron pulses assume the role of flying memory qubits . we propose an experimental setup which reduces the measurement of the time correlator to the measurement of currents and noise , exploiting the ( pulsed ) injection of electrons to cope with the challenge of performing short - time measurements .

1502.02938

within the quantum world , the question what quantities can be measured in an experiment is often a non - trivial one , e.g. , measuring time correlators ( with times @xmath0 ) requires finding the correct ordering of operators . concrete examples in mesoscopic physics and quantum optics are the measurements of charge correlators @xcite or full counting statistics @xcite and that of photon correlators @xcite . the question is usually resolved by including the measurement apparatus in the description and its internal workings decide upon the form of the measured correlator . examples are the ampremeter , double - dot detector , and spin counter used in refs . , , and , or the different photodetectors introduced by glauber and by mandel @xcite . these detectors then act back on the system , thereby influencing the measurement outcome , i.e. , the specific form of the correlator . e.g. , a weak measurement as used in ref . , see also ref . , leads to symmetrized ( @xmath1 ) and antisymmetrized ( @xmath2 ) correlators weighted with different detector response functions @xcite , while a strong measurement produces a projected correlator @xmath3 . these different forms of measured correlators can be derived @xcite by invoking the von neumann projection postulate @xcite , telling how to restart the system after the first measurement at @xmath4 ( or after previous measurements at times @xmath5 in a @xmath6-th order correlator ) . the measurement with a weakly coupled detector can be treated perturbatively , with the von neumann projection excerted on the detector and no back action on the system @xcite . in a strong measurement with a large system detector coupling , the projection formally can be applied directly to the system , therefore producing a maximal back action @xcite . schematic illustration of different measurement procedures for a two - time ( @xmath4 , @xmath7 ) correlator : ( a ) strong measurement described by von neumann projection acting directly on the system at time @xmath4 and providing a projected correlator @xmath8 after readout at @xmath9 ; for a weak measurement , the von neumann projection at time @xmath4 acts on the weakly coupled detector . ( b ) repeated measurement without von neumann projection at @xmath4 : unitary co - evolution of system and quantum memories which are entangled at times @xmath4 and @xmath9 and final readout after @xmath9 . the coupling strength between the system and the quantum memories determines the degree of entanglement.,width=302 ] in this paper , we invoke the principle of deferred measurement @xcite known from quantum information theory , where it can be used for quantum computing , and apply it to the problem of repeated measurements , specifically , of time - correlators . we replace the von neumann projection by entangling the measured system with co - evolving quantum memories , see fig.[fig : comp ] , thereby ( effectively ) expanding the hilbert space of the total system in every measurement step . the desired correlator then is derived from a final measurement of all the quantum memories by invoking born s rule @xcite . hence the entire system plus memories undergoes a unitary quantum evolution until the very end , where the born rule takes us from the quantum to the classical world . the new scheme captures the cases of weak and strong measurement within a unique formalism by merely changing the degree of entanglement between the system and the quantum memory . in the limits of weak and strong entanglement , we reproduce the results previously derived via use of the projection postulate . no simple physical form for the time - correlators could be found so far in the intermediate coupling regime . describing a measurement by entangling the system with a detector and including a ( dissipative ) bath in the evolution of the density matrix is a concept that has been well developed over the past two decades @xcite . here , we extend the idea of system detector entanglement to the case of repeated measurement . thereby , the quantum memories evolve coherently during and after the information transfer from the system due to entanglement , with the ( dissipative ) measurement deferred to the very end of the process . the proposed scheme for the measurement of time - correlated observables finds an interesting application in mesoscopic physics . in particular , measuring the charge @xmath10 dynamics of a quantum dot with the help of a nearby quantum point contact is a classic problem by now@xcite . in such a setup , single electron pulses @xcite assume the role of flying qubit memories which are either transmitted or reflected by the quantum point contact ( qpc ) , depending on the dot s charge state , see fig.[fig : meso_setup ] . analyzing the charge transmitted across the qpc then provides the desired information on the dot s charge correlator . making use of recent developments in electron quantum optics @xcite , we propose a setup , see fig . [ fig : exp ] , that shifts the task of resolving short times , typically done on the level of the detection , to the proper timing of electron pulses and gate operations . we briefly sketch the main idea of the paper : consider a system observable @xmath11 with an eigenbasis @xmath12 that is to be measured . we start with a system state at the initial time @xmath13 @xmath14 and a quantum memory in the initial state @xmath15 and have them transiently interact at time @xmath4 with the help of an externally controlled interaction ( to be identified with a quantum detector ) . the system and memory then become entangled , @xmath16 |\phi_\mathrm{in}^{\scriptscriptstyle ( 1 ) } \rangle \to \sum_n \psi_n(\tau_1 ) |n\rangle |\phi_n^{\scriptscriptstyle ( 1)}\rangle,\end{aligned}\ ] ] where @xmath17 denote memory states after interaction with the system in state @xmath18 and we assume a negligible evolution of the system during the time of interaction . evolving the system to the later time @xmath9 and entangling it with a second memory , we obtain the state @xmath19 with @xmath20 the matrix elements of the system propagator @xmath21 and @xmath22 . the unitary evolution of the memory states @xmath23 preserves the system information gained at times @xmath0 . at time @xmath24 we measure the memory observables @xmath25 and @xmath26 with discrete spectra @xmath27 . making use of born s rule , we find the probability distribution function @xmath28 for the measurement outcomes @xmath29 and @xmath30 on the two memory observables ; this probability distribution contains the desired information on the system s two - time correlator . the specific relation between the distribution function @xmath28 of measurement outcomes and the correlator of system observables @xmath31 and @xmath32 depends on the system detector coupling and the observables measured on the memories ; we will show below how to extract the well known ( anti-)symmetrized and projected correlators from the probabilities @xmath28 in the limits of weak and strong measurements . in the end , by making use of quantum memories which store the information acquired from the system at the quantum level at earlier times @xmath4 and @xmath9 until the final time @xmath33 , we have avoided the intermediate readout which requires the use of the projection postulate . hence the entire measurement process follows a unitary quantum evolution until the transition to the classical world is done via born s rule . in the following , we will first derive the general framework describing the deferred measurement of a correlator with the intermediate von neumann projection replaced by a system detector entanglement ( sec . [ sec : uni ] ) . in section [ sec : gen_wc ] , we discuss the limit of weak measurement and use qubits as quantum memories to arrive at a simple relation between the measurement outcome on the qubits and the ( anti- ) symmetrized correlators of the system . in sec . [ sec : gen_sc ] , we first discuss a strong measurement at strong coupling using qudit memories and then invoke ( weakly coupled ) qubit registers to show that both types of strong measurements produce the projected time correlator of the system . an illustration of our formalism is given in sec . [ sec : charge ] , where we describe the measurement of the charge correlator in a mesoscopic setting , specifically , the two - time charge correlator of a quantum dot ( qd ) as measured by a quantum point contact ( qpc ) . [ sec : charge_exp ] describes a possible experimental implementation and in sec . [ sec : con ] we summarize our results . we consider the situation where a two - time correlator of a system operator @xmath11 is measured with the help of two quantum memories ; the system s initial state @xmath34 is given by eq . ( [ eq : psi ] ) , while the memories are described by initial states @xmath35 , @xmath36 ( see below for the discussion of an open system described by a density matrix @xmath37 ) . the memories interact with the system at times @xmath38 during a small time intervall @xmath39 . after this interaction , the resulting memory states @xmath40 depend on the system state @xmath18 , where @xmath41 describes the time evolution of the quantum memories during the interaction with the system ( we assume a trivial evolution of the free memories ) . the system state @xmath42 is assumed to remain unchanged during the time @xmath39 of the individual interaction events , see sec . [ sec : finitewidth ] for an extended discussion of this point . after the second interaction event at @xmath9 , the wave function @xmath43 of the system is entangled with the states @xmath44 of the memories and the combined wave function @xmath45 reads @xmath46 with @xmath47 . the quantum memories are supposed to keep their system information after their interaction . at time @xmath33 , we measure the operators @xmath48 and @xmath26 on the first and second memory , respectively . denoting the ( discrete ) eigenvalues and eigenstates of @xmath49 by @xmath50 and @xmath51 , we rewrite the memory states @xmath52 . applying born s rule to the final state ( [ eq : psif ] ) provides us with the probability distribution @xmath53 where @xmath54 is the projector onto the eigenstate @xmath29 of the @xmath55-th memory , i.e. , @xmath56 . making use of the unitarity condition @xmath57 ( rendering the evolution @xmath58 in eq . ( [ eq : psif ] ) irrelevant ) , we obtain the probabilities @xmath59 the above expressions ( [ eq : pabdef ] ) and ( [ eq : pab ] ) for the probabilities @xmath60 provide us with the desired information on the two - time correlator of the system . they are easily generalized to the case of open systems by introducing the combined system plus bath ( the open system ) density matrix @xmath61 and evolve it in time including the subsequent entanglement with the quantum memories : starting from the initial density matrix @xmath62 describing the open system plus memories at time @xmath63 , we proceed as in the case of isolated systems by conditioning the time evolution of the memory states on the corresponding system states and obtain the final density matrix at time @xmath64 @xmath65 with the open system s density matrix @xmath66 at time @xmath4 , its reduced part @xmath67 , and the reduced operators @xmath68 with @xmath69 the evolution operator of the open system . note that here , the outgoing states @xmath70 and @xmath71 are conditioned on the system states @xmath72 and @xmath73 at times @xmath4 and @xmath9 . we define the probabilities @xmath28 as @xmath74\end{aligned}\ ] ] with the trace taken over both the open system and the memory states . calculating this expression with the final density matrix eq . , we obtain @xmath75 , \nonumber\end{aligned}\ ] ] with the remaining trace taken over the bath degrees of freedom . this result is the direct generalization of to the case of open systems . the expressions ( [ eq : pabdef ] ) and ( [ eq : pab ] ) as well as ( [ eq : pabdef_o ] ) and ( [ eq : pab_o ] ) constitute the basic formulas which we will further develop in the following sections . indeed , as expressed in terms of evolution amplitudes of system and detectors , it is difficult to appreciate the physical meaning and content of these results . in order to make progress , we consider next the two cases of weak and strong measurements . given a _ weak system detector coupling _ , the most direct way to find the probabilities @xmath28 in terms of physically transparent quantities is to start from eq . ( [ eq : pabdef ] ) and evaluate this expression perturbatively in the linear system detector coupling @xmath76 , where the time - dependent coupling @xmath77 acts on the @xmath55-th memory during a time @xmath78 around @xmath0 . the unperturbed evolution of the memories is described by the hamiltonian @xmath79 and we make use of the interaction representation . we go over to irreducible quantities by subtracting the uncorrelated contribution , @xmath80 with @xmath81 and latexmath:[$p_\beta^{\scriptscriptstyle ( 2 ) } = \langle \psi_f^{\scriptscriptstyle ( 2)}| { \hat{p}}_\beta^{\scriptscriptstyle ( 2 ) } involving a single entanglement at time @xmath4 or @xmath9 with only one memory , respectively . the quantity @xmath83 can be obtained by a simple summation of @xmath28 , @xmath84 and @xmath85 for a time - independent problem ( otherwise , the determination of @xmath86 necessitates a second measurement ) . note that the sum over the first index of @xmath28 already includes correlations , see eq . ( [ eq : pab ] ) , and hence @xmath87 . the task then is to evaluate the irreducible expression @xmath88 with the expectation value to be taken over the initial system state @xmath89 , @xmath90 refers to the irreducible part , and the time evolution operator reads @xmath91,\ ] ] with @xmath92 denoting time - ordering . evaluating to lowest relevant order in the coupling , we find @xmath93 , \hat{h}_\mathrm{sd}(\tau '' ) ] | \psi \rrangle \nonumber\\ & = \frac{(-i)^2}{\hbar^2 } \int_{\tau_{\rm in}}^{\tau } \!\!d\tau ' \int_{\tau_{\rm in}}^{\tau ' } \!\!d\tau '' \bigl [ \llangle \psi| { \hat{o}}(\tau ' ) { \hat{o}}(\tau '' ) |\psi\rrangle \ , \langle \phi_\mathrm{in}^{\scriptscriptstyle ( 1 ) } |{\hat{p}}_{\alpha}^{\scriptscriptstyle ( 1)}(\tau_f ) { \hat{b}}^{\scriptscriptstyle ( 1)}(\tau '' ) |\phi_\mathrm{in}^{\scriptscriptstyle ( 1)}\rangle \ , \langle \phi_\mathrm{in}^{\scriptscriptstyle ( 2 ) } |{\hat{p}}_{\beta}^{\scriptscriptstyle ( 2)}(\tau_f ) { \hat{b}}^{\scriptscriptstyle ( 2)}(\tau ' ) |\phi_\mathrm{in}^{\scriptscriptstyle ( 2)}\rangle \nonumber\\ \noalign{\vspace{-8pt } } & \hspace{97pt}- \llangle \psi| { \hat{o}}(\tau '' ) { \hat{o}}(\tau ' ) |\psi\rrangle \ , \langle \phi_\mathrm{in}^{\scriptscriptstyle ( 1)}|{\hat{b}}^{\scriptscriptstyle ( 1)}(\tau '' ) { \hat{p}}_{\alpha}^{\scriptscriptstyle ( 1)}(\tau_f ) |\phi_\mathrm{in}^{\scriptscriptstyle ( 1)}\rangle \ , \langle \phi_\mathrm{in}^{\scriptscriptstyle ( 2)}i |{\hat{p}}_{\beta}^{\scriptscriptstyle ( 2)}(\tau_f ) { \hat{b}}^{\scriptscriptstyle ( 2)}(\tau ' ) |\phi_\mathrm{in}^{\scriptscriptstyle ( 2)}\rangle \nonumber\\ \noalign{\vspace{-4pt } } & \hspace{97pt}- \llangle \psi| { \hat{o}}(\tau ' ) { \hat{o}}(\tau '' ) |\psi\rrangle \ , \langle \phi_\mathrm{in}^{\scriptscriptstyle ( 1)}i |{\hat{p}}_{\alpha}^{\scriptscriptstyle ( 1)}(\tau_f ) { \hat{b}}^{\scriptscriptstyle ( 1)}(\tau '' ) |\phi_\mathrm{in}^{\scriptscriptstyle ( 1)}\rangle \ , \langle \phi_\mathrm{in}^{\scriptscriptstyle ( 2)}|{\hat{b}}^{\scriptscriptstyle ( 2)}(\tau ' ) { \hat{p}}_{\beta}^{\scriptscriptstyle ( 2)}(\tau_f ) i |\phi_\mathrm{in}^{\scriptscriptstyle ( 2)}\rangle \nonumber\\ \noalign{\vspace{-4pt } } & \hspace{97pt}+ \llangle \psi| { \hat{o}}(\tau '' ) { \hat{o}}(\tau ' ) |\psi\rrangle \ , \langle \phi_\mathrm{in}^{\scriptscriptstyle ( 1)}|{\hat{b}}^{\scriptscriptstyle ( 1)}(\tau '' ) { \hat{p}}_{\alpha}^{\scriptscriptstyle ( 1)}(\tau_f ) |\phi_\mathrm{in}^{\scriptscriptstyle ( 1)}\rangle \ , \langle \phi_\mathrm{in}^{\scriptscriptstyle ( 2)}|{\hat{b}}^{\scriptscriptstyle ( 2)}(\tau ' ) { \hat{p}}_{\beta}^{\scriptscriptstyle ( 2)}(\tau_f ) |\phi_\mathrm{in}^{\scriptscriptstyle ( 2)}\rangle \bigr ] , \label{eq : pirrexpr}\end{aligned}\ ] ] where we made sure that the first memory interacts with the system at the earlier time @xmath94 . for a slow system dynamics and exploiting that @xmath95 only for @xmath96 , we can replace @xmath97 and @xmath98 . we make use of the standard definitions for the symmetrized and anti - symmetrized irreducible correlators ( with @xmath99 $ ] and @xmath100 denoting the usual commutator and anti - commutator ) @xmath101\bigr\rrangle,\end{aligned}\ ] ] to arrive at the final result @xmath102 with the detector response functions @xmath103| \phi^{\scriptscriptstyle ( j)}_\mathrm{in } \rangle . \label{eq : isdet}\end{aligned}\ ] ] the symbols @xmath104 and @xmath105 address symmetrized and anti - symmetrized quantities ( or equivalently , up to factors of 2 , real- and imaginary parts ) . in a situation where the full information @xmath28 can be extracted from the memories , the individual correlators @xmath106 and @xmath107 can be obtained from ( [ eq : pabweak ] ) by combining two different probabilities , e.g. , using @xmath108 and @xmath109 one obtains @xmath110/ d^{\scriptscriptstyle ( 1)}\ , \mathcal{i}s^{\scriptscriptstyle ( 2)}_{{\rm det},\beta } , \\ \label{eq : isoof } \mathcal{i}s^{\rm irr}_{{o}{o } } & = [ -\mathcal{i}s^{\scriptscriptstyle ( 1)}_{{\rm det},\bar\alpha } p^{\rm irr}_{\alpha\beta } + \mathcal{i}s^{\scriptscriptstyle ( 1)}_{{\rm det},\alpha } p^{\rm irr}_{\bar\alpha\beta } ] / d^{\scriptscriptstyle ( 1)}\ , \mathcal{i}s^{\scriptscriptstyle ( 2)}_{{\rm det},\beta},\end{aligned}\ ] ] with @xmath111 . alternatively , one may have preferential access to combinations of probabilities @xmath28 ( see sec . [ sec : charge_exp ] for an example ) or make use of specific detector properties , see below and the appendix for examples . a generic choice for the quantum memories are qubit devices that couple to the system via the hamiltonian @xmath112 , where the coupling @xmath113 is switched on during a short time @xmath39 around @xmath0 . assuming initial states @xmath114/\sqrt{2},\end{aligned}\ ] ] a system residing in a state @xmath18 with eigenvalue @xmath115 will rotate the qubit around the @xmath116-axis by @xmath117 with @xmath118 . the response functions ( [ eq : rsdet ] ) and ( [ eq : isdet ] ) involve integrals of the type @xmath119 and we obtain the response functions @xmath120 choosing @xmath121 , i.e. , polarizing the second memory along the @xmath122-axis , we can directly find the correlator @xmath123 ( @xmath124 ) from the memory correlator @xmath125 by choosing @xmath126 ( @xmath127 ) . alternatively , we may use the results ( [ eq : rsoof ] ) and ( [ eq : isoof ] ) and @xmath128 to find ( we choose @xmath129 , @xmath130 , and @xmath131 ) @xmath132 note that the sums @xmath133 and @xmath134 contain very different types of information , once a correlation , the other time only a mean value , as discussed in more detail in sec . [ sec : charge_exp ] . hence , we find that the delayed measurement of the two quantum memories provides the symmetrized and anti - symmetrized correlators ( [ eq : rsoo ] ) and ( [ eq : isoo ] ) . finally , we note that a weak linear coupling between the system observable and the detector / memory variable canonically conjugated to the detector readout provides a more effective entanglement . such a von neumann like interaction allows to produce a strong entanglement and a strong measurement even for weak coupling if sufficient time is available for the entanglement process@xcite . we are now going to show that a _ strong system detector coupling _ naturally leads to projective correlators . consider a situation where the system operator @xmath11 is measured via entanglement with two quantum memories . we first consider an operator @xmath11 with a non - degenerate spectrum and comment on the general case in the end . a strong coupling between the system and the memory implies that the memory states @xmath135 after interaction with the system in state @xmath18 are fully distiguishable , i.e , @xmath136 the observables @xmath49 distinguish between memory eigenstates @xmath51 and we assume a one - to - one relation with the evoluted memory states @xmath44 , @xmath137 with @xmath138 for @xmath139 ( otherwise , the observable @xmath49 measures linear combinations of eigenstates of @xmath11 and thus is not suitable for a measurement of this observable ) . under these ( strong coupling ) conditions , the amplitudes @xmath140 reduce to @xmath141 ( for @xmath142 , @xmath143 for the second memory ) . in order to describe the strong coupling situation it is favorable to proceed with the expression ( [ eq : pab ] ) and we obtain the result @xmath144 the right hand side of the above expression is nothing but the projected system correlator @xmath145 with the projected density matrix @xmath146 @xmath147 and the projectors @xmath148 onto different system states , see eq.([eq : psi ] ) . the projected correlators @xmath149 are easily combined into the desired two - time correlator @xmath150 @xmath151 \\ \nonumber & = \sum_{nm } o_n o_m s^{p}_{{p}_{n}{p}_{m}}(\tau_{21})\\ \nonumber & = \sum_{nm } o_n o_m p_{\alpha_n \beta_m}(\tau_{21}),\end{aligned}\ ] ] thereby establishing the general result relating the system correlator to the memory readings . a further simplification can be achieved if the eigenvalues of @xmath49 and @xmath11 obey a linear relation @xmath152 , in which case we obtain the simple result @xmath153 directly relating the projected system correlator to the memory correlator @xmath154 . these results can easily be generalized to the case where the observable @xmath11 involves a degenerate spectrum : the system evolution of the memories is conditioned on the eigenvalue @xmath115 rather than the state @xmath18 of the system . degenerate eigenstates produce equal evolutions of the memories , implying that @xmath155 has to be replaced by @xmath156 and @xmath157 is substituted by @xmath158 with the projection operator @xmath159 . with these replacements , the above results remain valid also for a degenerate spectrum . hence , the perfect entanglement between the system and the memories arising due to strong coupling is equivalent to a von neumann projection applied to the system . while no back action is apparent on the level of the system dynamics , the strong back action of this maximal entanglement with the memories manifests itself in a strong change of the system s density matrix when tracing over the memories . the realization of a strong measurement as described above requires equal or more memory states @xmath160 than system states @xmath18 , hence quantum memories with dimensionality @xmath161 or qudits are required . besides the obvious difficulty in realizing qudit memories , their coupling to the system in order to serve as a measurement device is equally nontrivial . as an alternative , we discuss a measurement scheme involving qubit registers , instead . such an alternative setup implementing a strong measurement involves a weak system detector coupling but invokes multiple measurements . we replace the strongly coupled qudit memories by weakly coupled qubit registers with @xmath162 qubits each , probing the system state close to @xmath4 and @xmath9 within a short time interval @xmath39 . for a strong measurement , @xmath162 is chosen sufficiently large to distinguish the different eigenvalues @xmath115 of the system . assuming the system not to change during the interaction time @xmath163 with one register , the final state of system and memory registers is of the same form as in eq . with the outgoing individual memory states @xmath164 and @xmath165 replaced by the outgoing register states @xmath166 . making use of born s rule , we obtain the probabilities @xmath167 for finding @xmath168 ( @xmath169 ) qubits of the first ( second ) register in states @xmath170 ( @xmath171 ) , @xmath172 with @xmath173 and @xmath174 . introducing the conditional probability @xmath175 for measuring @xmath168 of the @xmath162 qubits in the state @xmath29 after interaction with the system in state @xmath18 , we can separate the system- and detector response in the above equation , @xmath176 p_\alpha^\mu(n , n ' ) p_{\beta}^\nu(m)\end{aligned}\ ] ] with the ` off - diagonal ' conditional probabilities @xmath177 ( note that @xmath178 ) . the conditional probabilities @xmath179 depend only on the eigenvalue @xmath115 of the state @xmath18 ( and not on the state @xmath18 itself ) . we then have to distinguish two cases : ( i ) all system eigenvalues are non - degenerate , i.e. , @xmath180 for @xmath181 , ( ii ) there are degenerate system states . in the non - degenerate case ( i ) and for a strong measurement , the probability distributions @xmath179 and @xmath182 for different @xmath183 do not overlap as functions of @xmath168 ( this is the very definition of this measurement being a strong one ) and the ` off - diagonal ' elements @xmath184 for @xmath185 are suppressed , as follows from the relation @xmath186 . we then can simplify the expression ( [ eq : pabmnrp ] ) to the form @xmath187 . the register correlators ( replacing the distribution functions @xmath28 ) take the form @xmath188 assuming again a linear system qubit coupling @xmath189 that rotates the qubits by an angle @xmath190 around the @xmath116-axis , the evolution @xmath191 produces the memory states @xmath192 , where we again assume initial states polarized in the @xmath193-plane , see ( [ eq : q_in ] ) . the probabilities @xmath194 for an individual qubit to reside in state @xmath195 are given by @xmath196 with the qubits initially polarized along the @xmath122-axis , i.e. , @xmath197 , we define the register s ` magnetizations ' @xmath198 = 2 j \delta\vartheta\ , o_n,\end{aligned}\ ] ] where we have made use of ( [ eq : prob_q ] ) in the last equation . the combination @xmath199 then involves the product of register polarizations @xmath200 and using the relation @xmath201 , see eq.([eq : spoo ] ) , we can relate the system correlator to the register correlators via @xmath202 note the normalization @xmath203 , which follows from replacing @xmath204 $ ] by @xmath205 = \langle \mu \rangle_0 + \langle \mu \rangle_1 = j$ ] . in the degenerate case ( ii ) , the distribution functions separate only for states with different eigenvalues @xmath206 , implying that @xmath207 , while for degenerate eigenvalues with @xmath208 , @xmath209 . the probabilities @xmath210 then are given by @xmath211 p_\alpha^\mu(n ) p_{\beta}^\nu(m).\end{aligned}\ ] ] on the other hand , the projected density matrix @xmath157 appearing in the system correlator @xmath212 involves the projectors @xmath213 , such that the result ( [ eq : spooreg ] ) remains unchanged . summarizing , we have seen that the measurement scheme invoking entanglement with quantum memories and their delayed measurement provides us , in the weak- and strong measurement limits , with the same results as obtained via the traditional route using an intermediate von neumann projection . furthermore , these results have been obtained within a unified description starting from the same initial formula in the form ( [ eq : pabdef ] ) or ( [ eq : pab ] ) . the above general theoretical considerations are rather non - trivial to implement in a practical situation , as the preparation , entanglement , and measurement of many quantum memories is often a non - trivial task . the implementation of these ideas is less demanding , though still challenging , when considering specific examples . indeed , quantum memories for delayed measurement are naturally provided in a scattering geometry , where individual scattered particles take the role of flying qubits . in the following , we focus on a specific example in mesoscopic physics , the measurement of a charge correlator of a quantum dot by scattering electrons in a nearby quantum point contact . we first analyze the situation for a simple two - state system with charges @xmath214 , @xmath215 , and then extend these considerations to arbitrary charge states . we consider a classic problem @xcite , the charge @xmath10 dynamics of a quantum dot ( qd ) ( attached to leads or coupled to another dot in an isolated double - dot system ) measured by a nearby quantum point contact ( qpc ) , see fig . [ fig : meso_setup ] and ref . for a recent discussion of this system . here , we want to characterize the dot s dynamics by its two - time charge correlator . in this system , the measurement is executed by the electrons which are transmitted across / reflected by the qpc with probabilities that depend on the dot s charge state . for the present discussion , it is convenient to view the qpc current as a sequence of individual electron pulses ; during recent years , this theoretical idea @xcite has progressed to an experimental reality @xcite , opening the new field of electron quantum optics @xcite . the quantum memories then can be viewed as flying qubits , individual electron pulses arriving at the qpc at times @xmath4 and @xmath9 that probe the charge state of the qd through the capacitive coupling between the qd and the qpc , see fig . [ fig : meso_setup ] . quantum dot system ( qd ) measured by a capacitively coupled quantum point contact ( qpc ) : single - electron pulses incident on the qpc from the left ( at @xmath13 ) are either transmitted ( with amplitude @xmath216 ) or reflected ( amplitude @xmath217 ) ; the outgoing lippmann - schwinger wave functions describe flying qubits without own dynamics and serve as quantum memories . two pulses separated in time by @xmath218 are needed to measure the two - time correlator of the dot s charge . after scattering at the qpc , the two electrons ( flying qubits ) are entangled with the quantum dot system and carry information on its dynamics . simultaneous detection of the two scattered electrons ( at @xmath33 ) , e.g. , a distance @xmath219 away with both positions on the right of the qpc , provides information on the two - time charge correlator . , width=264 ] the memory states @xmath220 are the two scattered states where the electron is reflected ( @xmath221 ) or transmitted ( @xmath222 ) , i.e. , the outgoing state is given by @xmath223 and @xmath224 depending on the charge state of the system . we assume well separated single - electron pulses and an evolution of the scattered waves @xmath225 emanating from the qpc at times @xmath0 that preserves the corresponding system information , in particular , @xmath226 . this allows us to envision an individual detection of the electrons @xcite . in the final readout , the flying qubits are detected on the right or left side of the qpc , telling whether the two electrons have been transmitted ( with probability @xmath227 ) , reflected ( @xmath228 ) or mixed ( @xmath229 and @xmath230 ) . these probabilities then contain the information about the two - time charge correlator of the qd . formally , such a final - state analysis corresponds to measuring the ( charge ) operators @xmath231 , with the projectors @xmath232 providing the transmitted ( @xmath233 ) or reflected ( @xmath234 ) components of the @xmath55-th electron . the eigenvalues @xmath235 depend on the measured charge , e.g. , @xmath236 and @xmath237 if the transmitted charge is measured on the right of the qpc , while @xmath238 and @xmath239 if the reflected charge is measured on the left . both types of measurements provide us with the same probability distributions ( [ eq : pabdef ] ) or ( [ eq : pab ] ) with @xmath240 . in the following , we first analyze the case of a qd with binary charge states @xmath241 and @xmath242 and eigenvalues @xmath243 , @xmath244 , @xmath245 , see eq . ( [ eq : psi ] ) . second , we generalize the discussion to a qd with multiple charge states as described by the charge operator @xmath246 . while two - state quantum memories ( qubits ) are always sufficient for a complete description when the measurement is weak , for a strong measurement , a qd with multiple charge states will require the use of qubit registers , i.e. , finite trains of electron pulses . for the case of a weak measurement , we can make immediate use of the general results eqs . ( [ eq : pabweak ] ) , ( [ eq : rsoof ] ) , and ( [ eq : isoof ] ) by replacing @xmath247 and choosing the values @xmath248 equal to @xmath249 and @xmath250 . it remains to determine the detector response functions ( [ eq : isdet ] ) and ( [ eq : rsdet ] ) . we consider a linear system detector coupling of the form @xmath251 with @xmath252 the charge on the qpc . furthermore , we parametrize the scattering matrices for the system in states @xmath241 and @xmath242 by @xmath253 , @xmath254 , @xmath255 , @xmath256 , with small corrections @xmath257 , @xmath258 , and @xmath259 . in order to find the detector response functions ( [ eq : rsdet ] ) and ( [ eq : isdet ] ) we replace @xmath260 by @xmath261 and determine the integral @xmath262 ( we drop the memory index @xmath263 as the electrons are scattered by the same qpc ) . to lowest order in @xmath264 , this can be written in the form @xmath265 with @xmath266 and @xmath267 describing the dynamics of the detector in the heisenberg representation in the absence and presence of a charge on the dot , respectively . furthermore , we have used that @xmath268 with @xmath269 and @xmath270 in the interaction representation . the initial memory states @xmath271 evolve under @xmath272 and @xmath273 according to @xmath274 and hence @xmath275 , where we have used that @xmath276 . expanding @xmath277 in @xmath257 , @xmath278 , and @xmath279 , we find that @xmath280 and taking real and imaginary parts , we arrive at the final results @xmath281 using these detector response functions in the general expressions ( [ eq : rsoof ] ) and ( [ eq : isoof ] ) one easily arrives at the system correlators ( we choose @xmath282 , @xmath283 , and @xmath284 ) , @xmath285 alternatively , the qpc can be tuned to deliver the individual system correlators @xmath286 or @xmath287 . indeed , using a detector with high transmission , e.g. , a qpc with energetic ( @xmath288 ) single - electron pulses @xmath289 , @xmath290 the qpc barrier , one easily finds that ( although @xmath291 and therefore @xmath292 measures @xmath287 , while @xmath293 measures @xmath294 . when the detector predominantly reflects particles , e.g. , for low energy single electron pulses with @xmath295 , the situation is reverse , @xmath296 ( @xmath286 ) is measured by @xmath297 ( @xmath292 ) , see the appendix for further details on the qpc detector response . for a quantum dot with multiple charge states we have to require that the expansion of the scattering amplitudes @xmath298 and @xmath299 of the qpc detector scale linearly in the charge @xmath300 of the dot , i.e. , @xmath301 , @xmath302 , and @xmath303 . a straightforward calculation then shows that the results ( [ eq : det_r ] ) and ( [ eq : det_i ] ) for the detector response functions as well as the final results ( [ eq : rs_c ] ) and ( [ eq : is_c ] ) remain unchanged . when performing a strong measurement of a quantum dot with a binary charge it is sufficient to invoke individual electron pulses as quantum memories . for a strong dot qpc coupling , we require a one - to - one relation between the presence of a charge on the dot and the outcome of the measurement , i.e. , @xmath304 and @xmath305 and @xmath306 . this is achieved by tuning the qpc such as to generate a unique scattering outcome with @xmath307 , @xmath308 and @xmath309 , @xmath310 , i.e. , the presence of a charge @xmath244 on the dot reflects the qpc electron back to the left . in this case , it is the reflection probability @xmath311 that directly traces the charge @xmath10 and according to ( [ eq : pab ] ) , we have to determine the expression @xmath312 with @xmath313 , we obtain the simple result @xmath314 ( see also ( [ eq : panbm ] ) with @xmath315 and @xmath316 ) and since @xmath317 , we find the projected correlator ( see eq . ( [ eq : spoo ] ) ) @xmath318 similarly , the probability to find no charge on the dot in either of the two measurements is @xmath319 , while the mixed results are @xmath320 and @xmath321 . the strong measurement of the charge correlator for a multi - charge quantum dot quite naturally involves trains of electron pulses @xcite , with the number @xmath162 of electrons in each train sufficiently large to distinguish the different charge eigenvalues @xmath300 of the dot . the separation @xmath78 between electron pulses within a train has to be sufficiently long in order to allow for their separate detection ( i.e. , counting ) , while the train duration @xmath322 must remain small on the scale @xmath323 of the dot s dynamics . when going over from qubit registers to electron trains scattered at the qpc we replace the ` magnetization ' ( [ eq : mn ] ) by the disbalance between reflected and transmitted electrons , @xmath324 = j[r - t + 2 q_n \delta r],\end{aligned}\ ] ] where we have assumed a linear qpc characteristic with a reflection probability scaling linearly with the dot s charge @xmath300 , @xmath325 . operating the qpc at the symmetry point @xmath326 , we can determine the combination @xmath327 and relate this quantity to the projected charge correlator @xmath328 , @xmath329 operating the qpc away from the symmetric point , one has to determine the weighted sum @xmath330 instead and divide by @xmath331 to arrive at the projected correlator @xmath332 . above we have assumed an instantaneous ( within a short time @xmath78 ) entanglement between the system and the memory states , requiring that both the width @xmath333 of the wave - packet and the scattering time @xmath334 at the qpc satisfy @xmath335 , where @xmath336 denotes the characteristic time scale of the system . here , we allow for a spread in time of the detector s electron wave function and drop the condition @xmath337 , i.e. , we assume that @xmath338 while the scattering event itself remains fast , @xmath339 . in general terms , this corresponds to a measurement which probes the system sharply ( @xmath340 ) during some finite time ( @xmath341 ; longer measurement times @xmath342 do not provide meaningful results ) . let us suppose that the @xmath55-th wave - packet incident on the qpc around @xmath0 is described by the wave function @xmath343 which is normalized ( @xmath344 ) and peaked at the time @xmath0 . assuming instantaneous scattering , we obtain the final state @xmath345 with @xmath346 and @xmath347 and the outgoing memory states @xmath348 that have been scattered at the qpc at time @xmath349 . making use of @xmath350 , we obtain the smeared probabilities @xmath351 with @xmath28 given by eq . . the finite width wave - packets enter as integration kernels . while for sharp wave - packets , the entanglement between system and memories ( i.e. , the ` measurement ' ) takes place at times @xmath4 and @xmath9 , for broader wave - packets , the entanglement arises within a finite time @xmath333 around @xmath4 and @xmath9 with distribution functions @xmath352 and @xmath353 . as a consequence , for the case of strong coupling where the entanglement gives rise to projective measurements , the times of projection are not fixed but distributed with the distribution functions above . note that the result is only valid for negligible scattering time @xmath354 in the qpc . if the scattering time @xmath354 is finite compared to the system time @xmath336 , the effect of interaction can not be accounted for by the scattering matrices @xmath355 of the qpc depending on the system state , but the interaction during the scattering has to be treated in more detail . the limit of @xmath356 considered above has also implications for the resulting back - action : while the back - action for @xmath356 only consists of dephasing , i.e. , a suppression of off - diagonal elements of the system s density matrix , for finite @xmath354 , the back - action of the measurement on the system goes beyond pure dephasing and alters the system s dynamics . the study of higher - order correlators is straightforward , e.g. , to measure a third - order charge correlator , we send three electrons scattering from the qpc at times @xmath357 and obtain the probabilities @xmath358 describing electrons transmitted across ( @xmath359 ) or reflected from ( @xmath360 ) the qpc . for weak coupling , its irreducible part can be recast in the form @xmath361 with @xmath362 , the detector responses @xmath363 and @xmath364 , and the third - order correlators @xmath365_{\sigma'}]_{\sigma}\rrangle\end{aligned}\ ] ] with the constants @xmath366 and @xmath367 and @xmath99_-=[\cdot,\cdot]$ ] resp . @xmath99_+=\{\cdot,\cdot\}$ ] ( note that ( anti-)symmetrized charges in @xmath368 ( encoded in @xmath369 ) relate to opposite detector response functions ( encoded by @xmath370 ) . this result agrees with the one in ref . obtained with the help of the von neumann projection postulate and shows that only _ keldysh time - ordered charge correlators _ are measurable . our general concept of deferred repeated measurements has been formulated with quantum memories , e.g. , qubits , qubit registers , and qudits , see sec.[sec : uni ] . in our application of these general considerations , the measurement of a charge correlator with the help of a quantum point contact ( sec . [ sec : charge ] ) , the quantum memories have been replaced by scattered electrons ( flying qubits ) . it then is natural to seek for an experimental implementation , where the final measurement of the quantum memories , i.e. , the scattered electrons , can be cast into a measurement of currents and noise rather than an individual detection of qubit states . such an implementation is proposed below . by now , several experiments have demonstrated the controlled generation of individual electron pulses @xcite and even the detection of such pulses via qubit detectors seems to be in reach @xcite . a setup particularly well suited within the current context is the one of fletcher _ @xcite , involving a single electron pump and a time - correlated detector setup in a quantum hall device , see fig . [ fig : exp ] for an illustration . pairs of electron pulses of width @xmath371 ps are injected at a typical rate @xmath372 and are well suited to probe the dynamics of a dot with a system time @xmath373 ns . the detector involves a dynamically switchable barrier ( dsb in fig . [ fig : exp ] ) that can be switched on or off with a nanosecond precision in time . schematic illustration of a quantum hall device with a single electron pump ( left ) , the qpc qd setup ( middle ) and a detector arrangement ( right ) , inspired by refs .. pairs of electrons are injected by the single electron pump with a time delay @xmath374 and scattered by the quantum point contact with scattering coefficients @xmath217 and @xmath216 depending on the charge state of the quantum dot . the dynamically switchable barrier@xcite ( dsb ) separates the first and second electron ( flying qubits @xmath142 and @xmath375 ) into arms 1 and 2 . closing the static switchable barrier ( ssb ) and measuring the currents @xmath376 and @xmath377 provides the anti - symmetrized charge - charge correlator @xmath378 . opening the static switchable barrier ( ssb ) , having the two particle streams along the trajectories 1 and 2 interfere at the 50/50 beam splitter , and measuring the current noise @xmath379 as a function of the tunable delay @xmath380 in the detector 3 provides the symmetrized correlator @xmath381.,width=302 ] we propose an experiment implementing a weak measurement of a charge correlator along the lines of sec . [ sec : charge_wc ] , where electron pairs with a tunable delay time in the range @xmath382 nanoseconds are injected by the single electron pump ( with injection currents in the @xmath383 pa regime ) and scattered at the qpc probing the quantum dot qd . the transmitted electrons arrive at the detector dsb after a delay time @xmath384 . the detector barrier dsb is switched on at a time @xmath385 in order to let the first electron pass along the path @xmath386 and deflect the second electron along @xmath387 , see figure [ fig : exp ] . with the static barrier ( ssb ) closed , the measured currents @xmath388 and @xmath389 resulting from the transmission of the first and second electron , respectively , can be related to the probabilities @xmath28 via @xmath390 the two sums @xmath391 and @xmath392 in the above equations are fundamentally different in a subtle way : summing over the second particle @xmath30 produces the trivial probability latexmath:[$p_\alpha^{\scriptscriptstyle ( 1 ) } = \langle \psi_f ( independent ) transmission ( @xmath282 ) or reflection ( @xmath234 ) of the first electron by the qpc , @xmath394 . hence , the first equation ( [ eq : i1 ] ) reduces to @xmath395 and thus contains only information about the mean charge on the dot . on the contrary , summing over the first particle @xmath29 _ does not _ generate @xmath396 as the interaction of the first electron at the earlier time @xmath4 introduces nontrivial correlations , see eq . ( [ eq : pab ] ) . this becomes apparent when going over to irreducible probabilities @xmath397 and using the normalization @xmath398 . then , the second equation ( [ eq : i2 ] ) becomes @xmath399 which includes information about the dot s charge correlator . assuming a time - independent mean charge on the dot , we have @xmath400 and using the general result ( [ eq : pabweak ] ) , the measured currents are easily transformed to provide the antisymmetrized correlator @xmath401 note that the evaluation of the irreducible probability @xmath402 with the help of ( [ eq : pabweak ] ) indeed provides a vanishing result , @xmath403 , explicitly demonstrating that the sum @xmath404 contains no correlations . in order to find the symmetric correlator @xmath405 one has to measure a time correlator on the transmitted channel . this can be conveniently done with the help of a hong - ou - mandel type splitter as implemented in the experiment of bocquillon _ et al . _ @xcite and sketched in fig . [ fig : exp ] . in this experiment , the dynamically switchable barrier ( dsb ) again splits the two electrons in each pair to propagate along the paths 1 and 2 @xmath406 , respectively . the static barrier ( ssb ) is left open , such that the two particles interfere in the splitter . measuring the current noise @xmath407 in channel 3 as a function of mutual delay @xmath380 ( tuned via an additional gate in loop 1 , see figure ) then provides all information needed to construct @xmath405 . we can calculate the evolution of the state through the hong - ou - mandel setup using the wave function @xmath408 , i.e. , the individual components are not normalized and only @xmath409 . due to the orthogonality of these four components , the particle numbers @xmath410 , @xmath411 , emerging from our hong - ou - mandel splitter can be analyzed term by term . in particular , the particle number fluctuations @xmath412 in channel 3 involve single - particle and two - particle contributions , @xmath413 and @xmath414 , hence , @xmath415 and the same result holds true for the @xmath416 scattering component . while there is no contribution from @xmath417 , the one originating from @xmath418 depends on the time delay @xmath380 . let @xmath419 and @xmath420 denote the two electron wave - packets propagating along the incoming paths @xmath386 and @xmath387 of the hom - interferometer . as shown in ref . , we can relate the number fluctuation in the component @xmath418 to the overlap of wavefunctions as @xmath421 where @xmath422 . for a large time delay between the first and second electron , these do not interfere , @xmath423 , such that the fluctuations are just the double of ( [ eq : dn32 ] ) , @xmath424 compensating the original time delay @xmath425 ( such that the two electrons arrive simultaneously at the splitter ) , we have @xmath426 since the pauli exclusion forces the two particles to propagate to different channels . the final result @xmath427 and @xmath227 , @xmath428 when the delay between electron pulses is not compensated and @xmath429 when the time delay is properly compensated , @xmath425 . in a next step , we express the charge fluctuations in channel @xmath430 through the irreducible current - current correlator,@xcite @xmath431 where the time - window @xmath432 is centered around the arrival time of the wave - packets at the detector @xmath430 and has a width of the order of @xmath433 , with @xmath434 the rate of pair injection by the pump . the particle number fluctuations then can be expressed through the low - frequency current noise @xmath435 , @xmath436 with @xmath437 . neglecting the frequency dependence of the noise at small @xmath438 , @xmath439 , we obtain @xmath440 and hence @xmath441 using eq . , we can derive an alternative ( more symmetrized ) version of eq . which involves just the combinations @xmath442 and @xmath443 , @xmath444 defining the detector parameters @xmath445 and going over to irreducible probabilities ( with @xmath446 independent of time ) , we obtain the final result @xmath447 note that in this setup , the final measurement of qubit memories does not require fast or time resolved detection schemes , but merely relies on the measurement of average currents and low - frequency noise . this is due to the fact , that all timing tasks are realized by the properly time - delayed electron pulses in the incoming channel and the dynamically switchable gate ( dsb ) which separates the electron pairs ; both elements have been realized in an experiment@xcite . hence , the new measurement scheme , combined with novel elements from electron quantum optics , allows to shift the ( difficult ) timing issues in the measurement of a time - correlator from the detector to the source . a _ strong coupling _ between the quantum dot and the quantum point contact provides us with a projected correlator . for the simplest case of a binary charge on the dot with values @xmath448 , the projected charge correlator is given by eq . , @xmath449 . making use of the normalization @xmath450 and the result eq . ( [ eq : prts ] ) , we find that @xmath451 in conclusion , we have applied the principle of deferred measurement to the problem of repeated measurement and have derived physical expressions for the two- and multi - time correlators . the measurement involves the inclusion of quantum memories which are entangled with the system at specific times @xmath0 where the system observable is to be probed . the expanded system plus memories undergoes a unitary evolution until the very end , where the result is extracted via application of born s rule to the memories . the measured probabilities @xmath28 ( see eq . ( [ eq : pab ] ) ) or memory correlators @xmath452 ( see eq . ( [ eq : sab ] ) ) then can be combined to extract the desired system correlators . the limits of weak and strong measurements provide the standard ( anti- ) symmetrized and projected time correlators previously obtained by invoking the ( non - unitary ) von neumann projection . the general results have been illustrated by using qubits and qubit registers as quantum memories . our analysis sheds new light on the problem of repeated measurement and illustrates the usefulness of qubits as sensitive measurement devices . although our paper s main results are rather on the conceptual side , one could imagine an implementation of such a deferred measurement in an experiment . a system that naturally lends itself for a realization of these ideas is the classic mesoscopic setup which probes the charge of a quantum dot through a quantum point contact . the individual scattered electrons in the qpc can be understood as flying qubits which are either transmitted or reflected , with amplitudes depending on the charge state of the quantum dot . in particular , the qubit register required in the strong measurement of a dot with a multi - valued charge is easily implemented in terms of finite trains of electrons . we have applied our formalism to this situation and derived the corresponding expressions for a weak and strong measurement . the experimental implementation of these ideas requires a system control that can only be met with a modern quantum engineering approach . recent developments in electron quantum optics provide controlled single electron pulses and allow for their time resolved manipulation / detection on a sub - nanosecond time scale @xcite . using the setup of ref . as a base and augmenting it with a hong - ou - mandel type analyzer @xcite , we propose to include a quantum point contact and a quantum dot in order to realize the principle of repeated measurement by a deferred measurement of quantum memories . as a final remark , one may appreciate the relation of the deferred measurement principle to everett s idea of a multiverse @xcite . rather than applying a projection after the first measurement and pursuing a single further evolution ( of the system = ` universe ' ) , the principle of a deferred measurement involving the system s entanglement with a quantum memory enhances the overall dimensionality , e.g. , for a qubit memory the dimensionality is doubled ( with two ` universes ' evolving in parallel ) . it is then only the final measurement which determines which evolution ( i.e. , which ` universe ' ) actually has been realized . we thank renato renner for discussions and acknowledge financial support from the swiss national science foundation through the national center of competence in research on quantum science and technology ( qsit ) , the pauli center for theoretical studies at eth zurich , and the rfbr grant no.14 - 02 - 01287 . a quick overview is provided by the example of a _ @xmath453-function scatterer : _ expressing the strength of the scatterer @xmath454 for the two charge states by @xmath455 and @xmath456 , an incoming state with wave vector @xmath457 is transmitted with amplitude @xmath458 , @xmath459 . expanding the transmission @xmath460 and phase @xmath461 , we find the modifications @xmath257 and @xmath462 ( for a symmetric scatterer ) in the scattering characteristic of the qpc upon charging the dot @xmath463 in the limit of a large incoming energy , i.e. , @xmath464 , we find @xmath465 and @xmath466 and hence @xmath467 ; with @xmath468 and @xmath469 , we have @xmath470 and therefore @xmath471 . for a small incoming energy @xmath472 , we obtain @xmath473 and @xmath474 and using @xmath475 and @xmath476 , we find @xmath477 . when @xmath478 , all response functions are of the same order . alternatively , we can consider a _ single electron transistor ( set ) _ with the level position @xmath479 affected by the capacitive coupling and depending on the dot s charge state @xmath18 , i.e. , @xmath480 and @xmath481 . the transmission coefficient is given by @xmath482 , where @xmath483 is the level width . again expanding @xmath484 and @xmath485 for small @xmath486 , we find @xmath487 ^ 2},\\ \delta\theta = \delta \chi & = -\frac{\gamma \delta k_{\rm res } } { ( k - k_{\rm res})^2+\gamma^2}.\end{aligned}\ ] ] for incoming electrons on resonance with the level , i.e. , @xmath488 , we obtain @xmath489 and @xmath490 , such that @xmath467 and using @xmath468 and @xmath491 , we find that @xmath492 and @xmath493 , such that @xmath467 and using @xmath494 and @xmath495 , we find @xmath496 , all response functions are of the same order . a more realistic description for the _ quantum point contact ( qpc ) _ is achieved by considering a parabolic scattering potential @xmath497 where the offset @xmath498 is the qpc barrier height when the dot is in the charge state @xmath18 . here , we assume a quasi - classical description and consider the two limits of electrons with energy @xmath499 resp . @xmath500 , see fig . [ fig : qpc_qc ] . using the kemble formula @xcite , we obtain the transmission @xmath501)$ ] , where we have chosen @xmath502 . for weak coupling @xmath503 , we obtain the shift ( we define the energy scale @xmath504 ) @xmath505 which is suppressed exponentially for @xmath506 and @xmath295 due to an exponentially small reflection or transmission . the change in phase at large energies @xmath506 is determined by the transmission phase accumulated in the region @xmath507 $ ] ; within a quasi - classical description , this is given by ( @xmath508 ) @xmath509.\end{aligned}\ ] ] expanding this result for small @xmath510 , we obtain the change in phase @xmath511 , @xmath512 given the exponential suppression of @xmath257 at large energies @xmath506 , we find that @xmath513 and for large transmission we have @xmath514 . in the opposite regime of small energies @xmath295 , we determine the change in phase ( within quasi - classics ) from the phase of the reflection amplitude , @xmath515 where the reversal point @xmath516 is characterized by @xmath517 . to leading order in @xmath510 we find that @xmath518 once more , it follows that @xmath519 due to the exponential suppression of @xmath520 and the response functions respect the order latexmath:[$|\mathcal{r}s_\mathrm{det , r}| \gg |\mathcal{i}s_\mathrm{det , r / t}|\gg functions are of similar magnitude . summarizing , we find that a scatterer with large transmission is characterized by the response functions satisfying latexmath:[$|\mathcal{r}s_\mathrm{det , t}|\gg |\mathcal{i}s_\mathrm{det , r / t}|\gg m. field , c.g . smith , m. pepper , d.a . ritchie , j.e.f.frost , g.a.c . jones , and d.g . hasko phys . lett . * 70 * , 1311 ( 1993 ) ; l.m.k . vandersypen , j.m . elzerman , r.n . schouten , l.h . willems van beveren , r. hanson , and l.p . kouwenhoven , appl . lett . , 4394 ( 2004 ) ; s. gustavsson , r. leturcq , b. simovi c , r. schleser , t.ihn , p. studerus , k. ensslin , d.c . driscoll , and a.c . gossard , phys . 96 , 076605 ( 2006 ) . g. feve , a. mahe , j .- berroir , t. kontos , b. placais , d.c . glattli , a. cavanna , b. etienne , and y. jin , science * 316 * , 1169 ( 2007 ) ; j. dubois , t. jullien , f. portier , p. roche , a. cavanna , y. jin , w. wegscheider , p. roulleau , and d.c . glattli , nature * 502 * , 659 ( 2013 ) .

previous studies determining which astronomical photometry software is best suited for a particular dataset are usually focused on speed , source classification , and/or meeting a sensitivity requirement . for faint objects in particular , the priority is given to maximizing signal - to - noise . photometry of moving targets offers additional challenges ( i ) to aperture photometry because background object contamination varies from image to image , and ( ii ) to routines that build a psf model from point sources in the image because trailed field stars do not perfectly represent the psf of the untrailed target . here , we present the results of testing several photometry algorithms ( tphot , daophot , dophot , apt , and multiple techniques within source extractor and iraf s phot ) on data for a faint , slow - moving solar system object with a known light curve . we find that the newly - developed tphot software most accurately and precisely reproduces the object s true light curve , with particular advantages in centroiding , exclusion of contaminants from the target s flux , and fitting flux in the wings of the point - spread function .

1305.5586

photometry is the practice of measuring the flux and corresponding uncertainty of astronomical objects . many different photometry approaches have been developed , each of them optimized for a particular set of goals , and choosing which approach to adopt is a difficult task . the science goals , intrinsic nature of the target(s ) , and instrumental properties dictate which of the dozens of available photometry algorithms might be most effective . some scientists , such as those with high - volume computing needs like the sloan digital sky survey group , develop their own codes , comparing them to popular algorithms ( _ e.g. _ , * ? ? ? * ; * ? ? ? * ; * ? ? ? other groups try to optimize existing software for a specific task . for example , @xcite tested several algorithms for meeting lsst science requirements , and @xcite assessed the performance of two algorithms in crowded fields . photometry packages currently distributed for public use can be roughly divided into three groups : those that perform aperture photometry , those that construct a model point - spread function ( psf ) from field objects , and those that blend the two . it is now generally accepted that low signal - to - noise ( s / n ) object magnitudes are most accurately recovered through psf - fitting since the errors are dominated by background uncertainty ( _ e.g. _ , * ? ? ? * ; * ? ? ? conversely , high - s / n objects are best represented by aperture photometry , which are more forgiving than psf - fitting techniques of out - of - focus frames and intrinsic morphological complexities ( which may be time variable ) in the psf . in addition to accurately measuring flux , a second challenge to photometry is absolute calibration , which converts the flux measured in instrumental units ( or instrumental magnitudes ) into true apparent magnitudes . absolute calibration takes into account the zeropoint ( the scaling factor in magnitude units ) , the atmospheric extinction , and the color conversion from those of the filters used to the standard color system . we shall not discuss calibration of this dataset in this paper , which focuses only on the initial flux measurements . moving targets are typically comets and asteroids , which by virtue of their size , distance , and low reflectivity , are very faint ; thus , observations of the primary science targets are typically low s / n , making optimal sky determination critical . unless imperfections in the optics distort the field , fixed point sources in the field should have exactly the same psf that can most likely be described by a fitting function . moving targets , however , will produce a trailed psf that may not be approximated by a function , with the amount of trailing dependent on the object s proper motion . no publicly available photometry software was designed to handle trailed psfs . we conducted the first study that quantitatively determines which available photometry algorithm optimizes accuracy and precision for faint moving targets . we familiarized ourselves with 15 different aperture or psf - fitting algorithms and tested which one most accurately and precisely reproduced the true light curve of a moving object with a known rotational light curve . one way of testing a technique s accuracy is to use it on data with a known trend and measure the root - mean - square ( rms ) of the residuals against a model . we acquired data for the trans - neptunian object 1996 to@xmath0 as part of a campaign to determine the object s light curve . as the purpose of this paper is simply to use this dataset to explore methodology , we will not comment on the campaign itself , nature of the target , or the details of the model and instead defer that information to an upcoming publication ( hainaut _ et al . _ , _ private comm . _ ) . the data were obtained on september 22 , 2011 using the university of hawaii ( uh ) 2.2-m telescope with the kron - cousins @xmath1-band filter and the tek ( @xmath2 ) ccd camera , which has a pixel scale of 0.219@xmath3/pixel . the sky was not photometric due to occasional cirrus clouds . we measured a median seeing of 0.95@xmath3 fwhm , which varied by @xmath4 over the course of the 1996 to@xmath0 observations . the focus was checked before beginning the 1996 to@xmath0 observations , but we could not change the focus later as the night progressed due to technical difficulties . consequently , a slightly triangular - shaped psf indicative of poor focus was present in two of our images ( fig . [ fig : images ] ) . exposure times varied between 900 and 1100 seconds ( the median being @xmath5 ) to achieve our s / n goal of @xmath6 , which is still too faint to have well - defined psf wings . this was done in an attempt to achieve a constant s / n under changing transparency and seeing conditions . non - sidereal guiding was used so that the target was not trailed . the target s motion averaged -2.7@xmath7hr in r.a . and -0.8@xmath7hr in dec . , corresponding to @xmath8 of trailing per frame for stars , or a theoretical psf length - to - width ratio ( or trailing aspect ) of @xmath9 for field stars and 1.0 for the target . the best - fit models using the tphot psf - fitting algorithm , however , give trailing aspects of 1.5 for the field stars and 1.1 for the target ; the difference between the theoretical and actual trailing aspect is likely attributable to telescope guiding problems ( fig . [ fig : contours ] ) . we stress that this amount of trailing is considered minimal compared to typical rates of inner solar system objects @xcite . the medians of the peak and background counts were @xmath10 and @xmath11 , respectively , and the median flux within a 7-pixel aperture was 23000 @xmath12 600 counts . we calculate a medan s / n of @xmath13 using the tphot algorithm ( discussed in section [ sec : tphotap ] ) . figure [ fig : images ] shows both 1996 to@xmath0 s track along the sky and the target - centered sub - sections of the ten 1996 to@xmath0 images . the observations interleaved 1996 to@xmath0 with another target over 8.12 hours , until 1996 to@xmath0 reached airmass 2.0 . preliminary results from hainaut _ et al._@xmath14(private comm . ) give a light curve amplitude of @xmath15 magnitudes and a rotation period of @xmath16 hours . the model light curve reproduces all data obtained as part of a multi - year , multi - telescope , multi - solar phase angle observational campaign to determine the light curve of 1996 to@xmath0 . our observations correspond to rotation phases ( @xmath17 ) of @xmath18 , or 87.5% of the full rotational light curve . one data point was consistently fainter than the model by @xmath19 magnitudes regardless of the technique used , likely due to a bad pixel that was not identified during construction of a bad pixel mask , so we excluded it from analysis ( fig . [ fig : images ] ) . we prepared the images for analysis using iraf s `` ccdproc '' package @xcite , applying bad pixel , bias , and overscan corrections and flat - fielding with dithered twilight sky frames . to exclude field stars from the bad pixel mask construction , we median - combined the brightest twilight flats to produce a stacked bright flat , then repeated that process for a stacked faint flat in the same filter . we fed the ratio of the stacked bright to the stacked faint flat to the iraf `` ccdmask '' task , which computes a bad pixel mask from a ratio image , then used the `` fixpix '' task within ccdproc to interpolate over bad pixels . we noted that in two images , a bad pixel was relatively close to the target 8 pixels from the target centroid in frame 3 ( fig . [ fig : images ] ) , and 9 pixels from the target centroid in the frame that was excluded from analysis ( see section [ sec : obs ] ) . the bad pixel regions were visually and quantitatively comparable to the local sky after correction ( as it should be ) , so their proximity to the target was of no consequence to our photometry . afterward , we fit a polynomial to the overscan for line by line bias subtraction and trimmed the data . we then combined 20 bias frames taken at the beginning and end of the night to produce a stacked bias frame for subtraction . we acquired seven @xmath1-band twilight sky flats at the beginning of the night and six at the end , all below the limit of the linearity regime for the chip . however , because the dust pattern changed multiple times within the 1996 to@xmath0 observation window , the twilight flats were not always representative of the field . consequently in the flattened images , residuals from improperly - corrected dust donuts can be seen in a third of our images , possibly influencing background determination for field stars in affected areas . the residual dust donuts were not close enough to the target to affect the target photometry from any of the algorithms . lastly , we corrected for cosmic rays using the `` cosmicray '' task within iraf s `` crutil '' package . we found that a threshold of @xmath20 relative to noise in @xmath1-band images effectively removed most cosmic rays without falsely flagging real sources . the entire reduction procedure from bad pixel to cosmic ray correction did not differ between photometry algorithms . after reduction , we measured the effective gain and read noise by doing a linear least squares fit to the square of the sky noise over the sky , all in units of adu . the slope of the fit is the inverse of the gain , and the y - intercept is the square of the read noise in adu . we measured a gain and read noise of @xmath21 e@xmath22adu and @xmath23 e@xmath24 , respectively . field source detection for all photometry algorithms was done using source extractor ( sextractor ) , which identifies point sources as a user - defined minimum number of adjacent pixels ( sharing either a border or a corner ) that are above a specified detection threshold . if the intensity distribution of a cluster of source pixels shows a distinct saddle point , the surrounding peaks are considered to belong to two separate objects . sextractor then measures the shape and centroid of each object , eliminates bright object artifacts , and further cleans the source list . because we find this rigorous detection method very effective at identifying , de - blending , and broadly classifying all field sources , we used it to produce a coordinate list for field stars . we restricted field stars for further analysis to those that had s / n @xmath2510 within an aperture diameter of @xmath26fwhm , as preliminarily determined by sextractor s `` mag_best '' routine . to correct for extinction and occasional clouds for each algorithm , we applied an offset correction for each frame to bring the field stars brightnesses to the same level . this correction was computed by removing the median offset in bright field star magnitudes between frames . we recognize that systematic errors - a bias in the way the photometry was performed , not in the data itself - may be present due to over- or under - sampling of the background , non - optimal criteria for rejecting contaminants , and/or source flux threshold being too low or high . to test and correct for systematic errors , after removing the shift between frames for cloud correction , we subtracted each star s median magnitude for the night and fit a gaussian to the distribution of differential star magnitudes in that frame . the hwhm of the gaussian fit ( @xmath27 ) is the same as the total error ( @xmath28 ) and should be a sum in quadrature of the statistical and median systematic error for each target ( @xmath29 and @xmath30 , respectively ) . therefore , @xmath31 . if @xmath32 , then we considered @xmath33 . we note that this is the systematic error for field stars , not the untrailed target . because the trailed field stars cover a larger number of pixels , the probability of contamination and erroneous background sampling is higher than for the target therefore , these systematic errors may be overestimated for the target . figure [ fig : syserr ] shows a typical gaussian fit to the differential magnitude distribution . we performed monte carlo simulations of random gaussian distributions with different sample sizes and found that fitting greater than 30 field stars gave a @xmath34% chance of getting a real and reasonable solution . we found that @xmath27 varied depending on the absolute magnitude range of the stars used , with brighter stars giving a smaller @xmath27 than fainter stars . therefore , to determine the systematic error most representative of the target , we considered only the @xmath35 field stars closest in magnitude to the target ( r - magnitude @xmath36 ) . lastly , to fit the data to the model , we tested relative offsets in increments of 0.0001 magnitudes within 0.1 magnitudes of the median - subtracted target magnitudes and used the offset that gave the lowest rms residuals against the model . aperture photometry packages usually differ in their definition of aperture size and shape and/or in the way the background is determined , but they are popular largely because they are not sensitive to irregular psf shapes . determining the background correctly is one of the fundamental challenges of photometry . @xcite found that the flux measured within an aperture radius is extremely sensitive to correct background measurement and subtraction . if the sky is overestimated , then too much background is subtracted , and as the aperture size is increased , the sum of counts within the aperture decreases rather than yielding a constant flux . this effect is less prevalent for brighter objects , whose psf wings are more distinct from the noise out to larger radii . it is therefore recommended not to use a bright star alone to determine which aperture radius captures a large percentage of the source flux ( @xmath37% ) because a faint object s flux at that radius may already be significantly contaminated by background . @xcite also found that different background measurement methods ( _ e.g. _ , median vs. mean , etc . ) yield different accuracies . they found that a weighted mean is more accurate for fainter objects but underestimated uncertainties for brighter objects . in addition to proper background subtraction , variation in the psf across the frame caused by optical aberrations is an important consideration with aperture photometry . most aperture photometry algorithms , save sextractor , do not allow for variation in aperture size across a chip , leaving it difficult to choose the best constant aperture radius for the entire frame . in a detailed study using the hst wfpc-2 ccd camera , @xcite found that radial variation in the images was significant ( @xmath38% ) , and @xcite found up to a 15% difference in fwhm across the ccd cameras for the sloan digital sky survey ( sdss ) . although these are extreme examples of wide field imagers with large distortions , it is important to check for spatial variations . we saw no such distortion in our images to a fraction of a percent , so no spatially - variant aperture was necessary . popular aperture photometry algorithms that were compared include the phot task within the iraf / apphot package and source extractor ( sextractor ) . we also implement the tphot aperture photometry technique , optimized in centroiding and reducing background contamination . ours is the first publication of the tphot software . lastly , the aperture photometry tool ( apt ) was developed as an educational tool , and because of its ease of use , possible similarity in quality to sextractor , and application to a wider populace , we include it in our study @xcite . the phot task is widely used for traditional aperture photometry @xcite . an identical task can be found within the daophot package . depending on the sky in the vicinity of the target , there are several opportunities to change the way the background is measured within phot , four of which are tested here ( each on all 10 images ) with iraf version 2.14.1 . phot gives the option of rejecting sky annulus pixels that are above or below a user - defined sigma - threshold , which we chose to be 2.5 . photometry for both the target and the field stars were done with these settings . to determine the aperture radius , we measured a curve - of - growth on several stars of different magnitudes spread over the chip and identified the aperture radius that contained 99.5% of the light , with weighting given to brighter stars . in general , we found that this aperture only varied by @xmath39% between bright unsaturated stars and fainter stars at s / n @xmath6 , indicating that the background was measured accurately . we repeated this process for all 10 images used and took the largest frame - specific aperture ( 18 pixels , or @xmath40 ) as the fixed , unweighted aperture radius for all images . the aperture chosen was used for both the trailed stars and the target . one concern in choosing a photometric aperture radius for moving targets is that the target psf should be narrower than trailed field stars . consequently , the aperture chosen from field star curves of growth may be larger than the target s 99.5% flux radius , causing an increased contribution from sky noise and potential faint background source contamination . if the target moves along a track free of significant background contaminants within the sky annulus ( see fig . [ fig : images ] ) , then a target - centered sky annulus best represents the sky in the photometric aperture . for fainter point sources , the psf wings are indistinguishable from random noise beyond the aperture containing 99.5% of the target flux ( assuming the ccd is flat ) , so the choice of inner radius of the sky annulus is somewhat arbitrary , so long as it is outside the aperture . for bright point sources , however , a small amount of noticeable source flux may be in the region outside the aperture , affecting the magnitude uncertainty . to be sure we were clear of the psf wings of these bright stars , we defined the inner radius of the annulus at @xmath41 , or 28 pixels ( 10 pixels greater than the aperture ) . we chose the outer bounds of the sky annulus to be @xmath42 ( 50 pixels ) to ensure good background sampling . if the sky annulus centered on the target contains a significant amount of unavoidable contamination from background sources , one can manually measure the sky background away from the target . a manually - defined sky annulus would also be necessary for active comets to avoid dust trails . we used a sky aperture of the same area as the centered sky annulus in section [ sec : photcentered ] and measured the sky 10 - 15 times in a radially - isotropic distribution within @xmath43 ( 80 pixels ) of the target for every frame , avoiding overlap with the photometric aperture or contaminants . the final sky and sky noise levels used in the photometry for each frame were the average of the individual measurements in that frame . manually measuring the background at several locations off - center from the photometric aperture is too time - intensive to execute for every field star in every frame . thus , for field stars only , we used phot s non - interactive centered sky annulus technique described in section [ sec : photcentered ] . since the systematic error is empirically determined from the dispersion in differential field star magnitudes , the systematic error per frame was the same as determined in the centered sky annulus method . decreasing the photometric aperture excludes some sky noise and nearby contamination but may also exclude source flux . we used bright field stars with well - defined psf wings to estimate the aperture correction , or amount of source flux lost by using a smaller aperture . we then added back the residual magnitude estimated from our aperture correction to the magnitudes measured with the smaller aperture for the target . with moving targets , we risk that the untrailed target may not be accurately represented by the aperture correction given by trailed field stars , which have a wider psf than the target . the choice of the small aperture size is somewhat arbitrary , so long as it is consistent from frame to frame , so we chose a small aperture radius of @xmath45 ( 10 pixels ) to avoid the cosmic ray hit seen in frame 1 ( fig . [ fig : images ] ) . we measured the aperture correction between the small aperture and a large aperture at @xmath46 ( 18 pixels ) , using only field stars with @xmath28@xmath47 , to ensure good flux sampling in the psf wings . field stars that gave aperture corrections more than 2 standard deviations from the mean for that frame were excluded from the computation , leaving on average 16 field stars per frame to determine the aperture correction . the standard deviation of these stars aperture corrections ( @xmath48 ) was used as the error on the frame s aperture correction , which was added in quadrature to @xmath29 and @xmath30 to compute @xmath28 ( _ i.e. _ , @xmath28@xmath49@xmath29@xmath50@xmath30@xmath51 ) . to measure the sky , we used the sky annulus settings described in section [ sec : photcentered ] . there are sometimes faint background contaminants close to the target that render inaccurate sky determination or add to the target flux . these background sources may not be obvious in individual frames , but can be seen in a stacked image . to correct for these contaminants , we constructed a median - combined sky composite ( to remove the moving target ) , scaled the sky composite to match the mode of an individual frame s background , and subtracted the scaled , aligned composite from each frame using the iraf `` imarith '' task . the sky composite was made from 8 of the 10 science frames , with frames 8 and 9 excluded because the target s location did not change enough relative to other frames to remove all target flux from the median sky composite . the subtracted images should be free of faint sources , but residuals from subtraction close to the centroid of brighter sources are often apparent , caused by inaccurate alignment before subtraction and/or the changes in seeing / image quality or non - linear regime of the detector . we also found that inaccurate alignment before subtraction led to substantial inaccuracies in background measurement , so we used dozens of bright field stars to compute a shift as accurately as possible . figure [ fig : skysub]b shows one of the individual frames used in our analysis after imarith sky subtraction . sky subtraction takes place _ before _ performing photometry , so to assess the improvement to the results , we apply the same photometry technique used on frames without sky subtraction and compare the outcomes . we chose to implement the aperture correction photometry method ( section [ sec : photapcor ] ) because it is believed to produce more accurate results than the centered and off - center sky annulus methods described in sections [ sec : photcentered ] and [ sec : photoffcenter ] . field star photometry was performed on the frames before sky subtraction , and photometry on the target as performed after the subtraction . section [ sec : photimarith ] describes how sky subtraction can improve photometry by removing background contaminants , but constructing a sky composite is a sensitive function of determining the correct sub - pixel shift between stars in different frames and scaling the composite to match each frame . the isis software package @xcite , which internally determines shifts , median - combines frames , and subtracts the composite from individual images , was designed to do the best possible job of sky composite construction and subtraction . the main differences between using imarith and using isis to build a sky composite are that : ( 1 ) the isis composite must be built from frames with the best seeing , and ( 2 ) before subtracting the composite , isis convolves it with a kernel solution for the individual frame , matching the fwhm of the sources in the composite to the same sources in the individual frame . we made the isis sky composite from the five frames with the best seeing , with a mean seeing of @xmath52 compared to the remaining five frames mean seeing of @xmath53 . isis can also perform differential photometry after sky subtraction . although @xcite found that using isis instead of traditional aperture photometry did not improve photometry results for star clusters , @xcite noted an improvement by up to a factor of 20 in isis photometry over dophot . however , we were not able to fully adapt the isis software for moving objects , even when shifted stamps showing the object at the same position were manually constructed and fed into the algorithm . therefore , we could not make use of isis s photometry capabilities for our moving target ; instead , we only used it for sky subtraction . figure [ fig : skysub ] shows that isis offers considerable improvement over imarith in removing background sources . the standard deviation within an aperture radius of @xmath54 ( or 18 pixels ) at the location of subtracted field stars was reduced by a factor of @xmath55 by using isis rather than imarith to perform sky subtraction . source extractor ( hereafter sextractor ) is a photometry algorithm that was designed to quickly identify and classify field objects and to perform aperture photometry on both extended and point sources . we used sextractor 2.5.0 . one of the main differences between sextractor and other aperture photometry algorithms is that sextractor constructs a background map for the entire image rather than measure the background within a defined annulus for each object . the background map is locally determined over a defined mesh size , and pixels above a 3@xmath56 threshold within the mesh are iteratively discarded from the background computation until no pixels above this threshold remain . if the field is considered crowded ( _ i.e. _ , the sky noise drops less than 20% per iteration ) , then the final background within the mesh is computed in a different way from an uncrowded background @xcite . the final background map is a bi - cubic - spline interpolation over all regions . by visual comparison between the background map and an individual frame , we found that a mesh size of @xmath57 pixels and a smoothing - region size of 3 gave the most accurate background map . sextractor also offers flexibility in the aperture shape . there are five types of apertures available within sextractor , all of which were tested in this study : mag_aper , mag_auto , mag_iso , mag_isocor , and mag_best . the mag_aper option lets the user define a fixed circular aperture to place over all identified sources in a frame . we chose an aperture radius at @xmath26hwhm ( which theoretically encompasses 99.994% of the object s flux ) for each frame . using the mag_auto aperture choice will internally determine an aperture at the kron radius , defined as the radius that contains 90% of a source s flux , for each identified field source @xcite . the aperture s ellipticity and position angle with respect to a row is then computed via the second order moment on each object . mag_iso draws an aperture around all adjacent pixels above a defined detection threshold , making it highly flexible for irregularly - shaped psfs and extended sources . we chose a recommended detection threshold of 1.5@xmath56 above the background @xcite . the mag_isocor aperture type is a crude method of determining an aperture correction to the mag_iso magnitudes , assuming the object s psf is an axisymmetric gaussian @xcite . lastly , mag_best is a combination of mag_auto and mag_isocor ; when the neighboring source contamination reaches 10% , mag_isocor measurements are returned , otherwise mag_auto measurements are returned . the tphot routine has three basic components : ( 1 ) triggering on a potential object , ( 2 ) aperture photometry , and ( 3 ) psf fitting photometry . the trigger function identifies all pixels that are a local maximum out to a specified radius ( nominally 5 pixels ) and that exceed a threshold level ( specified as an absolute level or as a s / n above sky level and noise ) . tphot performs photometry on all triggered objects , but this these calculations may be rejected according to a number of tests , including inadequate s / n , a fitted centroid that moves from the brightest pixel , a profile that is too broad or narrow , etc . tphot offers the option of using an external file of object positions . the uncertainty in the photometry is based on a noise model that is implemented using two parameters , a gain ( e@xmath24/adu ) and a bias level . the `` bias '' level is subtracted from the image and the noise is then taken to be the square root of the number of electrons . the `` bias '' level can be negative if the image preprocessing has already subtracted a sky level , and it can be augmented according to read noise variance . the aperture photometry function of tphot performs a fit for the sky level near an object , and then calculates the total flux within a specified aperture , less the sky contribution . for maximum accuracy the aperture should be 1.5 fwhm in radius @xcite . for robustness in the presence of neighboring objects , tphot starts by calculating the median flux in annular rings around the object of interest . the sky level is determined by fitting the outermost rings beyond a specified radius with a constant plus @xmath58 object profile . the object flux is calculated as the sum of pixels within the aperture radius , less the sky level . this algorithm has proven to be very accurate and robust for bright objects and fields that may be somewhat crowded . the aperture photometry tool ( version 2.1.9 , hereafter apt ) is a relatively new algorithm intended as both a professional and an educational tool capable of performing straight - forward and accurate photometry @xcite . its main feature is its graphical user interface , encouraging visualization of the process and visual checks on aperture and sky annulus choices . because of its simplicity , apt was not designed for or tested on crowded fields . the user defines the sky annulus and the radius , ellipticity , and position angle of the photometric aperture . there are also options to remove the median or mean background , perform photometry in batch mode , define an aperture correction , and include / exclude manually - selected pixels from the aperture radius . apt only accepts integer input pixels , then upon the user s request , it internally recomputes the centroid to 0.01 pixel resolution . the apt help manual states that this sub - pixel centroiding may perform poorly , especially in more crowded fields . the user can also define rejection limits ( in counts ) above and/or below which sky annulus pixels are excluded from background determination . performing photometry manually for every field source was impractical because we had over 100 objects in our target list , so we chose to operate in batch mode , which meant using the same aperture and sky annulus settings across the entire frame . we chose a fixed circular photometric aperture of 18 pixels , inner - sky radius at 28 pixels , and outer - sky radius at 50 pixels ( the same settings used in section [ sec : photcentered ] ) . we also chose to perform median background subtraction and make use of the pixel exclusion feature to remove the cosmic ray hit in frame 1 ( fig . [ fig : images ] ) . we used the apt - computed sky level and sky sigma in the target s sky annulus to determine the @xmath59 rejection threshold for background determination in each frame . psf - fitting algorithms differ amongst themselves primarily in the way the model is stored for later use @xcite . some fit an analytic function ( gaussian , lorentzian , or a combination thereof ) while others determine an empirical psf model through numerical interpolation . some psf - fitting algorithms also offer a spatially - variable solution in case the psf varies across the chip . because psf - fitting routines theoretically exclude more background contamination from the source flux than an aperture , they offer a potential advantage in accuracy and precision of faint objects @xcite . a few comparison studies of psf - fitting algorithms have been published . @xcite found that daophot outperforms dophot by a factor of 1.5 - 4 in accuracy for fixed targets , but @xcite and @xcite noted that daophot can be susceptible to background source contamination . however , the usefulness of these algorithms for moving target photometry has not been explored . the fact that trailed field stars are used as templates for the psf model presents a problem to moving targets in that the model will not exactly represent the untrailed target . we investigate the accuracy and precision of the popular psf - fitting algorithms daophot and dophot and the profile - fitting capabilities of the new tphot algorithm . daophot is one of the most popular photometry algorithms . this software package operates within iraf and exploits the strengths of both analytic and empirical approaches to psf - fitting . first , an analytic function is fit to manually - inspected template stars without overlapping psfs , and errant pixels in the source psf ( either cosmic ray hits or bad pixels ) are down - weighted and local background is determined . six different types of analytic functions can be fit - gaussian , lorentzian , 2 modified lorentzians , and 2 gaussian @xmath44 lorentzians @xcite . the model that returns the lowest scatter is subtracted from each template star , and the average residuals are stored in a two - dimensional lookup table to account for any flux lost in the analytic model . the user can specify whether or not the psf varied across the chip , in which case daophot will compute additional lookup tables representing the linear or quadratic change in flux across the frame . to ensure that flux is not artificially enhanced or decreased due to spatially - variant lookup tables , the net volume of the higher - order tables is forced to be zero . the option for a spatially - variable model solution have sometimes been implemented with difficulty , even failing altogether as they did for us @xcite . for each field source , the model is scaled down to the flux within a defined fit radius ( usually the fwhm ) , and the psf model is iteratively fit to field sources . we altered the many input free parameters as recommended by the daophot reference guide , using the daoedit task to determine and define each image s fwhm , background , and sky noise @xcite . we defined the inner - sky radius as @xmath26fwhm , the outer - sky radius as @xmath60fwhm , and the radius within which the model computed for template stars is computed as @xmath61fwhm@xmath62 , following the @xcite suggestions . because the target was moving and we were comparing photometry for individual frames , we were unable to use daophot s allframe task , which performs photometry on a stacked image @xcite . dophot @xcite was developed as an alternative fitting algorithm to daophot , because determining the appropriate settings for each image and computing the model psf for daophot is very time - consuming . dophot requires minimal user interaction and optimizes speed ; we used version 4.1 . the user inputs estimates for the background , seeing , gain and read noise , which are used by the algorithm to identify sources above a defined threshold . these sources are fit with an analytical power - law function to determine a best - fit model , which is then subtracted from the frame . noise is added back to the black sky patches left after star subtraction to ensure that the patches resemble the sky noise . the detection threshold is then lowered to fit a model to fainter sources , which are then subtracted . the brighter sources already identified are added back into the frame and refit , now without possible contamination from faint sources . the improved model parameters are saved , and the magnitude , background , and poisson noise close to the source is recorded . these iterations continue for lower and lower thresholds ( _ i.e. _ , fainter objects ) until no additional stars are detected . a final model weighted by s / n is computed from all objects that are unambiguously identified as point sources , then all of the objects detected in the image are refit with the improved model . we used daoedit s background and fwhm measurements ( see section [ sec : daophot ] ) for the dophot input background and seeing estimates , respectively . otherwise , we preserved the default setting as recommended by the dophot manual @xcite . after performing an aperture photometry calculation ( section [ sec : tphotap ] ) , tphot also performs a psf fit to each potential object . using the position and crude fwhm from the aperture fit , tphot does a least - squares fit of an elliptical `` waussian '' ( gaussian truncated at the @xmath63 term ) profile . tphot can be requested to perform profile fits of two parameters ( flux and sky ) , four parameters ( plus position ) , or seven parameters ( plus psf shape ) . it also can fit a trailed profile , substituting trail length for major axis . the resulting major axis @xmath64 , minor axis @xmath65 , and peak value @xmath66 can be multiplied to obtain a value @xmath67 that is proportional to the total flux of the waussian . depending on the application , tphot may be run multiple times . it is most commonly run once at a relatively high s / n threshold to obtain aperture photometry and psf profile shape parameters for bright stars . it is then rerun at a low s / n threshold and a four parameter fit , forcing all stars to a common psf profile . the @xmath67 value provides accurate , relative photometry for faint objects as well as bright ones , and the aperture photometry of bright stars from the previous iteration provides net fluxes . in this image there are a comparable number of galaxies and stars ; a different galactic latitude or magnitude limit would be different . we did not attempt any automatic star - galaxy classification for this application . instead , we chose the objects whose major axis a falls below the median , and then use the median values of those objects for @xmath64 , @xmath65 , and @xmath68 as representative of stellar psfs . a mode of all values less than the median @xmath64 , @xmath65 , and @xmath68 would have also been suitable . measuring the change in aperture photometry of stars between different frames then correcting for the change put all images on the same photometric scale . the rms for these comparisons was less than 0.01 magnitude . although the target was trailed , its trailing amount was small enough relative to the seeing that we did not bother to use the trailed profile function of tphot . we were not able to test all existing photometry algorithms . @xcite find that the photo software used on the sloan digital sky survey performs high - quality photometry on both extended sources and field stars , but it is not portable or flexible and was thus not available to install . three other prominent photometry packages , isis , momf , and dia , could not be modified for use on data for a moving target , though they did offer promising improvements to our photometry results @xcite . these packages all perform photometry on fixed objects in median - sky subtracted images , identifying low - level deviations from the nightly median . the main advantage of these methods over traditional sky - subtraction is the process of convolving a `` best - seeing '' sky composite with the fwhm of individual frames before subtraction , so that the psfs match and no residual background source flux remains on the chip after subtraction , with the exception of saturated pixels ( see fig . [ fig : skysub ] and section [ sec : photisis ] ) . because the change in magnitude is measured from sky - subtracted images , no absolute magnitude calibration is offered in any of these methods . these difference - image techniques may also present a challenge to moving targets in that the fwhm would not be the same for the target as the trailed stars , so the convolution would be incorrect and residual flux would always be present . section [ sec : photisis ] details how we were unable to adapt the promising isis software for our moving target . we assumed we would experience the same difficulties implementing the other difference image techniques , so we did not attempt to run them . the potential of these difference - frame algorithms to optimize accuracy and precision begs that they be made adaptable to moving objects . for each algorithm , we use the rms of the residuals against the light curve model as a diagnostic for magnitude accuracy , and the @xmath69 statistic as a measure of how reasonable the magnitude uncertainties are . section 2 explains the model s origin . figure [ fig : summary ] plots the results of all algorithms against the model light curve , shown in order of least to greatest rms of the residuals . we also report @xmath30 to determine how much systematic bias is incurred in each algorithm and @xmath28 ( calculated using the @xmath30 determined for each frame and the target - specific @xmath29 reported by each algorithm ) , which assesses the precision of the results . if the magnitudes themselves are accurate and all systematic errors have been accounted for , then @xmath69@xmath70 . if @xmath69@xmath71 , then the errors ( including systematics ) are overestimated , but if @xmath69@xmath72 , either the magnitudes are inaccurate or @xmath28 is still underestimated . table [ tab : stats ] lists the rms of the residuals , the @xmath69 , @xmath28 , and @xmath30 for each implementation . figure [ fig : summary ] and table [ tab : stats ] show that the widely - used aperture photometry algorithm phot generally performed poorest of all other algorithms tested , with or without subtracting a sky composite with imarith . we saw no change in the results by using a manually - placed , off - center sky aperture as opposed to a centered sky annulus , presumably as long as the sky measurements were made relatively close to the target ( fig . [ fig : summary]m vs. fig . [ fig : summary]n ) . aperture correction gave an improvement in rms by a factor of 2.5 over using an aperture that encompassed 99.5% of the flux , making it comparable to sextractor results . since the sky was measured in phot in the same way before or after sky subtraction , the improvement in rms must be caused by a more accurate accounting of source flux . the two phot algorithms that involved subtracting a sky composite to remove faint background contamination before performing photometry ( sections [ sec : photcentered ] and [ sec : photoffcenter ] ) generally rendered a larger light curve amplitude than the model , which increased the rms of the residuals and @xmath69 ( fig . [ fig : summary]f and fig . [ fig : summary]o ) . visual inspection shows that the isis - processed images contain one exceptionally deviant data point at @xmath73 . this data point corresponds to frame 1 in figure [ fig : images ] , where a cosmic ray hit that was missed in initial cosmic ray correction is seen close to the target . it may be that the cosmic ray hit compromised the photometry more so in the isis - treated frames than in results from other realizations of phot , though there is no clear explanation for why this would be the case . also , the cosmic ray removal routine that we used obviously did not correct for all hits because it was designed to flag point source cosmic ray hits ( _ e.g. _ , single pixels ) rather than grazing hits such as the one in fig . [ fig : images ] . thus , the routine may have left additional faint contaminants near the target psf and affected the accuracy of the photometry . only after removing the anomalous data point at @xmath73 from analysis did isis sky - subtraction render a rms comparable to the other phot aperture correction algorithms . subtracting the sky with imarith before applying an aperture correction offered no improvement in rms , likely because contaminants were already sufficiently removed from source flux measurements by using a smaller aperture . we speculate that if the background within the sky annulus contained more faint sources , then performing imarith sky subtraction would offer further improvement in accuracy before using aperture correction . the @xmath69 @xmath74 for all phot algorithms , especially the two algorithms involving sky subtraction ( table [ tab : stats ] ) , comes from incorrect magnitude measurement , background determination , and/or noise models . to test whether or not the background was correctly determined and subtracted , we repeated the centered sky annulus method ( section [ sec : photcentered ] ) with successively larger apertures ; if the background was overestimated , then the magnitude measured would become steadily fainter toward larger radii , and vice versa @xcite . we found that the magnitude stayed the same , only the magnitude uncertainty changed using larger aperture radii , implying that background subtraction was done correctly . phot calculates the magnitude uncertainty from the photon noise within the aperture and the standard deviation within the sky annulus , and it includes a term to account for uncertainty in the background level . since we find that background subtraction was done properly , we assume that the standard deviation measurement within the sky annulus is also correct . with sky subtraction , we found that the standard deviation of the background was reduced by a factor of 2 , which consequently gave a smaller magnitude uncertainty and even larger @xmath69 . if the aperture contained contamination from background sources , however , then both the source flux and aperture s photon noise would be inaccurate , the magnitude more so . such an effect would manifest as slightly large @xmath28 and even larger rms , which is what we see in the phot results . the factor of @xmath75 improvement in rms by using a smaller photometric aperture but measuring the background in the same way provides further evidence that background contamination within the aperture is significant . @xcite compared apt to sextractor s mag_aper , noting overall agreement between the two algorithms . we do not find the same result here , possibly indicative of apt s limitations toward fainter objects . table [ tab : stats ] shows that apt s residual rms is roughly twice that of mag_aper , though the @xmath69 is slightly larger due to the slightly smaller @xmath28 and much larger rms . to investigate the cause of the large rms , we tested for correct background determination by checking that the source flux stayed constant in several aperture radii beyond the original 18-pixel radius . we saw that as we increased the aperture size , the source flux continued to decrease , indicating that the background was overestimated . incorrect background determination can lead to inaccurate measurement of source flux and background uncertainty , which then increase the rms and @xmath69 . some photometry inaccuracy may stem from incorrect sub - pixel shifts , as noted in the apt manual @xcite . like phot , apt also appears to do a poor job of systematically excluding faint background contamination from the aperture . we found that apt measured the largest target flux within a 10-pixel radius of all aperture photometry algorithms , even with the background overestimated , implying background contamination . the dophot results are given in figure [ fig : summary]l and table [ tab : stats ] . the comparably large rms shows that dophot did not do a good job of reproducing the model , performing worst of the psf - fitting algorithms but slightly better than phot s non - aperture correction algorithms . magnitude uncertainties are also relatively large compared to the algorithms tested here . we suspect that the main sources of error for both magnitudes and uncertainties are that ( 1 ) dophot psf - fitting fails at accurately modeling the psf of a trailed field star , and ( 2 ) dophot is too inclusive in psf model computation . if a field contains many irregularly - shaped , compromised , and/or extended sources , then the final model may be significantly affected . due to dophot s limited user interaction , we could not further explore the role of potentially incorrect background determination , insufficient uncertainty estimates , and/or contamination in increasing the rms and @xmath69 . the @xmath30 was also the largest of all algorithms tested , indicating that a substantial amount of systematic bias was incurred . table [ tab : stats ] shows that sextractor improved the rms by a factor of @xmath76 over phot ( except for the aperture correction algorithm ) and @xmath77 over apt and dophot , implying better magnitude accuracy . of the sextractor aperture choices , mag_isocor delivered the lowest rms but relatively unchanged @xmath28 , meaning it offered an improvement in accuracy but not in precision . mag_isocor determines an aperture correction to the mag_iso magnitude measurements ( fig . [ fig : summary]h ) , and the aperture correction technique was already shown in section [ sec : discphot ] to be more accurate than using a large aperture to add up source photons , so it is unsurprising that mag_isocor provided an improvement in rms and @xmath69 over mag_iso ( table [ tab : stats ] ) . @xcite noted that in isophotal mode , the position and shape of sources detected in sextractor may be systematically inaccurate toward fainter magnitudes , but given the success of mag_iso and mag_isocor here , our target must not have been faint enough to be affected . the mag_aper choice ( fig . [ fig : summary]j ) gave the smallest @xmath69 , owing to it having a relatively large @xmath28 and a minimally different rms . this finding suggests that a fixed circular aperture can include significant amounts of background , adding more background uncertainty to the final error . the mag_auto aperture option ( fig . [ fig : summary]g ) offered essentially no improvement in accuracy but improvement in precision by a factor of @xmath78 over mag_aper , implying that it was more effective at excluding pixels that do not contain source flux . the mag_best results were almost identical to the mag_auto results , meaning that the field was relatively uncrowded ( table [ tab : stats ] ) . despite sextractor s general success over other aperture photometry algorithms ( except tphot ) at accurately reproducing the model light curve , the @xmath69 was comparably large . a smaller rms but larger @xmath69 suggests that the magnitude uncertainties are underestimated . pointed out that the uncertainty in the local background estimate ( which is notably complex ) was not included in the final reported magnitude . we therefore assume that this exclusion made the magnitude uncertainties in our experiments artificially and significantly small . because the degree to which the magnitude uncertainties are artificially small is unknown , it is difficult to determine whether or not inaccuracy of the sextractor magnitudes also contribute to the large @xmath69 value . the daophot results are given in figure [ fig : summary]c and table [ tab : stats ] , which shows that daophot is the third - most accurate algorithm tested , behind the two tphot algorithms . the @xmath28 is comparably small , indicating good isolation of source flux , though the small @xmath69 value may indicate that the magnitude uncertainty is still overestimated . @xcite noted that systematic errors stem from inadequate correction factors in the wings of the model psf , and that daophot consistently underestimated the uncertainties by @xmath79% , which is inconsistent with our results . we suspect that the underestimation they noted was for brighter objects , where the intricacies of the true psf may be poorly reproduced by the model . the @xmath30 in our results was relatively small , suggesting minimal systematic bias compared to other algorithms . the somewhat small amount of trailing in the template stars did not seem to affect the accuracy of the psf model s fit to the target . we hypothesize that because the trailing was within the expected seeing disc , daophot s analytic functions had no difficulty computing a model that was applicable to both the trailed template stars and the untrailed target . we tried using daophot to build a model from moving - object frames where the stars showed more pronounced trailing ( @xmath80 , or trailing aspect of @xmath81 ) , but the @xmath69 fit of the model to the untrailed target was @xmath82 per frame , clearly indicating that the template stars were too trailed to reasonably match the target s psf . in contrast , the average @xmath69 of daophot fits to the 1996 to@xmath0 data was 1.84 . @xcite found that performing aperture photometry within daophot ( the same as aperture photometry with phot described in section [ sec : phot ] ) rendered more accurate results than daophot s psf - fitting routine because the flux of faint objects is increasingly underestimated toward fainter magnitudes . because our results do not reflect this flux inaccuracy , we assume that 1996 to@xmath0 was bright enough ( s / n @xmath83 from daophot ) for this not to be a factor . results from the tphot psf - fitting and aperture photometry algorithms are shown in figures [ fig : summary]a and [ fig : summary]b , respectively , and in table [ tab : stats ] . the low rms value shows that tphot produced the most accurate results of any algorithm tested . the tphot psf - fitting algorithm returned the lowest @xmath28 of all algorithms tested , implying excellent precision , but @xmath69 @xmath84 suggests that the errors may be very slightly underestimated . visual inspection of the results in fig . [ fig : summary]a shows that the uncertainties on magnitudes are consistent with the model , except for one data point at @xmath85 ( corresponding to frame 3 in fig . [ fig : images ] ) , which is the likely cause of the large @xmath69 . we can find no reason to exclude this point from analysis unless the model is poorly constrained at this particular rotation phase or the target psf contains unavoidable deviant pixels . removing this point gives @xmath69 @xmath86 , exactly what it should be if the data accurately represent the model , so either the questionable data point is compromised in some way or tphot is reporting inaccurately . given the success of tphot at reproducing the model at all other phases , we assume the problem is not with the algorithm itself and rather with the data , possibly due to a bad pixel missed during bad pixel mask construction . despite the target psf not appearing especially radially - isotropic in the images ( _ e.g. _ , some frames were out of focus ; fig . [ fig : images ] ) , tphot did not exclude asymmetric flux within the circular aperture annuli by computing the median . tphot s aperture photometry gave a @xmath69 @xmath87 for all data and @xmath69 @xmath88 if we exclude the data point at @xmath85 , despite a relatively small @xmath28 , meaning the magnitudes and uncertainties are a good representation of the model . we also note that the low @xmath30 implies that the tphot aperture photometry technique incurs little systematic bias compared to all algorithms explored ( except daophot , which has a similarly low mean @xmath30 ) . for these reasons , we have no reason to suspect any errors in the tphot aperture photometry routine . the tphot algorithm produces the best photometry results for our faint moving target , improving rms very minimally over daophot , and by a factor of @xmath89 over sextractor , @xmath90 over apt , @xmath91 over dophot , and @xmath92 over iraf s phot . tphot s success is likely thanks to its careful treatment of the centroid , exclusion of contamination within the photometric aperture , and use of an iteratively - refined waussian function . an updated version of tphot is currently under construction to better handle significantly trailed objects . daophot produced the next - best results , giving the same accuracy as tphot s aperture photometry algorithm ( largely due to careful accounting of source flux ) but overestimated errors . while offering a distinct advantage in speed , sextractor generally underestimates magnitude uncertainties , likely because of exclusion of error in the background determination . the mag_isocor aperture technique renders the most accurate and precise magnitude measurements amongst the sextractor options because as an aperture correction technique , it more carefully isolates target flux and excludes contaminants . iraf s phot algorithm can only accurately measure magnitudes of a moving target if an aperture correction technique is used ( without sky subtraction ) , though it returns median uncertainties @xmath93 times larger than what could be achieved with tphot . although the background was measured correctly , phot did not exclude background contaminants from within the aperture , giving inaccurate photometry . the apt algorithm , while fairly flexible and excellent as a visualization and educational tool , was outperformed by all other software except dophot and a few phot algorithms . apt struggles with both accurately measuring the background and with systematically excluding contaminants from within the photometric aperture . we suggest that apt is better suited for photometry on brighter objects . dophot performed slightly worse in both accuracy and precision compared to apt . we deduce that its fitting routine was not able to as accurately produce a psf model possibly because it included too many field sources in its construction of a model psf and/or it was very sensitive to trailed psfs . at some point , trailing will be significant enough for the psf peak to be two - dimensional ( _ i.e. _ , a line rather than a point ) . in these cases , the ideal aperture would be the same shape of the target psf to exclude as much background contamination as possible . techniques that only allow a circular aperture such as tphot , sextractor s mag_aper , and phot will incorporate large amounts of background contamination with an aperture that encompasses the extent of the trailed psf . the elliptical aperture allowed by apt and sextractor s mag_auto and mag_best may be a good approximation to the target shape if trailing is minimal ( _ e.g. _ , our data ) , but at some point , trailing may be significant enough for an ellipse s axes to extend well beyond the psf . the only aperture photometry technique that we tested that places no restrictions on target shape are sextractor s mag_isocor and mag_iso , which we consequently predict will be most suitable of the techniques tested here for aperture photometry of significantly trailed objects . these findings could be more statistically robust if we had more images of 1996 to@xmath0 , which would eliminate the dependency on a relatively small sample size of n @xmath94 . as such , our results are only applicable to data sets similar to ours for slow - moving objects with s / n @xmath13 ( according to uncertainties given by our most precise algorithm ) , mostly because no other images for a moving object with a well - determined light curve described by a shape model were available to us at the time . it would be useful to test these algorithms on other objects with ( 1 ) brighter targets to determine at what point aperture photometry starts to outperform psf - fitting because of more well - defined intricacies in the psf morphology that are not reproduced by analytic functions , and ( 2 ) a range of trailing aspects to determine the aspect beyond which photometric accuracy becomes significantly compromised . @xcite tested the latter for a 2-dimensional symmetric gaussian function and a square aperture 3@xmath95fwhm , finding a @xmath88 and 0.5 magnitude loss , respectively , at a trailing length of 3@xmath95fwhm , but this experiment should be repeated for popular algorithms . psf - fitting routines suffer similar challenges to accurately representing the target psf as aperture photometry . techniques that fit a radially - isotropic function to the data will fail in recovering the true psf shape , but no profile - fitting technique that we tested makes this assumption of radial isotropy . allowing for an elliptical analytic function is a first - order approximation to representing the true psf shape of minimally - trailed data , and all of the psf - fitting techniques we tested have this feature . however , rather than a single function calculated at the centroid , a trailed psf might be best represented by the sum of a series of analytic functions calculated at different increments along the direction of motion , such as a trailed gaussian given in its analytic form by @xcite . however , this technique has not yet been adapted into software available to the public . existing psf - fitting algorithms can not calculate a series of different functions for a single trailed source , so we suspect that in the event that an ellipse is no longer a good approximation to the psf shape and until trail - fitting algorithms are made available , all psf - fitting packages will fail at accurately representing the psf . therefore , of all the techniques tested here , sextractor s mag_iso and mag_isocor are likely the only ones capable of performing accurate photometry on significantly trailed psfs . this material is based upon work supported by the national aeronautics and space administration by nasa grant nos . nag5 - 4495 , nnx07a044 g , nnx07af79 g , and from the national science foundation through grant ast-1010059 . partial support for this work was provided by national science foundation grant ast-1009749 . initial image processing in this paper has been performed using the iraf software . iraf is distributed by the national optical astronomy observatories , which is operated by the association of universities for research in astronomy , inc . ( aura ) under cooperative agreement with the national science foundation . , c. ( 2000 ) . . in l. szabados and d. kurtz , editors , _ iau colloq . 176 : the impact of large - scale surveys on pulsating star research _ , volume 203 of _ astronomical society of the pacific conference series _ , pages 5055 . , . , lupton , r. h. , schlegel , d. , boroski , b. , adelman - mccarthy , j. , yanny , b. , kent , s. , stoughton , c. , finkbeiner , d. , padmanabhan , n. , rockosi , c. m. , gunn , j. e. , knapp , g. r. , strauss , m. a. , richards , g. t. , eisenstein , d. , nicinski , t. , kleinman , s. j. , krzesinski , j. , newman , p. r. , snedden , s. , thakar , a. r. , szalay , a. , munn , j. a. , smith , j. a. , tucker , d. , and lee , b. c. ( 2004 ) . , * 325 * , 583589 . , r. r. , rebull , l. m. , gorjian , v. , masci , f. j. , fowler , j. w. , grillmair , c. , surace , j. , mattingly , s. , jackson , e. , hacopeans , e. , hamam , n. , groom , s. , teplitz , h. , mi , w. , helou , g. , van eyken , j. c. , law , n. m. , dekany , r. g. , rahmer , g. , hale , d. , smith , r. , quimby , r. m. , ofek , e. o. , kasliwal , m. m. , zolkower , j. , velur , v. , walters , r. , henning , j. , bui , k. , mckenna , d. , and kulkarni , s. r. ( 2012b ) . , * 124 * , 764781 . , r. h. , ivezic , z. , gunn , j. e. , knapp , g. , strauss , m. a. , and yasuda , n. ( 2002 ) . . in j. a. tyson and s. wolff , editors , _ society of photo - optical instrumentation engineers ( spie ) conference series _ , volume 4836 of _ society of photo - optical instrumentation engineers ( spie ) conference series _ , pages 350356 . , m. and schechter , p. l. ( 1989 ) . in p. j. grosbl , f. murtagh , and r. h. warmels , editors , _ eso / st - ecf data analysis workshop _ , volume 31 of _ european southern observatory conference and workshop proceedings _ , pages 6983 . , p. b. ( 1992 ) . . in d. m. worrall , c. biemesderfer , and j. barnes , editors , _ astronomical data analysis software and systems i _ , volume 25 of _ astronomical society of the pacific conference series _ , page 297 . , d. ( 1986 ) . . in d. l. crawford , editor , _ society of photo - optical instrumentation engineers ( spie ) conference series _ , volume 627 of _ society of photo - optical instrumentation engineers ( spie ) conference series _ , page 733 . ) is given as @xmath28 in the upper right corner , along with the target s magnitude , range of magnitudes of the comparison stars , number of comparison stars within this range , and computed systematic error ( @xmath30 ) . the systematic error calculation is described in more detail in section [ sec : processing ] . the differential field star magnitudes are well fit by a gaussian , rendering reasonable systematic error computations . [ fig : syserr],width=624 ] and [ sec : photisis ] ) . the moving target has a dashed circle placed around it in each panel . the size and orientation is the same for all panels . a ) the original frame before sky subtraction . b ) the residuals of composite subtraction performed with iraf s imarith task , which subtracted a median - combined reference sky image from individual frames . the negative spots are residuals left from field stars after subtraction and are likely caused by inadequate alignment and/or composite scaling before subtraction . c ) sky subtraction residuals using isis software , which convolves a `` best - seeing '' sky composite with a spatially - variable kernel representative of the individual frame s fwhm in order to match the image quality before subtraction . isis subtraction shows a noticeable improvement over imarith subtraction in fully removing background sources . [ fig : skysub],width=796 ] llllll & panel & & & mean & mean + technique & in fig . [ fig : summary ] & rms & @xmath69 & @xmath28 & @xmath30 + tphot , fit photometry & a & 0.030 & 1.436 & 0.031 & 0.012 + tphot , ap . photometry & b & 0.031 & 1.488 & 0.032 & 0.008 + daophot & c & 0.031 & 0.561 & 0.044 & 0.010 + sextractor , mag_isocor & d & 0.058 & 2.501 & 0.045 & 0.027 + phot , ap . corr . & e & 0.060 & 1.855 & 0.049 & 0.021 + phot , ap . , imarith subtraction & f & 0.062 & 3.296 & 0.039 & 0.018 + sextractor , mag_auto & g & 0.063 & 2.985 & 0.047 & 0.023 + sextractor , mag_iso & h & 0.064 & 5.062 & 0.046 & 0.032 + sextractor , mag_best & i & 0.064 & 3.008 & 0.050 & 0.028 + sextractor , mag_aper & j & 0.068 & 1.342 & 0.101 & 0.066 + apt & k & 0.125 & 1.858 & 0.085 & 0.026 + dophot & l & 0.141 & 2.145 & 0.090 & 0.082 + phot , off - center sky annulus & m & 0.159 & 2.413 & 0.085 & 0.048 + phot , centered sky annulus & n & 0.161 & 2.655 & 0.089 & 0.048 + phot , ap . , isis subtraction & o & 0.255 & 14.840 & 0.048 & 0.025 +

[ sec : abstract ] this paper introduces the notion of second minimal-periodic orbit of the continuous map on the interval according as if is a successor of the minimal period of the map in sharkovski ordering . we pursue classification of second minimal-orbits in terms of cyclic permutations and digraphs . it is proved that there are 9 types of second minimal orbits with accuracy up to inverses . the result is applied to the problem on the distribution of periodic windows within the chaotic regime of the bifurcation diagram of the one - parameter family of unimodal maps . it is revealed that by fixing the maximum number of appearances of the periodic windows there is a universal pattern of distribution . in particular , the first appearance of all the orbits is always a minimal orbit , while the second appearance is a second minimal orbit . it is observed that the second appearance of 7-orbit is a second minimal 7-orbit with type 1 digraph . the reason for the relevance of the type 1 second minimal orbit is the fact that the topological structure of the unimodal map with single maximum is equivalent to the structure of the type 1 piecewise monotonic endomorphism associated with the second minimal 7-orbit . yet another important report of this paper is the revelation of the universal pattern dynamics with respect to increased number of appearances . abdulla/ ) . ]

1610.00814

let @xmath2 be a continuous endomorphism , and @xmath3 be a non - degenerate interval on the real line . let @xmath4 be an @xmath0th iteration of @xmath5 . a point @xmath6 is called a periodic point of @xmath5 with period @xmath7 if @xmath8 , @xmath9 for @xmath10 . the set of @xmath7 distinct points @xmath11 is called the orbit of @xmath12 , or briefly @xmath7-orbit or periodic @xmath7-cycle . in his celebrated paper @xcite , sharkovski discovered a law on the coexistence of periodic orbits of continuous endomorphisms on the real line . [ thm : sharkovskii ] @xcite let the positive integers be totally ordered in the following way : @xmath13 if a continuous endomorphism , @xmath14 , has a cycle of period @xmath0 and @xmath15 , then @xmath5 also has a periodic orbit of period @xmath7 . this result played a fundamental role in the development of the theory of discrete dynamical systems . a conceptually novel proof was given in @xcite . following the standard approach , we characterize each periodic orbit with cyclic permutations and directed graphs of transitions or _ digraphs_. consider @xmath7-orbit : @xmath16 if @xmath17 for @xmath18 , with @xmath19 , then @xmath20 is associated with cyclic permutation @xmath21 let @xmath22 be the order reversing permutation @xmath23 then , given a cyclic permutation @xmath24 , it s inverse is obtained as @xmath25 . in the sequel @xmath26 means either @xmath27 $ ] or @xmath28 $ ] . let @xmath29 $ ] . the digraph of @xmath7-orbit is a directed graph of transitions with vertices @xmath30 and oriented edges @xmath31 if @xmath32 . the inverse digraph of @xmath7-orbit is a digraph associated with inverse cyclic permutation @xmath33 . equivalently , inverse digraph is obtained from the digraph of @xmath7-orbit by replacing each @xmath34 with @xmath35 . proof of the sharkovskii s theorem significantly uses the concept of _ minimal orbit_. @xmath0-orbit of @xmath5 is called minimal if @xmath0 is the minimal period of @xmath5 in sharkovski s ordering . digraph of the @xmath7-orbit contains the red edge @xmath36 if @xmath37 . the structure of the minimal orbits is well understood @xcite . minimal odd orbits are called stefan orbits , due to the following characterization : [ thm : stefan ] @xcite the digraph of a @xmath38 minimal odd orbit has the unique structure given in figure [ fig : minodddigraph ] up to an inverse . similar characterization of @xmath39-orbits ( @xmath40 ) is given in @xcite . [ thm : abdulla13 ] @xcite the digraph of a minimal @xmath39-orbit ( @xmath40 ) has one of four types up to their inverses ( type i is shown in figure [ fig : min2timesodd ] ) . the main idea of the constructive proof of @xcite is based on the fact that each half of the minimal @xmath39-orbit is minimal @xmath41 orbit of @xmath42 . therefore , the digraph of the minimal @xmath39-orbit is designed as one of the possible four `` unions '' of two stefan digraphs of @xmath42 . the result of theorem [ thm : abdulla13 ] can be generalized as follows : [ thm : odd2n ] the digraph of any minimal @xmath43-orbit , @xmath40 , has one of @xmath44 types up to their inverses . furthermore , each digraph is strongly simple and can be constructed from the digraphs of two minimal and strongly simple @xmath45-orbits in @xmath42 . the main goal of this paper is the characterization of second minimal odd orbits . [ thm : secondmin ] an @xmath0-orbit , @xmath46 , of @xmath5 is called second minimal if @xmath0 is the successor of the minimal orbit of @xmath5 in the sharkovskii ordering . for example , if map has a second minimal @xmath1-orbit , then it has a minimal @xmath47-orbit , but no @xmath48-orbit . our main result reads : [ thm : secondmin7 ] the second minimal @xmath1-orbit has one of @xmath49 possible types up to their inverses . the associated cyclic permutations are listed in table [ tab : allsecmin7 ] ; digraphs and piecewise linear representatives are demonstrated in appendix 1 . .all second minimal @xmath1 cycles [ cols="^,^,^ " , ] the following few pages contain parameter values that we used to construct the tables and figures in this document . below is a key outlining the table headers . * * parameter * : the numeric value of the parameter for the associated map * * p * : the super stable periodic orbit corresponding to the above parameter value * * a * : the appearance number of the periodic orbit corresponding to the parameter value p0.30p0.05p0.05 0.5 & 1.0 & 1.0 + 0.8090169943749475 & 2.0 & 1.0 + 0.8746404248319252 & 4.0 & 1.0 + 0.8886602156922059 & 8.0 & 1.0 + 0.8916668449640671 & 16.0 & 1.0 + 0.8925435410844437 & 176.0 & 1.0 + 0.8925462153481506 & 144.0 & 1.0 + 0.8925493727315035 & 176.0 & 2.0 + 0.8925536426542862 & 112.0 & 1.0 + 0.8925582633384074 & 176.0 & 3.0 + 0.8925618374268403 & 144.0 & 2.0 + 0.8925652869739683 & 176.0 & 4.0 + 0.892573585 & 80.0 & 1.0 + 0.892582971578052 & 176.0 & 5.0 + 0.8925860303078793 & 144.0 & 3.0 + 0.8925889559052782 & 176.0 & 6.0 + 0.8925934776392263 & 112.0 & 2.0 + 0.8925977476162251 & 176.0 & 7.0 + 0.8926000233456378 & 144.0 & 4.0 + 0.8926020087836917 & 176.0 & 8.0 + 0.8926169497695979 & 176.0 & 9.0 + 0.8926273 & 48.0 & 1.0 + 0.8926273095473648 & 144.0 & 5.0 + 0.8926424478626298 & 144.0 & 6.0 + 0.8926468116175647 & 176.0 & 10.0 + 0.8926516782349787 & 176.0 & 11.0 + 0.8926574315669168 & 144.0 & 7.0 + 0.8926586161442951 & 176.0 & 12.0 + 0.8926614251928092 & 112.0 & 3.0 + 0.8926652649929492 & 144.0 & 8.0 + 0.8926842268307457 & 112.0 & 4.0 + 0.8926913356572714 & 144.0 & 9.0 + 0.892695290032331 & 144.0 & 10.0 + 0.8926999134097414 & 144.0 & 11.0 + 0.8927022801053289 & 112.0 & 5.0 + 0.8927050057536571 & 144.0 & 12.0 + 0.8927172512571973 & 112.0 & 6.0 + 0.8927293163954468 & 112.0 & 7.0 + 0.8927391046311725 & 112.0 & 8.0 + 0.892745143089132 & 112.0 & 9.0 + 0.8927531275154216 & 88.0 & 1.0 + 0.892765612708715 & 72.0 & 1.0 + 0.8927803546566632 & 88.0 & 2.0 + 0.8928002912804274 & 56.0 & 1.0 + 0.8928218666978042 & 88.0 & 3.0 + 0.8928385565717427 & 72.0 & 2.0 + 0.8928546647294011 & 88.0 & 4.0 + 0.89286570826588 & 112.0 & 10.0 + 0.8928758764598548 & 112.0 & 11.0 + 0.892893412 & 40.0 & 1.0 + 0.8929185769659241 & 112.0 & 12.0 + 0.8929372464436822 & 88.0 & 5.0 + 0.892951531075995 & 72.0 & 3.0 + 0.892965194521167 & 88.0 & 6.0 + 0.8929863118792093 & 56.0 & 2.0 + 0.8930062532268376 & 88.0 & 7.0 + 0.8930168810677759 & 72.0 & 4.0 + 0.8930261525483736 & 88.0 & 8.0 + 0.8930959299227967 & 88.0 & 9.0 + 0.893144323 & 24.0 & 1.0 + 0.8932150514348008 & 72.0 & 5.0 + 0.8932354347086963 & 88.0 & 10.0 + 0.8932581641927477 & 88.0 & 11.0 + 0.8932657490534088 & 88.0 & 12.0 + 0.8932850347611261 & 72.0 & 6.0 + 0.8933036890878607 & 56.0 & 3.0 + 0.8933216260408074 & 72.0 & 7.0 + 0.8933978510203553 & 72.0 & 8.0 + 0.8934102079449422 & 56.0 & 4.0 + 0.8934213197439284 & 72.0 & 9.0 + 0.8934434098714304 & 72.0 & 10.0 + 0.8934618867418858 & 72.0 & 11.0 + 0.8934834850666612 & 72.0 & 12.0 + 0.8934945442626503 & 56.0 & 5.0 + 0.8935645266299209 & 56.0 & 6.0 + 0.8936209165390675 & 56.0 & 7.0 + 0.8936666925819085 & 56.0 & 8.0 + 0.893694946890496 & 56.0 & 9.0 + 0.8937323239550504 & 44.0 & 1.0 + 0.8937906969816722 & 36.0 & 1.0 + 0.8938595743690791 & 44.0 & 2.0 + 0.8939527169808105 & 28.0 & 1.0 + 0.8939585734139139 & 56.0 & 10.0 + 0.8940534775173637 & 44.0 & 3.0 + 0.8941313720345442 & 36.0 & 2.0 + 0.8942065586523593 & 44.0 & 4.0 + 0.894258151388425 & 56.0 & 11.0 + 0.894387542 & 20.0 & 1.0 + 0.8945920053348253 & 44.0 & 5.0 + 0.8946586324286321 & 36.0 & 3.0 + 0.8947223418891009 & 44.0 & 6.0 + 0.894820829737213 & 28.0 & 2.0 + 0.8949138449982476 & 44.0 & 7.0 + 0.8949634233198132 & 36.0 & 4.0 + 0.8950067008708713 & 44.0 & 8.0 + 0.8953321587557187 & 44.0 & 9.0 + 0.8955574589550901 & 12.0 & 1.0 + 0.8955574589550902 & 36.0 & 5.0 + 0.8958862652026675 & 36.0 & 6.0 + 0.8959812147446011 & 44.0 & 10.0 + 0.8960872008248462 & 44.0 & 11.0 + 0.8961224931207159 & 44.0 & 12.0 + 0.896187724634684 & 44.0 & 13.0 + 0.8962125114340377 & 36.0 & 7.0 + 0.8962994172610155 & 28.0 & 3.0 + 0.8963829388112704 & 36.0 & 8.0 + 0.8967376720178901 & 36.0 & 9.0 + 0.8967952385115885 & 28.0 & 4.0 + 0.8968470365678113 & 36.0 & 10.0 + 0.8969500795862201 & 36.0 & 11.0 + 0.8970359452655111 & 36.0 & 12.0 + 0.8971364940346715 & 36.0 & 13.0 + 0.8971878607732342 & 28.0 & 5.0 + 0.897349960937811 & 16.0 & 2.0 + 0.8975119706948392 & 28.0 & 6.0 + 0.897773472481103 & 28.0 & 7.0 + 0.8979846837132985 & 28.0 & 8.0 + 0.8981144866729309 & 28.0 & 9.0 + 0.898285536828268 & 22.0 & 1.0 + 0.8985555274564071 & 18.0 & 1.0 + 0.8988759584501568 & 22.0 & 2.0 + 0.8993095495931392 & 14.0 & 1.0 + 0.8993368709151569 & 28.0 & 10.0 + 0.8997801682018646 & 22.0 & 3.0 + 0.900145977896958 & 18.0 & 2.0 + 0.9004988403933514 & 22.0 & 4.0 + 0.900739199236906 & 28.0 & 11.0 + 0.9009615610934926 & 28.0 & 12.0 + 0.9013464593838434 & 10.0 & 1.0 + 0.9023089359519223 & 22.0 & 5.0 + 0.9026237070977975 & 18.0 & 3.0 + 0.9029255119168045 & 22.0 & 6.0 + 0.9033912545246447 & 14.0 & 2.0 + 0.9038307061198744 & 22.0 & 7.0 + 0.9040647384776517 & 18.0 & 4.0 + 0.904267999021263 & 22.0 & 8.0 + 0.904899707989313 & 16.0 & 3.0 + 0.9058068890614333 & 22.0 & 9.0 + 0.9068893823788807 & 6.0 & 1.0 + 0.907823200062555 & 12.0 & 2.0 + 0.9084892763171067 & 18.0 & 5.0 + 0.908943552643572 & 22.0 & 10.0 + 0.909446366503746 & 22.0 & 11.0 + 0.9095331442064777 & 16.0 & 4.0 + 0.909616877636151 & 22.0 & 12.0 + 0.9099232980258403 & 22.0 & 13.0 + 0.9100404763587777 & 18.0 & 6.0 + 0.9104560684602736 & 14.0 & 3.0 + 0.9108571928810814 & 18.0 & 7.0 + 0.9113101549515245 & 16.0 & 5.0 + 0.9117622003804752 & 10.0 & 2.0 + 0.9122196907504082 & 16.0 & 6.0 + 0.9125698915821074 & 18.0 & 8.0 + 0.9128454558604397 & 14.0 & 4.0 + 0.9130921658581117 & 18.0 & 9.0 + 0.9133219453957768 & 16.0 & 7.0 + 0.913580229382562 & 18.0 & 10.0 + 0.9137869897027185 & 12.0 & 3.0 + 0.9140000120689893 & 18.0 & 11.0 + 0.9142572545642417 & 16.0 & 8.0 + 0.9144843071853473 & 18.0 & 12.0 + 0.9147365695291755 & 14.0 & 5.0 + 0.9150310389753986 & 18.0 & 13.0 + 0.9155481259216439 & 8.0 & 2.0 + 0.9156307240370415 & 16.0 & 9.0 + 0.9163686307982645 & 14.0 & 6.0 + 0.9167444770891021 & 16.0 & 10.0 + 0.9170937820054763 & 12.0 & 4.0 + 0.9174080707508858 & 16.0 & 11.0 + 0.9176716828336033 & 14.0 & 7.0 + 0.9179438449410644 & 16.0 & 12.0 + 0.918252061590967 & 10.0 & 3.0 + 0.9185479687517751 & 16.0 & 13.0 + 0.9187709570364896 & 14.0 & 8.0 + 0.9189599787371455 & 16.0 & 14.0 + 0.9191588201786222 & 12.0 & 5.0 + 0.9193383506312537 & 16.0 & 15.0 + 0.9194727761973615 & 14.0 & 9.0 + 0.9195830311115901 & 16.0 & 16.0 + 0.9196435795904915 & 27.0 & 1.0 + 0.9196439467055643 & 25.0 & 1.0 + 0.919644981018353 & 23.0 & 1.0 + 0.9196457756192704 & 31.0 & 1.0 + 0.9196459733265683 & 33.0 & 1.0 + 0.919646584854938 & 29.0 & 1.0 + 0.9196478947752951 & 21.0 & 1.0 + 0.9196561008171046 & 19.0 & 1.0 + 0.9196791956839688 & 17.0 & 1.0 + 0.9197000311872442 & 33.0 & 2.0 + 0.9197006174971487 & 29.0 & 2.0 + 0.9197013110760547 & 27.0 & 2.0 + 0.9197027892812369 & 25.0 & 2.0 + 0.9197070458081346 & 19.0 & 2.0 + 0.9197149184463773 & 23.0 & 2.0 + 0.919720432954781 & 31.0 & 2.0 + 0.9197223223802037 & 21.0 & 2.0 + 0.9197440854758494 & 15.0 & 1.0 + 0.9198001163269032 & 29.0 & 3.0 + 0.9198006805909831 & 31.0 & 3.0 + 0.9198015834347826 & 21.0 & 3.0 + 0.9198060374454656 & 25.0 & 3.0 + 0.9198101479894252 & 23.0 & 3.0 + 0.9198106458924652 & 33.0 & 3.0 + 0.9198220811388089 & 17.0 & 2.0 + 0.9198265854125903 & 27.0 & 3.0 + 0.919864618020104 & 19.0 & 3.0 + 0.9199002311171905 & 29.0 & 4.0 + 0.9199010748389 & 31.0 & 4.0 + 0.919902361355556 & 23.0 & 4.0 + 0.9199062616676469 & 27.0 & 4.0 + 0.919909788780055 & 25.0 & 4.0 + 0.9199257014200635 & 13.0 & 1.0 + 0.9199604232996532 & 33.0 & 4.0 + 0.9199641405971246 & 21.0 & 4.0 + 0.9199901616901687 & 19.0 & 4.0 + 0.9200000703977084 & 27.0 & 5.0 + 0.9200028110917958 & 23.0 & 5.0 + 0.9200076215315786 & 25.0 & 5.0 + 0.9200103594866011 & 33.0 & 5.0 + 0.9200106468302367 & 31.0 & 5.0 + 0.9200112670767221 & 29.0 & 5.0 + 0.9200132410614315 & 21.0 & 5.0 + 0.9200389918663741 & 17.0 & 3.0 + 0.9200856001137521 & 19.0 & 5.0 + 0.9201002143603496 & 33.0 & 6.0 + 0.9201016145225464 & 29.0 & 6.0 + 0.9201028782109886 & 25.0 & 6.0 + 0.9201090481341733 & 21.0 & 6.0 + 0.9201105233194542 & 31.0 & 6.0 + 0.9201219025506991 & 23.0 & 6.0 + 0.9201255900181393 & 27.0 & 6.0 + 0.9201422993862319 & 15.0 & 2.0 + 0.9202001209047099 & 33.0 & 7.0 + 0.9202005092526497 & 31.0 & 7.0 + 0.9202028502440535 & 19.0 & 6.0 + 0.9202076590267333 & 27.0 & 7.0 + 0.9202106868115649 & 25.0 & 7.0 + 0.9202121549955485 & 29.0 & 7.0 + 0.9202163098045701 & 23.0 & 7.0 + 0.9202278551537337 & 21.0 & 7.0 + 0.9202587007006321 & 17.0 & 4.0 + 0.9203000082728384 & 33.0 & 8.0 + 0.9203010086311395 & 29.0 & 8.0 + 0.9203057998429245 & 23.0 & 8.0 + 0.9203127589957107 & 25.0 & 8.0 + 0.9203165520944553 & 27.0 & 8.0 + 0.9203227908438997 & 19.0 & 7.0 + 0.920342369174139 & 31.0 & 8.0 + 0.9203571728536436 & 21.0 & 8.0 + 0.9204296685342831 & 11.0 & 1.0 + 0.9204767917194359 & 31.0 & 9.0 + 0.9204782548259948 & 27.0 & 9.0 + 0.9204811789142139 & 25.0 & 9.0 + 0.9204886765851206 & 23.0 & 9.0 + 0.920491177506046 & 29.0 & 9.0 + 0.9205005434280713 & 27.0 & 10.0 + 0.9205016326349753 & 31.0 & 10.0 + 0.9205060640270746 & 21.0 & 9.0 + 0.9205100093913368 & 33.0 & 9.0 + 0.9205152878078745 & 25.0 & 10.0 + 0.9205204912373594 & 29.0 & 10.0 + 0.9205229230057148 & 23.0 & 10.0 + 0.9205419557953939 & 19.0 & 8.0 + 0.9206005784482749 & 31.0 & 11.0 + 0.9206016043882802 & 27.0 & 11.0 + 0.9206035067349837 & 29.0 & 11.0 + 0.9206106074930887 & 17.0 & 5.0 + 0.9206233780780297 & 25.0 & 11.0 + 0.9206307692696445 & 23.0 & 11.0 + 0.9206445592873718 & 21.0 & 10.0 + 0.9206503319730334 & 33.0 & 10.0 + 0.9206722630021638 & 19.0 & 9.0 + 0.920701250060699 & 21.0 & 11.0 + 0.9207028636238962 & 33.0 & 11.0 + 0.9207058521120631 & 27.0 & 12.0 + 0.9207091690061839 & 25.0 & 12.0 + 0.9207108947869364 & 29.0 & 12.0 + 0.9207162655093767 & 23.0 & 12.0 + 0.9207419764574405 & 15.0 & 3.0 + 0.9207611508956879 & 31.0 & 12.0 + 0.9208003677980954 & 31.0 & 13.0 + 0.9208013027396837 & 25.0 & 13.0 + 0.9208044399266484 & 27.0 & 13.0 + 0.9208098673955014 & 19.0 & 10.0 + 0.9208136718855724 & 29.0 & 13.0 + 0.9208200470149717 & 33.0 & 12.0 + 0.9208239900768528 & 23.0 & 13.0 + 0.9208357735776028 & 21.0 & 12.0 + 0.9208669721008041 & 17.0 & 6.0 + 0.9209010909575819 & 33.0 & 13.0 + 0.9209014409363525 & 31.0 & 14.0 + 0.9209022018802908 & 29.0 & 14.0 + 0.9209065639725457 & 25.0 & 14.0 + 0.9209125024919976 & 23.0 & 14.0 + 0.9209159028698781 & 27.0 & 14.0 + 0.9209285533727417 & 19.0 & 11.0 + 0.9209609162728946 & 21.0 & 13.0 + 0.9210201003306971 & 13.0 & 2.0 + 0.9210523200855003 & 29.0 & 15.0 + 0.9210565730461244 & 25.0 & 15.0 + 0.921060891466536 & 31.0 & 15.0 + 0.9210642268502021 & 23.0 & 15.0 + 0.9210684773689786 & 27.0 & 15.0 + 0.9210811677775855 & 21.0 & 14.0 + 0.9210906606157544 & 33.0 & 14.0 + 0.9211005911749909 & 33.0 & 15.0 + 0.9211045739676469 & 25.0 & 16.0 + 0.9211102289158498 & 29.0 & 16.0 + 0.9211152051517735 & 19.0 & 12.0 + 0.9211323778639362 & 23.0 & 16.0 + 0.9211358613331826 & 27.0 & 16.0 + 0.9211463306470374 & 21.0 & 15.0 + 0.9211500728400285 & 31.0 & 16.0 + 0.9211820874871725 & 17.0 & 7.0 + 0.9212003053975469 & 33.0 & 16.0 + 0.9212007128291876 & 29.0 & 17.0 + 0.9212015158659378 & 31.0 & 17.0 + 0.9212029872773682 & 23.0 & 17.0 + 0.9212066360711628 & 27.0 & 17.0 + 0.9212095811496441 & 25.0 & 17.0 + 0.9212170738871973 & 21.0 & 16.0 + 0.9212463984436541 & 19.0 & 13.0 + 0.921300240890266 & 33.0 & 17.0 + 0.9213006615814194 & 29.0 & 18.0 + 0.9213014802403597 & 31.0 & 18.0 + 0.9213026978247993 & 25.0 & 18.0 + 0.92130631826283 & 27.0 & 18.0 + 0.9213262606625509 & 15.0 & 4.0 + 0.9213588721853865 & 23.0 & 18.0 + 0.9213762639622086 & 21.0 & 17.0 + 0.9214035603610099 & 27.0 & 19.0 + 0.9214107783908108 & 19.0 & 14.0 + 0.9214153297683447 & 31.0 & 19.0 + 0.9214200902652883 & 33.0 & 18.0 + 0.9214216285362573 & 25.0 & 19.0 + 0.9214288350417134 & 23.0 & 19.0 + 0.9214310694267652 & 29.0 & 19.0 + 0.9214438552596709 & 21.0 & 18.0 + 0.9214845277144422 & 17.0 & 8.0 + 0.9215001291474537 & 33.0 & 19.0 + 0.9215010171790381 & 25.0 & 20.0 + 0.9215049782513212 & 27.0 & 20.0 + 0.9215095294630585 & 23.0 & 20.0 + 0.9215112716081109 & 31.0 & 20.0 + 0.9215121500634174 & 29.0 & 20.0 + 0.9215262948579038 & 21.0 & 19.0 + 0.9215608336278994 & 19.0 & 15.0 + 0.9216002756249053 & 33.0 & 20.0 + 0.9216008621175658 & 25.0 & 21.0 + 0.9216017202012348 & 31.0 & 21.0 + 0.921606307688425 & 23.0 & 21.0 + 0.9216126226488097 & 27.0 & 21.0 + 0.921753409712203 & 31.0 & 22.0 + 0.9217577448964835 & 29.0 & 21.0 + 0.9217611380044505 & 33.0 & 21.0 + 0.921804045233538 & 9.0 & 1.0 + 0.9218590865036999 & 33.0 & 22.0 + 0.9218641957928685 & 31.0 & 23.0 + 0.9218709393905855 & 29.0 & 22.0 + 0.9219718987947105 & 33.0 & 23.0 + 0.9220111472734216 & 33.0 & 24.0 + 0.9220160672052569 & 29.0 & 23.0 + 0.9220166608744411 & 27.0 & 22.0 + 0.9220183435162941 & 25.0 & 22.0 + 0.9220208907672919 & 31.0 & 24.0 + 0.9220230674037606 & 23.0 & 22.0 + 0.9220360043243131 & 21.0 & 20.0 + 0.9220694892396009 & 19.0 & 16.0 + 0.9221006939816137 & 33.0 & 25.0 + 0.9221048175786382 & 21.0 & 21.0 + 0.9221120027015463 & 29.0 & 24.0 + 0.9221139584364046 & 25.0 & 23.0 + 0.9221218006952228 & 23.0 & 23.0 + 0.922126355265872 & 27.0 & 23.0 + 0.9221480960567324 & 17.0 & 9.0 + 0.922159726448752 & 31.0 & 25.0 + 0.9222002247854527 & 33.0 & 26.0 + 0.9222005785028199 & 29.0 & 25.0 + 0.9222011980332242 & 31.0 & 26.0 + 0.9222057411189049 & 23.0 & 24.0 + 0.9222129968030404 & 25.0 & 24.0 + 0.9222166052355004 & 27.0 & 24.0 + 0.9222244298335882 & 19.0 & 17.0 + 0.9222597382550419 & 21.0 & 22.0 + 0.9223138550089415 & 15.0 & 5.0 + 0.9223381697320739 & 29.0 & 26.0 + 0.9223393802222187 & 27.0 & 25.0 + 0.922342562851247 & 25.0 & 25.0 + 0.9223501969457514 & 23.0 & 25.0 + 0.9223519582560188 & 31.0 & 27.0 + 0.9223665449604025 & 21.0 & 23.0 + 0.922399214446316 & 19.0 & 18.0 + 0.922406402226228 & 27.0 & 26.0 + 0.9224096209205783 & 25.0 & 26.0 + 0.9224112875747574 & 29.0 & 27.0 + 0.9224161022741247 & 23.0 & 26.0 + 0.9224297139943903 & 21.0 & 24.0 + 0.9224411422767945 & 31.0 & 28.0 + 0.9224669462877152 & 17.0 & 10.0 + 0.9225049540612161 & 21.0 & 25.0 + 0.9225111577683214 & 29.0 & 28.0 + 0.9225127813206834 & 25.0 & 27.0 + 0.9225156274141197 & 27.0 & 27.0 + 0.9225191563794206 & 23.0 & 27.0 + 0.92253694125618 & 19.0 & 19.0 + 0.922542458896398 & 31.0 & 29.0 + 0.9226176193928061 & 31.0 & 30.0 + 0.9226464425273783 & 13.0 & 3.0 + 0.9226985119519096 & 27.0 & 28.0 + 0.9227008178525822 & 25.0 & 28.0 + 0.9227022407363636 & 29.0 & 29.0 + 0.922703450274448 & 27.0 & 29.0 + 0.9227069440391705 & 23.0 & 28.0 + 0.9227218838495027 & 21.0 & 26.0 + 0.92275449593336 & 19.0 & 20.0 + 0.9227598793106112 & 31.0 & 31.0 + 0.9228004001233571 & 31.0 & 32.0 + 0.9228055063846784 & 25.0 & 29.0 + 0.9228208374473782 & 17.0 & 11.0 + 0.9228318593515585 & 29.0 & 30.0 + 0.9228424021108256 & 23.0 & 29.0 + 0.9228457697436099 & 27.0 & 30.0 + 0.9228555496778423 & 21.0 & 27.0 + 0.9228834258451859 & 19.0 & 21.0 + 0.9229008001875197 & 31.0 & 33.0 + 0.9229053068625388 & 25.0 & 30.0 + 0.9229128243166754 & 21.0 & 28.0 + 0.9229176329559665 & 27.0 & 31.0 + 0.9229222384020115 & 29.0 & 31.0 + 0.9229271234309651 & 23.0 & 30.0 + 0.9229611968821689 & 15.0 & 6.0 + 0.9230008998938598 & 29.0 & 32.0 + 0.9230025440384575 & 25.0 & 31.0 + 0.923005640632976 & 27.0 & 32.0 + 0.9230106288246114 & 21.0 & 29.0 + 0.9230134345245601 & 31.0 & 34.0 + 0.9230246352291566 & 23.0 & 31.0 + 0.9230414320248848 & 19.0 & 22.0 + 0.923108777775757 & 17.0 & 12.0 + 0.9231208724973803 & 31.0 & 35.0 + 0.9231213450951792 & 29.0 & 33.0 + 0.9231257848675283 & 25.0 & 32.0 + 0.923132894975511 & 23.0 & 32.0 + 0.9231366730398138 & 27.0 & 33.0 + 0.9231477692758365 & 21.0 & 30.0 + 0.9231797780922061 & 19.0 & 23.0 + 0.9232022555173477 & 29.0 & 34.0 + 0.9232037853402316 & 25.0 & 33.0 + 0.9232113433701622 & 21.0 & 31.0 + 0.9232159122244098 & 27.0 & 34.0 + 0.9233553354446671 & 11.0 & 2.0 + 0.9234364494843212 & 29.0 & 35.0 + 0.9234875471606739 & 25.0 & 34.0 + 0.9234952282874029 & 27.0 & 35.0 + 0.923505076081172 & 21.0 & 32.0 + 0.9235127679201179 & 25.0 & 35.0 + 0.9235155521191672 & 27.0 & 36.0 + 0.9235372401278704 & 19.0 & 24.0 + 0.9236101303997307 & 17.0 & 13.0 + 0.9236248828050347 & 27.0 & 37.0 + 0.9236279066755088 & 25.0 & 36.0 + 0.9236491052579816 & 21.0 & 33.0 + 0.9236796161121292 & 19.0 & 25.0 + 0.9237031728677695 & 25.0 & 37.0 + 0.9237111298598573 & 21.0 & 34.0 + 0.9237162552841238 & 27.0 & 38.0 + 0.9237638529359127 & 15.0 & 7.0 + 0.9238051113914373 & 27.0 & 39.0 + 0.9238079477611446 & 25.0 & 38.0 + 0.9238158785374112 & 21.0 & 35.0 + 0.9238460988672327 & 19.0 & 26.0 + 0.9239112916557798 & 17.0 & 14.0 + 0.9239255388317077 & 27.0 & 40.0 + 0.9239488116172666 & 21.0 & 36.0 + 0.9239796821838245 & 19.0 & 27.0 + 0.9240118642346291 & 21.0 & 37.0 + 0.9240168986630016 & 27.0 & 41.0 + 0.9240988760780329 & 13.0 & 4.0 + 0.9241659057686024 & 27.0 & 42.0 + 0.9241874953823742 & 21.0 & 38.0 + 0.9242056864820108 & 27.0 & 43.0 + 0.9242202029150759 & 19.0 & 28.0 + 0.9242516437364504 & 21.0 & 39.0 + 0.9242901169978515 & 17.0 & 15.0 + 0.9243274730242318 & 21.0 & 40.0 + 0.9243566357281391 & 19.0 & 29.0 + 0.924440039748957 & 15.0 & 8.0 + 0.9244925837164232 & 21.0 & 41.0 + 0.9245216932198678 & 19.0 & 30.0 + 0.9245488951118092 & 21.0 & 42.0 + 0.9245822242032017 & 17.0 & 16.0 + 0.9246362004773152 & 19.0 & 31.0 + 0.9248291259456144 & 16.0 & 17.0 + 0.9254422883844889 & 7.0 & 1.0 + 0.9255708476075072 & 14.0 & 10.0 + 0.9261810653663104 & 16.0 & 18.0 + 0.9263695549335671 & 19.0 & 32.0 + 0.926422205650506 & 17.0 & 17.0 + 0.926480708392207 & 19.0 & 33.0 + 0.9265617938148366 & 15.0 & 9.0 + 0.9266443295176789 & 19.0 & 34.0 + 0.9267072296096274 & 17.0 & 18.0 + 0.9267719927018105 & 19.0 & 35.0 + 0.9268990591234583 & 13.0 & 5.0 + 0.9270246251963171 & 19.0 & 36.0 + 0.9270866574862666 & 17.0 & 19.0 + 0.9271462242414003 & 19.0 & 37.0 + 0.9272250976785409 & 15.0 & 10.0 + 0.92730477980184 & 19.0 & 38.0 + 0.9273649682044417 & 17.0 & 20.0 + 0.9274246529676042 & 19.0 & 39.0 + 0.927643837998337 & 11.0 & 3.0 + 0.9278623135222064 & 19.0 & 40.0 + 0.9279189741782298 & 17.0 & 21.0 + 0.9279752930366489 & 19.0 & 41.0 + 0.9280495878140084 & 15.0 & 11.0 + 0.9281221075383218 & 19.0 & 42.0 + 0.9281752606494361 & 17.0 & 22.0 + 0.9282295925464943 & 19.0 & 43.0 + 0.92834448677078 & 13.0 & 6.0 + 0.928460602487508 & 19.0 & 44.0 + 0.9285154193462041 & 17.0 & 23.0 + 0.9285688582630766 & 19.0 & 45.0 + 0.9286415302414859 & 15.0 & 12.0 + 0.9287139150577448 & 19.0 & 46.0 + 0.928765286070779 & 17.0 & 24.0 + 0.9288127175465508 & 19.0 & 47.0 + 0.9289968814034584 & 16.0 & 19.0 + 0.9292792068508622 & 9.0 & 2.0 + 0.9295822427339708 & 16.0 & 20.0 + 0.9298079717433595 & 17.0 & 25.0 + 0.9299278057808813 & 15.0 & 13.0 + 0.9300482202563531 & 17.0 & 26.0 + 0.9302132331790187 & 13.0 & 7.0 + 0.9303744281142803 & 17.0 & 27.0 + 0.9304878779751862 & 15.0 & 14.0 + 0.9306020957994455 & 17.0 & 28.0 + 0.9308659010883394 & 11.0 & 4.0 + 0.9311266981784078 & 17.0 & 29.0 + 0.931234015921366 & 15.0 & 15.0 + 0.9313370849667829 & 17.0 & 30.0 + 0.9314786851600155 & 13.0 & 8.0 + 0.9316127851041636 & 17.0 & 31.0 + 0.9316990324815639 & 15.0 & 16.0 + 0.931774933086075 & 17.0 & 32.0 + 0.931881041671243 & 16.0 & 21.0 + 0.9320175871084143 & 14.0 & 11.0 + 0.9321798504486302 & 16.0 & 22.0 + 0.9324535237980285 & 12.0 & 6.0 + 0.9327763934958844 & 16.0 & 23.0 + 0.9330334577962366 & 14.0 & 12.0 + 0.9333315218062391 & 16.0 & 24.0 + 0.9336982614806218 & 17.0 & 33.0 + 0.9347287282426712 & 5.0 & 1.0 + 0.9354619793333402 & 10.0 & 4.0 + 0.936004218370594 & 15.0 & 17.0 + 0.9365481090303157 & 17.0 & 34.0 + 0.9368470960303923 & 16.0 & 25.0 + 0.9371045687350086 & 14.0 & 13.0 + 0.9373251521199856 & 16.0 & 26.0 + 0.9375492280855212 & 17.0 & 35.0 + 0.9376188083061392 & 12.0 & 7.0 + 0.9376870171464302 & 17.0 & 36.0 + 0.9378666057635101 & 16.0 & 27.0 + 0.9380019767238563 & 14.0 & 14.0 + 0.938117088017013 & 16.0 & 28.0 + 0.9382016815414446 & 17.0 & 37.0 + 0.9382639401482904 & 15.0 & 18.0 + 0.9383328148965207 & 17.0 & 38.0 + 0.938448694765481 & 13.0 & 9.0 + 0.938569177987284 & 17.0 & 39.0 + 0.9386465789836986 & 15.0 & 19.0 + 0.9387244096568935 & 17.0 & 40.0 + 0.9388708890819253 & 16.0 & 29.0 + 0.9389691251075292 & 11.0 & 5.0 + 0.9390695421382402 & 16.0 & 30.0 + 0.9392151036120271 & 17.0 & 41.0 + 0.9392926094377416 & 15.0 & 20.0 + 0.9393698233226494 & 17.0 & 42.0 + 0.9394928977047089 & 13.0 & 10.0 + 0.9396153225473503 & 17.0 & 43.0 + 0.939690024842712 & 15.0 & 21.0 + 0.9397608915679229 & 17.0 & 44.0 + 0.9398638787402191 & 16.0 & 31.0 + 0.9400376250602329 & 14.0 & 15.0 + 0.9403137986767697 & 9.0 & 3.0 + 0.94059942667533 & 14.0 & 16.0 + 0.9407653946677715 & 16.0 & 32.0 + 0.9408613682291624 & 17.0 & 45.0 + 0.9409270154265458 & 15.0 & 22.0 + 0.9409953129169278 & 17.0 & 46.0 + 0.9411078240980755 & 13.0 & 11.0 + 0.941219364734741 & 17.0 & 47.0 + 0.9412866882425523 & 15.0 & 23.0 + 0.9413532521023238 & 17.0 & 48.0 + 0.9414673057559139 & 16.0 & 33.0 + 0.9415765699409272 & 11.0 & 6.0 + 0.9416867792979688 & 16.0 & 34.0 + 0.9417978781941158 & 17.0 & 49.0 + 0.9418617516317436 & 15.0 & 24.0 + 0.9419252584019139 & 17.0 & 50.0 + 0.9420287698786108 & 13.0 & 12.0 + 0.9421288573690431 & 17.0 & 51.0 + 0.942185810613247 & 15.0 & 25.0 + 0.9422379080085772 & 17.0 & 52.0 + 0.9423072956256685 & 16.0 & 35.0 + 0.942405302536878 & 14.0 & 17.0 + 0.9425207052272616 & 16.0 & 36.0 + 0.9426527498373356 & 17.0 & 53.0 + 0.9427557608876059 & 12.0 & 8.0 + 0.9428743623740117 & 17.0 & 54.0 + 0.9430870153972655 & 16.0 & 37.0 + 0.9435535472252282 & 7.0 & 2.0 + 0.9436340122846377 & 14.0 & 18.0 + 0.9440868779686737 & 16.0 & 38.0 + 0.9442804882592601 & 17.0 & 55.0 + 0.944400441261753 & 12.0 & 9.0 + 0.9445048407326306 & 17.0 & 56.0 + 0.9446182118866014 & 16.0 & 39.0 + 0.9447209484546417 & 14.0 & 19.0 + 0.9448077454709249 & 16.0 & 40.0 + 0.9448674903684716 & 17.0 & 57.0 + 0.9449126279511749 & 15.0 & 26.0 + 0.9449612825801651 & 17.0 & 58.0 + 0.9450488345622481 & 13.0 & 13.0 + 0.9451381264415866 & 17.0 & 59.0 + 0.9451899365694643 & 15.0 & 27.0 + 0.9452412487095767 & 17.0 & 60.0 + 0.9453222087248859 & 16.0 & 41.0 + 0.9454345985777954 & 11.0 & 7.0 + 0.9455471174411741 & 16.0 & 42.0 + 0.9456259644051741 & 17.0 & 61.0 + 0.9456755743657617 & 15.0 & 28.0 + 0.9457253447634651 & 17.0 & 62.0 + 0.9458115547312089 & 13.0 & 14.0 + 0.9458967766826185 & 17.0 & 63.0 + 0.9459443478465995 & 15.0 & 29.0 + 0.9459893788372576 & 17.0 & 64.0 + 0.9460505092541333 & 16.0 & 43.0 + 0.9461476071342457 & 14.0 & 20.0 + 0.9462966096162295 & 16.0 & 44.0 + 0.9464449481328123 & 9.0 & 4.0 + 0.9465988685328639 & 16.0 & 45.0 + 0.9467408391711305 & 14.0 & 21.0 + 0.9468310268519844 & 16.0 & 46.0 + 0.9468864578791083 & 17.0 & 65.0 + 0.9469269490048738 & 15.0 & 30.0 + 0.9469691298705613 & 17.0 & 66.0 + 0.9470443898694233 & 13.0 & 15.0 + 0.9471188724630772 & 17.0 & 67.0 + 0.9471603042130635 & 15.0 & 31.0 + 0.9472008999953558 & 17.0 & 68.0 + 0.9472621657601605 & 16.0 & 47.0 + 0.9473591815013696 & 11.0 & 8.0 + 0.9474541321755662 & 16.0 & 48.0 + 0.9475106957943875 & 17.0 & 69.0 + 0.9475469696260843 & 15.0 & 32.0 + 0.9476448977552586 & 13.0 & 16.0 + 0.9477034426138526 & 17.0 & 70.0 + 0.947733921617124 & 15.0 & 33.0 + 0.9477966857540892 & 16.0 & 49.0 + 0.9478462851458759 & 14.0 & 22.0 + 0.9479006922691185 & 16.0 & 50.0 + 0.9479516033174422 & 17.0 & 71.0 + 0.9480053188877896 & 12.0 & 10.0 + 0.9480607439004328 & 17.0 & 72.0 + 0.9481160741636439 & 16.0 & 51.0 + 0.9481796300838617 & 14.0 & 23.0 + 0.9482435763382607 & 16.0 & 52.0 + 0.9482944689150308 & 17.0 & 73.0 + 0.9483413732621206 & 15.0 & 34.0 + 0.948408642108592 & 17.0 & 74.0 + 0.9485164441889147 & 10.0 & 5.0 + 0.9486285476412364 & 17.0 & 75.0 + 0.9486987975767931 & 15.0 & 35.0 + 0.9487487242954044 & 17.0 & 76.0 + 0.9488036471569236 & 16.0 & 53.0 + 0.9488743597460321 & 14.0 & 24.0 + 0.9489462845704794 & 16.0 & 54.0 + 0.949009332533322 & 17.0 & 77.0 + 0.9490780804877298 & 12.0 & 11.0 + 0.9491473662644617 & 17.0 & 78.0 + 0.9492107277102163 & 16.0 & 55.0 + 0.9492826687713001 & 14.0 & 25.0 + 0.9493514336511595 & 16.0 & 56.0 + 0.9494014907276402 & 17.0 & 79.0 + 0.9494428882786827 & 15.0 & 36.0 + 0.9495891122434322 & 13.0 & 17.0 + 0.9497094552611501 & 17.0 & 80.0 + 0.9497890939253142 & 15.0 & 37.0 + 0.9498857281645797 & 17.0 & 81.0 + 0.9501927359686673 & 8.0 & 3.0 + 0.9502236731934814 & 16.0 & 57.0 + 0.9505347418045125 & 17.0 & 82.0 + 0.9506352651907028 & 15.0 & 38.0 + 0.9507189563549525 & 17.0 & 83.0 + 0.9508487313327707 & 13.0 & 18.0 + 0.9509575188358351 & 17.0 & 84.0 + 0.9510088573843923 & 15.0 & 39.0 + 0.9510544395866831 & 17.0 & 85.0 + 0.9511093923830842 & 16.0 & 58.0 + 0.9511864043035728 & 14.0 & 26.0 + 0.9512673274850394 & 16.0 & 59.0 + 0.9513363596128612 & 17.0 & 86.0 + 0.9514221837472476 & 12.0 & 12.0 + 0.9515089240660483 & 17.0 & 87.0 + 0.9515790930715127 & 16.0 & 60.0 + 0.9516631530453461 & 14.0 & 27.0 + 0.9517463040461553 & 16.0 & 61.0 + 0.951808892592533 & 17.0 & 88.0 + 0.9518652770653788 & 15.0 & 40.0 + 0.9519383831319697 & 17.0 & 89.0 + 0.9521244725098057 & 10.0 & 6.0 + 0.9523162411492646 & 17.0 & 90.0 + 0.9523905365647094 & 15.0 & 41.0 + 0.9524481238036833 & 17.0 & 91.0 + 0.9525120069820141 & 16.0 & 62.0 + 0.9525975616461875 & 14.0 & 28.0 + 0.9526840291164363 & 16.0 & 63.0 + 0.9527546685675844 & 17.0 & 92.0 + 0.9528475463736816 & 12.0 & 13.0 + 0.9529398694971234 & 17.0 & 93.0 + 0.9530080745662677 & 16.0 & 64.0 + 0.9530891576691004 & 14.0 & 29.0 + 0.9531651619914416 & 16.0 & 65.0 + 0.9532174331812081 & 17.0 & 94.0 + 0.9532603361848672 & 15.0 & 42.0 + 0.9533072546905522 & 17.0 & 95.0 + 0.9534043196366604 & 13.0 & 19.0 + 0.9535088561007298 & 17.0 & 96.0 + 0.9535658791619818 & 15.0 & 43.0 + 0.9536237513110958 & 17.0 & 97.0 + 0.9537070876960089 & 16.0 & 66.0 + 0.9539187227353836 & 11.0 & 9.0 + 0.9541535736642798 & 16.0 & 67.0 + 0.9542576473407822 & 17.0 & 98.0 + 0.9543353288341542 & 15.0 & 44.0 + 0.9544165158088548 & 17.0 & 99.0 + 0.9545820065186768 & 13.0 & 20.0 + 0.9547587014626097 & 17.0 & 100.0 + 0.954852987166387 & 15.0 & 45.0 + 0.9549467352368605 & 17.0 & 101.0 + 0.9550714036503227 & 16.0 & 68.0 + 0.9552867504518042 & 14.0 & 30.0 + 0.9555838961047767 & 16.0 & 69.0 + 0.9559280829047855 & 17.0 & 102.0 + 0.9579685138208287 & 3.0 & 1.0 +

we describe an algebraic proof of the well - known topological fact that . the fundamental group of appears in our approach as the center of a certain finite group defined by generators and relations . the latter is a factor group of the braid group , obtained by imposing one additional relation and turns out to be a nontrivial central extension by of the corresponding group of rotational symmetries of the hyperoctahedron in dimension .

1607.05876

the set of all rotations in @xmath5 forms a group denoted by @xmath1 . we may think of it as the group of @xmath6 orthogonal matrices with unit determinant . as a topological space it has the structure of a smooth @xmath7-dimensional submanifold of @xmath8 . the group structure is compatible with the smooth one in the sense that the group operations are smooth maps , so it is a lie group . the space @xmath1 when @xmath9 has a fascinating topological property there exist closed paths in it ( starting and ending at the identity ) that can not be continuously deformed to the trivial ( constant ) path , but going twice along such a path gives another path , which is deformable to the trivial one . for example , if you rotate an object in @xmath10 by @xmath11 along some axis , you get a motion that is not deformable to the trivial motion ( i.e. , no motion at all ) , but a rotation by @xmath12 is deformable to the trivial motion . further , a rotation by @xmath11 along any axis can be deformed to a rotation by @xmath11 along any other axis . we shall call a _ full rotation _ in @xmath5 any motion that corresponds to a closed path in @xmath1 , starting and ending at the identity . thus , it turns out that there are two classes of full rotations : topologically trivial , i.e. , deformable to the trivial motion , and topologically nontrivial . every nontrivial full rotation can be deformed to any other nontrivial full rotation . two consecutive nontrivial full rotations produce a trivial one . for any topological space , one can consider the set of closed paths starting and ending at some fixed point , called _ base - point_. two closed paths that can be continuously deformed to each other , keeping the base - point fixed , are called _ homotopic_. one can multiply closed paths by _ concatenation _ , i.e. , take the path obtained ( after appropriate reparametrization ) by traveling along the first and then along the second . there is also an inverse for each path the path traveled in reverse direction . these operations turn the set of homotopy classes of closed paths ( with a given base - point ) into a group and it is an important topological invariant of any topological space . it was introduced by poincar and is called the _ first homotopy group _ or the _ fundamental group _ of the space , denoted by @xmath13 . thus , the property of @xmath1 stated above is written concisely as @xmath14 . this specific topological property in the case @xmath15 plays a fundamental role in our physical world . to the two homotopy classes of closed paths in @xmath16 correspond precisely two principally different types of elementary particles : bosons , with integer spin , and fermions , with half - integer spin , having very distinct physical properties . the difference can be traced to the fact that the complex ( possibly multicomponent ) wave function determining the quantum state of a boson is left unchanged by a rotation by @xmath11 of the coordinate system while the same transformation multiplies the wave function of a fermion by @xmath17 . this is possible since only the modulus of the wave function has a direct physical meaning , so measurable quantities are left invariant under a full rotation by @xmath11 . however , as discovered by pauli and dirac , one needs to use wave functions having this ( unexpected ) transformation property for the correct description of particles with half - integer spin , such as the electron . the careful analysis showed ( @xcite ) that the wave function has to transform properly only under transformations which are in a small neighborhood of the identity . a `` large '' transformation such as a rotation by @xmath11 can be obtained as a product of small transformations , but the transformed wave function need not come back to itself there may be a complex phase multiplying it . from continuity requirements it follows that if one takes a closed path in @xmath16 which is contractible , the end - point wave function must coincide with the initial one . therefore , a rotation by @xmath12 should bring back the wave function to its initial value and so the phase factor corresponding to a @xmath11-rotation can only be @xmath17 . what we have just described is the idea of the so - called _ projective _ representations of a lie group , which we have to use in quantum physics . as we see on the example of @xmath16 , they exist because the latter is not simply - connected , i.e. , @xmath18 is not trivial . projective representations of a non - simply - connected lie group are in fact representations of its covering group . in the case of @xmath16 this is the group @xmath19 of @xmath20 unitary matrices with unit determinant . topologically , this is the three - dimensional sphere @xmath21 ; it is a double cover of @xmath16 and the two groups are locally isomorphic and have the same lie algebra . the standard proof that @xmath0 when @xmath15 uses substantially lie theory . a @xmath22 homomorphism @xmath23 is exhibited , which is a local isomorphism of lie groups . this is the double covering map in question , sending any two antipodal points of @xmath19 ( i.e. , @xmath21 ) to one point in @xmath16 . the case @xmath24 reduces to the above result by applying powerful techniques from homotopy theory . there are several more or less easy geometric methods to unveil the nontrivial topology of @xmath16 . among them , a well - known demonstration is the so - called `` dirac s belt trick '' in which one end of a belt is fastened , the other ( the buckle ) is rotated by @xmath12 . then without changing the orientation of the buckle , the belt is untwisted by passing it around the buckle ( see , e.g. , @xcite for nice java applets and explanations ) . a refinement of `` dirac s belt trick '' was proposed in @xcite where an isomorphism was constructed between homotopy classes of closed paths in @xmath16 and a certain factor group of the group @xmath25 of pure braids with three strands . this factor group turns out to be isomorphic to @xmath3 . the idea is fairly simple and is based on the following experiment : attach the ends of three strands to a ball , attach their other ends to the desk , perform an arbitrary number of full rotations of the ball to obtain a plaited braid . then try to unplait it without further rotating the ball . as expected , braids that correspond to contractible paths in @xmath16 are trivial , while those corresponding to noncontractible paths form a single nontrivial class . while the method of @xcite is simple and easy to visualize , it has the disadvantage that it does not lend itself to a generalization to higher dimensions . ( a geometric braid in @xmath5 is always trivial when @xmath24 . ) the present paper takes a different , more algebraic approach . we study a certain discrete ( in fact finite ) group of homotopy classes of paths in @xmath1 , starting at the identity and ending at points which are elements of some fixed finite subgroup of @xmath1 . it turns out that it is convenient to use the finite group of rotational symmetries of the hyperoctahedron ( the polytope in dimension @xmath4 with vertices @xmath26 ) . each homotopy class contains an element consisting of a chain of rotations by @xmath27 in different coordinate planes . these simple motions play the role of generators of our group . certain closed paths obtained in this way remain in a small neighborhood of the identity ( in an appropriate sense , explained later ) and can be shown explicitly to be contractible . thus , certain products among the generators must be set to the identity and we get a group defined by a set of generators and relations . interestingly , the number of ( independent ) generators is @xmath28 and the relations , apart from one of them , are exactly artin s relations for the braid group @xmath2 . in this way we obtain for each @xmath4 a finite group , which is the quotient of @xmath2 by the normal closure of the group generated by the additional relation . when @xmath15 the order of the group turns out to be 48 and it is the so - called _ binary octahedral group _ which is a nontrivial extension by @xmath3 of the group of rotational symmetries of the octahedron . we may think of the former as a double cover of the latter and this is a finite version of the double cover @xmath23 . the next groups in the series have orders 384 , 3840 , 46080 , etc . ; in fact the order is given by @xmath29 . note that these groups have the same orders as the coxeter groups of all symmetries ( including reflections as well as rotations ) of the respective hyperoctahedra , but they are different . this is analogous to the relationship between @xmath30 and @xmath1 on the one hand , and @xmath31 and @xmath1 on the other . it turns out that the subgroup of homotopy classes of closed paths , i.e. @xmath32 , in each case either coincides or lies in the center of the respective group and is isomorphic to @xmath3 . factoring by it , the respective rotational hyperoctahedral groups are obtained . we believe that these are new presentations of all rotational hyperoctahedral groups and their double covers . we will consider paths in @xmath1 starting at the identity . in other words , we have continuous functions @xmath33\rightarrow so(n)$ ] , subject to the restriction @xmath34 . because the target space is a group , there is a natural product of such paths , i.e. if @xmath35 and @xmath36 are paths of this type , we first translate @xmath35 by the constant element @xmath37 to the path @xmath38 . then we concatenate @xmath36 with @xmath38 . thus , by @xmath39 we mean the path : + @xmath40 for each @xmath41 we denote by @xmath42 the path given by @xmath43 the set formed by continuous paths in @xmath1 obviously contains an identity , which is the constant path . the product in this set is not associative since @xmath44 and @xmath45 are different paths ( due to the way we parametrize them ) , even though they trace the same curve . also , @xmath42 is by no means the inverse of @xmath41 . in fact , there are no inverses in this set . however , the set of homotopy classes of such paths ( with fixed ends ) is a group with respect to the induced operations . further , we will use @xmath41 to denote the homotopy class of @xmath41 . now @xmath42 is the inverse of @xmath41 . recall that in algebraic topology one constructs the universal covering space of a non - simply connected space by taking the homotopy classes of paths starting at some fixed point ( base - point ) . within a given homotopy class , what is important is just the end - point and so the covering space is locally homeomorphic to the initial one . however , paths with the same end - point belonging to different classes are different points in the covering space . thus , one effectively `` unwraps '' the initial space . we see that the group we have defined is none other than the universal covering group of @xmath1 . in the case @xmath15 this is the group @xmath19 ; in general it is the group denoted by @xmath31 . our aim is of course to show that the subgroup corresponding to closed paths is @xmath46 or equivalently that the covering map , which is obtained by taking the end - point of a representative path , is @xmath47 . since the full group of homotopy classes of paths starting at the identity is uncountable and difficult to handle , the main idea of this paper is to study a suitable discrete , in fact finite , subgroup by limiting the end - points to be elements of the group of rotational symmetries of the hyperoctahedron in dimension @xmath4 . in what follows we denote this group by @xmath48 . ( one could perhaps take any large enough finite subgroup of @xmath1 , but using the hyperoctahedral group seems the simplest . ) the further study of @xmath48 requires algebra and just a couple of easy results from analysis and geometry . we state these here and leave the proofs for the appendix . by a _ generating path _ @xmath49 we will mean a rotation from @xmath50 to @xmath27 in the coordinate plane @xmath51 . every homotopy class of paths in @xmath1 starting at the identity and ending at an element of the rotational hyperoctahedral group contains a representative which is a product of generating paths . each vertex of the hyperoctahedron lies on a coordinate axis either in the positive or negative direction and determines a closed half - space ( of all points having the respective coordinate nonnegative or nonpositive ) to which it belongs . let us call _ local closed paths _ those closed paths in @xmath1 for which no vertex of the hyperoctahedron leaves the closed half - space to which it belongs initially . local closed paths are contractible . a closed path consisting of generating paths is contractible if and only if the word representing it can be reduced to the identity by inserting expressions describing local closed paths . we consider three - dimensional rotations separately since the essential algebraic properties of @xmath48 are present already here and in a sense the higher - dimensional cases are a straightforward generalization . it is worth recalling some facts about the group of rotational symmetries of the octahedron in three - dimensions . the octahedron is one of the five regular convex polyhedra , known as platonic solids . it has six vertices and eight faces which are identical equilateral triangles . each vertex is connected by an edge to all other vertices except the opposite one . we may assume that the vertices lie two by two on the three coordinate axes . we will enumerate the vertices from 1 to 6 as follows : @xmath53 , @xmath54 , @xmath55 , @xmath56 , @xmath57 and @xmath58 . the octahedron is the dual polyhedron of the cube and so they have the same symmetry group . the analogous statement is true in any dimension . we find it convenient to think of a spherical model of the octahedron the edges connecting the vertices are parts of large circles on the unit sphere ( figure [ octah ] ) . + -10 mm the octahedron has three types of rotational symmetries belonging to cyclic subgroups of different orders : + _ type 1 _ : three cyclic subgroups of order @xmath59 generated by rotations by @xmath27 around the three coordinate axes . each such symmetry preserves the two vertices lying on that axis while permuting the rest . + _ type 2 _ : six cyclic subgroups of order @xmath60 , generated by rotations by @xmath61 around axes connecting the centers of six pairs of opposite edges . + _ type 3 _ : four cyclic subgroups of order @xmath62 generated by rotations by @xmath63 around axes connecting the centers of four pairs of opposite faces . + counting the number of nontrivial elements in each subgroup and adding @xmath64 shows that the group of rotational symmetries of the octahedron has order @xmath65 . as each symmetry permutes the six vertices , it is obviously a subgroup of the symmetric group @xmath66 . it is well - known and easy to see experimentally that the group of rotational symmetries of the octahedron is faithfully represented as the group of permutations of the four pairs of opposite faces . therefore , it is isomorphic to @xmath67 . all symmetric groups can be realized as finite reflection groups and thus finite coxeter groups of type @xmath68 @xcite . we have @xmath69 and in the case at hand the group is @xmath70 the full symmetry group ( including reflections ) of the tetrahedron . the case @xmath15 is an exception . when @xmath24 the respective rotational hyperoctahedral group is not a coxeter group . for any @xmath4 however , it is a normal subgroup of index @xmath60 of the respective full hyperoctahedral group which is a coxeter group of type @xmath2 ( not to be confused with the braid group on @xmath4 strands , for which the same notation is used ) . -10 mm , title="fig:",width=377 ] -10 mm for pedagogical purposes we choose to make a short description of the full hyperoctahedral group as a symmetry group of the hyperoctahedron , which is generated by reflections . starting with the octahedron in dimension 3 , we see that it is preserved by reflections with respect to planes perpendicular to the vectors @xmath71 ( the six vertices ) and @xmath72 ( the middles of the twelve edges ) . the vectors just listed are called _ roots _ and they satisfy two essential properties which are axioms for a _ root system _ : 1 ) each reflection determined by a root maps the root system onto itself , and 2 ) the root system contains together with each root its negative but no other multiples of that root . among the 18 roots , we can choose ( not uniquely ) 3 roots , called _ simple roots _ with the following properties : 3 ) they span the whole space in which the root system lives , and 4 ) each root is either entirely positive linear combination of the simple ones or is entirely negative linear combination of those . one possible choice of simple roots is ( see figure [ roots ] ) @xmath73 one can check that indeed every root is a linear combination of these three with either entirely positive or entirely negative coefficients . for example notice that the angle between @xmath75 and @xmath76 is @xmath77 or @xmath78 , the angle between @xmath76 and @xmath79 is @xmath80 or @xmath63 , while the angle between @xmath75 and @xmath79 is @xmath81 or @xmath27 . this is always the case any two simple roots form either obtuse angle or are orthogonal . the angle between two roots is further restricted by the fact that the product of two reflections corresponding to two different roots is a rotation at an angle linked to the angle between the roots . more precisely , if the angle between @xmath82 and @xmath83 is @xmath84 and the corresponding reflections are denoted as @xmath85 and @xmath86 , then @xmath87 is a rotation by @xmath88 . since this rotation must have some finite order @xmath89 , we have @xmath90 . the whole finite reflection group is determined by the positive integers @xmath91 for each pair of simple roots . note that because all @xmath82 are reflections and thus have order two , the diagonal entries @xmath92 always . for the octahedron from the angles between simple roots we get @xmath93 , @xmath94 , and @xmath95 . the information is traditionally encoded in the so - called _ coxeter dynkin diagram _ where each simple root is represented by a node , two nodes are connected by an edge if @xmath96 and @xmath89 is written as a label below the corresponding edge if @xmath97 . the coxeter dynkin diagram for the octahedral group is shown on figure [ dynkin ] . and @xmath2,title="fig:",width=113 ] and @xmath2,title="fig:",width=113 ] the generalization to higher @xmath4 is pretty straightforward . the roots will be all @xmath98 vertices and the middles of all @xmath99 edges of the respective hyperoctahedron . it is easy to figure out that one can choose the following @xmath4 simple roots : @xmath100 , and calculating the integers @xmath89 , we see that the group has a coxeter dynkin diagram as in figure [ dynkin ] . the groups of this type are denoted by @xmath2 , where @xmath4 is the number of simple roots and is called the _ rank_. alternatively , the full hyperoctahedral group can be thought of as the _ wreath product _ @xmath101 . the wreath product in this special case is the semidirect product of the product of @xmath4 copies of @xmath102 with @xmath103 , where @xmath103 acts on the first factor by permuting its components . more precisely , let @xmath104 and let @xmath105 . the symmetric group @xmath103 , considered as a permutation group of @xmath106 , acts naturally by automorphisms on @xmath107 , namely , if @xmath108 @xmath109 now , the wreath product @xmath101 is just the semidirect product @xmath110 . the @xmath4 pairs of opposite vertices determine @xmath4 mutually orthogonal ( non - oriented ) lines in @xmath5 . a rotation in the @xmath51th plane by @xmath27 permutes the @xmath111th and @xmath112th lines . an arbitrary symmetry can be realized by an arbitrary permutation of the lines plus possible reflections with respect to the hyperplanes perpendicular to the @xmath4 lines . this explains the structure of the full hyperoctahedral group as a wreath product . one advantage is that it is easy to see that the order of the group is @xmath29 . the rotational hyperoctahedral group forms a normal subgroup of index 2 in the full hyperoctahedral group . in terms of orthogonal matrices , this is the subgroup of matrices with unit determinant . note that each rotation is a product of two reflections . coming back to the rotational octahedral group ( i.e. the case @xmath15 ) , we observe that it is generated by three elements , denoted by @xmath113 , @xmath114 and @xmath115 , of _ type 1 _ rotations by @xmath27 around the three coordinate axes . they permute the vertices of the octahedron and using standard notations for permutations in terms of cycles ( omitting the trivial 1-cycles ) we can write : @xmath116 obviously , these three elements have order 4 , and generate the respective cyclic groups of _ type 1 _ but they also generate all symmetries of _ type 2 _ and _ type 3 _ : the six rotations by @xmath61 around the respective rotational axis ( as described above ) in terms of @xmath113 , @xmath114 and @xmath115 are @xmath117 , @xmath118 , @xmath119 , @xmath120 , @xmath121 , and @xmath122 . the eight rotations by @xmath123 around the respective rotational axis ( as described above ) in terms of @xmath113 , @xmath114 and @xmath115 are @xmath124 , @xmath125 , @xmath126 , @xmath127 , and their inverses . we consider the discrete group @xmath48 generated by the generating paths @xmath128 , @xmath129 , and @xmath130 , treated as homotopy classes . ( we have @xmath131 . ) local closed paths built out of the generators @xmath132 and their inverses must be set to identity . apart from trivial cases where an @xmath132 is followed by its inverse , we have a family of paths for which each vertex either goes around the edges of a single triangular face of the octahedron or moves along an edge and comes back . inspecting all possible ways in which such `` triangular '' closed paths can be built , we see that each one is represented by a word of four letters . each word contains either @xmath132 or @xmath133 for each @xmath111 . no word contains twice a given letter but it may contain a letter together with its inverse . in that case this letter and its inverse conjugate one of the other two letters . here is a list of all identities that follow : @xmath134 ( the path traced by vertex @xmath64),width=302 ] very few of these 24 identities are actually independent . first , we notice that one of the generators , e.g. @xmath135 can be expressed as a combination of the other two and their inverses , in several different ways . for example , if we use the first identities in the last two rows we get @xmath136 , from which follows @xmath137 while if we use @xmath138 , and also @xmath139 , we obtain : @xmath140 all other identities are consequences of these three . therefore , @xmath48 is presented as a group generated by two generators and a set of relations , one of which ( equation [ artin ] ) is precisely artin s braid relation for the braid group with three strands @xmath141 . in other words @xmath48 is the quotient of @xmath141 by the normal closure of the subgroup generated by the additional relations ( [ add ] ) . actually , only one of the identities ( [ add ] ) is independent : [ second]the second identity in ( [ add ] ) follows from the first one and artin s braid relation ( [ artin ] ) . using artin s relation twice it is almost immediate that @xmath142 now using the first of the identities ( [ add ] ) inside the expression above we get @xmath143 -10 mm that @xmath144 follows from @xmath145,title="fig:",width=302 ] * note : * as shown by artin @xcite , the braid group @xmath2 can be thought as a group of isotopy classes of geometric braids with @xmath4 strands or as a group generated by @xmath28 generators satisfying what came to be called artin s braid relations . when @xmath15 there is just one relation ( [ artin ] ) between the two generators . the geometric picture has the advantage of being more intuitive and providing us with ways to see identities , which can then be shown algebraically . thus , for example , the proof above has a geometric version which is easy to visualize ( figure [ braiding ] ) . the same situation will be in place when we consider @xmath24 . our group @xmath48 will be generated by @xmath28 generators satisfying the standard braid relations plus some additional ones . using the geometric picture will allow us to arrive at conclusions which are difficult to see algebraically . * corollary : * _ the group @xmath48 has a presentation _ @xmath146 [ r4 ] the order of @xmath35 and @xmath36 in @xmath48 is eight . using relations ( [ add ] ) we conclude that @xmath147 and similarly @xmath148 . therefore @xmath149 next , one can write @xmath150 putting together the results in the last two equations we see that @xmath151 and similarly @xmath152 , which imply @xmath153 . from this we can tell that the order of @xmath35 and @xmath36 is at most 8 . to conclude that it is exactly 8 ( and not for example 4 ) is not a trivial problem . one way to do this is to perform the todd coxeter algorithm ( see e.g. ken brown s short description @xcite ) . if the group defined by a finite set of generators and relations is finite , the algorithm will ( in theory ) close and stop and will produce the order of the group and a table for the action of the generators on all elements . we used the simple computer program graciously made available by ken brown on his site . the resulting table showed that the order of @xmath48 is 48 and the central element @xmath154 is not trivial . the results are represented graphically in the so - called cayley graph ( figure [ cayley ] ) where dotted lines represent multiplication by @xmath35 and solid lines by @xmath36 ( the dotted lines wrap around horizontally ) . , width=377 ] we know that taking the end - point of any element @xmath155 gives a homomorphism from @xmath48 onto the group of rotational symmetries of the octahedron and the latter can be considered as a subgroup of the symmetric group @xmath66 . so we have a homomorphism @xmath156 . in particular , @xmath157 and @xmath158 . to prove that the fundamental group of @xmath16 is @xmath159 , we need to show that the kernel of the homomorphism described above is @xmath159 , because the kernel consists of those elements of @xmath48 which are classes of closed paths in @xmath16 . in other words , these are the motions that bring the octahedron back to its original position . it will be helpful to come up with a way to list all elements of @xmath48 as words in the generators @xmath35 , @xmath36 and their inverses , i.e. , find a _ canonical form _ for the elements of @xmath48 . [ elements ] any @xmath160 can be written uniquely in one of the three forms : + 1 . @xmath161 ( 32 elements ) , + 2 . @xmath162 ( 8 elements ) , + 3 . @xmath163 ( 8 elements ) , + where @xmath164 and @xmath165 . any element of @xmath48 can be obtained by multiplying the identity by a sequence of the generators and their inverses either on the right or on the left . we are using right multiplication . the idea is to show that when multiplying an element @xmath160 , which is written in the form 1 , 2 or 3 by any of the two generators of @xmath48 or their inverses , we get again an expression of these three types . since @xmath166 it is enough to check the above for positive powers . the proof is a direct verification using artin s braid relation and the identity @xmath167 ( and the symmetric one with @xmath35 and @xmath36 interchanged ) . thus , e.g. , multiplying the first expression by @xmath35 we get @xmath168 , if @xmath4 is even and either expression 2 or 3 , if @xmath4 is odd . similarly , multiplying expression 3 by @xmath35 we get @xmath169 while multiplying it by @xmath36 gives @xmath170 uniqueness is proven by inspection . let us show as an example that @xmath171 . indeed , the assumption that the two are equal leads to the following sequence of equivalent statements : @xmath172 the last identity is apparently wrong since no power of @xmath35 can be equal to @xmath36 . the kernel of the homomorphism @xmath173 is isomorphic to @xmath159 . we study how the images under @xmath174 of the elements of @xmath48 permute the numbers @xmath175 . first , @xmath176 leaves @xmath64 in place , then @xmath177 sends @xmath64 to @xmath59 , while @xmath178 sends @xmath64 to @xmath62 . finally , recalling that @xmath157 we see that @xmath64 is not in the orbit of @xmath59 or @xmath62 under the action of @xmath176 , so there is no way that @xmath179 or @xmath180 can bring @xmath64 back to itself . next we look at @xmath181 . recall that because @xmath182 and @xmath183 we have @xmath184 and @xmath185.unless @xmath186 , which happens if and only if @xmath187 , one of the numbers @xmath188 will be sent to @xmath64 . then , since @xmath64 is fixed by @xmath189 , we see that @xmath181 can not be trivial unless @xmath190 . in that case it is obvious that @xmath189 can be the identity only when @xmath191 . thus , @xmath192 since @xmath193 is a central element of order two , the group generated by it , @xmath194 , belongs to the center @xmath195 . it is not obvious a priori that there are no other central elements . we make the following observation : the center of the braid group @xmath2 is known to be isomorphic to @xmath196 . it is generated by a full twist of all @xmath4 strands @xmath197 ( using @xmath198 for the generators of @xmath2 ) . since @xmath48 is obtained from @xmath2 by imposing one more relation , any central element in @xmath2 will be central in @xmath48 , although there is no guarantee that we will obtain a nontrivial element in @xmath48 in this way . in addition , there may be central elements in @xmath48 which come from non - central elements in @xmath2 . when @xmath15 the above argument ensures that @xmath199 is a central element in @xmath48 . the calculation in proposition [ r4 ] shows that @xmath200 so we do not get any new central element , different from the one we have already found . when @xmath201 , however the analogous calculation , using the braid relations plus the additional relations as in equation [ add ] , gives @xmath202 this is another central element , different from @xmath203 . taking the product of the two we obtain a third central element @xmath204 . therefore when @xmath201 , @xmath195 contains ( in fact coincides with ) the product of two copies of @xmath159 . this obviously generalizes to any even @xmath4 the element @xmath205 is central , as can be checked explicitly . the difference between even and odd dimensions can be traced back to the difference between rotation groups in even and odd dimensions . as we shall see shortly , when we factor @xmath48 by @xmath206 we obtain the rotational hyperoctahedral group . in odd dimensions this has trivial center , which follows for example from schur s lemma and the fact that the matrix @xmath207 is not a rotation . however , in even dimensions reflection of all axes , given by @xmath208 , is a rotation and therefore the corresponding rotational hyperoctahedral group has a nontrivial center . according to the general construction , we expect that when we factorize @xmath48 by the kernel of the covering map , which is nothing but @xmath209 , we should obtain the rotational octahedral group . this can also be established directly . denoting by @xmath210 the quotient , we have a presentation for it : @xmath211 the order of @xmath210 is 24 . we can list its elements using the same expressions as in proposition [ elements ] , except that now the two integers @xmath4 and @xmath89 run from @xmath50 to @xmath62 . an easy calculation shows that @xmath210 contains nine elements of order 2 , eight elements of order 3 , and six elements of order 4 . among all 15 classified groups of order 24 , the only one that has this structure is the symmetric group @xmath67 , which on the other hand is the group of the octahedral rotational symmetries . given the structure of @xmath48 that can be visualized through its cayley graph , we conclude that it is a nontrivial extension by @xmath159 of @xmath67 described by a non - split short exact sequence @xmath212 the non - isomorphic central extensions by @xmath159 of @xmath67 are in one - to - one correspondence with the elements of the second cohomology group ( for trivial group action ) of @xmath67 with coefficients in @xmath159 and the latter is known to be isomorphic to the klein four - group @xmath213 . the identity in cohomology corresponds to the trivial extension @xmath214 ( more generally , a semidirect product is also considered trivial ) . the other three elements of the cohomology group classify the three non - isomorphic nontrivial extensions , namely the binary octahedral group @xmath215 , the group @xmath216 of nonsingular @xmath20 matrices over the field with three elements , and the group @xmath217 of @xmath20 matrices with unit determinant over the ring of integers modulo 4 . [ 2o ] @xmath48 is isomorphic to the binary octahedral group @xmath215 . * note : * by construction , we expect the group @xmath48 to be a subgroup of @xmath19 and the covering map @xmath218 to be a restriction of the covering map @xmath23 . the fact that @xmath48 with the presentation ( [ biocta ] ) is isomorphic to @xmath215 is mentioned in section 6.5 of @xcite . the gap i d of @xmath48 is [ 48 , 28 ] . the binary octahedral group , being a subgroup of @xmath19 , can be realized as a group of unit quaternions . in fact it consists of the 24 hurwitz units @xmath219 and the following 24 additional elements : @xmath220 to prove the isomorphism it is enough to find two elements in @xmath215 which generate the whole group and which satisfy the same relations as the defining relations of @xmath48 . obviously we are looking for elements of order 8 . let , e.g. , @xmath221 . it is a simple exercise in quaternion algebra to prove that @xmath222 . similarly , we check that @xmath223 , or equivalently that @xmath224 . indeed , since @xmath225 and since @xmath226 and @xmath112 anticommute , @xmath227 a direct verification further shows that the whole @xmath215 is generated by @xmath228 and @xmath229 . the 24 hurwitz units , when considered as points in @xmath230 , lie on the unit sphere @xmath21 and are the vertices of a regular 4-polytope the 24-cell , one of the exceptional regular polytopes with symmetry the coxeter group @xmath231 . the set is also a subgroup of @xmath215 the binary tetrahedral group , denoted as @xmath232 . the second set of 24 unit quaternions can be thought of as the vertices of a second 24-cell , obtained from the first one by a rotation , given by multiplication of all hurwitz units by a fixed element , e.g. @xmath233 . the convex hull of all 48 vertices is a 4-polytope , called disphenoidal 288-cell . it may be instructive to consider the symmetric group @xmath67 with its presentation given by equation ( [ s4 ] ) and try to construct explicitly all non - isomorphic central extensions by @xmath159 . this means that the three relations in the presentation of @xmath67 must now be satisfied up to a central element , belonging to the ( multiplicative ) cyclic group with two elements @xmath234 : @xmath235 the element @xmath236 in the first relation can be absorbed by replacing @xmath35 by @xmath237 , so artin s braid relation remains unchanged . we are left with four choices for @xmath238 and @xmath239 : + 1 . the choice @xmath240 leads to the trivial extension as a direct product @xmath214 . the choice @xmath241 leads to the already familiar group @xmath242 3 . the choice @xmath243 leads to the group @xmath216 with presentation @xmath244 4 . the choice @xmath245 leads to the group @xmath217 with presentation @xmath246 or , after simplifications , to @xmath247 the proofs of points 3 and 4 above repeat the logic of the proof of proposition [ 2o ] . for the group @xmath216 we can make the following identifications @xmath248 and then check that they generate the whole @xmath216 and satisfy the respective relations . similarly , for @xmath217 we can set @xmath249 notice that there is an essential difference between the extensions described in 2 and 3 , and the extension in 4 . the extensions in 2 and 3 have presentations as the presentation of @xmath67 ( equation [ s4 ] ) with the same set of generators and some relation removed . this is not the case with @xmath217 where , if we want to keep the form of the relations , we need three generators . notice also that for @xmath215 and @xmath216 the order of @xmath35 and @xmath36 becomes 8 ( it is 4 in @xmath67 ) , while in @xmath217 the order of @xmath35 and @xmath36 remains 4 . ( it may be worth pointing out that the properties of the braid group @xmath2 ensure that in any of its factors the order of any two ( standard ) generators has to be the same . ) , width=188 ] the groups @xmath215 and @xmath216 are so - called _ stem extensions _ of @xmath67 , i.e. , the abelian group by which we extend is not only contained in the center of the extended group , but also in its commutator subgroup . one can check explicitly that for @xmath215 and @xmath216 the corresponding central element @xmath154 can be written as a commutator , while in the case @xmath217 the center is generated by @xmath199 and it is not in the commutator subgroup . if one looks at the cayley graph of @xmath215 ( figure [ cayley ] ) , one can see that it really looks like a ( topological ) double cover of the corresponding cayley graph of @xmath67 ( figure [ cayley24 ] ) . for this reason stem extensions are called covering groups and are discrete versions of covering groups of lie groups . it is intriguing that while @xmath16 has @xmath19 as its unique ( double ) cover , the finite subgroup @xmath67 has one additional double cover , namely @xmath216 , which does not come from lifting @xmath16 to @xmath19 . finally , we may notice that in the presentations of @xmath215 and @xmath216 there is no condition imposed on the order of the central element . its order comes out to be 2 automatically , which means that @xmath67 does not admit bigger stem extensions and the two groups considered are maximal stem extensions , i.e. , _ schur extensions_. the @xmath4-dimensional case is an easy generalization of the three - dimensional one . we consider products of generating paths @xmath49 in @xmath1 as defined in section 2 . if we take a closed path of the form @xmath251 , where all four indices are different , it is clear that this is contractible as these are rotations in two separate planes and at the level of homotopy classes the generators @xmath49 and @xmath252 will commute . when one of the indices coincides , we have a motion that takes place in a 3-dimensional subspace of @xmath5 and we may use the algebraic relations we had in the previous section . in particular , if we consider the elements @xmath49 , @xmath253 and @xmath254 , we can express one , e.g. the third , in terms of the other two , just as we expressed @xmath255 as the conjugation of @xmath256 by @xmath257 . at this point , it seems convenient to choose a different notation , where @xmath258 . with this notation we have @xmath259 , @xmath260 , etc . in this way , all @xmath49 are products of the @xmath28 generators @xmath132 and their inverses . we have @xmath261 when @xmath262 . further , since local closed paths that correspond to rotations in any 3-dimensional subspace must be set to identity when passing to homotopy classes , we have identities analogous to the ones in equation [ i d ] for any two generators @xmath132 and @xmath263 . in particular , artin s braid relation is satisfied for any @xmath264 : @xmath265 the other relation must also hold : @xmath266 the fact that there are no additional relations follows from the observation that any contractible closed path in @xmath1 which is a product of generating paths can be written as a product of `` triangular '' local closed paths of the type described in section 3 ( see appendix ) . the properties of the braid group lead to some interesting restrictions on the type of additional relations that can be imposed on the generators . in particular , if @xmath267 is some relation involving the first two generators and their inverses , then it follows that @xmath268 ( translation ) and @xmath269 ( symmetry ) will be satisfied automatically . these have a simple geometric explanation . for example , the first relation involves the first three strands of the braid , so if we want to prove the relation for @xmath36 and @xmath135 , we flip the first strand over the next three , apply the relation and then flip back the former first strand to the first place . the symmetry property follows from the first and the fact that the braid group has an outer automorphism @xmath270 which correspond to `` looking at the same braid from behind . '' it is also clear that the order of all generators in the factor group will be the same . therefore , for any @xmath4 the group @xmath48 generated by the generating paths , up to homotopy , has presentation @xmath271 the group @xmath48 for arbitrary @xmath4 has many features in common with the case @xmath15 . in particular , all @xmath132 have order @xmath272 and @xmath273 is central . the todd - coxeter algorithm when run on the computer gives the following results for the order of @xmath48 when @xmath274 respectively 48 , 384 , 3840 , 46080 . in fact , we have @xmath275 , which will be shown next . as mentioned already in section 3 , the full hyperoctahedral group in dimension @xmath4 , which is the coxeter group @xmath2 , has the same order , but @xmath48 is a different group it is a non - trivial double cover of the orientation - preserving subgroup of the full hyperoctahedral group . a canonical form of the elements of @xmath48 can be defined inductively as follows : + let @xmath276 denote a word which contains no @xmath277 with @xmath278 . then we will say that @xmath279 is in canonical form if it is written as @xmath280 with @xmath281 being in canonical form and @xmath282 being an expression of one of the types : + 1 . @xmath283 , + 2 . @xmath284 , + 3 . @xmath285 , + as an example , let us list all elements of @xmath48 in the case @xmath201 , by multiplying all canonical expressions containing @xmath35 and @xmath36 ( see proposition [ elements ] ) with all the expressions as above , with @xmath286 . we have @xmath287 ( 128 elements ) , @xmath288 ( 32 elements ) , @xmath289 ( 32 elements ) , @xmath290 ( 32 elements ) , @xmath291 ( 32 elements ) , @xmath292 ( 32 elements ) , @xmath293 ( 8 elements ) , @xmath294 ( 8 elements ) , @xmath295 ( 8 elements ) , @xmath296 ( 8 elements ) , + @xmath297 ( 32 elements ) , @xmath298 ( 8 elements ) , @xmath299 ( 8 elements ) , @xmath300 ( 8 elements ) , @xmath301 ( 8 elements ) . the following is a straightforward generalization of proposition [ elements ] . any element of @xmath48 can be written uniquely in the canonical form defined above . the number of elements is @xmath29 . the idea is to show that by multiplying an element in canonical form on the right by any @xmath277 , one gets another element that can be brought to a canonical form as well . there is nothing conceptually different from the proof of proposition [ elements ] and we skip the details . in order to calculate the order of @xmath48 , we notice that the elements in canonical form of type @xmath279 are obtained by all possible products of all elements of type @xmath281 with the @xmath302 different expressions of type @xmath282 . starting with @xmath303 where we have @xmath304 and remembering that @xmath111 runs from @xmath64 to @xmath28 we get @xmath305 the group @xmath48 consists of homotopy classes of paths in @xmath1 starting at the identity and ending at an element of @xmath1 which is a rotational symmetry of the hyperoctahedron in @xmath4 dimensions ( also called @xmath4-orthoplex or @xmath4-cross polytope ) . in particular we have a homomorphism @xmath306 because we have a permutation of the @xmath98 vertices of the hyperoctahedron ( we take left action of the group on the set ) . since the fundamental group of @xmath1 consists of the homotopy classes of closed paths , we investigate the kernel of @xmath174 . @xmath307 we choose to enumerate the @xmath98 vertices of the hyperoctahedron as @xmath308 + @xmath309 taking @xmath310 to denote the two opposite vertices on the first axis , @xmath311 on the second axis , etc . then the element @xmath312 permutes cyclically only the elements @xmath313 leaving the rest in place . consider an element @xmath314 in canonical form . by looking at the three possible expressions for @xmath315 we see that the leftmost letter is @xmath316 to some power , possibly preceded ( on the right ) by a sequence of @xmath132s with decreasing @xmath111 . since in the word @xmath315 only @xmath317 moves @xmath4 and @xmath318 , we have @xmath319 , unless @xmath320 . now , because @xmath321 does not contain @xmath316 , the number @xmath4 is not the image of any number @xmath322 under the action of @xmath323 . therefore we have @xmath324 , unless @xmath325 . proceeding in this way we see that @xmath326 if and only if @xmath327 . finally , as @xmath328 ( cyclic permutation of @xmath329 ; the rest fixed ) , the only possible cases are when @xmath330 . therefore , @xmath193 is the only nontrivial element of @xmath331 and it has order @xmath60 . we obtained a series of finite groups @xmath48 from the braid groups @xmath2 by imposing one additional relation , namely @xmath332 . these groups are nontrivial double covers of the corresponding rotational hyperoctahedral groups and have order @xmath29 . it is quite obvious that we obtain a second , nonisomorphic series of double covers if we impose the relations @xmath333 and @xmath334 , instead . note that for the second series we need the additional condition , which then implies that @xmath273 is central . when @xmath15 the two groups are the two schur extensions of the base group . this is perhaps the case also for arbitrary @xmath4 . _ proof of lemmas 1 and 2 . _ it is helpful to introduce a function , measuring the ( square of a ) `` distance '' between two points on @xmath1 . let @xmath335 be written as @xmath6 orthogonal matrices and define @xmath336 it is a simple exercise to show that this is a positive - definite symmetric function on @xmath337 . notice that @xmath338 is the cosine of the angle between the image of the standard coordinate basis vector @xmath339 under the action of @xmath340 and its image under the action of @xmath341 . the function @xmath342 does not satisfy the triangle inequality and is not a true distance , but this causes no difficulties in our considerations . further we write the proof for @xmath16 for brevity . it is obvious that the same method works in general . let @xmath343 be written as a @xmath344 orthogonal matrix : @xmath345 the `` distance '' of @xmath340 to the identity ( writing just one argument ) is : @xmath346 thus @xmath347 and @xmath348 implies @xmath349 . the gradient of @xmath350 is a 9-dimensional vector field , which we choose to write as a @xmath344 matrix . we have @xmath351 @xmath16 is a 3-dimensional submanifold of @xmath352 consisting of the points satisfying six algebraic equations , ensuring that the @xmath344 matrix @xmath340 is orthogonal . in vector form these are equivalent to the statement that the ( column ) vectors @xmath353 form an orthonormal basis : the respective gradients of these six functions are @xmath355 @xmath356 it is well known and an easy exercise that for each @xmath343 these six vectors are linearly independent and their orthogonal complement is precisely the tangent space of @xmath16 . ( this is in fact how it is shown that these six equations in @xmath352 define indeed a three - dimensional submanifold . ) we want to show that under our assumptions @xmath357 has a nonzero tangential component . indeed , supposing that @xmath357 is in the span of the six vectors above leads to @xmath358 which in turn is equivalent to the three vector equations @xmath359 taking the scalar product of the first equation with @xmath360 yields @xmath361 , while taking the scalar product of the second equation with @xmath362 yields @xmath363 . in a similar way we see that @xmath364 for any @xmath365 . therefore the matrix @xmath340 must be symmetric and being also orthogonal its square is the identity . the eigenvalues can only be @xmath64 and @xmath17 but the latter is excluded by the assumption that under the transformation corresponding to @xmath340 the coordinate axes do not leave the closed half - space they belong to initially . thus that the tangential component of @xmath366 defines a vector field on @xmath16 which will be nonzero for any @xmath367 where @xmath33\rightarrow so(3)$ ] is a local closed path . this means that the flow along this vector field defines a homotopy from @xmath368 to the identity of @xmath16 . this proves the first part of lemma 2 . let us denote by @xmath369 the respective rotational hyperoctahedral group in dimension @xmath4 . taking an arbitrary path @xmath370\rightarrow so(n)$ ] with @xmath34 and @xmath371 , we want to construct another path , homotopic to the first one , which is a product of generating paths @xmath132 . we can proceed as follows : if @xmath372 $ ] the smallest @xmath373 for which the `` distance '' from @xmath368 to some @xmath374 becomes equal to the `` distance '' to @xmath375 , we take a product of generating paths @xmath376 with @xmath377 and @xmath378 . note that the transformation @xmath379 leaves every vertex of the hyperoctahedron in the closed half - space determined by it ( as defined in section 2 ) or , equivalently , the angle between the standard basis vector @xmath380 and its image under @xmath379 does not exceed @xmath27 for any @xmath111 and any @xmath373 . indeed , as we consider a continuous path in @xmath1 starting at the identity , we may think of the motion of the @xmath4 points on the @xmath381dimensional sphere ( the images of the vectors @xmath380 under @xmath368 ) . initially all points remain in some respective adjacent @xmath381-cells ( these are @xmath381simplices ( spherical ) ) forming the spherical hyperoctahedron . if the point @xmath111 is to leave the closed half - space to which it initially belonged , it must reach the @xmath382-dimensional boundary opposite to @xmath380 for some @xmath373 . ( at the same time there will be at least one more point @xmath383 belonging to a boundary of an @xmath381-cell since two points can not belong to the interior of the same @xmath381cell ) . but then it follows that there will be a vertex @xmath112 the angular distance to which , from @xmath111 is less than or equal to @xmath27 . this implies that the element @xmath384 giving rotation by @xmath27 in the @xmath51th plane is not further to @xmath368 than the `` distance '' between @xmath368 and the identity . the above argument shows that even though the element @xmath113 above does not determine the product @xmath376 uniquely , it is unique up to homotopy . next , we take @xmath385 as the smallest @xmath386 for which the `` distance '' from @xmath368 to some @xmath114 becomes equal to the `` distance '' to @xmath113 . there is an element @xmath387 in @xmath388 , such that @xmath389 and @xmath387 satisfies the same property as @xmath113 above . we take a product of generating paths @xmath390 with @xmath391 and @xmath392 . proceeding in this way and taking the product of products of generating paths we produce a path ( after renumbering ) @xmath393 with @xmath394 and @xmath395 . the elements @xmath396 which we have to use at each step may not be unique and we will have to make a choice but the end result will lead to homotopic paths . to show that the path we have constructed is homotopic to the original path @xmath41 we can use the following trick for each fixed @xmath373 take @xmath397 . the path @xmath398 is closed and is local by the construction of @xmath399 . therefore it is homotopic to the identity by the first part of lemma 2 . with this lemma 1 is proven . finally , we need to show that the expression corresponding to any contractible closed path consisting of generating paths can be reduced to the identity by inserting in it words giving local closed paths . more precisely , the words that must be inserted give triangular closed paths as described in section 3 . this is essential as we must be sure that there are no additional relations other than the ones as in equations [ artin ] and [ add ] . first we carry out the proof for local closed paths consisting of generating paths ( these are contractible because of locality ) by induction on the length of the word representing the path . a single - letter path @xmath49 can not be closed . a two - letter path can only be closed if the word is @xmath400 . a three - letter path can not be closed as can easily be seen considering all possibilities . for example , if we want to try to close the path starting with @xmath401 , we have to add @xmath402 on the right so that vertex 1 goes back to 1 but @xmath403 is not closed since 2 goes to -3 and 3 goes to -2 . ( we adopt enumeration of the @xmath98 vertices with @xmath404 where @xmath405 denotes the vertex opposite to @xmath111 . the element @xmath49 moves vertex @xmath111 to vertex @xmath112 , vertex @xmath112 to vertex @xmath405 and leaves all other vertices in place . ) considering four - letter paths , it is easy to check that they can be closed either if they consist of a product of two closed two - letter paths or if they involve only rotations in a three - dimensional subspace spanned by some three axes @xmath111 , @xmath112 and @xmath226 . these are precisely what we called triangular closed paths in section 3 . in general notations it turns out that all such closed paths are words that are cyclic permutations of the following four expressions : @xmath406 notice that setting these expressions to one gives conjugation identities , e.g. @xmath407 + @xmath408 , etc . suppose now that the statement we want to prove is valid for all words with length @xmath98 and consider a word of @xmath409 letters . suppose that the first letter on the right is @xmath49 . we can move @xmath49 to the left across any @xmath252 with @xmath226 , @xmath410 different from @xmath111 and @xmath112 , as rotations in two such planes commute . if @xmath49 gets next to a word @xmath411 we are allowed to cancel the two and obtain a word of length @xmath98 and we are done . the possibility to obtain @xmath412 is excluded by locality since anything to the right leaves vertex @xmath111 invariant and @xmath413 sends @xmath111 to @xmath405 . there are four additional possibilities in which @xmath49 gets next to a letter with which it does not commute . these are @xmath414 , @xmath415 , @xmath416 and @xmath417 . using the identities following from setting the expressions in equation ( [ triang ] ) to one , we replace the above combinations by products of two letters in which the index @xmath111 appears only in the letter on the left . here are the actual identities that can be used : @xmath418 next , we continue moving the letter involving @xmath111 to the left . notice that the index @xmath111 may now appear as the second index , so we may need to use the additional identities @xmath419 eventually the letter involving the index @xmath111 will reach the left end of the word and there will be no letters involving @xmath111 to its right . such a path can not be closed as it does not leave vertex @xmath111 in place . therefore , the only possibility is that the letter involving @xmath111 gets canceled in the process and the length of the word becomes @xmath98 . now consider a general closed path @xmath41 consisting of generating paths and homotopic to the identity . in general the homotopy from the identity to @xmath41 need not pass through paths consisting of generating paths but we can modify it so that it does . indeed , if @xmath420 is the homotopy from the identity to @xmath41 , i.e. @xmath421 , @xmath422 , for each @xmath423 we can homotope @xmath420 to the nearest closed path consisting of generating paths , as described earlier in this section . for @xmath423 small enough the nearest will be @xmath375 . for some @xmath424 the construction will yield a path @xmath399 consisting of generating paths , such that @xmath399 is homotopic to @xmath425 . furthermore , the construction is such that @xmath399 is local closed path as any vertex remains in some @xmath381-cell without leaving it . then for some @xmath426 we will get @xmath398 which is nearest to @xmath427 . the path @xmath398 is not local but @xmath428 ( usual product of paths by concatenation ) is local . since the words giving @xmath399 and @xmath428 can be reduced to the identity by inserting words giving triangular closed paths , the same will be true for @xmath398 . proceeding in this way , after a finite number of steps we show that the word for the closed path @xmath41 has the same property . 9 e. artin , theory of braids , _ math . ann . _ * 48 * ( 1 ) ( 1947 ) 101126 . v. bargmann , on unitary ray representations of continuous groups , _ math . ann . _ * 59 * no . 1 ( 1954 ) 146 . k. brown , _ the todd - coxeter procedure _ , lecture , cornell univ . , ny , 2013 http://www.math.cornell.edu/~kbrown/7350/toddcox.pdf . h. s. m. coxeter , w. o. j. moser , _ _ g__enerators and relations for discrete groups . ergebnisse der mathematik und ihrer grenzgebiete , vol . 14 , springer - verlag , 1980 . g.egan , applets gallery ( 2000 ) http://gregegan.customer.netspace.net.au/applets/21/21.html . j. humphreys , _ reflection groups and coxeter groups _ , vol . 29 , cambridge university press , 1992 . b. palais , quicktime movies of bob palais demonstrating the belt trick and the plate trick , http://www.math.utah.edu/~palais/links.html . v. stojanoska , o. stoytchev , touching the @xmath159 in three - dimensional rotations , _ mathematics magazine _ * 81 * no . 5 ( 2008 ) 345357 e. wigner , on unitary representations of the inhomogeneous lorentz group , _ math . _ * 40 * no . 1 ( 1939 ) 149204 .

galaxy evolution depends strongly on the physics of the interstellar medium ( ism ) . motivated by the need to incorporate the properties of the ism in cosmological simulations we construct a simple method to include the contribution of non - thermal components in the calculation of pressure of interstellar gas . in our method we treat three non - thermal components - turbulence , magnetic fields and cosmic rays - and effectively parametrize their amplitude . we assume that the three components settle into a quasi - steady - state that is governed by the star formation rate , and calibrate their magnitude and density dependence by the observed radio - fir correlation , relating synchrotron radiation to star formation rates of galaxies . we implement our model in single cell numerical simulation of a parcel of gas with constant pressure boundary conditions and demonstrate its effect and potential . then , the non - thermal pressure model is incorporated into _ ramses _ and hydrodynamic simulations of isolated galaxies with and without the non - thermal pressure model are presented and studied . specifically , we demonstrate that the inclusion of realistic non - thermal pressure reduces the star formation rate by an order of magnitude and increases the gas depletion time by as much . we conclude that the non - thermal pressure can prolong the star formation epoch and achieve consistency with observations without invoking artificially strong stellar feedback . [ firstpage ] galaxies : evolution hydrodynamics cosmic rays ism : general ism : magnetic fields

1311.1206

both star formation and stellar feedback play crucial roles in galaxy evolution . star formation leads to stellar feedback , which in turn is assumed to regulate the star formation rate and prevent galaxies from turning all their gas into stars over less than a @xmath0 ; moderate star formation rates are implied from low and high redshift observations . in contrast , pure hydrodynamic simulations and semi - analytic models of galaxy formation tend to predict high gas densities within galaxies . these densities cause the gas to cool very efficiently and supersede the density threshold required for star formation @xcite . for purely hydrodynamic simulations , unrealistically strong stellar feedback is often necessary to regulate the star formation rate in galaxies @xcite . several approaches have been attempted to regulate simulations of star formation during galaxy evolution . within the framework of pure hydrodynamics , the most basic feedback mechanism is usually thermal feedback , or injection of some fraction of the supernova energy into the gas . since this energy is injected into dense , cold gas it cools efficiently and typically has small overall effect on galaxy evolution @xcite . momentum feedback is added by explicitly injecting momentum to the material that surrounds star forming regions @xcite . while the efficiency of such models is slightly better than thermal feedback @xcite , the ejected gas is almost always highly supersonic and kinteic energy is converted to thermal radiation very efficiently through shocks , unless the feedback is injected effectively over very large volumes @xcite . attempts have also been made to inject the energy into a warm component of a two - phase gas , effectively delaying the cooling until energy is transferred from the diffuse gas to a denser gas which immediately cools @xcite . an additional technique sometime used has been to increase the efficiency of the feedback by releasing the feedback energy in bursts rather than spread out @xcite . however , in all these recipes , the cooling of a parcel of gas near the plane of the disc still occurs with isobaric boundary conditions set by the weight of the atmosphere on top of it . once the gas cools , it contracts over a crossing time to regain its pressure . since the cooling rate scales as @xmath1 ( where @xmath2 is the gas density ) this leads to runway cooling . in order to increase the efficiency of supernova feedback , the cooling of injected energy is artificially delayed @xcite , and momentum feedback efficiency in enhanced by preventing it from interacting with its immediate , dense environment @xcite . in this paper we revisit the conjecture that non - thermal processes contribute to the total pressure . with this additional pressure , gas can reach hydrostatic equilibrium with a considerably lower gas density that naturally predicts lower star formation rate and bypasses the need for unrealistic supernovae feedback . the non - thermal pressure does not depend on the temperature of the gas , and the gas cools isochorically rather than isobarically , further stabilizing the gas . we develop a simple , easy to use , parametric model which allows to study ( analytically and in simulations ) the effect of such non - thermal components on galaxy formation . the enhanced star formation problem is closely related to the general complication of modeling the interstellar medium ( ism ) gas . while the use of a standard , purely thermodynamic equation of state of an ideal gas is justified outside of galaxies in the igm , it becomes less appropriate to use in haloes ( haloes of galaxy clusters exhibit non - negligible magnetic fields ) and even more so in the ism of galaxies . this gas is highly multiphased , and consists of cold , warm and hot gas arranged within atomic and molecular clouds , filaments , and bubbles @xcite . complicated chemistry and dynamics , as well as radiation fields at multiple wavelengths affect the behaviour and interrelation between the different phases . moreover , the dynamics of the gas are strongly affected by non - thermal components , namely turbulence , magnetic fields and cosmic rays ( cr ) . stars , through their formation , evolution and destruction pump energy into the ism by stirring turbulence , emitting high energy particles ( cosmic rays ) and releasing radiation that heats and drives the gas @xcite . gravitational energy also powers turbulence @xcite and heats the gas ( see however * ? ? ? turbulence , could , in principle be accurately followed by pure hydrodynamics . however , modeling turbulence requires high resolution and realistic driving of the turbulence which is still an open question @xcite . the detailed modeling of these effects is the subject of intensive ongoing efforts ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? all these physical phenomena are determind by the relatively small ( @xmath3pc ) scale of observed giant molecular clouds ( gmcs ) , large eddies of turbulence @xcite , and tangled magnetic fields . it is prohibitively challenging to include all the aforementioned effects and small scales in cosmological simulations . effective equations of state for star forming gas are thus constructed , directly pressurizing the thermal component of the gas @xcite . an equation of state for subgrid turbulence has been proposed in @xcite . @xcite proposed an effective eos directly related to star formation rates , and @xcite incorporated turbulent pressure as a sub - grid model of the various phases motivated by @xcite . @xcite showed that sub - grid models for supersonic turbulence can have a large affect on dwarf galaxies . while these attempts artificially pressurize the gas ( as we propose below ) they do not prevent the gas from over - cooling isobarically and still require the unrealistically strong feedback described above . recent attempts @xcite to simulate the effects of cosmic ray pressure on star formation and winds in galaxies have shown that they are able to drive significant outflows and could be efficient in regulating the star formation of star forming galaxies . @xcite separately implemented a two fluid approximation for cosmic rays and gas for single - galaxy simulations , and propagated the cr as a diffusive component with constant diffusion coefficient . @xcite used an mhd code to simulate anisotropic diffusion that preferentially diffuses crs along magnetic fields . in all three implementations the crs are found to drive winds in some cases . these results are encouraging but the simulations are of single , ideal galaxies and the suggested methods can not be easily extended to the evolution phases of the galaxies and cosmological initial conditions . their physics is too detailed and the required resolutions are too fine to be practical in full hydrodynamic cosmological simulations . to date , no simulations exists that include all the physical processes which are important for pressurizing the gas , and no cosmological - scale simulation will likely be able to simulate all this physics in the forthcoming future . in view of all these complications , here we follow a simpler , alternative avenue , and construct an effective eos that mimics some of the main observed characteristics of the non - thermal pressure components . first , we take advantage of the fact that a scale separation roughly exists between the parsec - scale phenomena discussed in the previous paragraph and that of typical observed scales for vertical scale heights of discs which range between @xmath4 for the milky way @xcite and quiescent star forming galaxies to @xmath5 for starbursting high redshift galaxies @xcite . this implies that in the context of galaxy formation simulations , it should be sufficient to resolve the ism on scales much larger than @xmath61 parsec in order to reproduce galactic discs with realistic characteristics . correspondingly we use a coarse grained , effective modeling of the ism . our model bridges the gap between the parsec and kilo - parsec scales . second , we complement this scale separation with an effective equation of state that is straightforward to use . in principle , one can attempt to apply a rigorous treatment of the physical processes that constitute the non - thermal physics as sub - grid models . however , even if realistic such modules were constructed , there remains the problem of stipulating physically - consistant initial conditions : how to seed magnetic fields , how and when cosmic rays are generated and accelerated , and what drives turbulence and precisely on which scales . in addition , one has to relate the initial conditions to star formation which most likely drives these effects . hence , we opt for a simple , easy to use , pressure - density relation for the non - thermal eos which we develop below . a key feature of our implementation is the calibration of the eos by the observed relation between the fir radiation - a star formation indicator - and the radio radiation , which constrains the joint energy content of cosmic rays and magnetic fields . this novel approach provides a quantitative relation between the gas density and the magnitude of the non - thermal components . additional physical assumptions which are required to complete the specific parametrization of the effective eos are then limited to factors of order unity , rather than being arbitrary . the structure of this paper is as follows . in [ sec : eos ] we describe possible modifications for the equation of state of the gas that manifest some important aspects of the non - thermal pressure components . in [ sec : toymodel ] we demonstrate the effectivenes of such modified eos and and the importance of the non - thermal conmponents in general , using simple calculations of a point ( single - cell ) model , focusing on the regaulation of the overall star formation rate . [ sec : ramses ] describes the incorporation of the model into the hydrodynamic code _ ramses _ @xcite and the setup and results of isolated galaxy simulations with and without our model . in [ sec : summary ] we summarize and discuss our results . to incorporate non - thermal pressure we need some typical scale for its density dependence and amplitude . we will use the observed fir - radio relation for a typical value of magnetic field and . the model we describe in this paper assumes that ( i ) the non - thermal components are in equilibrium between themselves in the sense that energy can move quickly between them , and ( ii ) that this equilibrium does not depend on the magnitude of the energy . the first assumption , of strong coupling between the components , is justified because the timescales for interactions between the components are eddy turnaround time for the turbulence , alfvnic crossing time for magnetic fields and diffusion times for cosmic rays , all on scales much smaller than galactic or cosmologic scales . an eddy turnaround time for a @xmath7 eddy rotating at a typical ism speed of @xmath8 would take less than @xmath9 - much smaller than cosmological evolution timescales . the alfvnic crossing time is even shorter : the velocity for @xmath10 and @xmath11 is about @xmath12 the diffusion coefficients for the cosmic rays indicate an even smaller timescale for equilibrium across a @xmath13 with a typical diffusion coefficient of @xmath14 the diffusion over this scale - length will take @xmath15 . we stress that the second assumption is not necessary for the effect of the non - thermal pressure to be important , and that we use it below for the sake of simplicity . there are , nonetheless , strong qualitative arguments which motivate such strong coupling . compelling physical arguments can be made in favor of the interrelation between turbulence energy and tangled magnetic field energy and cosmic rays . turbulence and magnetic fields are naturally related since turbulent flow can increase the energy in magnetic fields ( by elongating and wraping the field lines ) , while large magnetic fields tend to rearrange and freeze the material in order to decrease the length of the flux tubes ( see * ? ? ? * for a discussion including the dependence on magnetic field geometry and turbulence velocity ) . either way , energy is naturally converted from one component to the other . we assume that this qualitative argument holds , even though the dominant coupling process and the precise energy distribution between the components might vary somewhat ( see discussion and references in * ? ? ? * who also assumes equal energies in turbulence and magnetic fields for starbursting galaxies ) . equipartition between magnetic and cosmic rays should also be natural ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * see however @xcite ) . cosmic rays are relativistic electrons and protons ( most likely accelerated during supernavae explosions ) , and travel along magnetic fields which confines them to the galaxy with some effective diffusion regulated by the fields . strong magnetic fields increase cosmic ray energy loses through synchotron radiation , and reduce the diffusion rate thus increasing the steady - state density of the cosmic rays in the galaxy . the simplest manifestation of such an approach is to assume equipartition between the three non - thermal components , so that the total energy density is three times that of each separate component . equipartition between magnetic fields and cosmic rays corresponds to a minimum of the total energy of cosmic rays and magnetic fields for a given measured synchrotron radio emission @xcite . in cases where the two are not in equipartition the combined non - thermal pressure due to the two components will be larger . finally , the cycle is completed by instabilities in the cr flows along magnetic fields giving rise to small scale turbulence @xcite . it is worth mentioning that an equipartition between these three components are also observed in intra - cluster medium ( icm ) where they each contribute about @xmath16 of the total pressure , while the bulk of the pressure comes from the thermal component of x - ray emitting gas . we emphasize that the `` equipartition anzatz '' is basically just a scaling parameter for the effective non - thermal eos ; it is not an essential component of our model and different scalings can be used ( see below ) . once the distribution of energy among the non - thermal components is determined , we can naturally continue to develop an equation of state for them . while a `` proper '' thermal equation of state relates all the thermodynamic variables to two independent variables ( for example , the density and internal energy ) , a non - thermal component is , by nature , a single parameter function . in the case of magnetic fields , for example , if a volume of space occupying magnetic fields is compressed , no heat is generated , and the process can be reversed . hence , the magnetic pressure and energy are functions of only one variable ( the field magnitude , @xmath17 ) which can in turn be related to only one thermodynamic variable . this is completely analogous to the equation of state of cold matter , which is commonly used in the analysis of compact objects . it is important that the concept of entropy does exist in multi - component non - thermal system , through the requirement for equilibrium . we note that the system is not closed , since energy is constantly pumped in by star formation and leaks out of the system by cosmic ray diffusion , electromagnetic radiation , reconnections , and dissipation of turbulence . hence the entropy of the non - thermal component can change while generating the equilibrium configuration . in other words a cold component is not a zero entropy system ( for example , we can not use the adiabatic relation between the work done on an element and the internal energy within it ) . again , in analogy with compact objects , this is the basis for deteminning the composition of high density matter in neutron stars , while allowing for energy loss through the emission of neutrinos . as mentioned above , in the absence of a first - princples model for the relation between magnetic fields , cosmic and turbulence , we simply assume that equipartition exists between these three components . this simplification allows us to evaluate the entire non - thermal pressure based on a relation between one of these components and the star formation rate , and dictating the energy density in the other two components by equating it to the first . in accordance with this approach , we base our derivation on relations between magnetic fields and the star formation rate . specifically , we model the dependence of the magnitude of magnetic field , @xmath17 , on the star formation rate , @xmath18 ( in mass per unit time per unit volume ) as a power law : @xmath19 . combined with a @xcite power law relating the star formation rate to the gas density , @xmath2 , i.e. , @xmath20 we have a simple power law term of @xmath21 for which @xmath22 , @xmath23 and @xmath24 are all constants , which along with the proportionality factors must be constrained from observations . once this assumption has been made , the total non - thermal volumetric energy arising from these power laws takes the form : @xmath25 , @xmath26 with the prefactor of @xmath27 originating from the contributions of the @xmath27 components in equipartition . we reiterate that the equipartition is not a necessary assumption for this model . any constant distribution between the components is consistent with the assumptions and can be readily used . the effective equation of state can easily be modified for further deviations from the``equipartition anzatz '' , and any non - thermal pressure that is a monotonic function of density can be incorporated in an analogous way . essentially , the only requirement that can not be simply generalized is that the non - thermal components are in quasi - steady state that depends on the gas density alone . that is , that the non - thermal processes are related , and settle down on timescales which are short with respect to the evolution of galaxies . we begin with the simplest model that incorporates the additional non - thermal pressure components . in this model we assume that the energy in the non - thermal components is completely determined by the local instantaneous star formation rate . stipulating this assumption the energy in the non - thermal components becomes a simple function of @xmath2 . the function is determined by various physical processes that contribute to the non - thermal pressure of the ism , which depend differently on the density of the gas , but eventually materializes through equipartition . at this point we will also assume that the gas is always in appropriate conditions so that star formation is turned on . assuming a power - law dependence of @xmath17 ( equation [ eq : b_rho_alphaprime ] ) and a schmidt law for the star formation , the non - thermal pressure in equilibrium , @xmath28 , is : @xmath29 where @xmath30 is a model dependent proportioanlity factor . here , and throughout the paper , we neglect the order - of - unity differences between pressure and energy density . magnetic field s pressure depends on the field s morphology and drops from @xmath31 for magnetic field in the disc s plane , to @xmath32 for isotropically tangled field . cr pressure equals @xmath33 of its energy density for non - relativistic particles . we show below that the while the detailed analysis changes somewhat , the qualitative effect of the non - thermal pressure remains unaffected . the total pressure at each point is the sum of the thermal and non - thermal pressure , @xmath34 . as an aside we mention that the sound speed of the gas is striaghtforward to calculate in this effective eos : for application in numerical codes , it is useful ( for setting the timesteps according to the courant conditions , for example ) to calculate the numerical speed of sound of gas : @xmath35 we note that it is not the physical speed of sound of the multi - component gas , which depends on the alfvnic velocity and largest eddy velocity in a non - trivial manner . the eos described in the previous subsection has the advantage of being extremely simple to implement since it requires a trivial addition to the ideal equation of state depending only on the gas density ; there is no need to specifically trace the non - thermal component . however , it suffers from undesired consequences that arise from the assumption that the non - thermal pressure traces the local instantaneuos star formation . this means that if star formation was to suddenly begin ( by passing the a threshold density , for example ) or to suddenly end ( perhaps if feedback blowing of the gas transfers it to a different thermodynamic regime where star formation is extinguished ) a sudden jump in the pressure will follow , and potentially create spurious shocks and disturbances . in addition , we know from observations that the magnetic field and cosmic ray vertical scale heights are considerably larger than those of the gas and star formation . @xcite estimates the magnetic scale height of the galaxy at @xmath36 based on rotation measures of pulsars @xcite and the cr scale height of @xmath37 based on observed abundance of secondary particles such as @xmath38 that are used to constrain various diffusion models for the galaxy @xcite . the magnetic field of distant galaxies is also observed via polarization and radio emission measurements at a vertical scale of a few kiloparsecs above the disc plane ( see , for example * ? ? ? * ) . the existence of non - thermal pressure based on radio observations and on a discrepancy between the vertical density and gravity profiles of the galaxy perpendicular to the plane at the local neighborhood was advocated by @xcite . the fact that magnetic fields and cosmic rays at regions that are far from star forming gas indicate that the coupling between the non - thermal components and star formation is more complicated than the simple assumptions in [ sec : ssnteos ] . specifically , this suggests that the coupling is not instantaneous , but has a finite response time as energy convects with the gas or diffuses through it . alternatively , there may be additional sources for turbulence , magnetic fields and cr which are dominant at @xmath39 altitudes above the disc ( see , for example , @xcite for extra - galactic driving of turbulence and @xcite for internal ism instability - driven turbulence ) . to address this complication we introduce another independent variable into the equation of state so that the amount of non - thermal energy responds to star formation over a finite time . we set @xmath40 ( equation [ eq : pnt ] ) as the equilibrium value of non - thermal pressure for a given stellar density , and add a time - dependent form of the actual non - thermal pressure , @xmath41 , which approaches @xmath40 through some temporal dependence . this adjustment expands the stationary eos and includes a time integrated function of the star formation rate in the non - thermal pressure , thus ensuring both temporal and spatial continuity even if star formation flicks on and off . keeping with the spirit of our model , we do not attempt to describe the physics of the relaxation of the non - thermal component , and settle for a parametric description . we note that this formulation also removes the numerical complications which arise from discontinuities ( in the latter sense , this additional term has a stabilizing effect similar to the von - neumann artificial viscosity which was introduced to help integrate over the non - smooth shock conditions ) . we accomplish this by paramerterizing the non - thermal heating and cooling rate . since the two must cancel each other for a steady - state star formation , @xmath40 must be an attractor of @xmath41 at a any given mass density : if @xmath42 is too large then there should be a net cooling and vice versa . non - thermal heating can be described by : @xmath43 where @xmath44 is the total energy injected into the gas per supernova , @xmath45 is the number of supernovae per solar mass of stars that are created , and with @xmath46 being the smaller than unity fraction of the supernova energy that ends up in the non - thermal components . non - thermal cooling is assumed to be responding to heating by the following characterization : @xmath47 where @xmath48 is a free parameter which essentially controls the response time of the non - thermal components to changes in the star - formation rate . the resemblance between the cooling term and the heating term arises from the requirement that the gas be in cooling / heating equilibrium at a star formation rate consistent with observations . combining these heating and cooling terms the time - dependent evolution of the non - thermal pressure at a constant gas density follows the simple form : @xmath49.\label{eq : p_nt_oft}\ ] ] stability requires that : @xmath50 so that @xmath51 ensures that the non - thermal pressure always approaches its asymptotic value for a steady star formation rate . the time dependent modification makes it possible to explicitly deal with a density threshold condition , as observed by @xcite . this condition cuts off star formation completely for gas densities below a critical value , @xmath52 . we note that most numerical codes apply such a threshold ( but for considerably lower threshold densities ) also in order to prevent spurious star formation from occurring outside of galaxies . below this threshold , the steady - state non - thermal pressure is expected to vanish . however , in reality , a non - star forming region can still maintain a steady state non - thermal pressure due to diffusion processes from neighboring regions @xcite , or to various other sources @xcite . we partly account for that within our `` single - cell '' framework by setting a finite decay rate for non - star - forming regions . in eq . ( [ eq : cooling1 ] ) , we can not set @xmath53 for @xmath54 , since the cooling rate then becomes ill defined . we remedy this by formulating @xmath40 as a function of @xmath2 ( rather than @xmath18 ) and redefining ( eq . ( [ eq : pnt ] ) ) as follows : @xmath55 using eq . ( [ eq : heating ] ) and eq . ( [ eq : cooling1 ] ) now assures that the heating turns off when no star formation occurs , and the cooling can proceed as the non - thermal pressure asymptotically approaches @xmath56 . combining eq . ( [ eq : p_nt_oft ] ) with eq . ( [ eq : sfr ] ) then yields : @xmath57.\label{eq : p_nt_dot}\ ] ] there is a noteworthy simplification in our model . in order to achieve coarse - grained steady state of ism gas two conditions must be met simultaneously : the total pressure must balance the external pressure , and the cooling must balance the heating . for purely thermal pressure , these two equations are solved by varying two parameters - the density and temperature of the gas . for gas with purely non - thermal pressure components of the type proposed in this work ( which is a reasonable approximation for many external pressures , see fig . [ fig : extp ] ) , the pressure is a function of density alone , and the pressure equilibrium and heating / cooling equilibrium generally do not have a simultaneous solution . we bypass this by relating the heating to the cooling in such a way that the observed relation is always achieved . more advanced models which include physically motivated cooling and heating will introduce more dynamic parameters , allowing the gas to reach steady state more naturally . we now demonstrate the proporties and applicability of our effective eos for non - thermal components with a point ( zero - dimensional ) model of the ism . in this model we evolve the conditions of a parcel of gas with isobaric boundary conditions , solving both thermal and non - thermal pressure components , in accordance with the models described in [ sec : eos ] . a quantitative implementation of our eos requires the speficiation of the model s free parameters . for the stationary non - thermal pressure , these are the proportionality coeffecient and power which relate gas density to the pressure in magnetic fields ( @xmath58 ) . even after applying our hypothesis of equaipartition among the non - thermal components , current uncertainties regarding the magnitude of magnetic fields in early galaxies , are quite large . in essence , @xmath59 may be treated as free parameters . some indication can , however , be gained from observed relations between star formation rates and synchrotron radiation @xcite . we chose to use the fits from equation a11 of @xcite of the form : @xmath60 where @xmath61 and @xmath62 are galactic gas column densities and galactic scale - heights of galaxies . these two parameters are fitted to observations using a one zone model for galaxies including the cosmic ray specra of primary and secondary rays , tracing self - consistently generation and evolution with effective diffusion coefficients @xcite . the power law coefficient @xmath63 , and normalization @xmath64 , are observationally constrained by the frc ( far ir - radio correlation * ? ? ? * ; * ? ? ? * ) and by local measurements of cr and radio observations at @xmath65 ( see * ? ? ? * and reference within ) . a model for the turbulent amplification rate of magnetic fields by star formation that recovers the frc and predicts its breakdown was also suggested by @xcite . stipulating a typical vertical height for the magnetic fields of @xmath5 @xcite one finds for the two fits suggested by @xcite : @xmath66 these values are consistent with with observed milky way values @xcite and with theoretical predictions @xcite . we convert those relations to non - thermal energy according to @xmath67 ( again , recalling that the factor of 3 arises from the equipartition assumption ) and find two similar ( but not identical ) realizations for the non - thermal eos : @xmath68 according to @xcite these fits reproduce the fir - radio relations of galaxies equally well . one could consider a case when the magnetic pressure scales linearly with density ( corresponding to the first fit ) , or to the star formation ( corresponding to a fit with @xmath69 or @xmath70 ) - but this parameter is not favored by the more detailed model there . as we point below , in our model , the @xmath71 fit is singular in the sense that its evolution can never derail it from static equilibrium once it has achieved . for this reason , it is worthwhile to keep both fits at this stage . it is encouraging to note that these values are in rough agreement with the values needed to support the weight of the gas at the plane of the galactic disc against its self gravity @xcite . other tracers of star formation could in principle be used to calibrate and constrain the non - thermal pressure terms . x - rays are a good tracer for star formation as young ob stars emit x - rays @xcite and at high redshifts @xcite . however , x - rays are converted by neutral gas efficiently into uv and optical , and are never important as radiation pressure . further , x - rays are not a direct tracer for other sources of pressure of in the gas such as the amplitude of magnetic fields , turbulence or cr . measurements of turbulence by emission line broadening could , in principle , be used to calibrate the relation between star formation rate and the non - thermal components . however , such measurements inevitably probe only some of the ism gas phases , and measurements of high - redshift turbulence @xcite are few and might be driven by infalling gas rather than by star formation @xcite . alternative direct measurements of magnetic fields either trace the large scale magnetic fields of galaxies ( polarization measurements ) or the total line - of - sight magnetic fields ( faraday rotation measures ) and are harder to correlate to total star formation than the synchrotron radiation used in this work . for the dynamic non - thermal eos several additional parameters are required to define equations eq . ( [ eq : heating ] ) and eq . ( [ eq : cooling1 ] ) . the supernova energy @xmath44 and the supernova rate @xmath45 can by taken from standard theories , but we do need to specify the parameters which control non - thermal heating and cooling , @xmath46 and @xmath48 . the fraction @xmath46 sets the fraction of supernovae energy invested as non - thermal energy , while @xmath48 sets the power governing the rate at which the non - thermal energy reaches its equilibrium values . these values are numerical by nature and should be set to allow the non - thermal energy to achieve equilibrium , while smoothing over pressure jumps arising from abrupt changes in the star formation rates . as we show below , even for a low @xmath46 of @xmath72 the relaxation times for a wide choice of @xmath48 are shorter than a @xmath73 this value indicates that for smooth galactic histories the calculated state of the gas should be close to the asymptotic conditions constrained by observations . we incorporate our model for non - thermal pressure in a single cell model by tracing the evolution of gas under isobaric boundary conditions . this represents a simplified behaviour of a single hydrodynamic cell embedded within a galaxy that evolves slowly and supports this cell with nearly constant external pressure . for the general , dynamic case we solve the ordinary differential equations for the thermal internal energy of the gas and for the non - thermal pressure : @xmath74+p_\text{nt}\frac{\dot{\rho}}{\rho}\label{eq : pntdot}.\end{aligned}\ ] ] the last term in the right hand side of both equations is the contribution of the density change , @xmath75 , ( a @xmath76 term for the energy equation ) . the density is derived self - consistently by requiring that @xmath77 we complement our model with an appropriate paramerterization of the star formation rate and the gas heating and cooling functions . star formation is modeled with a schmidt law corresponding to a convention of @xmath78 of the gas into stars every dynamic free - fall time of the gas : @xmath79 for the sake of simplicity we begin with this star - formation rate with no density cutoff ( we examine the implications of such a cutoff in section [ sec : cutoff ] ) . the equation of state for the ideal gas implies @xmath80 in this work we allow the gas to cool according to the rates @xmath81 from cloudy ( version 96b4 * ? ? ? * ) by interpolating from tables described in @xcite . the cooling and heating of the gas includes compton heating and cooling , redshift dependent uv heating and atomic and molecular cooling . the tables provide the total cooling and heating and particle number as a function of the redshift , metalicity , density and temperature . the temperature is related to the internal energy by integrating over the particle number dependent heat capacity as the number of particles changes by a factor of a few at recombination and at molecular formation . for supernova heating we assume that one supernova occurs for every @xmath82 of stars formed ( @xmath83 ; see eq . ( [ eq : heating ] ) ) , corresponding to a salpeter imf between @xmath84 and @xmath85 and supernovae occurring above @xmath86 @xcite . we use a standard value for the average total energy released per supernova , @xmath87 . in most of the simulations described below we impose external pressure boundary conditions of @xmath88 in rough correspondence with observed conditions in the plane of the galactic disc @xcite . for completeness we list the various definitions and default values for the coefficients in our equations in table [ tab : values ] . r r r | l parameter & units & value & definition + + + + @xmath89 & & 5/3 & adiabatic constant + @xmath90 & & 0 & redshift + @xmath91 & @xmath92 & 1 & metalicity + @xmath93 & & 0.05 & star formation efficiency + @xmath45 & & 1/160 & supernova per stellar mass formed + @xmath94 & erg & @xmath95 & supernova energy + @xmath96 & & 0.1 & fraction of energy injected to thermal component + + + + @xmath46 & & 0.2 & fraction of energy injected to thermal component + @xmath97 & & & density power law coefficient of non - thermal pressure ( eq . ( [ eq : pnt ] ) ) + @xmath30 & see eq . ( [ eq : pnt ] ) & & normalization of non - thermal pressure + @xmath48 & & & cooling behaviour ( eq . ( [ eq : cooling1 ] ) + + + + @xmath98 & @xmath99 & @xmath100 & external pressure + @xmath101 & @xmath99 & & initial non thermal pressure + @xmath102 & k & @xmath103 & initial gas temperature + [ tab : values ] time evolution of the temperature ( _ top panel _ ) and density ( _ bottom panel _ ) of the thermal - pressure - only models for varying supernovae efficiencies . the pressure boundary conditions is @xmath104 and the initial temperature of the gas is @xmath103k . , width=316 ] time evolution of the star formation rates and the depletion of gas into stars ( @xmath105 ) for the model calculated in figure [ fig : thermal_trho ] . , width=316 ] we start by demonstrating the properities of the single - cell simulation when only thermal pressure exists . we set the initial conditions with a temperature of @xmath106k , which corresponds to a density of @xmath107 figure [ fig : thermal_trho ] demonstrates how cooling of the gas causes the temperature of the gas to decrease ( _ top panel _ ) , and , correspondingly the isobaric boundary condition forces the density to scale as @xmath108 forcing a density increase ( _ bottom panel _ ) . the bump at @xmath109 corresponds to the steep decrease in the cooling function at @xmath110k as gas becomes neutral and collisional excitation of lines becomes unimportant beyond this point . greater densities enhance the star formation rate , as well as the resulting supernova feedback , and once the supernova feedback power balances the cooling rate the density and temperature of the gas become constant and the gas is converted into stars at a constant rate . this is can be seen clearly in fig . [ fig : thermal_rhodot ] , which depicts the specific star formation rate and the depletion of gas into stars . this depletion ( fig . [ fig : thermal_rhodot ] , _ bottom panel _ ) is calculated by noting that the mass in stars evolves as : @xmath111 where @xmath112 and @xmath113 are the total , gas and stellar mass in our volume element , and @xmath114 and @xmath2 the time - dependent volume and density of the element . in the single cell model @xmath115 is fixed . as gas is converted to stars the volume adjusts itself so the pressure corresponds to the external boundary condition . this single cell assumption is self - consistently addressed in the full hydrodynamic implementation shown below ( [ sec : ramses ] ) . initially , @xmath116 with @xmath117 and @xmath118 the initial volume and gas density , respectively . @xmath119 is an integral equation that can be evolved in time . the depletion of gas is then shown as : @xmath120 as is to be expected , once the density levels off at an equilibrium value , so does the specific star formation rate ( the volume of the element continues to decrease over time in order to maintain a constant gas density with a decreasing mass ) . our key observation is that without non - thermal components the gas achieves an equilibrium between the heating and the cooling after less than @xmath121 and then converts most of the gas into stars quickly after that . about @xmath122 of the gas is depleted during the first @xmath123 for low supernova efficiencies , and even when assuming perfect ( @xmath124 ) supernovae effeciencies , star formation has comsumed over one half of the gas by @xmath125 . absolute efficiency is certainly non - physical , since in reality most of the supernova energy gets converted into radiation that escapes the galaxy without contributing to pressure support of the gas . in any case , we conclude that in a thermal - pressure - only model , boundary conditions corresponding to the pressure in the mid - plane of the milky way leads to gas being converted into stars over a few @xmath123 regardless of the efficiency of the thermal feedback . it is also noteworthy that since the gas achieves rate equilibrium during the first @xmath126 , the initial conditions of the gas do not affect the depletion time . in figure [ fig : thermal2_rhodot ] we repeat the exercise with an initial temperature for the gas of @xmath127k , and the results are virtually unchanged , except that cooling / heating equilibrium is achieved after as little as @xmath128 . we note that the unphysical @xmath124 case reaches a different equilibrium point that exists on the molecular cooling branch at a lower temperature and leads to even faster gas depletion . same as figure [ fig : thermal_rhodot ] , but when the initial gas temperature is @xmath127k . , width=316 ] we now proceed to examine the behaviour of a parcel of gas with similar boundary conditions as in [ sec : thermal ] , but with additional non - thermal components , evolved according to eq . ( [ eq : pntdot ] ) . figure [ fig : ism_trho ] shows the evolution of a parcel of gas in terms of temperature , density and non - thermal pressure , again with a pressure boundary condition of @xmath104 and and initial temperature of @xmath103k . for the non - thermal pressure we use the parametrization described in eq . ( [ eq : powerlaw1 ] ) . [ fig : ism_rhodot ] describes the evolution of the specific star formation rate and gas depletion for the same model . in all the simulations here the thermal supernova feedback is turned on with efficiency @xmath129 as described in [ sec : thermal ] , and the fraction of supernova energy that is injected into the non - thermal component here is @xmath130 . the green line shows the stationary ( @xmath131 ) non - thermal eos described in [ sec : ssnteos ] ( calculated by replacing eq . ( [ eq : pntdot ] ) with eq . ( [ eq : pnt ] ) ) , and the blue , cyan and gray lines are for the dynamic non - thermal eos ( [ sec : dnteos ] ) with the relaxation power laws of @xmath132 and @xmath133 as indicated on the plot . the blue and cyan lines correspond to models where we arbitrarily set a zero initial non - thermal pressure , @xmath134 this initial condition results in initial density of @xmath135 as for the thermal case , whereas the gray line corresponds to an initial @xmath136 which is in its steady - state value for an initial density , @xmath137 . note that this value is only slightly below the initial value when non - thermal pressure is neglected . we note that the green and gray lines are identical : the power - law in eq . ( [ eq : powerlaw1 ] ) is @xmath138 and when the gas is in equilibrium ( @xmath131 ) its evolution according to eq . ( [ eq : pntdot ] ) is just @xmath139 so it can not evolve away from equilibrium once initially achieved . we shall show below that the static evolution deviates from equilibrium initial conditions case when @xmath140 ( equation [ eq : powerlaw2 ] ) . time evolution of the temperature ( top panel ) density ( middle panel ) and non - thermal pressure ( bottom panel ) of the gas . curves correspond to no non - thermal pressure ( red ) , the steady - state non - thermal pressure ( green ) , and dynamic non - thermal pressure ( blue , cyan and gray ) ; see text for detail - but note that in this case the green and gray lines overlap completely . in all calculations the pressure boundary conditions is @xmath104 and the initial temperature of the gas is @xmath103k.,width=316 ] time evolution of the star formation rates and the depletion of gas into stars ( @xmath105 ) for the models calculated in figure [ fig : ism_trho ] . , width=316 ] the two figures clearly demonstrate the distinct effect that non - thermal pressure has on the simulation . for the initial conditions we set , the gas is initially supported ( at least in part ) by thermal pressure . as the thermal energy is radiated away , temperature drops and the gas contracts , increasing the star formation rate . however , the inclusion of non - thermal pressure removes the relation of @xmath141 , and introduces another degree of freedom . the gas can then cool without a dramatic density increase , and so cooling does not necessarily lead to enhanced star formation . in the calculations with the dynamic eos the non - thermal component adjusts ( increases ) until a new stationary equilibrium is reached . this equilibrium consists of a balance between supernovae feedback and cooling both for the gas and the non - thermal energy components , each separately . note that in all the simulations presented here the gas settles into this steady state in a few @xmath126 . the shape of the the curves found with the dynamical eos also deserves some elaboration . since the asymptotic non - thermal pressure is similar in all these cases , all trajectories with non - thermal pressure converge to the same values . as gas contracts , its star formation and supernova rate increases , and , for the dynamic non - thermal eos , it takes some time for the non - thermal reservoir to fill . during this time the gas is actually under - pressurized with respect to its asymptotic values and the density is larger than its final value . this overshoot is readily seen in the temperature and density of the gas of fig . [ fig : ism_trho ] and in the specific star formation rate of fig . [ fig : ism_rhodot ] . the _ bottom panel _ of fig . [ fig : ism_trho ] shows the gradual and monotonic increase of @xmath42 . the timescale for converging to the asymptotic value is set primarily by @xmath142 and slightly depends on the relaxation power law @xmath143 in all the runs here the gas quickly settles into a steady state for which the cooling is balanced by the thermal feedback and the total pressure is divided between the thermal component and the non - thermal component . the distinct effect of non - thermal pressure is easily seen by comparing the evolution in all of these calculations to the case in which non - thermal components are neglected , similar to fig . [ fig : thermal_trho ] and fig . [ fig : thermal_rhodot ] ( shown for reference in figure [ fig : ism_trho ] and figure [ fig : ism_rhodot ] in red curves ) . the outstanding feature is the dramatic difference in the asymptotic equilibrium between the two cases : the additional source of pressure allows the gas to cool without a dramatic density increase . hence , the equilibrium density is some @xmath144 times lower for the non - thermal case , and the temperature is @xmath145 times higher . these different conditions lead to very different star formation rates and depletion times as can be observed in fig . [ fig : ism_rhodot ] . once the non - thermal component is included , the equlibrium star formation rate is four order of magnitudes lower ( corresponding to the schmidt law used here that indicates @xmath146 ) . accordingly , the mass depletion time for the non - thermal eos gas increases to @xmath147 as opposed to about @xmath123 when only the thermal component in the pressure is included . this increase in depletion times is related to the lower asymptotic density for this case . since the density approaches its asymptotic value much faster then the depletion time , most of the gas depletion occurs at the equilibrium density . hence the gas depletion time is @xmath148 . for the kennicutt schmidt relation we use here , this leads to a @xmath149 relation , so reducing the density by a factor of @xmath144 leads to a twenty fold increase of the depletion time . we also note the value of @xmath48 has a minor impact on the relaxation time scale in the dynamic models , which is ten to a few tens of myr ( as is to be expected , the model with @xmath150 has a shorter relaxation time than the one with @xmath132 ) . in order to examine the robustness of the effects of the non - thermal eos we repeat the calculations described here for ( i ) the parameters in eq . ( [ eq : powerlaw2 ] ) and ( ii ) for a non - thermal component weaker by a factor of @xmath27 and @xmath145 ; @xmath151 instead of eq . ( [ eq : powerlaw1 ] ) . the results are described in fig . [ fig : ism_powerlaw2 ] and fig . [ fig : ism_equipartition ] , respectively . same as figure [ fig : ism_rhodot ] , except that the non - thermal pressure is calculated with equation ( eq . ( [ eq : powerlaw2 ] ) ) . , width=316 ] time evolution of the star formation rates and the depletion of gas into stars ( @xmath105 ) for various magnitudes of the non - thermal pressure based on the fit in eq . ( [ eq : powerlaw1 ] ) ) normalized to the full equipartition magnitude , width=316 ] in contrast with the previous calculations , setting @xmath140 in eq . ( [ eq : powerlaw2 ] ) causes the dynamic equation of state to slightly deviate from the static equation of state even when both calculations begin with initial conditions of @xmath152 the actual difference in the evolution between these two cases is still small in comparison with the difference between them and calculations which begin with @xmath134 , for which the gas density overshoots significantly , and peaks after about 0.5myr ( the dynamic calculation with a finite initial non - thermal pressure , shown in gray , does over shoot with the same time scale , but at a much smaller amplitude ) . finally , fig . [ fig : ism_equipartition ] shows that even when the non - thermal pressure is reduced by a factor of @xmath145 , the gas depletion time is still a factor of five or so longer than the depletion time for any of pure thermal calculations with realistic thermal feedback effeciencies ( see [ sec : thermal ] ) . only a perfect thermal efficiency @xmath153 allows for a depletion time that is comparbale to the case when a weak non - thermal component is included . this result emphasizes that non - thermal pressure is far more efficient in delaying gas depletion to star formation than enahncing thermal feedback from supernovae . the time scale for relaxation in this weaked non - thermal pressure case is reduced , however , to one to a few myr . our main conclusion is that any significant non - thermal pressure will inevitably lead to a large change in the gas depletion time when compared to pure - thermal pressure models . we infer that this is a general consequence of non - thermal pressure , regardless of whether equipartition is assumed . same as figure [ fig : ism_trho ] , but with density dependent star formation cutoff introduced at @xmath154 . , width=316 ] the gas density ( _ top panel _ ) and the fraction of the non - thermal pressure from the total pressure ( _ bottom panel _ ) as a function of the external pressure boundary condition for the parameters described in table [ tab : values ] . below @xmath155 the density is always the cutoff density for star formation ( @xmath156 here ) . above this pressure the density increases as a power - law . , width=316 ] the existence of a star formation density threshold is predicted by @xcite and is present in essentially all numerical models of galaxy formation . it is typically implemented by invoking a single numerical value , set to compensate for the inability to simulate star formation and reach the necessary ( high ) densities . moreover , the thershold is applied to prevent spurious star formation outside of galaxies . in this subsection we study the response of our model to an inclusion of such a fiducial threshold . we use a threshold value of @xmath157 which is typical in cosmological simulations . we emphasize that this is a qualitative demonstration of the effect which is included self - consistently in any numerical simulation , such as those described below , in [ sec : ramses ] . figure [ fig : ism_sfcutoff ] describes the temporal evolution of a gas parcel for a star formation law that includes a sharp cutoff for densities below @xmath158 this threshold is higher than the equilibrium density calculated without the threshold ( fig . [ fig : ism_trho ] ) so there is no strict equilibrium solution ( i.e. static solution ) for this case in which the cooling balances the heating at all times . instead , we find that the qualitative behaviour of the system is such , that the response time of the non - thermal components creates a cycle in which star formation flickers on and off and the time - averaged heating balances the continuous cooling . by construction , the single cell model is clearly inadequate for a quantitative study of this duty cycle , because it coarse grains over the relevant spatial and temporal scales necessary . we do confirm numerically that our integration does indeed flicker . it is noteworthy that applying the density cutoff does not imply that the star formation in a full simulation in a galactic ism will occur at constant density . non - homogeneity in the ism is expected ( see for example * ? ? ? * ) and implies that the external pressure boundary conditions should vary in space and time . we demonstrate that this principle applies also in the case of non - thermal pressure by examining the dependence of the equilibrium density and of the star formation rate on the external pressure conditions . we vary the latter and solve the equlibrium density and non - thermal pressure for the non - thermal relations in eq . ( [ eq : powerlaw1 ] ) and eq . ( [ eq : powerlaw2 ] ) . our results are presented in fig . [ fig : extp ] . we find that the density cutoff imposes a transition that depends on the external pressures : for low external pressures , the equilibrium density settles at the cutoff density for star formation as described above . however , for @xmath159 star formation becomes possible and the equilibrium density is larger ( see the top panel of fig . [ fig : extp ] ) . we note that the non - thermal pressure dominates for practically any external pressure above this tansition value ( lower panel in fig . [ fig : extp ] ) , so the density in this regime essentially scales as @xmath160 . it is encouraging to note that for pressures that correspond to the plane of the galaxy ( @xmath161 ) the relative contribution of the thermal component is a few per - cent , which is in agreement with observations ( see fig . 2 of * ? ? ? * and accompanying text ) . we test our model by implementing it on isolated spiral galaxy simulations ran on _ ramses _ @xcite . in the following section we will describe in some detail the non - trivial aspects of our implementation ( [ sec : implementation ] ) , the simulations that were ran ( [ sec : ramses_ic ] ) , and describe the effects of non - thermal feedback on the star formation history and on the morphology of the resulting galaxies ( [ sec : ramses_results ] ) . we now describe the numerical methods we have used to solve for the euler equations in presence of a non - thermal energy components . the original equations have to be modified by adding to the total fluid energy the non - thermal energy and to the total pressure the non - thermal pressure . the modified equations now read @xmath162 @xmath163 @xmath164 @xmath165 the total fluid energy is now defined as @xmath166 and the total fluid pressure as @xmath167 where the thermal pressure is given by the eos of the thermal component @xmath168 and the non - thermal pressure by the eos of the non - thermal component @xmath169 for the thermal component only , we can also define the specific thermal energy @xmath170 as @xmath171 we see in the previous equation that the internal energy of the thermal component is obtained by subtracting from the total energy the other energy components , namely @xmath172 in the hydrodynamics solver , we have to modify several components of the code to add this non - thermal energy variable . first , the predictor step in our muscl scheme @xcite is augmented by an additional equation for the non - thermal pressure . second , in the same predictor step , the non - thermal pressure is added to the thermal pressure in the equation governing the velocity update . the next important correction is for the riemann solver , used to define the flux at cell interface , as a function of the left and right states interpolated with in space and time at the interface position . we have modified our various approximate riemann solver by just replacing the fluid energy and the fluid pressure by the total energy and the total pressure . the sound - speed in the augmented hyperbolic system of ( quasi- ) conservation laws has to be modified as @xmath173 we have tested successfully our new algorithm on simple shock tubes featuring the additional non - thermal energy . an important point we would like to stress is that in our simple model , shock heating occurs only for the thermal component . since in the previous set of equations , there is no sources of non - thermal energy at shock fronts , and no coupling between the two energies , the evolution of the non - thermal component is strictly adiabatic . for our non - thermal pressure model described in this paper we allow for one additional scalar component of non - thermal energy with @xmath174 and add a source term for it to the feedback routine according to the heating portion of eq . ( [ eq : p_nt_dot ] ) and a sink term to the baryonic cooling routine according to the cooling portion of that same equation . feedback in hydrodynamical simulations is typically ineffective in regulating star formation and reducing gas depletion times . to overcome that , a combination of methods , all `` pumped up '' to be as efficient as physically possible , is used . we apply here the standard tools used in the _ ramses _ runs of the agora @xcite isolated galaxy . our base - line simulation , `` standard sim '' uses the standard tools used in ramses which include several ad - hoc measures calibrated to prevent reaching high density and short depletion times . these methods include delayed cooling ( preventing cooling for a period of time after feedback energy injection to account for the adiabatic phase of the sedov - taylor explosion ) . additional methods applied is to increase the stochasticity of the process by allowing feedback to operate according to a poisson distribution with a typical mass scale of a giant molecular cloud ( gmc ) and , although not strictly motivated by feedback , incorporating a pressure floor for the ism gas that prevents it from reaching extremely cold and dense states which would imply very large star formation rates . we aim here to demonstrate that the parameters of these methods can be relaxed once our non - thermal feedback model is used . for comparison we consider two more simulations . in the first , delayed cooling is turned off and the simulation indeed exhibits an over - production of stars . in the third simulation we introduce our non - thermal feedback and demonstrate its ability to reduce star formation without the delayed cooling model . the three simulations are defined in table [ tab : sims ] . the simulations were ran using the agora low resolution initial conditions for a milky way - like galaxy . the setup consists of a dm halo of @xmath175 , stellar disk of @xmath176 and a gaseous disk of @xmath177 the maximal refinement level is @xmath178 , with a box size of @xmath179 and a maximal resolution of @xmath180 particle mass was @xmath181 for the stellar component and @xmath182 for the halo dark matter particles . execution time until a physical time of @xmath183 on @xmath184 cores on the icpl cluster took a few days . in all simulation the gmc mass is @xmath185 and the pressure floor is defined as a minimal temperature of @xmath186k . for the standard feedback simulation the delayed cooling timescale is @xmath187 for the non - thermal feedback simulation the amount of energy that is injected into the non - thermal component is half of the energy that is injected into the thermal part , the asymptotic value for the non - thermal component was set by eq . ( [ eq : powerlaw1 ] ) and the rate of energy dissipation was set to @xmath188 ( eq . ( [ eq : cooling1 ] ) ) . r r r name & delayed cooling & non - thermal pressure + + standard sim & on & off + weak sim & off & off + non - thermal sim & off & on + [ tab : sims ] density colormap of the central @xmath189 of our `` standard sim '' simulation at @xmath190 . _ top panel _ edge on , _ bottom panel _ face on . , width=316 ] density - temperature histograms of `` non - thermal sim '' . _ panel ( a ) _ presents the thermal component , and _ panel ( b ) _ the non - thermal component at time @xmath191 the temperature of the non - thermal component is defined as the non - thermal pressure divided by the density.,width=316 ] the spatial extent of the non - thermal pressure at @xmath191 the colormap represents the ratio between the non - thermal pressure to the total pressure.,width=316 ] some results of the simulations described in [ sec : ramses_ic ] are presented in fig . [ fig : mw2a]-fig . [ fig : sfrcomp ] . fig . [ fig : mw2a ] shows a head - on and edge - on view of `` standard sim '' . the simulation produces a stable rotating disc , and strong outflowing winds that are noticeable as the rough `` x '' shaped overdensity extending diagonally to the edges of our @xmath189 box . we note that in all these simulations there is no hot halo gas component , that is expected to strongly affect outflows . [ fig : histnt ] presents a density - temperature histogram of our `` non - thermal sim '' . the top and bottom panels show the thermal and non - thermal components . for the non - thermal component we define the effective temperature as the non - thermal pressure divided by the density . in the histograms , isobaric lines would be represented by lines with slope of @xmath192 ( @xmath193 ) and the pressure floor is the sharp cutoff diagonal through the lower right part of the plot with slope of @xmath194 the non - thermal temperature is typically higher than the thermal pressure . for the formulation of the non - thermal pressure used here ( eq . ( [ eq : powerlaw1 ] ) ) , this asymptotic temperature is independent of density and is simply @xmath195k , with @xmath196 the proton mass and @xmath197 boltzmann s constant . for high densities , when the non - thermal cooling is efficient enough for gas to relax into the asymptotic value , this value roughly coincides with the effective temperature for the non - thermal component . however , the scatter due to the stochasticity of the feedback process is very large . [ fig : ppnt ] shows an edge - on view of the galaxy with the colormap representing the fraction of the non - thermal pressure to the total pressure . it is evident that dynamically significant non - thermal pressure exists around the disk at a distance of a few kiloparsecs , consistent with magnetic field observations @xcite . density - temperature histograms of the three models for feedback at time @xmath191 _ panels ( a ) , ( b ) and ( c ) _ show results from `` standard sim '' , `` weak sim '' and `` non - thermal sim '' respectively . the black ellipses point to the star forming region on the n - t plot.,width=316 ] star formation histories ( _ top panel _ ) and star formation rates ( _ bottom panel _ ) for simulations with the three feedback models.,width=316 ] fig . [ fig : hists ] compares the thermodynamic state of the gas of the various simulations . stars are formed through the high density branch with a slope of @xmath33 , emphasized by the black ellipse in the plots . while this line is truncated in `` standard sim '' ( _ panel ( a ) _ ) , at @xmath198 is continues beyond @xmath199 for `` weak sim '' ( _ panel ( b ) _ ) which causes an increase in star formation ( see fig . [ fig : sfrcomp ] ) . the third panel is similar to the top panel of fig . [ fig : histnt ] . when non - thermal feedback is added , the amount of high - density gas is reduced . for the non - thermal case ( _ panel ( c ) _ ) it is evident that there is a considerable fraction of the gas at intermediate densities ( @xmath200 ) and @xmath201k temperature that is absent from the two purely thermal simulations . this is a result of the isochoric cooling process that occurs in that case ( corresponding to purely vertical thermodynamic trajectories on the histogram plot ) versus the more efficient isobaric cooling for the thermal cases ( corresponding to diagonal trajectories going down and to the right , with slope of @xmath192 , as explained above ) . the extra non - thermal pressure supports the gas and allows it to cool without contracting to the pressure floor . in `` standard sim '' ( _ panel ( a ) _ ) , most of the gas has been blown out of the galaxy to @xmath202k and @xmath203 which effectively shut down star formation ( see fig . [ fig : sfrcomp ] ) . the non - thermal pressure ( _ panel ( c ) _ ) affects intermediate densities and keeps the gas pressurized with pressure corresponding to @xmath204k , without blowing it out of the galaxy altogether . [ fig : sfrcomp ] shows the star formation history and star formation rates for the three simulations . the most efficient quenching occurs for `` standard sim '' and is a result of massive blowout of gas from the galaxy . `` weak sim '' depletes most of the gas rapidly , within the first @xmath205 of the simulation , demonstrating the overcooling catastrophe that occurs for insufficient feedback . `` non - thermal sim '' exhibits regulated star formation that produces stars at @xmath206 throughout the simulation . not surprisingly , our results indicate that non - thermal pressure can have a significant effect on star formation and prolong the depletion times of galaxies to the observed timescales . clearly further research will be required to quantify the relative importance of the non - thermal components to other feedback mechanisms that are applied . regardless , we emphasize that observational evidence , as well as robust theoretical motivation , point to the existence of this component and that , in one form or another , it should be incorporated in cosmological and galactic simulations . on galactic scales , the interstellar medium exists at quasi - static pressure that is required to support the atmosphere above it . in an equilibrium configuration , loss of pressure due to cooling processes is balanced by heating , which for typical disc galaxies at low redshifts , is dominated by stellar feedback . stellar feedback , through its dependence on star formation rate is related to the ism gas density . it is a well - known result that when only thermal pressure is considered in simulations , the resulting ism density constrained by the pressure and heating - cooling equilibria leads to relatively large star formation rates and short ( @xmath207myr ) gas depletion time scales . this is considerably faster than depletion times of @xmath208 inferred from observations @xcite . in this work we consider the contribution of non - thermal pressure components to this picture . non - thermal pressure consists of turbulence , cosmic rays and magnetic fields , and we examine their impact in an effective model . current cosmological simulations generally do not include the later two , and do not always resolve turbulence . we demonstrate that non - thermal pressure components can be instrumental in solving the depletion time discrepancy in two respects : they reduce the quasi - steady state density and the corresponding star formation rates and cooling times , and they stabilize the gas by adding longer relaxation times in cases where star formation flickers on and off . the regulating effect has been shown previously for cosmic rays @xcite and turbulence @xcite but the two were not considered together and in any case were not yet formulated in a way which is applicable to large scale cosmological simulations . to test our assumptions we construct a simplified physical model for which all the non - thermal components achieve a steady state that is solely a function of density . while simplistic , the advantage of such an approach is that it is readily applicable in numerical simulations . furthermore , we calibrate this density dependence by using the observed relations between the star formation rate for various galaxy observations and the synchrotron radiation , so that the magnitude of the effect is reasonably constrained . to study its effect we first implement it into a single - zone numerical model that traces the evolution of a parcel of star forming gas with varying physical conditions under isobaric boundary conditions that mimic the pressure confinement of the gas by the atmosphere around and above it . using this mode we find that for a given , realistic , thermal feedback the depletion times naturally grow from @xmath209 to @xmath210 in better agreement with observations , and that the coarse grained density of the gas is reduced by several orders of magnitude . then , the model is implemented into the hydrodynamical code _ ramses _ and we present three simulations of the same isolated milky - way - like galaxy with three different physical models . in the first we use some of the `` standard '' recipes generally used for feedback . using that model reduces star formation by blowing the gas to high temperatures and low densities , and expelling it from the galaxy . then , to demonstrate the problem we deliberately turn off one of the key feedback components - the delayed cooling and show that gas cools and accumulates at the numerical pressure floor with high densities that cause a depletion of the stars within @xmath211 - at odds with observed depletion times of @xmath212 in the third simulation we introduce the non - thermal model calibrated by the observed radio fir relation . for that model , the gas remains pressurized at intermediate temperatures and densities , reducing the non - physical low density gas of the first model , and the non - physical high density star forming gas of the second model . this model is effective in regulating star formation for a long period of time ( @xmath213 ) without blowing the gas out of the galaxy altogether . this work is a natural first step in incorporating non - thermal pressure components in galactic scale simulation . the next steps can , and should , pursue several avenues of research . the first is to better model the various non - thermal components , their internal interaction and their interaction with the thermal component and star formation . this would relax the assumption of equipartition , and replace the observational constraints with more physically motivated ones . for this step , calibration against results from ism - scale hydrodynamic simulations will be beneficial . a different avenue to pursue , in tandem or separately , is to use our _ ramses _ patch to run large scale numerical simulations and to demonstrate its effect and applicability on cosmic scales . cosmological simulations today generally do not include the magnetic fields and cosmic rays , and do not always resolve turbulence , and our approach allows to circumvent this difficulty by using a simple effective parametrization . such simulations will naturally include realistic boundary conditions for the ism , namely the halo gas , and allow us to study its interactions with the ism and its effect on winds . ultimately , once large scale cosmological simulations are possible with all the necessary physics , a toy sub - grid model can also be calibrated directly to those simulations and used as a cheaper approximation for them . we thank andrey kravtsov for making the cooling tables available to us for the single cell calculations . computational resources were provided through icpl ( http://icpl.huji.ac.il ) . r. a. , theuns t. , dalla vecchia c. , eke v. r. , frenk c. s. , jenkins a. , kay s. t. , peacock j. a. , pearce f. r. , schaye j. , springel v. , thomas p. a. , white s. d. m. , wiersma r. p. c. , 2009 , , 399 , 1773

the 2df ( two - degree field ) facility at the prime focus of the anglo - australian telescope provides multiple object spectroscopy over a field of view . up to 400 target fibres can be independently positioned by a complex robot . two spectrographs provide spectra with resolutions of between 500 and 2000 , over wavelength ranges of 440 nm and 110 nm respectively . the 2df facility began routine observations in 1997 . 2df was designed primarily for galaxy redshift surveys and has a number of innovative features . the large corrector lens incorporates an atmospheric dispersion compensator , essential for wide wavelength coverage with small diameter fibres . the instrument has two full sets of fibres on separate field plates , so that re - configuring can be done in parallel with observing . the robot positioner places one fibre every 6 seconds , to a precision of 0.3 arcsec ( 20 ) over the full field . all components of 2df , including the spectrographs , are mounted on a 5-m diameter telescope top - end ring for ease of handling and to keep the optical fibres short in order to maximise uv throughput . there is a pipeline data reduction system which allows each data set to be fully analysed while the next field is being observed . 2df has achieved its initial astronomical goals . the redshift surveys obtain spectra at the rate of 2500 galaxies per night , yielding a total of about 200,000 objects in the first four years . typically a b galaxy gives a spectrum with signal to noise ratio of better than 10 per pixel in less than an hour ; redshifts are derived for about 95 per cent of all galaxies , with 99 per cent reliability or better . total system throughput is about 5 per cent . the failure rate of the positioner and fibre system is about 1:10,000 moves or once every few nights and recovery time is usually short . in this paper we provide the historical background to the 2df facility , the design philosophy , a full technical description and a summary of the performance of the instrument . we also briefly review its scientific applications and possible future developments . [ firstpage ] instrumentation : spectrographs techniques : spectroscopic surveys galaxies : distances and redshifts large - scale structure of universe

astro-ph0202175

the value of equipping telescopes with a large field of view has been recognised for some time . in 1986 the royal astronomical society report `` review of scientific priorities for uk astronomical research 19902000 '' @xcite put a wide - field multi - object spectroscopic facility at the top of its priority list for new projects . subsequently the uk large telescope panel recommended that a wide - field survey facility be pursued in tandem with an 8 metre telescope project . the general scientific case for a wide - field spectroscopic facility on a 4 m telescope was two - fold : to provide spectra for large samples of objects found in the multi - colour imaging surveys from the uk , eso and oschin ( palomar ) schmidt telescopes ; and to generate targets for the coming generation of 8 - 10 m optical telescopes . the biggest specific science driver was to obtain redshifts of tens or even hundreds of thousands of galaxies and quasars , to elucidate the three - dimensional structure and evolution of the universe . other major projects required spectra for large samples of stars , to determine their kinematics and composition and hence the dynamical and chemical history of our galaxy ; for similar studies of the magellanic clouds ; and for detailed studies of star clusters and clusters of galaxies . one possibility was to provide a multi - fibre upgrade to the 3.9 metre anglo - australian telescope ( aat ) . this was particularly advantageous for two reasons . firstly , the optical design of the telescope ( a hyperboloidal primary mirror and relatively slow f - ratio of f/3.3 at prime focus ) enabled a wide field of 2 to be achieved using a large but straightforward optical corrector . secondly , the anglo - australian observatory ( aao ) already had extensive experience with multi - object fibre spectroscopy . this dates from the pioneering days of optical fibres and its brass plug - plate system focap @xcite and more recently the fully automated autofib fibre positioner system @xcite . at the end of 1988 the aat board ( aatb ) commissioned a full design study of a wide - field fibre - optic spectroscopic facility for the aat @xcite . following further detailed investigations into cost and budgets , and in the expectation of some additional funding from both the australian and uk governments , the aatb gave its approval to begin the two - degree field ( or 2df ) project in march 1990 . initially the direct budget allocated was a$2.25 m ( for components and certain specific tasks which could not be done in - house ) and it was expected to take 4 - 5 years to complete . since the project would dominate the aao s activities for several years , a project management committee was established with several external expert members . in developing the 2df facility the aao wished to build on the techniques already used for the fully automated autofib instrument in use at the cassegrain focus of the aat . since 2df would offer almost an order of magnitude increase in multiplex advantage over autofib and would be located at the more challenging prime focus , several technical problems had to be solved before final approval was given for commencement of the project . the technical problems were addressed in the initial design study reports and covered areas such as the accuracy required of the robotic positioner , the design of the fibre retraction systems , location and design of the fibre spectrographs and the requirement for a double fieldplate system to maximise observing time . the mechanical constraints were eased by the strength and rigidity of the telescope tube structure and the size of the dome , which meant that a large instrument could be housed at prime focus without clearance and flexure problems . commissioning of the 2df facility began with the new prime focus corrector in the latter part of 1993 . the instrument was officially declared open at a ceremony on 1995 november 20 , and the first spectroscopic data were obtained in mid-1996 . the facility began to provide scheduled scientific observations in september 1997 with almost full functionality . the project ran about 40 per cent over the original time estimate and 20 per cent over budget ( these two are linked , in that 2df could have been completed sooner had more funds been available ) . the effective total cost was subsequently estimated to have been about a$8 m , including all staff costs and overheads . the bulk of the design and construction work was eventually done in - house using the aao s facilities in sydney and coonabarabran , partly because many aspects involved innovative design features which could not be easily specified or contracted out , and partly to contain costs . progress reports describing the evolution of the design of 2df have been published @xcite and project updates have featured regularly in the quarterly aao newsletter . in this paper we provide a full technical description of the 2df facility and its performance . section [ overview ] gives an overview of the instrumentation and its relationship to other multiple object fibre spectroscopy systems . in section [ components ] , a detailed description of the individual components of the 2df facility is given . operation of the 2df facility is covered in section [ operations ] and actual performance is detailed in section [ performance ] . sections [ science ] and [ future ] describe the range of projects being done with 2df and some scientific possibilities for the future . up to date technical information and signal to noise calculators are available on the 2df www pages ( http://www.aao.gov.au/2df/ ) . the 2df facility was developed with the aim of providing the aat with a dedicated prime focus spectroscopy facility , with order of magnitude improvements over existing systems in terms of the field area and number of objects that it is possible to observe simultaneously ( the multiplex advantage ) . while 2df was designed from the outset to be a versatile common - user instrument , suitable for a wide range of astronomical projects on an international research facility , it was always clear that the dominant project would be to obtain redshifts and hence distances for a very large sample of galaxies ( a few times @xmath2 objects ) , to map out the three - dimensional structure of the relatively local universe . two separate but linked large redshift surveys where the initial main projects with 2df : one for 250,000 galaxies brighter than b@xmath3=19.45 with a median redshift of 0.1 @xcite and one for 30,000 colour - selected quasars covering redshifts up to 4 @xcite . this meant that 2df had to be optimised to obtain low dispersion spectra of a few thousand targets per night , over a wide wavelength range , for mainly non - stellar objects which were in the magnitude range @xmath4 and spread across more than 1000 square degrees of sky . these requirements are very well met by a system providing two sets of 400 fibres covering 3 square degrees of sky , and capable of reconfiguring one set of fibres in about an hour , which happens to be equal to the time needed to take adequate signal - to - noise ratio spectra of 19 - 20 magnitude galaxies . whenever choices had to be made on design features , instrument parameters or priorities , the driver was to maximise 2df s efficiency as a redshift engine . the requirements of other projects , such as taking higher resolution spectra of stars , doing longer integrations on fainter galaxies , or working on densely clustered targets , were incorporated where possible , but not if they involved any compromise of performance for the redshift surveys . previous multi - object fibre spectroscopy systems on 4 metre class telescopes have used smaller fields of view and significantly smaller multiplex advantage . for example the william herschel telescope s @xcite and wyffos @xcite instruments provide fibre spectroscopy of up to 150 objects over a 1 field ( 40 arcminutes unvignetted ) . the hydra instrument @xcite on the 3.5 m wiyn telescope uses approximately 100 optical fibres to cover an unvignetted field of 1 . earlier fibre instruments on the aat were focap with 50 fibres @xcite and autofib with 64 fibres @xcite , both covering a 40 arcminute field at the cassegrain focus . some smaller telescopes offer a larger field of view or larger multiplex advantage . for example , the 2.5 m dupont telescope at las campanas with a 2.1 field and up to 200 optical fibres @xcite has been used for the largest previous galaxy redshift survey . at the 1.2-m uk schmidt telescope the flair fibre system with 92 fibres covered a 6 field @xcite . flair has now been replaced by a semi - automatic successor : 6df @xcite with 150 fibres . the 2.5 m sloan digital sky survey ( sdss ) telescope has a 3 field of view and 640 optical fibres ; this is a dedicated telescope carrying out both multi - colour imaging and a spectrosopic survey of a million galaxies @xcite . multi - object systems are also being built for several of the new 8 - 10 m class telescopes ; these will be able to observe fainter objects but cover much smaller fields of view . multi - object fibre systems come in three varieties : plug plates or other off - line field preparation , for example the sdss fibre system ; ` fishermen around the pond ' type instruments with separate remotely controlled mechanisms to simultaneously place each fibre at a target position , for example mx @xcite and mefos @xcite ; and ` pick and place ' systems involving consecutive placement of fibres with a single complex robot . the most versatile are the single robot systems , but they have the drawback that placing the fibres consumes a significant amount of observing time . 2df gets around this drawback by having two independent sets of 400 fibres and two field plates , mounted back - to - back on a tumbler assembly within the instrument . while one set of fibres is being used for observations , the second set is being reconfigured for the next target field . in order to give an unvignetted field of over 2 with a flat focal plane , a special corrector lens system had to be designed and manufactured @xcite . any wide field multi - fibre system must be able to cope with two separate atmospheric refraction effects , the variable distortion of the field as the telescope tracks across the sky , and variation of distortion with wavelength , which turns point images into very low dispersion spectra as zenith distance increases . the first effect means that for any one target field configuration , there is a maximum exposure time after which some of the target objects move out of the fixed fibre apertures . if we specify that a target field must be reconfigured once the target objects have moved more than a third of the fibre diameter ( 0.7 arcsec in the case of 2df ) , this defines the maximum time available for the robotic positioner to reconfigure the next set of fibres . the effect of differential refraction is dependent on hour angle and declination as shown in fig . [ darfig ] . evidently a reconfiguration time of at most an hour is required for an efficient system which can access most of the sky with minimal loss of observing time . the second effect arises because the atmosphere is a refracting medium whose refractive index depends on wavelength . for example , at a zenith distance of 60 , the light from a point source with wavelengths between 3651100 nm is spread out over 4.2 arcsec , which would make it impossible to use 2 arcsec fibres for low dispersion spectroscopy over a wide wavelength range . 2df has an atmospheric dispersion compensator ( adc ) built into the front two elements of the corrector . these two elements are slightly prismatic and are automatically counter - rotated during observations to provide an equal but opposite dispersion , to counteract the atmosphere as the telescope tracks across the sky . the fibres feed a pair of spectrographs which are mounted at the top of the aat near the prime focus , to keep the fibres short and maximise the uv throughput . each spectrograph takes 200 spectra simultaneously , with resolutions of between 500 and 2000 , on tektronix 1024 pixel square ccd detectors . mechanisms inside the spectrographs switch the fibre feeds in phase with the tumbling of the field plates . all of the hardware making up the 2df facility , including the spectrographs and electronics racks , is mounted on a purpose - built top end ring allowing straightforward interchange with the other three original alternative top ends of the aat . in particular , the entire fibre system can be left assembled and available for maintenance when 2df is not scheduled on the telescope . this section contains technical descriptions of the main components of the 2df facility . at the end of this section table [ specs ] summarises the main instrument properties . the new 2df top end ring ( fig . [ terfig ] ) is a direct copy of the three original aat top end rings ( f/8 and f/15 + f/35 secondaries and f/3 prime focus ) @xcite . this allows a fast ( @xmath51 hour ) interchange between top ends using the semi - automated mechanisms built into the aat dome . at the heart of the 2df facility is the corrector lens system which provides the 2.1 diameter field of view at the aat prime focus . the development of a corrector was initiated with a design by c.g . wynne @xcite offering a 2 field with 1.5 arcsec images using a 4-element corrector . further work by d. jones and r.g . bingham emphasised the need for an atmospheric dispersion compensator , the importance of chromatic variation in distortion ( cvd ) and of the telecentricity of the optical design . a relatively flat focal surface was also a requirement . the atmospheric dispersion of uncorrected images when sampled with a small fixed aperture size ( an optical fibre ) will reduce the throughput of the system significantly , by an amount which varies strongly with wavelength and zenith distance . when combined with small positioning and astrometric errors this will place severe limits on the ability to flux calibrate the resulting data . in order to minimise this effect an atmospheric dispersion compensator built into the corrector optics must provide a variable amount of dispersion in the opposite direction to the atmospheric dispersion , for as large as possible a range of zenith distances . all of the initial designs ( except for a significantly aspheric design by bingham ) exhibited chromatic variation of distortion ( cvd ) to some extent . this effect causes off axis , broadband images to be spread radially by up to about 2 arcsec for the 3501000 nm wavelength range , with maximum effect at about 0.5 field radius . this is a smaller effect than that of atmospheric dispersion and is independent of zenith distance ; it determines the ultimate limit to spectro - photometric accuracy with 2df . the telecentricity of an optical corrector design defines how the principal ray of each cone of light reaches the focal plane of the telescope . for an ideal fibre system the principal ray should be orthogonal to the focal surface . if the input light cone is not perpendicular to the focal surface then even in the absence of fibre focal ratio degradation ( frd ) the effective focal ratio of the output beam is decreased . in initial designs the angle of the principal ray varied across the field ( usually increasing towards the edge of the field ) by as much as 4 from the normal to the focal plane . this variation in input angle is effectively the same as reducing the input focal ratio of light to the fibre from f/3.5 to f/2.3 . note that this is a much more severe effect that that of frd within the fibre itself , which is minimal when working at this input focal ratio . if the spectrograph collimator is oversized to allow for this decrease in focal ratio , then we will reduce the spectral resolution for a fixed spectrograph beam size and camera focal ratio . alternatively , if the collimator is sized correctly for the f/3.5 beam , fibres accepting light from near the edge of the field ( the worst affected ) will be severely vignetted . unfortunately , the non - telecentricity of the final optical design means that the principal ray varies by up to 4 from the orthogonal case . this effect is partly compensated for in the design of the fibre probes ( detailed in section [ fibres ] ) . a slight oversizing of the spectrograph collimator also reduces the effect of the non - telecentricity and any fibre frd . the design of corrector finally selected @xcite contains counter - rotating prismatic doublets as the first two lens elements ( see fig . [ corrfig ] ) . these provide atmospheric dispersion compensation for zenith distances of up to 67 . the prismatic lenses are designed to give zero deviation of the optical path . cvd is maintained below 1.0 arcsec across the field for the maximum bandbass used by the spectrographs . the cvd is zero at the centre and edges of the field and reaches its maximum value at 30 arcmin field radius . the 2df project is exceptional in that the scientific imperatives drove the design towards a large field and high multiplex advantage , while the fixed fibre size , realistic astrometric errors and relatively poor average site seeing ( median value 1.5 arcsec ) all reduce the importance of the absolute imaging performance of the corrector optics . in response to these criteria and the three design issues raised above , extensive system modelling was performed on the design @xcite . this allowed not only the quantitative assessment of performance issues , but also the evaluation of the cost and risk implications of each potential design as work proceeded . at 0.9 metre diameter , the corrector optics contain some of the largest refracting elements made for an astronomical telescope . the glass blanks for the corrector were manufactured by ohara ( japan ) . since some of the corrector lenses are of a deep meniscus shape , a technique known as slumping was used to avoid a large and expensive wastage of glass and prolonged grinding . instead of cutting a lens from a thick blank , a thinner glass blank was heated and allowed to soften and slump under gravity over a convex mould . this technique does have the risk of increased internal stress and hence variable refractive index within the slumped blank . tests before and after slumping showed that any variations in refractive index were within specification . the internal transmittance , particularly in the uv and blue , was also an important consideration in selection of the glass . the catalogue specification of uv and blue transmittance of bk7 and f2 glass types is significantly worse than that of ubk7 and llf6 glass . however , discussions with the glass manufacturers revealed that the actual melt values were likely to be substantially better than the catalogue specification . fig . [ glasstrans ] shows the corrector throughput using actual glass melt transmittance measurements , compared to the catalogue values for the glasses used in the 2df corrector . the glass blanks were figured , coated with a mgf@xmath6 quarter wavelength anti - reflection coating ( optimised for 500 nm ) and mounted by contraves ( usa ) . the two adc elements are prismatic doublet lenses with the bk7 and f2 elements in contact to reduce the number of air - glass surfaces . an optical coupling compound is used between each of the lenses making up the doublets . after optical alignment in the corrector cells using temporary adjustment screws , the glass elements were mounted using flexible silicone rubber which allows for the differential thermal expansion of the steel corrector housing and the glass . the two lens elements that form the atmospheric dispersion compensator are rotated using stepper motors . these automatically move the adc elements to the required position whenever the telescope is slewed and then track continuously during observations . a mechanical switch acts as an index mark for each adc element . step counting is used to measure the position angle of the dispersing element . each long slew of the adc elements takes up to 3 minutes and includes a pass through the index marks to ensure that lost steps in the stepper motor do not accumulate and contribute to incorrect positioning of the lens elements . the optical corrector was received from contraves as a complete unit and commissioned on the aat between july 1993 and october 1993 , by taking direct night sky images with photographic plates and both cooled and uncooled ccd detectors @xcite . these tests included verification of the broad band imaging performance and the operation of the atmospheric dispersion compensator , as well as the initial distortion mapping of the corrector and telescope optics . the focal plane of the telescope is populated by a total of 404 deployable optical fibre probes which may be moved to cover any part of the available field of view of the corrector . the fibre probes are divided into two types ; one for target objects and the other for guide stars . the 400 object fibres each consist of a single 8-m long optical fibre of core diameter 140 , corresponding to an average diameter of 2.1 arcsec on the sky ( the effective diameter decreases non - linearly from 2.16 arcs at the field centre to 2.0 arcsec at the edge of the field ) . the 4 guide fibre bundles each consist of a 4 m long coherent bundle of 7@xmath7100 core diameter fibres in which 6 fibres are arranged in a hexagon around a central fibre ( fig . [ guidefibres ] ) . each individual fibre probe can access an area extending from the edge of the focal plane to just beyond the centre of the focal plane , and can cover a sector with apex angle of 28 . for the object fibre probes there is sufficient overlap between adjacent fibres to allow full field coverage . the four guide fibre bundles can access a total of about 30 per cent of the focal plane . the guide fibres are arranged at the four cardinal points on each fieldplate . at the focal plane the incoming light is folded into the optical fibres using 92 prisms ( 1.2 mm on a side ) made from schott sf5 high refractive index glass , with the input face anti - reflection coated . the prism is glued to the polished end of the optical fibre using uv - curing cement , after being optically aligned with the fibre core . the 92 angle of the prism was chosen to be half way between the extreme ranges of the beam angle for the non - telecentric corrector design . the prism material is a high refractive index glass so that the fast focal ratio input beam is totally internally reflected on the prism hypotenuse . this removes the requirement to aluminise the external reflecting face which would result in lower efficiency . the optical fibre used is a high oh or wet fibre manufactured by polymicro technologies . this has the advantage of good blue throughput at some cost of additional oh absorption bands in the far red ( see fig . [ fibrethroughputfig ] ) . the optical fibre is a step index fibre with core diameter of 140 , cladding diameter of 168 and a polyimide protective buffer 198 in diameter . each fibre and prism assembly is held in the focal plane by a small steel button 4 mm long and 2 mm wide containing a rare earth magnet ( ndfeb ) in its base ( see fig . [ fibrefig ] ) . the magnet holds the button in place on a steel field plate located just behind the telescope focal plane . the button has a vertical fin or handle to allow the robotic gripper to grasp the button easily . particular care was taken in the design of the button , with extensive simulations of the design to minimise the impact of the fibre probe footprint on the success rate with which fibres could be assigned to objects @xcite . an overlarge fibre probe footprint would affect the success with which fibres could be assigned to target objects clustered on relatively small scales , thus potentially imposing an instrumental signature on the observations . the output ends of the optical fibres at the spectrograph slit are aligned together in multiples of 10 fibres which form a slitlet . the 10 fibres are aligned and cemented on to a brass block using an assembly jig to provide the necessary fibre to fibre separation . the fibre separation projects to 5 pixels at the detector . the brass block and optical fibres are then polished together as a unit . the slitlets are held in the focal plane of the spectrograph as a set of 20 slitlets to form the spectrograph ` slit ' of 200 fibres . the fibres are fanned out to optimise the light path through the spectrograph optics . unlike previous autofib - type fibre positioners , the 2df fibre probes do not have protective stainless steel tubes along their length in the focal plane area . these tubes had a number of uses , they protected the fibre against breakages , avoided any risk of fibre tangles and simplified the fibre allocation process . however , when the fibre probes are parked at the periphery of the focal plane the steel tubes must be accommodated outside of the fieldplate area which increases the overall size of the instrument . this was not a problem with previous cassegrain focus instruments on the aat , but would have led to loss of light at prime focus . a second disadvantage with steel tubes is that they severely restrict the use of fibre - fibre cross - overs in the fibre allocation process , thus considerably limiting the number of fibres that can be allocated to target objects . at the edge of the focal plane the fibres enter retractor units which keep the fibres straight on the fieldplate . each fibre is mounted on a pair of pulleys within a retractor unit and independently maintained under 30 g tension using constant force springs @xcite . after exiting from the retractor unit each bundle of 10 optical fibres is protected by a single ptfe tube and is routed through the tumbler rotation axis ( see fig . [ tumbler ] ) and across the telescope top end ring spider vanes to the periphery of the top end ring where the spectrographs are located . the tumbler rotation axis is accommodated by allowing the fibres to twist . each retractor unit contains 10 fibres or 11 fibres ( 10 target fibres and a guide bundle for the 4 retractor units per plate that also carry a guide fibre bundle ) to match the fibre slitlets . this allows for the exchange of a complete retractor unit for subsequent repair . a total of 40 retractor units are arranged around the circumference of each of the two fieldplates . this gives a total of 404 fibre probes on each of the fieldplates . ten complete retractor units are maintained as spare units . this system has the advantage of maintaining the fibres in a straight line between the button and edge of the focal plane without increasing the footprint of the fibre probe in the focal plane . a drawback is the complexity of the retractor units themselves and the high reliability required , since a failure of the retractor unit will cause a loop of slack fibre to protrude over the fieldplate which may lead to tangles . an additional benefit of the bare fibre approach is that it is possible for each fibre to cross over many other fibres on its way from the edge of the focal plane to its target object . this considerably reduces the restrictions on allocating fibre probes to objects , at the cost of extra software to ensure that no attempt is ever made to move a button while its fibre is overlaid by other fibres . the fibres are held in the focal plane of the telescope s prime focus by magnetic buttons ( as described in the previous section ) on a flat circular magnetic stainless steel field plate 560 mm in diameter . two of these plates are arranged back to back on a tumbler arrangement ( fig . [ tumbler ] ) . the tumbler can be rotated back and forth through 180 to bring either field plate into play at the focal plane of the telescope . the other field plate is then ready to be accessed by the robotic fibre positioner . embedded in each fieldplate are 21 reference marks in a regular grid pattern . these reference marks consist of polished optical fibres inserted into holes in the fieldplate . the other ends of the optical fibres are illuminated by light emitting diodes ( leds ) . these reference marks have accurately known positions within the fieldplate and are the primary reference frame for both the gripper and focal plane imager gantry coordinate systems . the use of polished optical fibres means that the gripper optics can image and centroid the reference marks reliably . a single fast robotic positioner @xcite is used to manipulate the individual optical fibre probes . once the robot is given a list of x , y positions ( converted from astronomical coordinates provided by the observer ) , the positioner control software works out an optimal way to reconfigure the fibres from their current locations to the required positions . this usually requires moving a number of the fibres out of the way ( i.e. parking them at the periphery of the field ) before moving the majority of the fibres directly to their target position . this is at least 50 per cent faster than the more simplistic approach of parking all of the fibres before moving each fibre to its new target position . the robot consists of a complex gripper unit mounted on an x - y cartesian gantry , mounted above the tumbler field plate at the top of the 2df central section ( fig . [ terfig ] ) . the gripper gantry has two x axes and a y axis cross - beam in an ` h ' configuration . the axes are driven by linear ac servo motors . the use of linear motors over the more conventional leadscrews allows faster and more accurate motor control and eliminates the effects of backlash . positional information is determined from independent linear optical encoders for each axis , with a resolution of 1.25 . all movements of the xy gantry are balanced by counterweights that move in the opposite direction to the gantry , resulting in zero net momentum at the prime focus when the positioner is operational . with these precautions there is no measureable effect on the tracking of the telescope due to the movement of the gripper gantry during observations . the gripper unit contains a rotational @xmath8 axis to allow the gripper to be aligned with the ( approximately radially aligned ) off - axis button handle , a z axis to raise and lower the gripper unit and a set of jaws to grasp the individual button handles . the handling of the fibre probes is monitored by a small video ccd camera and optics built into the gripper unit . the optical fibres are back - illuminated by light projected into the optical fibre from the spectrographs . the robotic positioner can ` see ' the illuminated end of the fibre and knows the physical offset to the button handle which it grasps . this system allows the gripper to monitor fibre placement at the new target position , measure any positioning errors and if necessary correct those errors if they are deemed to be too large @xcite . this has the advantage that the robot positions the fibre core at the required position regardless of any actual manufacturing or assembly errors in the fibre button assembly . note that for the guide fibre bundles only the central fibre in the coherent bundle of 7 fibres is back - illuminated during the positioning process . the gripper unit was built under contract by the university of durham to an aao design , following earlier development work for the initial 2df design study . the design of the gripper jaws used in this process is particularly important and went through several prototyping stages @xcite in order to achieve repeatability at a level better than the overall accuracy requirements . one of the pair of gripper jaws is fixed to the xy-@xmath8 gantry and forms a reference surface while the second jaw is movable . to grasp a fibre button the fixed jaw is positioned against the button handle using the xy-@xmath8 gantry , then the movable jaw is closed on to the handle to avoid knocking the fibre button unnecessarily . releasing the button requires the movable jaw to be moved away from the button handle slightly before the fixed jaw is backed away from the button handle using the xy-@xmath8 gantry . the robotic positioner is capable of configuring a full target field of 404 fibres in about an hour . this requires on average a total of 600 fibre movements to untangle the previous field and configure the new target field , at a speed of under 6 seconds per move . the average positioning error is 11 ( 0.16 arcsec ) with all fibres required to be placed within 20 ( 0.3 arcsec ) of the demanded position ; if necessary , the robot picks up and replaces the button until the position is within tolerance . the effect of any offset due to flexure between the gripper gantry and fieldplate is removed using the grid of reference marks embedded in each fieldplate . these are measured with the gripper ccd video camera before positioning the fibres on the fieldplate . this survey process is always performed after a tumble operation or after the telescope has been slewed to a new position , before any fibres are moved . with a fibre - linked spectrograph one has to balance the merits of using long fibres to feed a remotely located bench spectrograph , against using very short fibres and mounting the spectrographs close to the fibre positioner . each approach has its merits and drawbacks . the remotely mounted spectrograph has the advantage of providing a stable spectrograph and making the engineering much simpler but with the drawback of the lower uv throughput of long optical fibres ( 30 metres ) . using short fibres with a locally mounted spectrograph reduces the uv light losses due to the fibre length but increases the mechanical design difficulties and subjects the spectrograph to variable gravitational flexure , and ambient temperature variations , as the telescope tracks across the sky . the choice between a single monolithic spectrograph accepting all 400 optical fibres compared to a pair of smaller spectrographs each accepting 200 optical fibres was largely determined by the availability of tektronix 1024 pixel square ccd detectors . these limit each detector to 200 fibres spaced by 5 pixels on the detector . larger detectors were promised but not available during the predicted time frame of the instrument construction . limiting each spectrograph to 200 fibres also eased the optical design somewhat , since it required a shorter slit assembly . the decision was made to keep the fibre lengths as short as practicable by locating the spectrographs at the top end of the telescope @xcite . initially they were to be mounted above the robot positioner @xcite with an extremely short fibre length . the final design , however , was constrained by the available space envelope and the spectrographs were mounted at the edge of the top end ring as shown in fig . [ terfig ] . this resulted in a fibre length of 8 metres . the overall construction of the spectrographs is depicted in fig . [ spectrograph ] . the spectrograph design is based on an f/3.15 off - axis maksutov collimator generating a 150 mm beam . this allows the use of the existing @xmath9 mm cassegrain rgo spectrograph reflection gratings . the collimator is slightly oversized compared to the prime focus corrector focal ratio of f/3.5 , to allow for some focal ratio degradation in the optical fibres and further reduce any effect of the non - telecentricity of the corrector design . the off - axis design has the benefit of no central obstruction so the fibre slit and its associated shutter , filter wheels and slit interchange mechanism do not vignette the beam . the 200 optical fibres from the focal plane are arranged in a single line to form a pseudo - slit . a shutter is located immediately in front of the fibre slit . the shutter is moved in a direction perpendicular to the slit axis and therefore only requires a short travel to open and close over the slit width as defined by the fibre diameter of 140 . two filter wheels , each with four apertures , are provided behind the shutter . one aperture in each filter wheel is kept as a clear aperture leaving a maximum of six filter spaces of which three are currently used to hold order - sorting filters for high resolution observations . the filters themselves are 2 mm thick . filters currently available include gg495 , s8612 and rg630 . use of a filter requires a change in the spectrograph focus as the filters are located in the diverging beam behind the fibre slit and shutter . the spectrograph camera is located at a collimator - camera angle of 40 and is a fast , wide - field schmidt camera with a ccd at its internal focus @xcite . the main difficulty with the camera is that the pixel matching requirements of imaging such a large number of fibres onto such a small detector force it to have an f - ratio as fast as f/1.2 in the spatial direction ( f/1.0 in the spectral direction ) while retaining its wide @xmath10 field of view . the camera uses a severely aspheric front corrector plate , to give good performance over the full wavelength range and field angle required to image the spectra . the camera is fully evacuated with the aspheric schmidt corrector lens acting as a dewar window . each ccd is cooled by use of a cold finger attached to a cti cryodyne closed - cycle helium cooler . a helium compressor is mounted on the telescope in a gymbal mount and is used to drive both cryodynes . the cool - down time for the ccds is relatively long , about 3.5 hours from ambient temperature to their operating temperature of 170k . the performance of the cryostats is sufficiently high to maintain their performance for up to 21 days . in addition to the basic design of the spectrographs , the double buffering of the fibre positioner system and the back illumination of the fibres provide some extra engineering problems . each spectrograph actually accepts 400 fibres , 200 from each field plate , which must be located in the focal plane of the spectrograph when required . the non - observing fibres must be illuminated with a light source bright enough to be seen by the robotic positioner s gripper tv system since fibre positioning is done simultaneously with observing . these two requirements mean that each spectrograph must have a slit interchange mechanism , a bright light source for the fibre back - illumination , and a reliable means of completely shielding this light source from the rest of the optical path . the spectrograph is focussed by moving the detector within the cryogenic camera . both focus and tilt adjustments of the ccd are available and can be driven remotely . with the exception of the selection of the grating ( which has to be manually inserted before observations start ) , all functions of the spectrographs are remotely controlled . this includes slit changeover , filters , hartmann shutter , grating rotation and spectrograph focus . gratings can be automatically recognised by the spectrograph control system which reads magnetic barcodes present on all of the grating cells . the complete spectrograph configuration is inserted into the data fits header information to allow automatic pipeline data reduction . a second xy gantry , almost identical to the gripper gantry , is located immediately behind the final element of the optical corrector ( fig . [ terfig ] ) and in front of the observing fieldplate . instead of a gripper unit , this gantry carries a pair of ccd cameras . one camera is a simple video ccd camera which can only view the back - illuminated fibres and field plate , while the second is a peltier cooled princeton instruments ccd camera , facing in the opposite direction , which can view the sky when its gantry is appropriately positioned . this gantry system is known as the focal plane imaging ( fpi ) system and is the primary means of determining the relationship between ra , dec and x , y on the field plates . this sky viewing camera can also be used for target field acquisition and seeing measurements . the image scale of the focal plane imager ccd is 0.3 arcsec per pixel . the focal plane imager gantry also surveys the reference marks embedded in the field plates , in a similar manner to the gripper gantry , to correct for registration and flexure before centroiding the images of reference stars . 2df contains its own remotely controlled calibration systems . two white flaps may be inserted into the telescope beam below the corrector , blocking off sky light and forming a reflective screen . a variety of calibration lamps can be used to illuminate this screen , with the resulting scattered light travelling through the 2df corrector and illuminating the focal plane . two intensities of quartz lamps for fibre flats and a variety of hollow cathode arc lamps ( copper - argon , copper - helium and iron - argon ) are provided . requesting an arc or fibre flat exposure will result in the control system automatically inserting the reflective flaps and turning on the requested calibration lamps . at the end of the exposure the lamps are turned off and the flaps either removed from the telescope beam or left in place if further calibration exposures are required . this automation avoids accidentally leaving calibration lamps turned on and reduces observing overheads . all calibration exposures have appropriate fits header items added to the data frames as an aid to the pipeline reduction of the data . the 2df facility uses the aao drama software infrastructure @xcite to build a fully integrated control system across several computers and operating systems . the graphical user interface is written using tcl / tk and provides for control of all aspects of the positioner system , the spectrographs , adc and ccd control , all from one simple to use interface . the same interface allows the observer to control the telescope directly , for example when slewing to a new target field . the graphical user interface is provided by several windows spread across two computer screens with controls arranged by subsystem . individual parts of the 2df facility are controlled by software tasks which are running on the most appropriate control computer ( vxworks , solaris , vms ) and drama allows the many separate tasks to communicate and work together seamlessly . the two tektronix 1024 ccds are controlled using the standard aao controller hardware , the aao external memory ( xmem ) and observer software interface on a vax / vms computer system . with a slight adaptation , the observer software communicates with the rest of the 2df control system ( running on solaris and vxworks computer systems ) using drama to enable it to add all of the more specialised 2df fits header items from the 2df control system to the data files for archiving . the two ccds are operated and read out simultaneously , but using separate controllers and data links . on each fieldplate there are four guide fibre bundles which do not feed to the spectrographs . these guide fibres consist of a coherent bundle of seven individual fibres with six fibres in a hexagon pattern surrounding a central fibre ( fig . [ guidefibres ] ) . the separation of the individual fibres in a guide bundle at the input end is 1.8 arcsec . each guide bundle may be positioned on a guide star in the focal plane to allow for acquisition of the target field . the output from the four guide fibre bundles is collected by a quantex intensified tv system , with its output visible to the telescope night assistant and observer . a guide star is accurately centred when the central fibre of a guide fibre appears brightest on the tv display and the six surrounding fibres are uniformly illuminated . a minimum of two guide fibres are recommended for each target field , although it is preferable that all four are used in case of inaccurate stellar positions . the quantex tv system can detect stars down to a magnitude limit of @xmath11 in average conditions . the system was designed to include an option to rotate the field plates for target acquisition and , if necessary , during tracking . however , this has not yet been implemented . it turns out that the instrument is sufficiently stable , and the flexure is small and predictable , so that it is easier to remove any overall field rotation in the positioner software . often , larger errors would be introduced by attempting to use the fiducial stars to correct rotation , especially since the surface density of usable stars means that it is difficult to obtain enough stars close to the edge of the field to allow rotation to be determined . the pointing of the aat is normally good enough ( better than 2 arcsec ) that target field acquisition is a simple procedure of slewing the telescope and the guide stars will be visible somewhere in at least one of the guide fibres . fine acquisition is performed manually by offsetting the telescope by very small amounts until all four guide stars are well centred in their respective guide fibres . if for any reason the guide stars can not be seen after slewing the telescope or there is any ambiguity in the guide star acquisition ( for example due to a very crowded field in a globular cluster or due to poor guide star positions ) the focal plane imager ccd camera can be used to image a small section of sky around each guide star to verify acquisition . a simple offset will then place the guide stars onto the guide bundles , after which the focal plane imager can be removed from the focal plane . once a target field is acquired , guiding can be done in one of two ways . the normal guiding method is manual guiding using the excellent tracking of the aat and making small adjustments every 10 minutes or so . an automatic autoguider has also been developed which uses the video output from the quantex tv system to feed a video frame grabber . the digitised video can then be analysed to determine the average centroid of each guide fibre and determine the correct telescope offset to restore the telescope pointing . initially however , the extra calibration required and problems with the dimensional stability of the quantex tv image at different gain levels meant that autoguiding was not generally used . a simplified version of the autoguider system was implemented in 2001 under the control of the telescope night assistant and this is now in regular use . [ cols= " < , < " , ] since the wavelength information in these broadband measurements was insufficient , an improved relative throughput calibration was derived using about one hundred stars observed simultaneously with 2df ( fig . [ 2dfefficiency ] ) . the relative efficiency versus wavelength was determined by dividing the observed spectra by model spectra that were chosen to best match the stars broadband magnitudes . of course these efficiencies change with choice of grating , the best performance comes with the use of the low resolution 270r and 316r gratings where the blaze of the grating matches the peak of the ccd quantum efficiency . in this case the system efficiency peaks at 9 per cent . for more details and signal to noise calculations see the 2df www pages ( http://www.aao.gov.au/cgi-bin/2dfsn.cgi ) . while the main effort has gone into optimising the use of 2df for the large redshift surveys , i.e. for very large numbers of low dispersion observations of relatively bright targets , it is clear that 2df can be used in many other ways . some of the scientific programs are described in the following section . here we mention some technical factors relevant to other applications of 2df . the biggest questions are how faint can 2df observations be pushed , and what is the best way of maximising the signal to noise ratio in a given observing time ? this comes down to two further questions : what is the best way of doing sky subtraction , and for how long can observations be continued and still give a useful gain ? the redshift survey data , on which many of the performance figures quoted above are based , typically consist of sets of @xmath12s integrations on sets of galaxies to @xmath13 . the mean dark sky signal is approximately equal to 20 per cent or less of the signal for most galaxies . for such observations , the standard procedure is to allocate a fraction ( at least 5 per cent ) of the fibres to random sky positions , derive a mean sky and scale this for each fibre . the same technique will work for total integration times of several hours and should produce similar quality data for targets up to one or two magnitudes fainter . beyond that , the observations will be sky limited and the gains will go as the square root of the observing time , at best . two alternative techniques are classical beam switching , in which pairs of fibres are placed equal distances apart and each target is observed alternately in either fibre , and ` nod & shuffle ' @xcite where the image is moved rapidly ( ` shuffled ' ) between an active and a storage area on the ccd , exactly in phase with the nodding of the telescope between targets and an offset sky position . experiments with both methods were done with 2df in january 2001 . in theory , the ` nod & shuffle ' method should give the best results since the identical optical path is used for the target and sky observations , and because the switching is very fast compared with the timescale of night sky line variations . beam switching should be almost as good when time variations are not critical and has the advantage that the target is always being observed , whereas ` nod & shuffle ' is off - source for half the time . both types of switching are inefficient compared with conventional observing , in terms of the number of targets observable . in practice neither mode gave dramatic gains in the magnitude range accessible to 2df although sky subtraction can be improved to below 1 per cent using these methods . a serious drawback of any target - sky switching technique is that it depends on finding clean sky apertures at the same fixed distance from every target . this becomes harder as the number of targets increases . moreover , to avoid damaging the signal for an object at say magnitude 21 requires having no objects in the sky aperture which are brighter than magnitude 24 or 25 , well below the detection limit of the schmidt sky surveys or most wide field ccd images . essentially , it depends on whether the limit to the accuracy of sky subtraction is set by the photon statistics and faint object contamination in the sky , or by the systematic errors involved in transferring the sky from other fibres . in the case of 2df , it seems that the operational complications and target losses inherent in beam switching mean that it is usually better to try to minimise the systematic errors and to continue to use the mean sky approach . a more fruitful line may be to exploit the large numbers of objects observable with 2df to create mean spectra , rather than trying to increase the signal to noise ratio in each individual spectrum . one approach is to take large numbers of targets selected to be almost identical from other evidence ( e.g. photometry , or an automatic spectral classification scheme ) and combine their spectra ; a second is to subdivide samples according to one or more strong spectroscopic features , and then to combine the subsets to look for weaker features which may correlate with the strong features . either way , spectra can be created with signal to noise ratio levels corresponding to hundreds or even thousands of hours of observation . the current version of the configure software , which allocates fibres to targets , has been optimised for the galaxy redshift survey @xcite . this is characterised by having a mean target density of about 180 per square degree , fairly uniformly distributed across the sky , so that it was necessary to have an algorithm which could achieve a very high yield in allocating all 400 fibres when there were often only about 400 targets in the 2df field . the galaxy redshift survey is able to allocate fibres to 94 per cent of the original input target list @xcite . somewhat different constraints arise in other applications , e.g. in crowded globular star clusters , in deep fields from ccd images , or when there are many fewer than 400 targets . configure has several options for different ways of allocating fibres to targets and these can sometimes give substantially higher yields than running the default parameters . for example , in crowded compact fields , it is best to centre the targets and to limit the pivot angle through which fibres can move . however , the maximum yields in very compact fields are not high , due to the size and shape of the 2df buttons and the requirement to leave safe clearance between buttons and fibres . over 300 fibres can be allocated to targets within a 1 diameter field and approximately 100 fibres can be allocated to objects in a 20 arcmin diameter field , based on observations of dense globular clusters . for sparse fields with fewer than 200 targets , it is often more efficient to restrict the fibres to a single spectrograph , if only for ease of data reduction . for covering a wide range of magnitudes , e.g. in galactic open star clusters , the targets should be split into subsets covering only 2 or 3 magnitudes each . note that the reconfiguration time becomes very short for small samples of stars , so it is not inefficient to use 2df for observations of relatively small numbers of bright stars ( 50 stars take about 5 minutes to configure ) . other techniques have also been used with the configure software to send the light from blue objects to one spectrograph with a blue grating and red objects to the second spectrograph with a red optimised grating . a good example of this is given by glazebrook . 2df is not very efficient for doing short observations of single targets such as standard stars . there is a procedure for placing a star down any desired fibre in an existing configuration , but it is rather slow and cumbersome , and it depends on blind offsetting from a guide fibre to a target fibre . if it is desired to measure many such stars , a better procedure is to prepare a special configuration with just a few fibres in a simple pattern at the field centre . for long observations of single targets where accurate centering is important , it is necessary to have one or more bright guide stars in the same way as for a full 2df configuration . this applies especially to faint ` targets of opportunity ' where it is essential to supply a target position , sky positions and a nearby guide star . a key design feature of 2df , which has been little exploited up to now , is the ability to keep track of differential atmospheric refraction by setting up the same configuration on both field plates , but corrected for different hour angles . to observe the same field all night one should tumble between the two plates every hour or two , depending on telescope attitude ( see fig . [ darfig ] ) . however , setting up such configurations is not trivial because there are small but significant geometrical differences between the two 2df field plates . these mean that a configuration which is valid on one plate will fail on the other because of button and fibre collisions , and because limits on fibre extensions and pivot angles are exceeded . these effects can be largely avoided by setting up the starting configuration with extra clearance around the buttons and reduced limits on the pivot angle . the minimum clearance , set by the 2df hardware , is 400 ; setting this to 800 and setting the maximum pivot angle to 10 instead of the default 14 largely avoids collisions when switching field plates or changing the configuration ha . there may be a decreased target yield since the fibres can not be placed so close together , but in practice such losses are generally only a few per cent . while the main science driver for the design of 2df has been the large redshift surveys @xcite , the inherent flexibility in the design of 2df means that it is a unique facility for many other types of astronomy . a diverse range of both galactic and extra - galactic astronomy problems have already been tackled using the unique area coverage and multiplex advantage of 2df . some of these are described briefly in this section . some example spectra obtained with 2df are shown in fig . [ spectra ] . gilmore , wyse , norris and freeman are using 2df to measure radial velocities to an accuracy of 10 km@xmath14 for several thousand f / g main sequnce stars at distances of 3@xmath157 kpc from the sun in several sightlines through the galaxy @xcite . these data will allow the statistical study of the kinematics and metallicity distributions of stars in the galactic thick disk and halo . initial results include a new population of low metallicity stars with disk - like kinematics , possibly representing debris from a merged satellite . spectroscopy of large samples of faint stars in globular clusters have been obtained by cannon , croke , da costa and norris ( private comm . ) . their objective is to compare the chemical abundance variations seen in red giant stars with unevolved main sequence stars to determine whether these variations are due to self - enrichment during evolution or are primordial . in the case of 47 tucanae , the spectra show clear separation into cn - strong and cn - weak stars . combining the spectra for 50100 stars in each class yields very high signal to noise ratio spectra with more than @xmath16 counts per pixel , making it possible to detect extremely weak features and to look for correlations with other abundance parameters . hatzidimitriou have been using 2df to study carbon stars in the magellanic clouds . two observing runs yielded over 2000 carbon star spectra in both clouds . these spectra have been used to map in fine detail the rotation and velocity dispersion across the lmc . the data are also being used to classify and determine chemical abundances for the carbon stars . evans and howarth ( private comm . ) are using 2df to undertake a spectroscopic survey of massive stars in the smc from an unbiased sample of bright blue field stars obtained from apm photometry . over 4000 intermediate resolution spectra have already been obtained and used to generate an observational h - r diagram . this can be compared with population synthesis models to investigate the field star initial mass function for the smc . drinkwater are carrying out a complete , unbiased survey of the fornax cluster by obtaining 2df spectra for all objects down to b@xmath17=19.7 in a 12 square degree region centred on the cluster . the goals of this project are to determine cluster membership for a complete sample of objects ( especially dwarf galaxies ) , study the cluster dynamics , detect previously unrecognised compact galaxies in the cluster and field , and to study background galaxies and quasars and foreground galactic stars . a deep narrowband [ oiii ] imaging survey of the virgo cluster has revealed a large population of emission line objects that could be either planetary nebulae associated with the intra - cluster medium or high redshift emission line galaxies . during a 5 hour 2df exposure freeman obtained 47 emission - line detections of which 23 turned out to be true intra - cluster planetary nebulae with detection of both the 4959 and 5007 lines , and another 16 in the outer regions of m87 . freeman also found 8 ly@xmath18 emitters at z@xmath19 3.1 with equivalent widths w@xmath20(ly@xmath18 ) @xmath21 150 . willis have compiled a sample of 485 early - type galaxies with redshifts with 0.3@xmath22 z @xmath22 0.6 . this represents the largest sample of luminous field galaxies at intermediate redshift . these data are being used to study the evolution of galaxy clustering on large scales , for an investigation of the fundamental plane for luminous galaxies as a function of environment , and for a survey for strong gravitational lensing . in this paper we have described a unique multi - object optical spectroscopy facility ( 2df ) available as a common user instrument at the anglo - australian telscope . we are currently performing experiments using 2df with techniques such as charge shuffling and telescope nodding @xcite to improve the sky subtraction of extremely faint target objects . future plans include replacing the two spectrographs with bench mounted spectrographs using volume phase holographic ( vph ) grating @xcite technology . the vph gratings have the advantage of offering higher efficiency than conventional reflecting diffraction gratings . upgraded detectors , improved anti - reflection coatings and new optical fibre materials @xcite will allow us to use longer optical fibres and still gain in overall throughput and provide higher resolutions . the advantage of the bench mounted spectrographs will be in their thermal and mechanical stability . we also plan to use the flexibility of the fieldplate tumbler system to provide one or two fibre feeds for integral field spectroscopy at the currently unused 90 positions of the 2df tumbler unit . the authors would like to acknowledge the many aao technical staff who have made 2df a reality and astronomy support staff of the aao who make 2df available to the general user . this project has only been possible with extensive support from both the australian and uk scientific communities and the long - term backing of the aat board . bailey j.a . , glazebrook k. , offer a.r . , taylor k. , 2001 , mnras , submitted barden s.c . , armandroff t. , muller g. , rudeen a.c . , lewis j. , groves l. , 1994 , in crawford d.l . , craine e.l . , eds , proc . spie 2198 , instrumentation in astronomy viii , p. 87 barden s.c . , arns j.a . , colburn w.s . , 1998 , in dodorico s. , ed . spie 3355 , optical astronomical instrumentation , p. 866 cannon r.d . , 1997 , in kontizas e. , eds , proc . iau commission 9 , widefield spectroscopy 40 , p. 33 cannon r.d . , 2000 , aao newsletter , 92 , 14 colless m. , , 2001 , mnras , in press croom s.m . , smith r.j . , boyle b.j . , shanks t. , loaring n. , miller l. , lewis i.j . , 2001 , mnras , 322 , l29 drinkwater m. , , 2000 , a&a , 355 , 900 farrell t.j . , bailey j.a . , shortridge k. , 1995 , in shaw r.a . , and hayes j.j.e . , eds , asp conf . ser . 77 , astronomical data analysis and systems iv , astron . , san francisco , p. 133 freeman k.c . , 2000 , in combes f. , mamon g.a . , charmandaris v. , eds , asp conf . 197 , dynamics of galaxies ; 15th aip meeting , astron . , san francisco , p. 389 gilmore g. , wyse r.f.g . , 2001 , in deiters s. , fuchs b. , just a. , spurzem r. , wielen r. , eds , asp conf . 228 , dynamics of star clusters and the milky way , astron . , san francisco , in press glazebrook k , bland - hawthorn j. , farrell t.j . , waller l.g . , barton j.r . , lewis i.j . , 1999 , aao newsletter , 90 , 11 glazebrook k. , bland - hawthorn j. , 2001 , pasp , 113 , 197 glazebrook k. , tsvetanov z. , zheng w. , hoversten e. , chiu k. , bridges t. , boyle b. , 2001 , aao newsletter , 98 , 4 gray p. m. , taylor k. , 1990 , in crawford d.l . , ed . spie 1235 , instrumentation in astronomy vii , p. 709 gray p. m. , 1983 , in boksenberg a. , crawford d.l . , eds , proc . spie 445 , instrumentation in astronomy v , p. 57 gray p. m. , taylor k. , parry i. r. , lewis i. j. , sharples r. m. , 1993 , in gray p.m. , ed . , asp conf . ser . vol . 37 , fibre optics in astronomy ii , astron . , san francisco , p. 145 guerin j. , bellenger r. , dreux m. , felenbok p. , fernandez a. , rousset g. , schmidt r. , 1993 , in gray p.m. , ed . , asp conf . ser . vol . 37 , fiber optics in astronomy ii , astron . , san francisco , p. 138 hatzidimitriou d. , cannon r.d . , croke b.f . , morgan d.h . , 2000 , in combes f. , mamon g.a . , charmandaris v. , eds , asp conf . 197 , dynamics of galaxies ; 15th aip meeting , astron . , san francisco , p. 347 hill j.m . , lesser m.p . , 1986 , in crawford d.l . spie 627 , instrumentation in astronomy vi , p. 303 hg , e. , fabricius , c. , makarov , v. v. , bastian , u. , schwekendiek , p. , wicenec , a. , urban , s. , corbin , t. , wycoff , g. , 2000 , a&a , 357 , 367 jenkins c.r . , gellatly d.w . , bingham r.g . , worswick s.p . , 1993 , in gray p.m. , ed . , asp conf . ser . vol . 37 , fiber optics in astronomy ii , astron . , san francisco , p. 209 jones d. j. a. , 1993 , in gray p.m. , ed . , asp conf . 37 , fibre optics in astronomy ii , astron . , san francisco , p. 355 jones d. j. a. , 1994 , applied optics , 33 , 7362 lewis i. j. , parry i. r. , 1990 , in crawford d.l . , ed . spie 1235 , instrumentation in astronomy vii , p. 745 lewis i. j. , parry i. r. , sharples r. m. , taylor k. , 1993 , in gray p.m. , ed . , asp conf . ser . vol . 37 , astron . , san francisco , p. 249 lewis i.j . , glazebrook k. , taylor k. , 1998a , in arribas s. , mediavilla e.,watson f.g . , eds , asp conf . ser . vol . 152 , fibre optics in astronomy iii , astron . , san francisco , p. 71 lewis i.j . , glazebrook k. , taylor k. , 1998b , in dodorico s. , ed . spie 3355 , optical astronomical instrumentation , p. 828 parker q.a . , watson f.g . , miziarski s. , 1998 , in arribas s. , mediavilla e. , watson f.g . , eds , asp conf . ser . vol . 152 , fibre optics in astronomy iii , astron . , san francisco , p. 80 parry i.r . , gray p.m. , 1986 , in crawford d.l . , ed . , proc . spie 627 , instrumentation in astronomy vi , p. 118 parry i. r. , sharples r. m. , lewis i. j. , gray p. m. , 1993 , in gray p.m. , ed . , asp conf . ser . vol . 37 , fibre optics in astronomy ii , astron . , san francisco , p. 36 parry i.r . , lewis i.j . , sharples r.m . , dodsworth g.n . , webster j. , gellatly d.w . , jones l.r . , watson f.g . , 1994 , in crawford d.l . , craine e.l . , eds , proc . spie 2198 , instrumentation in astronomy viii , p. 125 royal astronomical society report , 1986 , review of scientific priorities for uk astronomical research , 1990 to 2000 , london , ras roeser s. , bastian u. , 1988 , a&as , 74 , 449 sadler e.m . , harrison s. , lee s. , 1991 , aao observing guide um1.4 shectman s.a . , 1993 , in gray p.m. , ed . , asp conf . ser . vol . 37 , fibre optics in astronomy ii , astron . , san francisco , p. 26 shltz g.f . , vydra j. , lu g. , fabricant d. , 1998 , in arribas s. , mediavilla e. , watson f.g . , eds , asp conf . ser . vol . 152 , fibre optics in astronomy iii , astron . , san francisco , p. 20 smith g. , lankshear a. , 1998 , in dodorico s. , ed . spie 3355 , optical astronomical instrumentation , p. 905 taylor k. , gray p.m. , 1990 , in barr l.d . , , proc spie 1236 , advanced technology optical telescopes iv , p. 290 taylor k. , gray p.m. , 1993 , in gray p.m. , ed . , asp conf . ser . vol . 37 , fiber optics in astronomy ii , astron . , san francisco , p. 379 taylor k. , gray p.m. , 1994 , in crawford d.l . , craine e.l . , eds , proc . spie 2198 , instrumentation in astronomy viii , p. 136 taylor k. , bailey j.a . , wilkins t. , glazebrook k. , 1996 , in jacoby g.h . , barnes j. , eds , asp conf . vol 101 , astronomical data analysis and systems v , astron . , san francisco , p. 195 taylor k. , cannon r.d . , watson f.g . , 1996 , in ardeberg a. , ed . , proc spie 2871 , optical telescopes of today and tomorrow , p. 145 watson f.g , gray p.m. , oates a.p . , lankshear a. , dean r.g . , 1993 , in gray p.m. , ed . , asp conf vol 37 , fibre optics in astronomy ii , astron . , san francisco , p. 171 wilcox j.k . , 1993 , in gray p.m. , ed . , asp conf . ser . vol . 37 , fiber optics in astronomy ii , astron . , san francisco , p. 51 willis j.p . , hewett p.c . , warren s.j . , 2001 , mnras , 325 , 1002 wynne c.g . , 1989 , mnras , 165 , 47p york d.g . , 2000 , aj , 120 , 1579

electron transport experiments on two lateral quantum dots coupled in series are reviewed . an introduction to the charge stability diagram is given in terms of the electrochemical potentials of both dots . resonant tunneling experiments show that the double dot geometry allows for an accurate determination of the intrinsic lifetime of discrete energy states in quantum dots . the evolution of discrete energy levels in magnetic field is studied . the resolution allows to resolve avoided crossings in the spectrum of a quantum dot . with microwave spectroscopy it is possible to probe the transition from ionic bonding ( for weak inter - dot tunnel coupling ) to covalent bonding ( for strong inter - dot tunnel coupling ) in a double dot artificial molecule this review on the present experimental status of double quantum dot studies is motivated by their relevance for realizing solid state quantum bits .

cond-mat0205350

quantum dots are man - made sub - micron structures in a solid , typically consisting of 10@xmath0 - 10@xmath1 atoms and a comparable number of electrons @xcite . in semiconductor quantum dots all electrons are tightly bound , except for a small number of free electrons , which can range from zero to several thousands . for the quantum dot devices considered in this review , the starting point for fabrication is formed by a heterostructure consisting of different semiconductor materials ( gaas / algaas ) . the free electrons are strongly confined to the interface between gaas and algaas , forming a 2-dimensional electron gas ( 2deg ) . confinement in the other two dimensions is accomplished by locally depleting the 2deg , via etching techniques or metal gate electrodes . the resulting structure is weakly coupled to source and drain electrical contacts by tunnel barriers , connecting it to the outside world . + an important element of electronic transport through quantum dots is coulomb blockade @xcite . an extra electron can only be added to the dot , if enough energy is provided to overcome the coulomb repulsion between the electrons . next to this purely classical effect , the confinement in all three directions leads to quantum effects that strongly influence electronic transport at low temperature . in particular it leads to the formation of a discrete ( 0d ) energy spectrum , resembling that of an atom . this and other similarities have therefore lead to the name artificial atoms for quantum dots @xcite . + the next logical step after studying individual quantum dots is to study systems of more than one dot . where single quantum dots are regarded as ` artificial atoms ' , two quantum dots can be coupled to form an ` artificial molecule ' . depending on the strength of the inter - dot coupling , the two dots can form ionic - like ( weak tunnel coupling ) or covalent - like bonds ( strong tunnel coupling ) . in the case of ionic bonding the electrons are localized on the individual dots . the binding occurs , because a static redistribution of electrons leads to an attractive coulomb interaction . weakly , electrostatically coupled quantum dots with negligible inter - dot tunnel conductance are covered by orthodox coulomb blockade theory @xcite . in the case of covalent bonding , two electron states are quantum - mechanically coupled . the main requirement for covalent binding is that an electron can tunnel many times between the two dots in a phase - coherent way . here the electron can not be regarded as a particle that resides in one particular dot , but it must be thought of as a coherent wave that is delocalized over the two dots . the bonding state of a strongly coupled artificial molecule has a lower energy than the energies of the original states of the individual dots . this energy gain forms the binding force between the two dots . + the theoretical possibility to perform certain tasks in a much more efficient way using a ` quantum computer ' instead of a ` classical computer ' , has stimulated the search for physical realizations of the basic building block of such a computer : the quantum bit . in principle , any quantum two - level system can be used as such a qubit . in particular , recent studies have put forward double quantum dots as interesting candidates for realizing qubits @xcite . the possible application of double quantum dot devices in quantum logic forms an important motivation for this work . + in this review we concentrate on electron transport through lateral double quantum dots coupled in series . all devices have been fabricated and all experiments have been performed at delft university of technology and ntt basic research laboratories . by now there exists an extensive literature on experimental studies of electron transport through double lateral quantum dots coupled in series , and lateral double dots coupled in parallel . vertical double quantum dot structures fall outside the focus of this review . in vertical structures , the characteristics of the tunnel barriers are set by the growth parameters of the heterostructure , limiting the experimental tunability . besides that , the gate geometry used in these devices , makes it difficult to address dots independently . + as a first step to understanding double dot systems we introduce the stability diagram @xcite , or honeycomb diagram , in section [ stability ] . it is a convenient tool in the analysis of double dot transport properties . resonant tunneling experiments discussed in section [ restun ] , show that the resonant widths are only determined by the lifetime of the discrete energy states , independent of the electron temperature . in section [ magnetiz ] we discuss level spectroscopy in a magnetic field . the double dot geometry offers sufficient energy resolution to probe intra - dot level repulsion . in section [ spectro ] we present microwave spectroscopy measurements on a quantum dot molecule . we illustrate the transition from a weakly coupled double dot to a strongly coupled double dot , by discussing a two - level system in section [ 2level ] . although being a clear simplification , the mapping of the double dot on a two - level system grasps much of the physics of the experiments presented in section [ spectro ] . irradiation with microwaves leads to photon assisted tunneling ( pat ) ( sections [ weakcpat ] and [ strongcpat ] ) , which turns out to be a powerful tool not only to reveal the character of the inter - dot coupling , but also to quantitatively determine the bonding strength . in this section we introduce the stability or honeycomb diagram that visualizes the equilibrium charge states of two serially coupled dots . we start with a purely classical description in which the influence of discrete quantum states is not taken into account yet @xcite ( 9 ) , namely the numerator , ( @xmath2 ) . for @xmath3 , this relation gives ( @xmath4 ) when it should read ( @xmath5 ) , and vice - versa . ] . the double dot is modeled as a network of resistors and capacitors ( fig@xmath6 [ network ] ) . the number of electrons on dot 1(2 ) is @xmath7 . each dot is capacitively coupled to a gate voltage @xmath8 through a capacitor @xmath9 and to the source ( s ) or drain ( d ) contact through a tunnel barrier represented by a resistor @xmath10 and a capacitor @xmath11 connected in parallel . the dots are coupled to each other by a tunnel barrier represented by a resistor @xmath12 and a capacitor @xmath13 in parallel . the bias voltage , @xmath14 , is applied to the source contact with the drain contact grounded ( asymmetric bias ) . in this section we consider the linear transport regime , i@xmath6e@xmath6 @xmath15 . if cross - capacitances ( such as between @xmath16 and dot 2 ) , other voltage sources and stray capacitances are negligible , the double dot electrostatic energy reads ( a full derivation is given in the appendix ) @xmath17 + where @xmath18 is the charging energy of the individual dot 1(2 ) and @xmath19 is the electrostatic coupling energy . the coupling energy @xmath19 is the change in the energy of one dot when an electron is added to the other dot . these energies can be expressed in terms of the capacitances as follows @xmath20 + here @xmath21 is the sum of all capacitances attached to dot 1(2 ) including @xmath13 : @xmath22 . note that @xmath18 can be interpreted as the charging energy of the single , uncoupled dot 1(2 ) multiplied by a correction factor that accounts for the coupling . when @xmath13 = 0 , and hence @xmath23 , eq@xmath6 ( [ eqddenergy ] ) reduces to @xmath24 + this is the sum of the energies of two independent dots . in the case when @xmath13 becomes the dominant capacitance ( @xmath25 ) , the electrostatic energy is given by @xmath26 ^ 2}{2(\widetilde{c}_{1}+\widetilde{c}_{2 } ) } \label{eq4}\ ] ] + this is the energy of a single dot with a charge @xmath27 and a capacitance of @xmath28 , where @xmath29 is the capacitance of dot 1(2 ) to the outside world . thus , a large inter - dot capacitance @xmath13 effectively leads to one big dot . + the electrochemical potential @xmath30 of dot 1(2 ) is defined as the energy needed to add the @xmath7th electron to dot 1(2 ) , while having @xmath31 electrons on dot 2(1 ) . using the expression for the total energy eq@xmath6 ( [ eqddenergy ] ) , the electrochemical potentials of the two dots are @xmath32 @xmath33 + the change in @xmath34 if , at fixed gate voltages , @xmath35 is changed by one , @xmath36 , is called the _ addition energy _ of dot 1 and equals the charging energy of dot 1 in this classical regime . similarly , the addition energy of dot 2 equals @xmath37 , and @xmath38 . in the next section we will discuss the addition energy in the quantum regime , where also the spacing between discrete energy levels plays a role . + from the electrochemical potentials in eqs@xmath6 ( [ eqmu1 ] ) and ( [ eqmu2 ] ) we construct a charge stability diagram , giving the equilibrium electron numbers @xmath39 and @xmath40 as a function of @xmath16 and @xmath41 . we define the electrochemical potentials of the left and right leads to be zero if no bias voltage is applied , @xmath42 = @xmath43 = 0 . hence , the equilibrium charges on the dots are the largest values of @xmath39 and @xmath40 for which both @xmath34 and @xmath44 are less than zero . if either is larger than zero , electrons escape to the leads . this constraint , plus the fact that @xmath39 and @xmath40 must be integers , creates hexagonal domains in the @xmath45-phase space in which the charge configuration is stable . + for completely decoupled dots ( @xmath13 = 0 ) the diagram looks as in fig@xmath6 [ honeycombs](a ) . the gate voltage @xmath8 changes the charge on dot 1(2 ) , without affecting the charge on the other . if the coupling is increased , the domains become hexagonal ( fig@xmath6 [ honeycombs](b ) ) . the vertices of the square domains have separated into ` triple - points ' . when @xmath13 becomes the dominant capacitance ( @xmath25 ) , the triple - point separation reaches its maximum ( see fig@xmath6 [ honeycombs](c ) ) . the double dot behaves like one dot with charge @xmath46 , as seen from eq@xmath6 ( [ eq4 ] ) . + we are considering the linear regime of conductance , implying @xmath47 . in order to obtain a measurable current , the tunnel barriers need to be sufficiently transparent . at the same time , however , the tunnel barriers need to be sufficiently opaque to ensure a well - defined electron number on each dot . for double dots coupled in series , a conductance resonance is found when electrons can tunnel through both dots . this condition is met whenever three charge states become degenerate , i@xmath6e@xmath6 whenever three boundaries in the honeycomb diagram meet in one point . in fig@xmath6 [ honeycombs](d ) two kinds of such triple - points are distinguished , ( @xmath48 ) and ( @xmath49 ) , corresponding to different charge transfer processes . at the triple point ( @xmath48 ) , the dots cycle through the sequence @xmath50 , which shuttles one electron through the system . this process is illustrated by the counterclockwise path @xmath51 and the diagram of an electron sequentially tunneling from the left lead to the right in fig@xmath6 [ honeycombs](d ) . at the other triple - point ( @xmath49 ) , the sequence is @xmath52 , corresponding to the clockwise path @xmath53 in fig@xmath6 [ honeycombs](d ) . this can be interpreted as the sequential tunneling of a hole in the direction opposite to the electron . the energy difference between both processes determines the separation between the triple - points ( @xmath48 ) and ( @xmath49 ) , and is given by @xmath19 , as defined in eq@xmath6 ( [ eq2 ] ) . the dimensions of the honeycomb cell ( see fig@xmath6 [ spacings ] ) can be related to the capacitances using eqs@xmath6 ( [ eqmu1 ] ) and ( [ eqmu2 ] ) . from @xmath54 we obtain @xmath55 and similarly we can derive @xmath56 from @xmath57 we obtain @xmath58 and similarly we can derive @xmath59 however , for a full characterization of all capacitances in the system an analysis in the non - linear transport regime is required , as is discussed in section [ nonlinear ] . + above we assumed that @xmath16 and @xmath41 only couple directly to the respective dots . in practice , however , there is a finite cross - capacitance from one gate to the other . the respective cross - capacitances result in a change of the slope of the charge domain boundaries in the honeycomb diagram . from figs@xmath6 [ honeycombs](b ) and [ spacings ] it is clear that both kinds of triple points ( @xmath48 and @xmath49 ) form a square lattice . however , with finite cross - capacitances the positions of the triple points move to lower @xmath8 for increasing @xmath60 ( at constant @xmath14 ) . the discussion of the stability diagram so far has been completely classical . however , the strong confinement of electrons in the dots can lead to the formation of a discrete energy spectrum . to account for the quantized energy states in the dot , we need to incorporate their energies in the electrochemical potential . the electrochemical potential for adding an electron into energy level @xmath61 of dot @xmath62 is denoted by @xmath63 . within the constant interaction model , @xmath63 is the sum of the classical electrochemical potential @xmath64 and the single - particle energy @xmath65 : @xmath66 . in the classical regime we found that the addition energy ( the change in electrochemical potential needed to add an extra electron ) equals the charging energy @xmath67 ( for dot 1 ) or @xmath37 ( for dot 2 ) . in the quantum regime , the addition energy for the @xmath68th electron occupying discrete level @xmath69 , with the @xmath35th electron occupying discrete level @xmath61 , becomes + @xmath70 + similarly , we find @xmath71 for the addition energy of dot 2 . note that for a ( spin-)degenerate level @xmath72 can be zero . the dimensions of the honeycomb cell as given in eqs@xmath6 ( [ eqdeltavg ] ) and ( [ eqdeltavgm ] ) , and depicted in fig@xmath6 [ spacings ] for the classical regime , change as follows @xmath73 + the electronic configuration that gives the lowest possible total energy in dot 1(2 ) , is referred to as the dot 1(2 ) _ ground state_. any configuration with a higher total energy is referred to as an _ excited _ state . the electrochemical potential for adding the @xmath7th electron to the lowest unfilled energy level of the @xmath74-electron ground state is labeled @xmath75 \{@xmath76}. the electrochemical potential for adding the @xmath7th electron to a higher unfilled level of the @xmath74-electron ground state or to any unfilled level of an @xmath74-electron excited state is labeled @xmath77 , @xmath78 , ... \{@xmath79 , @xmath80 , ... } . + in fig@xmath6 [ linear_regime ] a schematic diagram is given , showing the electrochemical potentials in the leads and dots in the linear regime ( @xmath47 ) . the ground state electrochemical potentials @xmath81 and @xmath82 align within the small bias window , allowing an electron to tunnel from left to right . this is an example of an electron transfer process as depicted in fig@xmath6 [ honeycombs](d ) . note that an alignment of an arbitrary combination of electrochemical potentials in dot 1 and dot 2 does not necessarily lead to a current . for example , the alignment of @xmath81 and @xmath83 does not result in current through the double dot . in the linear regime electron transport occurs via ground states , whereas the excited states start to play a role in non - linear transport , as will be discussed in section [ nonlinear ] . in the following discussion of the linear regime the ground state of dot 1(2 ) is denoted by @xmath84 ( without the discrete level index ) . + a more detailed picture of a honeycomb cell in the linear regime , marked by the dashed square in fig@xmath6 [ honeycombs](b ) , is given in fig@xmath6 [ honeyelch ] . the configuration of the ground - state electrochemical potentials is given in schematic diagrams on some places in the stability diagram . the dashed lines , which are extensions of the solid lines forming the honeycomb cells , help to find the position of both electrochemical potentials on a certain place . a crossing of dashed lines ( as in the charge domains ( 0,1 ) and ( 1,0 ) ) indicates that two electrochemical potentials align , but does not result in a current through the double dot . before discussing some experimental stability diagrams , we introduce the two kinds of lateral double dot devices being studied in this review . for the first type , only metal gate electrodes are used to confine the electrons in the 2deg beneath . for the second type we use a combination of metal gates and dry etching to realize confinement . a scanning electron microscope ( sem ) image of the first device is shown in fig@xmath6 [ devices](a ) . metal gates are deposited on top of a gaas / algaas heterostructure with a 2deg 100 nm below the surface @xcite . applying a negative voltage to all gates depletes the 2deg underneath them and forms two quantum dots . current can flow from the large electron reservoir on the left via the three tunnel barriers induced by the gate pairs 1-f , 2-f , and 3-f to the reservoir on the right . the transmission of each tunnel barrier can be controlled individually by the voltage on gates 1 , 2 or 3 . a single quantum dot can be defined in the 2deg by applying only a voltage to gates 1 , 2 , i and f ( dot 1 ) or to gates 2 , 3 , ii and f ( dot 2 ) . in this way , the individual dots can be characterized and their properties compared to those of the double dot . + the second device is schematically shown in fig@xmath6 [ devices](b ) . first a channel is defined in the 2deg by focused - ion - beam ( fib ) or electron - cyclotron - resonance ( ecr ) etching of an al@xmath85ga@xmath86as / gaas modulation - doped heterostructure @xcite . a double quantum dot can be formed by applying negative voltages to gates gl , gc and gr . + all experiments have been performed in a dilution refrigerator with a base temperature of 10 mk . the effective electron temperature in the leads is higher and can vary between @xmath8740 and @xmath87100 mk . a significant source of heating is the noise coming from the measurement electronics . the filters , used to attenuate the noise , have to be effective over a very large band width . they consist of a distributed rc network , usually a thin resistive wire going through a conducting medium such as copper powder or silver epoxy . the filters are installed at low temperature to minimize the thermal noise of the resistors inside . the filters are integrated with the sample holder in such a way that all sample wires are carefully shielded once they are filtered . to effectively create a double quantum dot , all gate voltages need to be tuned properly . the cross capacitances between the various gate electrodes , make it difficult to vary just a single parameter without affecting the others . the stability diagram is of great value in setting up and characterizing a double quantum dot . figure [ exphoney ] illustrates the process of the creation of a double dot in a device similar to the one in fig@xmath6 [ devices](b ) . starting point is the creation of a single large dot formed by the outer tunnel barriers , gl and gr . the measured stability diagram ( fig@xmath6 [ exphoney](a ) ) resembles fig@xmath6 [ honeycombs](c ) . the successive stability diagrams are measured for increasingly negative voltages on the middle gate electrode , gc , thus reducing the coupling between the dots . it is clearly seen that going from fig@xmath6 [ exphoney](a ) to fig@xmath6 [ exphoney]f , the stability diagram gradually evolves into the characteristic honeycomb structure . the edges of the honeycomb cells are visible due to off - resonance current . at the edges of a honeycomb cell , the electrochemical potential of one of the dots aligns with its neighboring lead ( see fig@xmath6 [ honeyelch ] ) . by a process called co - tunneling @xcite , transport can still take place via an intermediate virtual state . co - tunneling processes are suppressed by increasing the tunnel barriers ( i.e. making the gate voltages more negative ) , as can be seen from fig@xmath6 [ exphoney ] . figure [ expfrank](a ) shows a detail of a stability diagram obtained in the device shown in fig@xmath6 [ devices](a ) . the edges of the honeycomb cells are indicated by dashed lines . the triple points within the black square are well separated , whereas the other ones are still grown together . to separate also those points outside , the gate voltage on the middle barrier has to be tuned towards more negative values . we assume that the bias voltage is applied to the left lead ( @xmath88 ) and that the right lead is grounded ( @xmath43 = 0 ) . the bias voltage is coupled to the double dot through the capacitance of the left lead , @xmath89 , and hence also affects the electrostatic energy of the system . the bias dependence can be accounted for by replacing @xmath90 with @xmath91 in eq@xmath6 ( [ eqddenergy ] ) , where @xmath92 is the capacitance of the left lead to dot 1 ( see appendix a ) . + the conductance regions at finite bias change from triple - points to triangularly shaped regions ( fig@xmath6 [ lbschem ] ) . the conditions @xmath93 , @xmath94 , and @xmath95 = 0 determine the boundaries of the triangular regions . the dimensions of the triangles @xmath96 and @xmath97 ( see fig@xmath6 [ lbschem ] ) are related to the applied bias voltage as follows @xmath98 + where @xmath99 and @xmath100 are the conversion factors between gate voltage and energy . combining eqs@xmath6 ( [ eqdeltavg ] ) , ( [ eqdeltavgm ] ) and ( [ alphas ] ) , we can calculate the values of the total capacitances @xmath101 and mutual capacitance @xmath13 . for sufficiently large bias voltages , multiple discrete energy levels can enter the bias window . in this case , not only ground states , but also excited states contribute to the conductance . for the illustrative case of two levels per dot , the four possible alignments of the electrochemical potentials are shown in fig@xmath6 [ non - lin - schematic ] . note that the electrochemical potentials are drawn for the situation where one electron is on the double dot and a second one is tunneling on to it . due to coulomb blockade , not less than one and not more than a total of two electrons is allowed on the double dot . the labeling of the electrochemical potentials , using the notation introduced in section [ linquant ] is straightforward , except for tunneling through the excited state of dot 1 in figs@xmath6 [ non - lin - schematic](c),d . although the second electron is tunneling into the lowest level available in dot 1 , this level is only accessible because dot 1 is in an excited state . for that reason , we choose the label @xmath102 ( instead of @xmath103 ) . the successive alignment of ground and excited states leads to resonances within the conductance triangles , as shown in fig@xmath6 [ qtriangles ] . the off - resonance conductance in the grey triangles is due to inelastic processes @xcite and co - tunneling @xcite . note that @xmath14 is so large that the two triangles partly overlap . figure [ expfrank](b ) shows the triple points within the black rectangle of fig@xmath6 [ expfrank](a ) at finite bias . the triangular regions are clearly visible as well as the resonances within the triangles . the growth of the triangular regions with increasing bias voltage is illustrated in fig@xmath6 [ mutriangles ] . whereas only the ground state resonance is observed in fig@xmath6 [ mutriangles](a ) , multiple resonances appear within the triangles of fig@xmath6 [ mutriangles](b ) . in this section we discuss resonant tunneling experiments through the double dot of fig@xmath6 [ devices](a ) with discrete energy levels @xcite . we show that , under appropriate conditions , the resonance widths are only determined by the lifetime of the discrete energy states , independent of the electron temperature in the leads . a small asymmetric deviation from the lorentzian resonance shape is attributed to inelastic tunnel processes . + the current - voltage ( @xmath104-@xmath14 ) curves of the single quantum dots in fig@xmath6 [ nijs2 ] provide two clear signatures for the presence of both coulomb blockade effects and discrete levels . at low bias voltages , the current through the dot is suppressed by the coulomb blockade @xcite . increasing the bias voltage lifts the blockade . the current shows a stepwise increase : each time when an additional level enters the bias window @xmath105 , an extra transport channel is opened and the current increases @xcite . hence , the voltage spacing of the current steps directly reflects the energy spacing of the levels . for the average level spacing , @xmath106 , we obtain @xmath107 = 125 @xmath108ev for dot 1 ( upper inset ) and @xmath109 = 225 @xmath108ev for dot 2 ( lower inset ) . the difference in these two energies reflects the different lithographic sizes of the two dots ( see fig@xmath6 [ devices](a ) ) . accounting for the depletion areas , we estimate that dot 1 has an effective diameter of 240 nm and contains about @xmath39 = 90 electrons , while dot 2 has an effective diameter of 200 nm and contains roughly @xmath110 = 60 electrons . using the fermi energy , @xmath111 , at bulk density , we estimate @xmath112 = 150 @xmath108ev and @xmath113 = 230 @xmath108ev . this is in good agreement with the estimates obtained from the @xmath104-@xmath14 curves . from the dimensions of the coulomb diamonds @xcite we obtain the charging energies @xmath114 for adding an electron to the dot : @xmath67 = 1.1 mev ( dot 1 ) and @xmath37 = 1.8 mev ( dot 2 ) . we focus on the role of the discrete levels and consider the charging energies as constant offsets in the transport conditions . figure [ nijs2 ] shows an @xmath104-@xmath14 curve of the double dot with all three tunnel barriers set in the weak - tunneling regime . the coulomb blockade suppresses the current through the double dot at low bias voltages . at larger bias the current shows sharp resonances . the spacing of the resonances is about 250 @xmath108ev . this is of the same order as the level spacing in the single dots . + the same resonances are seen when we sweep the gate voltage . figure [ nijs3 ] shows the current through the double dot versus the gate voltage on gate 1 , @xmath16 , with @xmath14 = 280 @xmath108ev . this corresponds to a vertical cut through a stability diagram as shown in fig@xmath6 [ mutriangles ] . the current shows three groups of sharp resonances separated by regions of zero current with a period @xmath115 = 9 mv in gate voltage @xmath16 . with only dot 1 formed , we observe coulomb oscillations as a function of @xmath16 with the same period @xmath115 ; each period thus corresponds to a change of one electron in dot 1 , while keeping the number of electrons on dot 2 , @xmath110 , constant . @xmath115 corresponds to the horizontal dimension of the honeycomb unit cell , as indicated in fig@xmath6 [ spacings ] in section [ linclass ] . when elastic tunnel processes are the dominant transport mechanism , the current through the double dot is resonantly enhanced only when two levels in dot 1 and 2 align , as explained in section [ nonlinquant ] . tuning the level alignment with @xmath14 or @xmath16 gives rise to the sharp resonances in figs@xmath6 [ nijs2 ] and [ nijs3 ] . resonant tunneling through the double dot is illustrated in the schematic potential landscape of the double dot in fig@xmath6 [ nijs1b ] . this figure shows a few of the levels in dot 1 ( levels 1 to 5 ) and dot 2 ( levels @xmath116 and @xmath117 ) . the electrostatic potentials @xmath118 and @xmath119 are tuned in such a way that transport through the double dot is possible only via the charge states @xmath120 . the finite bias voltage gives an electron from the left reservoir three choices to tunnel into dot 1 : it can tunnel to one of the unoccupied levels 3 , 4 or 5 . this changes the electrostatic potential @xmath118 by the charging energy @xmath67 ( the levels are drawn at the positions applicable after an electron has occupied one of them ) . when dot 1 relaxes to the ground state ( the incoming electron occupying level 3 ) , the electron can tunnel via level @xmath116 to the right reservoir . note that if dot 1 does not immediately relax to the ground state , but remains in an excited state , with an electron occupying either level 4 or 5 , electron transport through the double dot is temporarily blocked . the electron is ` trapped ' within dot 1 . only after relaxation to the ground state a next tunnel event can occur . if the relaxation rate is small on the scale of the tunneling rates through the barriers , the inclusion of levels 4 and 5 within the bias window could therefore lead to a _ decrease _ of the current through the double dot . on the other hand , in case of fast relaxation , the enhanced tunnel probability when also the levels 4 and 5 lie within the bias window , could lead to an increase in the current . note that next to intra - dot relaxation , also inelastic tunneling from level 4 or 5 to either level @xmath116 or @xmath117 can occur . this process is accompanied by emission of a boson ( usually phonons @xcite ) and contributes to the off - resonance current in fig@xmath6 [ nijs3 ] . the resonances in a particular group in fig@xmath6 [ nijs3 ] can be identified with the energy diagram of fig@xmath6 [ nijs1b ] . the first resonance occurs when level 3 aligns with level @xmath116 ( peak 3-@xmath116 ) . this corresponds to the rightmost peak in fig@xmath6 [ nijs3 ] . increasing @xmath121 by making @xmath16 more negative , brings transport off - resonance until level 2 aligns with @xmath116 ( peak 2-@xmath116 ) followed by the third peak 3-@xmath117 . continuing to sweep @xmath16 increases the energy of level 3 above the electrochemical potential @xmath42 of the left reservoir . this blocks transport and removes an electron from dot 1 permanently . the next group of resonances is observed when @xmath16 is changed by one coulomb oscillation period @xmath122 ( see fig@xmath6 [ nijs3 ] ) . note that the number of resonances decreases in the next two groups . sweeping @xmath16 also shifts the levels in dot 2 , due to a small cross - capacitance between gate 1 and 2 . transport is possible until level @xmath116 is shifted above @xmath42 ( @xmath16 @xmath123 -470 mv ) . + the level spacing is obtained by converting gate voltage to energy @xcite . this yields an energy separation of resonances 2-@xmath116 and 3-@xmath116 by 70 @xmath108ev , which is the energy separation of levels 2 and 3 . in the same way we find for levels @xmath116 and @xmath117 a separation of 200 @xmath108ev . both values are in good agreement with the typical values we found above . on increasing @xmath14 , we observe that the number of resonances in a particular group increases . the inset to fig@xmath6 [ nijs3 ] shows approximately 11 resonances as @xmath124 is swept . these observations are in agreement with the resonant tunneling picture of fig@xmath6 [ nijs1b ] : when @xmath14 is larger , more levels can align . generally , the relaxation rate to the ground state is not necessarily higher than the tunnel rate through the dot . in fig@xmath6 [ waaier ] the amplitude of the ground state resonance ( see lower curve ) clearly decreases with increasing bias voltage . at the same time new resonances appear , having a larger amplitude than the ground state resonance . this implies that in general transport through excited states can play a significant role . when a discrete level in dot 1 is at a distance much larger than the thermal energy , @xmath125 , from the electrochemical potential of the left lead , dot 1 acts as low - temperature - pass filter for dot 2 , such that only cold electrons contribute to the current @xcite . if the energy levels in the dots are separated by more than @xmath125 , the occupation of excited states becomes suppressed . this effectively leaves the dot at zero temperature . hence , the double dot geometry allows for an accurate measurement of the intrinsic line width of the discrete levels , which is not averaged by the fermi - dirac distribution of the electrons in the reservoirs . in other words , an energy resolution better than @xmath126 can be obtained . the line shape of a resonance is lorentzian when only elastic tunneling is important @xcite . in our geometry ( assuming a bias voltage sufficiently large such that electrons must tunnel from the left lead to the left dot to enter the system , and must tunnel from the right dot to the right lead to leave it again ) the current is given by @xmath127 + where @xmath128 is the energy difference between two discrete energy levels in the two dots , @xmath129 is the tunnel rate from the left lead to dot 1 , @xmath130 is the modulus of the tunnel coupling between the two dots and @xmath131 is the tunnel rate from dot 2 to the right lead . note that for elastic tunneling the resonance width is only determined by the lifetime of the 0d states and independent of temperature . + figure [ nijs4 ] shows a single resonance ( black dots ) . the right - hand side of the peak fits very well with the lorentzian line shape of eq@xmath6 ( [ lorentzian ] ) ( solid line ) , while the left - hand side shows a deviation from the lorentzian fit . the only free fit parameter is the full width at half maximum , fwhm = 5 @xmath108ev . from the maximum current and the width of the resonance we find with eq@xmath6 ( [ lorentzian ] ) a tunnel coupling @xmath130 @xmath132 0.2 @xmath108ev and a tunneling rate between the right dot and the right lead @xmath133ev . for comparison , we have fitted the resonance with a thermally broadened resonance @xmath134 cosh(@xmath135 ( dashed line ) @xcite . the top is fit very well for @xmath136 = 34 mk , but there is a large deviation in both tails of the resonance . on the right - hand side , the deviation can be accounted for by the lorentzian broadening . at the left - hand side , the deviation consists of two components . the first one is the same lorentzian broadening as observed on the right - hand side . the second one , however , is an asymmetric contribution only occurring on this side of the resonance . the asymmetric contribution to the current appears at the side where an electron tunnels from a higher to a lower electrochemical potential . upon reversing the sign of @xmath14 , we find that the asymmetry appears at the other side of the resonance . as shown in ref@xmath6 @xcite , this is due to inelastic tunnel processes . in such a process , an electron can tunnel inelastically and spontaneously emit its energy as a photon or a phonon . in this section we measure the energy evolution versus magnetic field , @xmath137 , of energy states near the fermi energy , @xmath111 , in the double quantum dot shown in fig@xmath6 [ devices](a ) . as a function of @xmath137 and the voltage on gate 3 , @xmath124 , we observe crossings and anti - crossings between coulomb peaks . the resolution is high enough that avoided crossings in the spectrum of a quantum dot can be resolved @xcite . to our knowledge , these are the only existing data revealing intra - dot level repulsion in a quantum dot system . + the experiments are performed in the weak coupling limit , such that mixing between quantum states in one dot with states in the other dot or in the leads is negligible ( see section [ spectro ] ) . we sweep the gate voltages over small ranges and focus on a particular charging transition ; i@xmath6c@xmath6 transitions between ( @xmath39 + 1,@xmath40 ) and ( @xmath39,@xmath138 + 1 ) only . since we discuss only one transition at a time , we can , for simplicity , leave out the coulomb energies from the discussion and concentrate on the alignment of discrete energy levels . + using the notation introduced in section [ linquant ] , we label the accompanying electrochemical potentials @xmath139 for dot 1 and @xmath140 for dot 2 ( or simply @xmath141 ) . the condition for tunneling between the lowest possible states , i@xmath6e@xmath6 from ground state to ground state , is @xmath142 = @xmath143 . to tunnel from the first excited state of dot 1 to the ground state of dot 2 , the condition becomes @xmath144 = @xmath143 . the changes in the electrochemical potential @xmath145 and @xmath146 are typically @xmath87 150 - 200 @xmath108ev . + figure [ bspectrofig](a ) shows a typical set of current traces for different magnetic fields while sweeping @xmath124 . the bias voltage @xmath14 = 1.2 mv is such that several discrete levels in each dot are between the electrochemical potentials of the two leads . this is similar as in the inset to fig@xmath6 [ nijs3 ] , but now measured for different magnetic fields . the _ change _ in coulomb peak position versus @xmath137 , @xmath147 , is proportional to the difference in the @xmath137-evolution of the electrochemical potentials @xmath148 and @xmath149 @xmath150 - [ \mu_{2,n}(b+\delta b ) - \mu_{2,n}(b ) ] \nonumber \\ = & \hspace{0.1 cm } \alpha \delta v_{g3}^{peak}(\delta b)\end{aligned}\ ] ] where @xmath116 is the conversion factor between @xmath124 and the electrostatic potential of dot 2 . note that if the states @xmath148 and @xmath149 have the same @xmath137-dependence , the coulomb peak position does not change . the energy resolution of @xmath151 is @xmath87 5 @xmath108ev , corresponding to @xmath152 50 mk . + the data in fig@xmath6 [ bspectrofig](a ) contain several interesting features . first , we observe crossings between different peaks as well as anti - crossings ( two are indicated by arrows ) . second , pairs of peaks exhibit the same @xmath137-dependence . these are general features that we observe at several charge transitions ( that is for several choices of @xmath39,@xmath40 ) . independent measurements on _ one _ of the individual dots also show states evolving in pairs below @xmath153 t. the observed pairing and ( anti-)crossing of the coulomb peaks in the double dot experiments can then be explained as shown schematically in fig@xmath6 [ bspectrofig](b ) . suppose two levels in one dot have an anti - crossing in their @xmath137-dependence . then two paired levels in the other dot , having the same @xmath137-dependence , both probe this anti - crossing . at the points where two coulomb peaks actually cross , two levels in dot 1 align with two levels in dot 2 simultaneously ( though only one electron can tunnel at a time , due to coulomb blockade ) . + for the interpretation of the data as schematically given in fig@xmath6 [ bspectrofig](b ) , tunneling through the excited state of dot 1 is a key ingredient . if dot 1 would relax to its ground state much faster than the tunnel rate through the barriers after an electron has tunneled onto it via @xmath154 , we would only observe the two lower traces in the rightmost diagram of fig@xmath6 [ bspectrofig](b ) . however , our data suggest a relatively slow relaxation rate between the excited and ground state of dot 1 . recent experiments on ( single ) quantum dots have shown that indeed relaxation times can be of the order of @xmath108s or longer when relaxation to the ground state involves electron spin flips @xcite . the condition for the relaxation rate from the excited state to the ground state in dot 2 is more subtle . the electron can tunnel onto dot 2 via @xmath155 and leave dot 2 either directly or after a relaxation process from @xmath155 to @xmath156 has occurred ( this second possibility requires @xmath157 ) . for transport through the double dot , the relaxation rate in dot 2 does not necessarily need to be slow as well . however , if the relaxation rate would be too high , the anti - crossing as shown in the middle diagram of fig@xmath6 [ bspectrofig](b ) would smear out . the constant level spacing @xmath158 in dot 1 ( see left diagram of fig@xmath6 [ bspectrofig](b ) ) could be explained by an exchange energy , e.g. when the upper level would correspond to a spin singlet state and the lower level to a spin triplet @xcite . in this section we present microwave ( 0 - 50 ghz ) spectroscopy experiments @xcite on double quantum dots for different coupling and microwave power regimes @xcite . we use photon assisted tunneling ( pat ) processes , as described in sections [ weakcpat ] and [ strongcpat ] , to measure the energy differences between states in the two dots of the devices shown in figs@xmath6 [ devices](a ) and [ devices](b ) . depending on the strength of the inter - dot coupling , the two dots can form ionic - like @xcite or covalent - like bonds @xcite . in the first case , the electrons are localized on individual dots , while in the second case , the electrons are delocalized over both dots . the covalent binding leads to a symmetric and anti - symmetric state , whose energy difference is proportional to the tunneling strength between the dots . + for the microwave experiments we make use of a coaxial cable . from room temperature to the 1k - pot , a 0.085 inch semi - rigid be - cu ( inner and outer conductor ) coaxial cable is used . from the 1k - pot to the mixing chamber , we use a 0.085 inch semi - rigid stainless steel ( inner and outer conductor ) coax . from the mixing chamber to the sample , various types of low attenuation semi - rigid or flexible coaxial cable can be used , since here the thermal conductivity is no longer a constraint . finally , the coaxial cable is capacitively coupled ( typically through a 10 pf capacitor ) to one of the gate electrodes of the sample , usually the center gate . this gate is capacitively coupled to both dots , and hence part of the incident power can generate a microwave oscillating potential across the center barrier . so far , we assumed a purely electrostatic coupling between both dots , whereas tunnel coupling was neglected . however , when electrons can tunnel coherently from one dot to the other at appreciable rates , the eigenstates become delocalized , extending over the entire double dot system . in principle , these are quantum - mechanical many - body states of the two coupled dots . it is very difficult to give a full description of such a many - body system . therefore , we discuss here the elementary case of a quantum - mechanical two - level system , which is quite adequate in grasping the physics of a tunnel - coupled double dot . basically , we only take into account the topmost occupied level in each dot and neglect the interaction with electrons in lower energy levels . + we consider a double dot consisting of two well - separated dots , described by a total hamiltonian @xmath159 @xcite , with eigenstates @xmath160 and @xmath161 , and eigenenergies @xmath162 and @xmath163 ( fig@xmath6 [ 2levelfig](a ) ) @xmath164 + we introduce a finite tunnel coupling between the levels in both dots described by the hermitian matrix @xmath165 , which for simplicity @xcite we assume to be purely non - diagonal @xmath166 + one obtains a new hamiltonian , @xmath167 , with delocalized eigenstates @xmath168 ( symmetric state ) and @xmath169 ( anti - symmetric state ) , and eigenvalues @xmath170 and @xmath171 @xmath172 the new eigenvalues can be expressed in terms of the eigenvalues of the uncoupled double dot and the tunnel matrix elements as follows @xmath173 + where @xmath174 and @xmath175 and @xmath176 . the eigenstates @xmath168 and @xmath177 in the basis of @xmath178 and @xmath179 are written @xmath180 + with @xmath181 . figure [ 2levelfig](b ) shows the eigenenergies of the coupled two - level system as function of @xmath128 . the renormalized energy difference , @xmath182 , is given by @xmath183 note that the effect of the coupling is stronger for small @xmath128 , i@xmath6e@xmath6 close to the crossing of the unperturbed energies @xmath162 and @xmath163 . where @xmath162 and @xmath163 cross ( @xmath128 = 0 ) , we have an anti - crossing of @xmath184 and @xmath185 , with @xmath186 = @xmath187 , the minimum bonding anti - bonding energy difference . for large @xmath128 , the eigenenergies of the coupled double dot approach the eigenenergies of the uncoupled dots , @xmath162 and @xmath163 . the general solution of the _ time - dependent _ schrdinger equation can be written in the form @xmath188 with eq@xmath6 ( [ eigenstates ] ) , @xmath189 can be expressed in terms of @xmath190 and @xmath191 . since @xmath192 and @xmath191 are not eigenstates of the total hamiltonian @xmath193 , they are no longer stationary states . if the system is in state @xmath192 at time @xmath194 = 0 ( @xmath195 ) the probability @xmath196 of finding it in the state @xmath197 at time @xmath194 is @xmath198 . \label{eqrabi}\ ] ] + equation ( [ eqrabi ] ) describes a coherent charge oscillation in the double dot system . if the inter - dot coupling is weak , electrons are strongly localized on the individual dots . in section [ linquant ] we saw that we expect a resonant current through the double dot system if @xmath199 . if we represent a weakly coupled double dot by a two - level system , we need the discrete energy levels @xmath162 and @xmath163 to align within the bias window . we will only consider the discrete , quantum contribution to the electrochemical potentials and therefore simply use the discrete level notation @xmath162 and @xmath163 instead of @xmath200 and @xmath201 . + an additional time - varying potential @xmath202cos@xmath203 can induce _ inelastic _ tunnel events when electrons exchange photons of energy @xmath204 with the oscillating field ( frequency , @xmath205 , is typically 1 - 75 ghz in our experiments ) . this inelastic tunneling with discrete energy exchange is known as photon assisted tunneling ( pat ) @xcite . pat through a single quantum dot with well resolved discrete 0d - states is reviewed in ref . pat is an invaluable spectroscopic tool for studying the energy spectra of quantum dots . a theoretical study of pat in double dots is given in refs@xmath6 @xcite . a voltage drop @xmath202cos@xmath206 across a tunnel barrier modifies the tunnel rate through the barrier as @xcite @xmath207 + here @xmath208 , and @xmath209 and @xmath210 are the tunnel rates at energy @xmath211 with and without an ac voltage , respectively . j@xmath212 is the square of the @xmath61th order bessel function of the first kind , evaluated at @xmath213 , which describes the probability that an electron absorbs ( @xmath214 ) or emits ( @xmath215 ) @xmath61 photons of energy @xmath204 . thus , the effect of the interaction between a single - electron state with a classical , oscillating field is that the energy state @xmath211 is split in a set of states @xmath216 ( see inset to fig@xmath6 [ besselfig ] ) . the power of pat as a spectroscopic tool lies in the fact that pat can only take place if the energy difference @xmath128 equals an integer number times the photon energy @xmath204 : @xmath217 , see fig@xmath6 [ weakcdiagrams ] . for the multiple photon processes ( @xmath218 ) to take place , the microwave power needs to be sufficiently large . + to use pat as a spectroscopic tool , we can make use of the configurations shown in fig@xmath6 [ weakcdiagrams ] . in the pumping configuration @xcite , the double dot is operated at zero bias voltage . absorption of a photon with energy @xmath219 leads to pumping of an electron from left to right ( fig@xmath6 [ weakcdiagrams](a ) ) or vice versa ( fig@xmath6 [ weakcdiagrams](b ) ) . the advantage of this configuration is that relaxation due to spontaneous emission does not contribute to the current . figure [ pathoney ] schematically shows how the honeycomb unit cell of fig@xmath6 [ honeyelch ] changes in the presence of a microwave field . + alternatively , the double dot can be operated in the large bias regime as depicted in fig@xmath6 [ weakcdiagrams](c),d . in this regime , in the case of weak coupling with @xmath220 , the dc pat current is given by @xcite @xmath221 + the current is composed of a number of satellite peaks , separated by the photon energy @xmath204 and all with width @xmath222 . note that the satellite peaks can become of the same order of magnitude as the main resonance , but that they have a smaller width than the main resonance . the pat experiment described below , is performed on a weakly coupled dot in the large bias regime . the double dot , shown in fig@xmath6 [ devices](a ) , is tuned such that only one level in each dot contributes to electron transport . the gate voltages are used to shift the level in dot 1 and in dot 2 . the resonance in the lowest trace in fig@xmath6 [ weakcoup1](a ) arises from the alignment of the two levels . the other traces are measured while applying a microwave signal . the satellite resonances are due to pat processes which involve the emission ( left satellite peak ) or absorption ( right satellite peak ) of one photon . figure [ weakcoup1](b ) shows that the energy separation of the satellite peaks from the main peak , @xmath128 , depends linearly on frequency between 1 and 50 ghz . as we will discuss below , this linearity implies that the tunnel coupling is negligible . the electrons are localized on the individual dots and they have an ionic bonding . the line proportional to @xmath223 is taken from data at higher microwave powers where electrons absorb or emit two photons during tunneling . + as the microwave power is increased , more satellite peaks appear corresponding to the absorption of multiple photons , which are observed up to @xmath61 = 11 ( see fig@xmath6 [ weakcoup2 ] ) . a high power microwave field strongly perturbs tunneling . this is reflected by the non - linear dependence of the peak heights on microwave power . in the right inset to fig@xmath6 [ weakcoup2 ] the peak heights of the main peak and the first four photon satellite peaks are shown , which agree well with the expected squared bessel function behavior shown in fig@xmath6 [ besselfig ] . the large bias configuration of figs@xmath6 [ weakcdiagrams](c),d was successfully employed to study pat in a weakly coupled double dot system . for the microwave spectroscopy of a strongly coupled double dot we will make use of the pumping configuration shown in fig@xmath6 [ strongcdiagrams ] . with increasing the coupling between the dots , the spontaneous emission rate from the higher level to the lower one increases as well . the advantage of the pumping configuration is that these processes are ` filtered out ' and do not contribute to the current . figure [ strongcdiagrams](a ) schematically shows the symmetric and anti - symmetric states in the double dot for @xmath224 . when microwave radiation is applied with a frequency such that @xmath225 , electrons are pumped from the left lead to the right lead and vice versa . since the weight of the symmetric and anti - symmetric wave function is distributed equally over both dots ( see fig@xmath6 [ strongcdiagrams](a ) ) , there is no net current . however , if we detune the levels ( @xmath226 ) the weight gets distributed asymmetrically as shown in figs@xmath6 [ strongcdiagrams](b),c and a net current is generated by applying microwave radiation matching @xmath227 . note that for frequencies @xmath228 no pat is possible . + multi - photon processes occur if the condition @xmath229 = @xmath230 ( @xmath218 ) is met . besides allowing for these higher order photon processes , a high power microwave field also renormalizes the tunnel coupling to a smaller value . the energy splitting @xmath231 now becomes @xmath232^{2}}. \label{eqdeltaestarhp}\ ] ] the experiments for strong inter - dot coupling are performed on the device shown in fig@xmath6 [ devices](b ) . to single out the current only due to microwaves , we operate the device as an electron pump driven by photons @xcite ( see the diagrams in fig@xmath6 [ strongcdiagrams ] ) . an electron is excited from the bonding to the anti - bonding state if the condition @xmath233 is fulfilled , or conversely @xmath234 figure [ strongchoney ] schematically shows the stability diagram for a strongly coupled double dot in the presence of a microwave field . here we assume that the symmetric and anti - symmetric states can only be occupied by a single electron . in other words , we assume spinless electrons . the triple points of the weakly coupled double dot , denoted by @xmath48 and @xmath49 , develop into a black and a white crescent , respectively . the length of these crescents increases with @xmath130 . + moving along the dotted line connecting the crescents from lower left to upper right , first the symmetric state aligns with the electrochemical potentials of the leads ( at the black crescent ) . current through the double dot is possible via the electron transfer process of fig@xmath6 [ honeycombs](d ) . at the black square , the anti - symmetric state aligns with the leads . however , current is blocked , since an extra electron is already added to the double dot and the charging energy @xmath19 is not available yet . at the white square , the electrochemical potential for adding the second electron to the symmetric state aligns with the leads . as we assumed single occupation of the delocalized states , current is blocked here as well . when arriving at the white crescent , the electrochemical potential for adding the second electron to the double dot in the ( empty ) anti - symmetric state becomes available . this enables the hole transfer process of fig@xmath6 [ honeycombs](d ) . + the black bars at a distance @xmath235 from the dotted line connecting the crescents , denote the places where @xmath236 and pat occurs . note that pat is only possible if the photon energy exceeds the coupling energy , @xmath237 . depending on the level configuration , pumping results in a negative or positive contribution to the current , as shown in fig@xmath6 [ toshipat](a ) ( here we choose @xmath238 for an electron moving from left to right ) . figure [ toshipat](b ) shows a part of the corresponding stability diagram between two triple points , clearly showing the energy regions of constant charge and extra transport lines due to pat . figure [ patpeaks ] shows measured current traces as a function of the uncoupled energy splitting @xmath128 , where from top to bottom the applied microwave frequency is decreased from 17 to 7.5 ghz in 0.5 ghz steps . at the highest frequencies , the distance between the pumping peaks is close to 2@xmath128 . however , the peak distance decreases faster than linearly as the frequency is lowered ; in fact the peaks follow the dotted hyperbola rather than the dashed straight lines . the distance goes to zero when the frequency approaches the minimum energy gap between bonding and anti - bonding states , @xmath239 . the coupling between the dots can be decreased by changing the gate voltage on the center gate to more negative values , or by applying a magnetic field perpendicular to the sample . in fig@xmath6 [ strongcouphp](a ) we plot half the spacing between the positive and negative satellite peaks as a function of frequency . the microwave power is kept as low as possible in order to meet the condition @xmath240 . in that case j@xmath241 and the general relation eq@xmath6 ( [ eqdeltaehp ] ) reduces to @xmath242 + different symbols correspond to different center gate voltage settings and magnetic fields . the solid lines are fits to eq@xmath6 ( [ eqdeltaelp ] ) . the good agreement with eq@xmath6 ( [ eqdeltaelp ] ) demonstrates the control over the formation of a covalent bonding between the two dots and that the condition @xmath240 is satisfied . we now discuss the case @xmath243 . as can be seen in fig@xmath6 [ besselfig ] , j@xmath244 deviates from 1 in this case and can not be neglected as before @xcite . in fig@xmath6 [ strongcouphp](b ) we show the power dependence for the case of a coupling of 60 @xmath108ev and a microwave frequency of 16 ghz , as indicated by the circle in fig@xmath6 [ strongcouphp](a ) ( similar results have been obtained for other couplings and microwave frequencies ) . the inset to fig@xmath6 [ strongcouphp](b ) shows the measured pat current as a function of @xmath128 for different powers . the absolute value of the microwave power at the position of the double quantum dot is unknown . therefore , we use a relative microwave power scale , which is expressed in terms of the attenuation of the microwave source signal . the positions of the pat peaks at the lowest power are indicated with two dashed lines . increasing the microwave power from the lowest value , the pat peak separation becomes larger , which is in agreement with eq@xmath6 ( [ eqdeltaehp ] ) . for higher powers , multi - photon processes can also take place , which result in extra current peaks . figure [ strongcouphp](b ) shows half the pat peak separation energy as function of the relative microwave power . the solid line is a fit with eq@xmath6 ( [ eqdeltaehp ] ) , @xmath205 = 16 ghz , @xmath187 = 60 @xmath108ev . because of the relative power scale , the fitting curve has been adjusted horizontally to give the best fit . we thus see that the microwave power effectively reduces the coupling between the dots . this is further illustrated by the vertical dotted line in fig@xmath6 [ strongcouphp](a ) at @xmath205 = 16 ghz . at -33 db the energy separation equals @xmath204 , which implies that the higher microwave power has effectively reduced the coupling between the dots . we can obtain an estimate of the power by noting that j@xmath244 has its first zero for @xmath213 = 2.4 and hence @xmath202 = 0.16 mv . by coupling two quantum dots in series , we obtain a system with fundamentally different behavior and possibilities in comparison to a single quantum dot . in this review we have discussed the superiority of a double quantum dot system in determining the intrinsic lifetime of quantum states and in probing intra - dot level repulsion . next to the added value as a spectroscopic instrument , the double dot manifests itself as an artificial molecule . by changing the inter - dot coupling , we have been able to tune the double dot from an ionic - like bonded to a covalent - like bonded molecule . + now that the ability to create and manipulate double quantum dots has been shown , the next challenge lies in the study and time - control of coherent phenomena in these systems . double quantum dots have been suggested as possible candidates for building blocks of a quantum computer @xcite . we have shown that it is indeed possible to coherently couple dots , and that one can induce transitions between the extended states . the next crucial step towards quantum logic gates is to show that the coherence of the superposition is preserved on time scales much longer than the time needed for manipulating the electron wave functions . the time - resolved measurement of coherent charge oscillations in double quantum dots using pulsed gate voltages , will be an essential step in determining the dephasing time in these systems . + in addition , the role of the electron spin and its relaxation and dephasing times ( @xmath245 and @xmath246 , respectively ) need to be characterized , as they are essential in the proposed quantum bit schemes based on coupled electron spins in double dots @xcite . the relaxation time @xmath245 in quantum dots has been experimentally shown to exceed 10 - 100 @xmath108s if the relaxation involves a spin - flip @xcite . the determination of the spin dephasing time @xmath246 , which is the most relevant timescale for quantum computing purposes , will be an experimental challenge for the near future . we conclude that our work on double dots so far , in combination with the results on the spin relaxation times in this kind of systems , forms a promising point of departure for further study on the suitability of double quantum dots as quantum coherent devices . we would like to thank r. aguado , s.m . cronenwett , d.c . dixon , s. godijn , p. hadley , c.j.p.m . harmans , r.v . hijman , y. hirayama , k. ishibashi , m.p . janus , k.k . likharev , f. mallens , j.e . mooij , yu.v . nazarov , t.h . oosterkamp , r.m . schouten , t.h . stoof , m.j . uilenreef , n.c . van der vaart and l. vandersypen for their help . we acknowledge financial support from the darpa grant number daad19 - 01 - 1 - 0659 of the quist program , the specially promoted research grant - in - aid for scientific research ; the ministry of education , culture , sports , science and technology in japan ; the dutch organization for fundamental research on matter ( fom ) ; the core research for evolutional science and technology ( crest - jst ) ; and the european union through a training and mobility of researchers ( tmr ) program network . in this appendix we derive the electrostatic energy of a single and double quantum dot system . before addressing these specific systems , we briefly discuss the method followed . hadley / set / electrostatics.html . ] consider a system consisting of @xmath247 conductors . a capacitance can be defined between each conductor and every other conductor as well as a capacitance from each of the @xmath247 conductors to ground . this results in a total of @xmath248 capacitors . the capacitor between node @xmath249 and node @xmath250 has a capacitance @xmath251 and stores a charge @xmath252 . the total charge on node @xmath249 is the sum of the charges on all of the capacitors connected to node @xmath249 @xmath253 here @xmath254 is the electrostatic potential of node @xmath249 and ground is defined to be at zero potential , @xmath255 . the charges on the nodes are linear functions of the potentials of the nodes so this can be expressed more compactly in matrix form @xmath256 where * c * is called the capacitance matrix . a diagonal element of the capacitance matrix , @xmath257 , is the total capacitance of node @xmath249 @xmath258 an off - diagonal element of the capacitance matrix is minus the capacitance between node @xmath249 and node @xmath250 , @xmath259 . the electrostatic energy of this system of conductors is the sum of the electrostatic energy stored on the @xmath260 capacitors and can be conveniently expressed using the capacitance matrix @xmath261 voltage sources can be included in the network by treating them as nodes with large capacitances to ground and large charges on them such that @xmath262 . in this case , it is numerically difficult to compute the inverse of the capacitance matrix since it contains large elements . however , it is not necessary to invert the entire capacitance matrix since the voltages on the voltage sources are already known . only the voltages on the other nodes need to be determined . these voltages can be determined by writing the relation between the charges and the voltages as @xmath263 here @xmath264 and @xmath265 are the charges and the voltages on the charge nodes , @xmath266 and @xmath267 are the charges and the voltages on the voltage sources , and the capacitance matrix has been expressed in terms of four sub - matrices . the voltages on the charge nodes are then @xmath268 and the electrostatic energy can be calculated with eq@xmath6 ( [ appenergy ] ) . we write the total charge @xmath269 on the dot as the sum of the charges on all the capacitors connected to the dot ( see fig@xmath6 [ appsd ] ) @xmath270 where @xmath271 is the total capacitance coupled to the dot , @xmath272 . the capacitance matrix @xmath273 only has one element . using eq@xmath6 ( [ appenergy ] ) and substituting @xmath274 , we find @xmath275 ^ 2}{2 c_1}\ ] ] where @xmath276 is the number of electrons on the dot when all voltage sources are zero , which compensates the positive background charge originating from donors in the heterostructure . we write the total charge @xmath277 on dot 1(2 ) as the sum of the charges on all the capacitors connected to dot 1(2 ) ( see fig@xmath6 [ appdd ] ) @xmath278 we can write this as @xmath279 where @xmath280 . the above expression in the form of eq@xmath6 ( [ voltage ] ) reads @xmath281 the electrostatic energy of the double dot system can now be calculated using eq@xmath6 ( [ appenergy ] ) . for the case @xmath282 and @xmath283 this becomes @xmath284 with @xmath285 +

we investigate the bose einstein condensation on non homogeneous non amenable networks for the model describing arrays of josephson junctions . the graphs under investigation are obtained by adding density zero perturbations to the homogeneous cayley trees . the resulting topological model , whose hamiltonian is the pure hopping one given by the opposite of the adjacency operator , has also a mathematical interest in itself . the present paper is then the application to the bose einstein condensation phenomena , of the harmonic analysis aspects , previously investigated in a separate work , for such non amenable graphs . concerning the appearance of the bose einstein condensation , the results are surprisingly in accordance with the previous ones , despite the lack of amenability . the appearance of the hidden spectrum for low energies always implies that the critical density is finite for all the models under consideration . first we show that , even when the critical density is finite , if the adjacency operator of the graph is recurrent , it is impossible to exhibit temperature states which are locally normal ( i.e. states for which the local particle density is finite ) describing the condensation at all . a similar situation seems to occur in the transient cases for which it is impossible to exhibit locally normal states describing the bose einstein condensation at mean particle density strictly greater than the critical density . indeed , it is shown that the transient cases admit locally normal states exhibiting bose einstein condensation phenomena . in order to construct such locally normal temperature states by infinite volume limits of finite volume gibbs states , a careful choice of the the sequence of the chemical potentials should be done . for all such states , the condensate is essentially allocated on the base point supporting the perturbation . this leads to . we prove that all such temperature states are kubo martin schwinger states for a natural dynamics . the construction of such a dynamics , which is a delicate issue , is also done .

1203.5522

the present paper is devoted to the phenomena of bose einstein condensation ( bec for short ) on non homogeneous networks obtained by adding density zero perturbations to homogeneous cayley trees . even if the unperturbed model already exhibit bec ( cf . @xcite ) , the arising picture for the perturbed situation deserves attention as we have already explained in @xcite , and we are going to study in details in the sequel . the present paper is then the second part of the previous one @xcite , in which the harmonic analysis aspects of the model were intensively studied . the reader is also referred to @xcite for some very interesting pivotal amenable models . we are going to investigate the physical application to the bec of the very fascinating picture which arises from the previous mathematical investigation . the physical underlying model concerns bardeen cooper boson pairs in networks describing arrays of josephson junctions . the formal hamiltonian is the quartic bose hubbard hamiltonian , given on a generic network @xmath4 by @xmath5 here , @xmath6 denotes the set of the vertices of the network @xmath4 , @xmath7 is the bosonic creator , and @xmath8 the number operator on the site @xmath9 ( cf . @xcite ) . finally , @xmath10 is the adjacency operator whose matrix element @xmath11 in the place @xmath12 is the number of the edges connecting the site @xmath13 with the site @xmath14 ( in particular it is hermitian ) . the reader is referred to the seminal paper @xcite in which the theory of the superconductivity ( called bcs theory in honor of the authors ) has been firstly discussed . it has been argued in @xcite that , in the case when @xmath15 and @xmath16 are negligible with respect to @xmath17 , the hopping term dominates the physics of the system . after neglecting such terms , putting for the resulting effective coupling constant @xmath18 , in principle different from @xmath17 , @xmath19 , and finally passing to the one particle space , the model under consideration becomes the so called pure hopping ones described by the _ pure hopping one particle hamiltonian _ which assumes the form @xmath20 here , @xmath10 is simply the adjacency of the fixed graph @xmath4 , acting on the hilbert space @xmath21 . the constant @xmath22 is added just in order to have @xmath23 and does not affect the analysis at all . the reader is referred to @xcite for the systematic investigation of the mathematical properties arising in non amenable cases considered here consisting of negligible additive perturbations of cayley trees . several interesting amenable models are treated in details in @xcite . the reader is also referred to @xcite for some promising experiments for the pure hopping model on the comb and star graphs ( see fig . [ figd ] ) pointing out an enhanced current at low temperature which might be explained by condensation phenomena . one of the first attempts to investigate the bec on non homogeneous amenable graphs , such as the comb graphs , was made in @xcite . in that paper , it was pointed out the appearance of a _ hidden spectrum _ in the low energy part of the spectrum , which is responsible of the finiteness of the critical density . in addition , the behavior of the wave function of the ground state , describing the spatial density of the condensate , was also computed . some spectral properties of the comb and the star graph were also investigated in @xcite in connection with the various notions of independence in quantum probability . in that paper , it was noted the possible connection between such spectral properties and the bec . the systematic investigation of the bec for the pure hopping model on a wide class of amenable networks obtained by negligible perturbations of periodic graphs , has been started in @xcite . the emerging results are quite surprising . first of all , the appearance of the hidden spectrum was proven for most of the graphs under consideration . this is due to the combination of two opposite phenomena arising from the perturbation . if the perturbation is sufficiently large ( surprisingly , in some examples , it is enough that the perturbation is indeed finite ) , the norm @xmath24 of the adjacency of the perturbed graph becomes larger than the analogous one @xmath22 of the unperturbed adjacency . on the other hand , as the perturbation is sufficiently small ( i.e. zero density ) , the part of the spectrum @xmath25 in the segment @xmath26 $ ] does not contribute to the density of the states . this allows to compute the critical density @xmath27 at the inverse temperature @xmath28 for the perturbed model by using the integrated density of the states @xmath29 of the unperturbed one , @xmath30 the resulting effect of the perturbed model exhibiting the hidden spectrum ( i.e. when @xmath31 ) is that the critical density is always finite . the bridge between the mathematics and physics can be easily explained as follows . consider the bose gibbs occupation number ( at the inverse temperature @xmath32 and chemical potential @xmath33 ) at low energies for the hamiltonian . after using taylor expansion , one heuristically gets @xmath34^{-1}=[\b((\|a\|-\m)i - a)]^{-1 } \equiv \frac1{\b}r_a(\|a\|-\m)\,.\ ] ] then the study of the bec is reduced to the investigation of the spectral properties of the resolvent @xmath35 , for @xmath36 , the latter being a very familiar object for mathematicians . due to the particular form , as the bec phenomena are connected with the spectral properties for small energies of the hamiltonian , we can reduce the matter to the investigation of the spectral properties of the adjacency operator @xmath10 of the network close to @xmath22 . another relevant fact connected with the introduction of the perturbation , and thus with the lack of homogeneity , is the possible change of the transience / recurrence character ( cf . @xcite , section 6 ) of the adjacency operator . it has to do with the possibility to construct locally normal states exhibiting bec . as explained in @xcite , the last relevant fact is the investigation of the shape of wave function of the ground state of the model , describing the spatial distribution of the condensate on the network in the ground state of the hamiltonian . from the mathematical viewpoint , this is nothing else the perron frobenius generalized eigenvector of the adjacency ( cf . it appears then clear that the physical and the mathematical aspects of the topological model based on the pure hopping hamiltonian are strongly related . summarizing , the new and very surprising phenomena are the following . first the finiteness of the critical density is connected with the appearance of the _ hidden spectrum _ above the zero of the energy . it can be computed in a quite simple way for many amenable and non amenable models by using the _ secular equation _ , see section 6 of @xcite and section 3 of @xcite . this leads to the fact that the bec can appear also in low dimensional cases @xmath37 . in addition , not for all the models with finite critical density , it is possible to exhibit locally normal states ( those for which the local particle density is finite in the infinite volume limit ) describing bec . due to non homogeneity , particles condensate also in the configuration space , that is the system undergoes a `` dimensional transition '' . this phenomenon is connected with the _ transience / recurrence _ character of the adjacency operator of the network ( the last describing the hamiltonian of the system ) . another new aspect is the difference between the geometrical dimension of the network and the growing of the wave function of the ground state . the last is called _ perron frobenius dimension _ @xmath38 for volume growing networks and is computed through the behavior of @xmath39norms of the finite volume perron frobenius eigenvectors , all normalized to @xmath40 at a common root . if the geometrical dimension @xmath41 is greater than @xmath38 , then it is impossible to exhibit locally normal states describing the bec , whose mean density of the particles is greater than the critical one . surprisingly , there are non homogeneous networks for which @xmath42 and @xmath43 ( i.e. the comb graph @xmath44 , see fig . [ figg ] ) . in addition , there are also simple non homogeneous graphs exhibiting infinite critical density , but for which it is yet possible to exhibit locally normal states describing bec . such a very intriguing situation is summarized as follows . is the comb shaped graph whose base point is @xmath10 , @xmath2 is the critical density , r / t denotes the recurrence / transience character of the adjacency , bec ( @xmath45bec ) denotes the existence of locally normal states exhibiting bec ( exhibiting bec at any mean density @xmath46 ) . ] [ cols="<,<,^,^,^,^,^,^",options="header " , ] a similar situation also happens for some non amenable networks . the networks under consideration in the present paper are density zero additive perturbations of exponentially growing graphs made of homogeneous cayley trees , see fig . the models treated in the present paper are the perturbations @xmath47 , @xmath48 , and @xmath49 , of the homogeneous cayley tree @xmath50 along a subtree isomorphic to @xmath51 , and @xmath52 respectively , see below , see also @xcite for the precise definition of such models and for further details . quite surprisingly , the emerging mathematical situation described in @xcite for such non amenable examples , is similar to that for the amenable graphs treated in @xcite . namely , we still find hidden spectrum , and the transience / recurrence character is determined by the transience / recurrence character of the negligible subgraph supporting the perturbation . finally , the infinite volume perron frobenius eigenvector is exponentially decreasing far from the base point supporting the perturbation . namely , concerning the wave function of the ground state , we have precisely the same situation arising for the amenable situations treated in @xcite ( see fig . [ figd ] ) . the first part of the present paper is devoted to describe general properties of ( bounded ) hamiltonians on networks equipped with an exhaustion such that the density of the eigenvalues of the underlying hamiltonian converges in the infinite volume limit to a cumulative function called _ integrated density of the states_. it includes as particular cases those arising by negligible perturbations of cayley trees as shown in @xcite . then we describe the general statistical properties of such models . in particular , a careful definition of the mean density and the possible amount of the bec condensate should be done , because of the non negligible effects of the boundary terms due to the possible lack of amenability . we pass to describe the existence of the dynamics associated to the pure hopping hamiltonian for the perturbed cayley trees . this is a delicate issue which is done in section [ dynkmss ] for all the cases under consideration . after that , by using the results in @xcite , in the transient case we write down the formula ( cf . ) for the natural class of states describing bec ( i.e. presenting a non negligible amount of condensate in the ground state wave function ) . such states are kms ( i.e. equilibrium ) states for the dynamics previously investigated . as for the previous amenable models treated in @xcite , such states are infinite volume kms states whose mean density coincides with the critical one . this simply means that the condensate is allocated on a subgraph whose volume growth is negligible with respect to that of the unperturbed one . we end the paper by investigating the infinite volume behavior of finite volume gibbs states . due to the lack of amenability , such an investigation presents some very delicate technical difficulties arising by the unavoidable boundary effects . modulo such difficulties , the emerging situation seems to be in accordance with that described in @xcite , and summarized as follows . first , relatively to the general recurrent situation , it is indeed impossible to construct locally normal states exhibiting bec at all , even if the critical density is finite . in addition , in the all transient cases under consideration , it seems to be impossible to exhibit locally normal states describing bec at the mean density @xmath53 , @xmath54 . finally , in the condensation regime where for the mean density @xmath1 of all such states @xmath55 , a careful choice of the sequence of the finite volume chemical potentials @xmath56 should be done . by considering the geometry of the graph @xmath4 and the @xmath39behavior of the ground state wave function @xmath57 , such a choice corresponds to the local density of the condensate @xmath58 in the infinite volume limit @xmath59 , see proposition [ 3a3c ] . this is nothing but the naive explanation of the fact that @xmath55 for the locally normal states describing the condensation regime in all the situations under consideration . a _ graph _ ( called also a _ network _ ) @xmath60 is a collection @xmath61 of objects , called _ vertices _ , and a collection @xmath62 of unordered lines connecting vertices , called _ edges_. usually , the edges describe the physical interaction between vertices . in our model describing the pure hopping interaction , multiple edges as well as self interactions , are also allowed . let @xmath63 be the collection of all the edges connecting @xmath64 with @xmath65 . we have @xmath66 . let us denote by @xmath67_{x , y\in x}$ ] , @xmath68 , the _ adjacency matrix _ of @xmath69 : @xmath70 as explained in @xcite , all the geometric properties of @xmath69 are encoded in @xmath10 . for example , @xmath69 is connected if and only if @xmath10 is irreducible . setting @xmath71 and @xmath72 we have @xmath73 , that is @xmath10 is bounded if and only if @xmath69 has uniformly bounded degree . for connected networks ( or on each connected component ) we can define the standard distance @xmath74 @xmath75 being the length of the path @xmath76 ( i.e. the number of the edges in @xmath76 ) . the adjacency operator considerably differs from the laplacian @xmath77 , whenever @xmath69 is non homogeneous . on the other hand , it was proven in theorem 7.6 of @xcite that the value of the critical density does not change under negligible perturbations on a fixed homogeneous graph , if the hamiltonian of the model is the ( opposite of the ) laplacian . thus , the main object for the investigation of the bec for the pure hopping model of inhomogeneous networks will be the adjacency operator . in the present paper , all the graphs are connected , countable and with uniformly bounded degree . in addition , we deal only with bounded operators acting on @xmath78 if it is not otherwise specified . let @xmath79 be a closed operator acting on @xmath78 , and @xmath80 the resolvent set of @xmath79 . as usual , @xmath81 denotes the _ resolvent _ of @xmath79 . fix a bounded matrix with positive entries @xmath79 acting on @xmath78 . such an operator is called _ positive preserving _ as it preserves the elements of @xmath78 with positive entries . if @xmath79 is selfadjoint , @xmath79 is _ positive _ if @xmath82 for each @xmath83 . a sequence @xmath84 is called a ( generalized ) _ perron frobenius eigenvector _ if it has positive entries and @xmath85 if such a vector is normalizable ( i.e. if it belongs to @xmath78 ) it is a standard @xmath78vector , otherwise it is a weight . in all the cases we simply call perron frobenius eigenvector such a sequence by dropping the word generalized . suppose for simplicity that @xmath79 is selfadjoint . it is said to be _ recurrent _ if @xmath86 otherwise @xmath79 is said to be _ transient_. it is shown in @xcite , section 6 , that the recurrence / transience character of @xmath79 does not depend on the base point chosen in computing the limit in . the perron frobenius eigenvector is unique up to a multiplicative constant , if @xmath69 is finite or when @xmath79 is recurrent , see e.g. @xcite . in general , it is not unique , see the end of the present section for the case relative to the cayley trees . we say that an operator @xmath79 acting on @xmath78 has _ finite propagation _ if there exists a constant @xmath87 such that , for any @xmath88 , the support of @xmath89 is contained in the ( closed ) ball @xmath90 centered in @xmath64 and with radius @xmath91 . it is easy to show that for the adjacency operator if @xmath10 on @xmath69 , then @xmath92 has propagation @xmath93 for any integer @xmath94 . let @xmath69 be an infinitely extended graph with an exhaustion @xmath95 which is kept fixed during the analysis . the definition of the _ the integrated density of the states _ ( cf . @xcite ) of a selfadjoint operator @xmath96 given in @xcite works for all the situations for which it is meaningful , and then is useful for the purposes of the present work . indeed , consider on @xmath97 the state @xmath98 @xmath99 being the selfadjoint projection onto @xmath100 . define for a bounded operator @xmath79 , @xmath101 where the domain @xmath102 is precisely the linear subspace of @xmath97 for which the limit in exists . in addition , define for a bounded selfadjoint operator @xmath79 , @xmath103 provided such a limit exists . the domain @xmath104 is precisely the linear subspace of the unital @xmath105algebra @xmath106 generated by @xmath79 , for which the limit in exists . notice that the definition of @xmath107 , @xmath108 might depend on the chosen exhaustion which we keep fixed during the analysis . suppose now that @xmath109 . then it provides a state on @xmath110 and , by the riesz markov theorem , a borel probability measure @xmath111 on @xmath112 . thus , there exists a right continuous , increasing , positive function @xmath113 satisfying @xmath114 such that @xmath115)=f_b(x)\,,\quad x\in\br\,.\ ] ] the previous described cumulative function @xmath113 is precisely the integrated density of the states ( ids for short ) associated to @xmath79 , provided it exists for the chosen exhaustion . when the graph is amenable and the operator @xmath79 is of finite propagation , the definition and some of the main facts relative to the ids considerably simplify as the boundary effects play no role in the infinite volume limit , see theorem 2.1 of @xcite . consider the graph @xmath116 such that @xmath117 , both equipped with exhaustions @xmath118 , @xmath119 such that @xmath120 , @xmath121 . the graph @xmath116 is a _ negligible _ or _ density zero perturbation _ of @xmath69 if it differs from @xmath69 by a number of edges for which @xmath122 where @xmath123 denotes the symmetric difference . to simplify matters , we consider only perturbations involving edges , the more general case involving also vertices can be treated analogously , see @xcite . suppose that @xmath124 that is @xmath125 admits the ids . ( proposition 1.3 of @xcite ) [ density0 ] let @xmath116 be a negligible perturbation of the tree @xmath69 . then @xmath126 , and @xmath127 if @xmath128 is sufficiently large , then it is possible to express the resolvent of @xmath116 in terms of the resolvent of @xmath69 by the krein formula . indeed , fix @xmath69 as the reference graph and define @xmath129 , @xmath130 , where @xmath131 is the perturbation , which eventually acts on @xmath132 . put , for @xmath133 , @xmath134 where @xmath135 is the selfadjoint projection onto @xmath136 . the krein formula assumes the form @xmath137 the resolvent formula holds true for @xmath138 , and extends to any simply connected subset containing the point at infinity of @xmath139 made of the @xmath140 for which @xmath141 is invertible . conversely , suppose that @xmath141 is not invertible for some positive number @xmath142 . then the norm @xmath143 of the perturbed graph @xmath116 is greater than that @xmath144 of the unperturbed one @xmath69 . it can be computed as @xmath145 fix now our set up which is a graph @xmath69 equipped with an exhaustion @xmath118 such that @xmath125 admits the ids . let @xmath116 be a negligible additive perturbation of @xmath69 as explained above . first of all , note that proposition [ density0 ] assures that @xmath146 admits the ids w.r.t . the exhaustion @xmath119 . in addition , we obviously have @xmath147 . define @xmath148 denote @xmath149 , @xmath150 , where @xmath151 is the ids of the operator @xmath79 . ( corollary 2.6 of @xcite ) [ 6 ] we have @xmath152 the ids of @xmath153 and @xmath154 exist by hypothesis and by proposition [ density0 ] , respectively . we get , again by proposition [ density0 ] , @xmath155 this leads to the assertion . suppose that @xmath156 . then @xmath157 in . it has a very precise physical meaning as an effective chemical potential ( cf . ) , see proposition 7.1 of @xcite . thus , the part of the spectrum @xmath158 $ ] does not contribute to the ids of @xmath146 . equivalently , the perturbed pure hopping hamiltonian @xmath154 presents the hidden spectrum , see proposition [ hyspe ] . the appearance of the hidden spectrum is then the combination of two different effects : the perturbation should be of density zero in order not to affect the ids ( proposition [ density0 ] ) , but it should be sufficiently large in order to increase the norm of the perturbed adjacency . our framework will be the investigation of the occurrence of the bec for perturbed cayley trees , whose mathematical aspects are investigated in detail in @xcite . namely , we consider the graphs @xmath47 with @xmath159 , and @xmath160 where @xmath161 + 1\ ] ] ( @xmath162 $ ] stands for the integer part of a positive number @xmath64 ) , and @xmath163 with @xmath164 . such networks consist of the cayley tree @xmath50 of order @xmath165 , suitably perturbed along subtrees as explained in fig . . one can prove ( cf . lemma 4.1 of @xcite , together with proposition 2.2 of @xcite ) that the adjacency of @xmath50 admits the ids w.r.t . the standard exhaustion made of all the balls of radius @xmath166 centered in a fixed root . in the above notations , the laplace transform @xmath167 $ ] ( or equivalently the _ one particle partition function _ in physical language ) of @xmath168 is given by @xmath169 this , together with proposition [ 6 ] , leads to @xmath170 , where @xmath116 is @xmath47 or @xmath163 . as @xmath171 , they all present hidden spectrum . we also have that the graphs @xmath47 and @xmath163 are all transient except @xmath172 , see @xcite . we briefly report the description of the perron frobenius eigenvectors @xmath173 given in @xcite for the graphs under consideration . let @xmath174 be the base point of the perturbed graph @xmath116 under consideration , supporting the perturbation . the exhaustion @xmath95 has an obvious meaning by considering the part @xmath175 of @xmath174 in @xmath176 , the last being the ball of radius @xmath166 of @xmath177 , centered in a root @xmath178 , kept fixed during the analysis . it was shown in @xcite that @xmath179 here , @xmath180 is the function given , @xmath181 is the distance given in , @xmath182 is the unique point on @xmath174 such that @xmath183 ( cf . lemma 4.2 of @xcite ) , and finally @xmath184 is the perron frobenius eigenvector of @xmath185 ( @xmath186 given by @xmath187 described in theorems 5.2 and 6.3 of @xcite for @xmath163 and @xmath47 , respectively . concerning the behavior of the @xmath39norms of the restrictions of @xmath173 , we have the following [ 3a3c ] for the perron frobenius eigenvector @xmath173 described by and , we get @xmath188 first of all we note that , in all the situations , @xmath189 this leads to @xmath190 by taking into account the form of @xmath184 , we easily obtain the following estimates . concerning @xmath49 we have @xmath191 , which leads to @xmath192 for the case @xmath47 , @xmath193 leads to @xmath194 , identically on @xmath174 . concerning the case @xmath195 , consider the function @xmath196 it is not difficult to see that @xmath197 and @xmath198 , @xmath199 . this leads to @xmath200 whenever @xmath201 . we finally get @xmath202 we end the present section by noticing that it is easily seen that the perron frobenius eigenvector is no longer unique , in general for infinitely extended networks , and in particular for the situation considered here . in fact , on @xmath177 , @xmath203 , @xmath204 , @xmath205 are perron frobenius eigenvectors for the adjacency , for any fixed root @xmath206 . the same happens for @xmath47 , @xmath195 where in it is enough to choose @xmath207 , @xmath208 is any fixed root , where @xmath209 the choice of the normalization at @xmath40 on a fixed root ( which is denoted by @xmath178 in our framework ) selects in a unique way the perron frobenius vector among the class previously described . for the networks @xmath210 and @xmath172 , the perron frobenius eigenvector for the adjacency is unique ( up to a multiplicative constant ) thanks to the fact that the adjacency is recurrent . it is expected that it is unique also for @xmath49 , thanks to the fact that @xmath211 has a unique end at infinity . in order to investigate the statistical properties of the pure hopping model describing bardeen cooper pairs on arrays of josephson junctions on non homogeneous networks we report some standard notions useful in the sequel . let @xmath212 be a dynamical system consisting of a non abelian @xmath105algebra and a one parameter group of @xmath213automorphism @xmath214 . . ] the state @xmath215 on the @xmath216algebra @xmath217 satisfies the kubo martin schwinger ( kms for short ) boundary condition at inverse temperature @xmath218 w.r.t the group of automorphisms @xmath219 if * for every @xmath220 , @xmath221 , @xmath222 are continuous ; * for each @xmath223 , @xmath224 where stands for the fourier transform and @xmath226 is the space of the infinitely often differentiable , compactly supported functions on @xmath227 . the following facts are well known . first , a kms is automatically invariant w.r.t . the automorphism group @xmath228 . second , the cyclic vector @xmath229 of the gelfand naimark segal ( gns for short ) quadruple @xmath230 is also separating for @xmath231 . denote @xmath232 its modular group . according to the definition of kms boundary condition , we have @xmath233 we refer the reader to @xcite , section 5.3 for equivalent formulations of the kms condition , proofs and further details . the @xmath105algebras considered here are those arising from the _ canonical commutation relations _ ( ccr for short ) . namely , let @xmath234 be a pre hilbert space equipped with the non degenerate inner product @xmath235 . denote @xmath236 its completion . consider the following ( formal ) relations , @xmath237 it is well known that the relations can not be realized by bounded operators . a standard way to realize them is to look at the _ symmetric fock space _ @xmath238 on which the annihilators @xmath239 and creators @xmath240 naturally act as unbounded closed , adjoint each other ( i.e. @xmath241 ) operators . this concrete representation of the ccr is called the _ fock representation_. an equivalent description for the ccr is to put on @xmath238 , @xmath242 and define the _ weyl operators _ @xmath243 . the weyl operators are unitary and satisfy the rule @xmath244 it is well known that the representation on @xmath238 of @xmath245 described above , is faithful . we adopt the _ weyl algebra _ as the definition of the ccr algebra , the last being the abstract @xmath105algebra @xmath245 generated by the operators @xmath246 satisfying the relations . sometimes , it is also useful to compute directly the expectation values of combinations of products of annihilators and creators ( equivalently field operators ) in a given state . let @xmath247 be a self adjoint operator acting on @xmath236 . suppose that @xmath248 . then the one parameter group of automorphisms @xmath249 defines a one parameter group of @xmath213automorphisms @xmath228 of @xmath245 by putting @xmath250 . the latter is called the one parameter group of _ bogoliubov automorphisms _ generated by @xmath251 . a representation @xmath76 of the weyl algebra @xmath245 is _ regular _ if the unitary group @xmath252 is continuous in the strong operator topology , for any @xmath253 . a state @xmath215 on @xmath245 is regular if the associated gns representation is regular . the _ quasi free _ states on the weyl algebra are those of interest for our purposes . such states @xmath0 are uniquely determined by the two point functions @xmath254 , @xmath255 . a standard textbook for ccr is @xcite ( cf . section 5.2 ) to which the reader is referred for proofs , literature and further details . let @xmath4 be any graph . to simplify notations , we denote the sums on the vertices @xmath256 of subgraphs @xmath257 , and of @xmath6 itself , directly by sums on @xmath174 or @xmath4 , respectively . consider a subspace @xmath234 of @xmath258 , which contains the indicator functions @xmath259 . a representation @xmath76 of the weyl algebra @xmath245 is said to be _ locally normal ( w.r.t . the fock representation ) _ if @xmath260 is quasi equivalent to the fock representation of @xmath261 . a state on @xmath245 is said to be locally normal if the associated gns representation is locally normal . a locally normal state @xmath215 does have finite local density @xmath262 even if the mean density might be infinite ( e.g. @xmath263 ) . let @xmath264 be an _ exhaustion _ of @xmath4 , that is a sequence of finite regions invading the graph @xmath4 , together with a sequence of states @xmath265 on @xmath266 . ( lemma 3.2 of @xcite ) [ add2 ] suppose that @xmath267 for some @xmath268 . then @xmath269 does not define any locally normal state on @xmath270 , where @xmath271 is any subspace containing the finite supported sequences . our purposes concerns the pure hopping model on the networks @xmath47 and @xmath49 or , more generally on a uniformly bounded connected network @xmath69 equipped with an exhaustion @xmath272 for which the adjacency admits the ids ( cf . @xcite ) . as previously explained , all the physical properties of the emerging quasi free multi particle model are encoded in the one particle pure hopping hamiltonian . namely , the graph @xmath69 will be one of such networks , @xmath116 a density zero additive perturbation of @xmath69 as described above , equipped with exhaustions @xmath118 , @xmath119 . we denote by @xmath4 one of such graphs , equipped with the finite volume exhaustion @xmath95 . the pure hopping hamiltonian , after normalizing such that the bottom of the energy is zero , assumes the form @xmath273 the finite volume hamiltonians will be @xmath274 acting on @xmath275 , @xmath276 being the selfadjoint projection onto @xmath275 . fix a general bounded positive hamiltonian @xmath277 admitting the ids for a fixed exhaustion as for the pure hopping one . such that @xmath278 is the bottom of the spectrum @xmath279 of the hamiltonian @xmath247 . ] consider the finite volume densities of the states @xmath280 relative to @xmath281 , and that @xmath282 relative to @xmath247 . define @xmath283 where the limit in the first equation exists by lemma 3.4 of @xcite , and that in the second one is meaningful directly by the definition of the ids . we always get @xmath284 . [ hyspe ] if @xmath285 we say that there is a _ ( low energy ) hidden spectrum _ , see e.g. @xcite . the following assertions can be shown for all the situations under consideration . first , @xmath247 is continuous in @xmath178 ( see also @xcite , proposition 3.9 ) in all the situation considered here . second , in the unperturbed situation the hidden spectrum does not occur ( by direct inspection of for @xmath177 , or by theorem 5.2 of @xcite in the case of periodic amenable graphs ) . if @xmath116 is any perturbation of @xmath69 , we have for the @xmath286 appearing in , @xmath287 the opposite of @xmath288 then plays the role of a chemical potential in determining the critical density of the perturbed graphs under consideration . indeed , consider the formula for the mean density at the inverse temperature @xmath28 of the particles ( cf . @xcite ) for the model with hamiltonian @xmath23 , and chemical potential @xmath289 , @xmath290 and the corresponding _ critical density _ @xmath291 by taking into account proposition [ 6 ] , we get @xmath292 this means @xmath293 . in addition , if there is hidden spectrum , @xmath294 is always finite and @xmath295 . it can happen that , after perturbing a graph whose critical density is infinite , we get finite critical density . as for the networks under consideration in the present paper , the unperturbed critical density is always finite ( cf . @xcite ) , the only effect of the appearance of the hidden spectrum on the critical density is that it decreases w.r.t . the unperturbed one . it is well known that the condensation regime is described by @xmath296 , see e.g. section 5.2 of @xcite , or section 3 of @xcite . indeed , one starts from the gibbs grand canonical ensemble at finite volume @xmath297 and fixed density . one fixes an inverse temperature @xmath32 together with the particle density @xmath45 , and determines the finite volume chemical potential @xmath298 by solving @xmath299 see e.g. fig . [ figg1 ] . , instead of the chemical potential . ] the infinite volume limit can be investigated by means of the reference exhaustion @xmath95 . to take into account also the very different situation appearing in non homogeneous situation , we start by fixing any general sequence of chemical potential @xmath300 which we can suppose to converge ( eventually passing to a subsequence ) to some @xmath301 . in the case when such a sequence is recovered by using , we put @xmath302 . since @xmath303 , we get @xmath289 . the finite volume state with chemical density @xmath304 at @xmath28 is described by the two point function @xmath305 where @xmath306 . thermodynamical states are then described as limits of the finite volume states as above . concerning the infinite volume limit of the finite volume density , we report a result ( cf . @xcite ) , whose proof relies upon the analogous one , proposition 3.8 of @xcite . it describes in a rigorous way how it is possible to define the amount of the condensate in the infinite volume limit for an arbitrary sequence of finite volume chemical potential @xmath307 . due to the general character of such a result , we suppose that @xmath247 is a bounded positive hamiltonian on a network @xmath4 equipped with an exhaustion @xmath308 such that @xmath247 admits the ids @xmath282 ( in the situation under consideration @xmath309 ) . denote as usual @xmath310 , and @xmath311 . to avoid the almost trivial situation @xmath33 , we restrict the matter to the condensation regime @xmath296 . [ ennz ] under the conditions listed above , the following assertions hold true for each sequence @xmath312 satisfying @xmath307 . * suppose that @xmath247 admits low energy hidden spectrum . then for each continuous mollifier @xmath29 which is identically @xmath178 in a neighborhood of @xmath178 , and @xmath40 in a neighborhood of @xmath313 , we have @xmath314 * suppose that @xmath315 . consider a sequence @xmath316 of increasing continuous mollifiers , all vanishing in a neighborhood of @xmath178 and converging almost everywhere ( w.r.t the measure determined by @xmath282 ) to @xmath40 . then @xmath317 in addition , if @xmath318 , one can replace for the continuous mollifiers @xmath316 , the condition of monotony with uniform boundedness . we start by noticing that @xmath319 , and all the @xmath320 are continuous in a common neighborhood of all the supports of the measures determined by the cumulative functions @xmath321 and @xmath322 . \(i ) by the existence of the hidden spectrum , we have @xmath323 since @xmath324 , uniformly on the support of @xmath29 , and the second summand goes to zero because @xmath325 for each continuous function @xmath326 in a common neighborhood of all the supports of the involved measures @xmath327 and @xmath328 . \(ii ) the same computation as before leads for each @xmath329 , @xmath330 the proofs follows either by the monotone or the dominated convergence theorem , as @xmath331 . thanks to the above result , the quantity @xmath332 is well defined and independent of the particular choice of the mollifiers as above . such a quantity describes the amount of the condensate in the ground state . indeed , if the sequence of chemical potentials are obtained through by fixing the mean density @xmath45 , we have @xmath333 @xmath33 if and only if @xmath334 . for any @xmath32 , @xmath33 , the sequence converges pointwise to a state @xmath0 , whose two point function is given by @xmath335 moreover , the _ density _ @xmath1 of the state @xmath0 , defined by @xmath336 satisfies @xmath337 . the transience of @xmath10 is a necessary and sufficient condition ( see proposition [ add4 ] and remark [ suftra ] ) for the existence of locally normal states on @xmath245 describing condensation phenomena , that is when for @xmath296 , for the pure hopping model . observe that @xmath338 if and only if @xmath339 . indeed , @xmath340 whenever @xmath33 , hence @xmath341 . as a consequence , @xmath342 , i.e. @xmath53 . conversely , @xmath53 implies @xmath296 , hence @xmath343 . observe that @xmath338 can be obtained only if @xmath344 . indeed , @xmath345 is a necessary condition for the occurrence of bec . observe that @xmath346 can occur also when the infinite volume limit of the two point function leads to non locally normal states i.e. @xmath347 . observe that it is possible to have locally normal states exhibiting a non trivial amount of condensate even if @xmath348 , see section 10 of @xcite , or sections [ dynkmss ] , [ dynkmss1 ] below . to simplify the notations , from now on we put @xmath349 throughout the paper if it is not otherwise specified . we also put @xmath350 , @xmath351 , and finally @xmath352 , @xmath247 being the hamiltonian of the model . let @xmath116 be any of the perturbed graphs under investigation . in order to study states describing local properties , and in particular those which are locally normal , we start to consider a subspace @xmath353 which contains all the canonical elements @xmath354 , and which is stable under the dynamics generated by @xmath355 . to this end , we define @xmath356 for the algebraic span ( no closure in @xmath357 ) of all the elements of the form @xmath358 . by construction @xmath359 . for our aims , we need also that the pairing @xmath360 is well defined , when @xmath173 is the perron frobenius vector , described in @xcite , for the adjacency matrix @xmath146 in each of the cases under consideration , see and . for @xmath361 we also define @xmath362 by putting @xmath363 , @xmath364 . [ rispf ] let @xmath365 be a function analytic in a neighborhood of @xmath366 such that @xmath367 , @xmath121 . then @xmath368 . as the adjacency matrix has positive entries , all the involved series have positive entries . thus , by the monotone convergence theorem we get @xmath369 the following functions , defined in @xmath370 $ ] , @xmath371 @xmath372 are useful in the sequel . if @xmath373 , it was shown in @xcite that , for the situation under consideration , the operator @xmath374 in describing the resolvent of @xmath146 through the krein formula assumes the form @xmath375 eventually acting on @xmath376 . on which is allocated the perturbation . ] in the cases under consideration when @xmath116 is @xmath47 , @xmath49 , @xmath377 ( where @xmath378 ) and @xmath379 , respectively . in all such situations , is proportional to the convolution by , and for @xmath380 $ ] it assumes the form @xmath381 provided at least that the series in converges pointwise . here , @xmath181 is the distance function given in , and @xmath382 and @xmath383 are the analytic continuations in @xmath384 $ ] of the corresponding objects given in and , respectively . it was shown in @xcite that is meaningful and defines an element of @xmath385 , at least when @xmath386 . now we show that the series in converges and defines an element in @xmath385 for @xmath387 such that @xmath388 is sufficiently large . [ rispf1 ] let @xmath382 and @xmath383 as above , and @xmath389 . then @xmath390 put @xmath391 . disregarding the constants and the quantities whose modulus is fixed , we have to maximize and minimize on circles of fixed radii less than @xmath40 , the functions @xmath392 where @xmath184 is in the open punctured disk @xmath393 of radius @xmath40 centered in the origin of the complex plane , and @xmath394 , see fig . [ figab ] . first of all we note that , for the analytic continuation @xmath395 of @xmath396 in the cut plane @xmath397\cup[1,+\infty)\big)$ ] , we have @xmath398 after some tedious computations , the functions @xmath29 , @xmath4 are expressed by functions @xmath326 , @xmath399 on the unit circle , given by @xmath400 here @xmath401 @xmath402 is fixed , and by , we can reduce the matter to @xmath403 . after some straightforward computations , we get that @xmath404 and @xmath405 are never zero for @xmath406 so the the functions attain the extrema either in @xmath178 or in @xmath407 . we get for @xmath326 , @xmath408 because @xmath409 . concerning @xmath399 we have , @xmath410 collecting together , we get with @xmath411 , @xmath412 and the proof follows . denote @xmath413 the closed disk centered in the origin of radius @xmath91 . [ koro ] there exists @xmath389 such that for @xmath414 in , we get @xmath415 for each @xmath416 . thus , for @xmath416 , the resolvent of the adjacency of @xmath116 can be written by the neumann series @xmath417 let @xmath418 be the decomposition of @xmath419 in positive elements . one has @xmath420 , @xmath421 . by taking into account and lemma [ rispf1 ] , we get @xmath422 as @xmath423 is continuous and decreasing ( cf . lemma 3.1 in @xcite ) with @xmath424 , there exists @xmath425 such that @xmath426 . collecting together , we have in the complement of @xmath427 , @xmath428 in any simply connected open subset of @xmath139 containing the point at infinity for which @xmath429 is invertible , @xmath430 assumes the form @xmath431 and the proof follows from the first half , by using the neumann expansion of @xmath432 . consider the function @xmath433 it is bounded on @xmath434 . in addition , @xmath435 . it provides the comparison between the resolvent of @xmath10 and the functional calculus of @xmath436 associated to the bose gibbs occupation number for the pure hopping hamiltonian . [ add4 ] under the above notations , the following assertions hold true . * @xmath146 is transient if and only if @xmath437 . * for @xmath438 we have @xmath439 . it is enough to prove the assertions for the generators @xmath440 of @xmath234 . @xmath441 by using the splitting described by we get that @xmath184 is in @xmath442 if and only if @xmath184 is in @xmath443 . let @xmath14 be any element of @xmath116 . obviously , @xmath444 and , by hypothesis , @xmath445 . so @xmath446 and @xmath146 is transient . for the reverse implication , we compute for @xmath447 , @xmath448 which is finite if @xmath146 is transient . thus , @xmath449 . @xmath450 fix @xmath451 , where @xmath91 is the number appearing in proposition , together with the circle @xmath452 of radius @xmath453 centered in the origin . by using the neumann expansion of @xmath432 on @xmath454 and the fact that @xmath455 itself is the convolution of the function @xmath456 ( cf . ( 7.6 ) in @xcite ) , we show that the corresponding resolvent @xmath430 is a combinations of convolutions and infinite ( absolutely converging ) sums of powers of the convolution by the function @xmath456 , see and . let @xmath454 be counterclockwise oriented . thus , thanks to proposition as explained above , together with lemma [ rispf ] , we get @xmath457 now we pass to describe a class of locally normal states exhibiting bec , which are indeed kms for the natural dynamics generated on @xmath245 , @xmath234 given in , by the bogoliubov automorphisms whose generator is the ( second quantization of ) the pure hopping hamiltonian . to this end , we consider the dynamical system @xmath458 , where @xmath245 is the @xmath105algebra made of the weyl ccr algebra on @xmath234 , and @xmath228 is generated by the bogoliubov automorphisms @xmath459 , @xmath460 which are well defined as @xmath461 by construction . by taking into account proposition [ add2 ] , the necessary condition for the existence of locally normal states exhibiting bec is the transience of @xmath146 . for the pure hopping model , the transience of the adjacency is indeed also sufficient as we are going to see . for the rest of the present section we limit our analysis when @xmath116 is either @xmath49 , @xmath462 , or @xmath47 where @xmath463 , and @xmath160 with @xmath464 given in , see theorems 4.1 , 5.3 , 6.5 of @xcite , together with proposition [ add2 ] . fix @xmath465 and define states on the weyl ccr algebra @xmath245 uniquely determined by the two point function @xmath466 the states @xmath467 are well defined ( cf . @xcite , section 5.2 ) because of proposition [ add2 ] . in the case of quasi free states on @xmath245 , it is also customary to provide the description directly in terms of their values on the weyl unitaries . we get ( cf . section 5.2.5 of @xcite ) , @xmath468 note that for quasi free states , and then in our situation , the kms condition can be checked directly in terms of @xmath469 , at least when @xmath470 is bounded and continuous . the boundedness automatically follows by the definition of the two point function thanks to transience . the continuity can be easily checked by lebesgue dominated convergence theorem by taking into account again proposition [ add2 ] . for each @xmath465 , the states @xmath467 given in are kms ( at inverse temperature @xmath349 ) for the dynamics generated by the bogoliubov automorphisms @xmath471 . it is easily seen by proposition [ add2 ] that @xmath472 , @xmath473 are bounded and continuous ( in @xmath474 ) . the proof follows by an elementary change of variables in as @xmath475 . the second addendum in describes the amount of particles which condensates in the ground state whose wave function is nothing but the perron frobenius weight @xmath173 . in the finite region @xmath476 , such a density of the condensate is roughly given by @xmath477 even if the model exhibits a non negligible amount of the condensate , @xmath478 as @xmath479 , see proposition [ 3a3c ] . formula explains what happens in the non homogeneous pure hopping model . particles condensate even in the configuration space . the amount of particles that the system can accommodate in the subnetwork made of the support of the perturbation is governed by the ratio between the volume growth of the original network and the @xmath39growth of the wave function of the ground state . in all the situations into consideration in the present paper , @xmath480 goes to @xmath178 as @xmath476 approaches to @xmath116 . this is due to the quite surprising fact that the perron frobenius weight decreases exponentially far from the perturbed zone ( cf . [ figd ] ) , see @xcite , or also @xcite for several amenable cases . this is the naive explanation that the mean density of the particles associated to all the states @xmath481 exhibiting bec is the critical one @xmath2 . we start with a simple general result explaining that , for the pure hopping model , the transience is a necessary and sufficient condition for the existence of locally normal states exhibiting bec if one disregards the dynamics . [ suftra ] let @xmath4 be a uniformly bounded degree connected network . suppose that its adjacency @xmath10 is transient . let @xmath482 denote the algebraic span of all the @xmath483 . then there exist locally normal states on @xmath484 exhibiting bec for the pure hopping hamiltonian @xmath485 . indeed , consider any perron frobenius eigenvector @xmath486 which exists by compactness , see e.g. proposition 4.1 of @xcite . fix @xmath487 and put @xmath488 by transience ( cf . ( i ) of proposition [ add4 ] which works for any network @xmath4 as above ) , is meaningful and defines a quasi free state on @xmath484 . such a state can be interpreted as exhibiting bec for the presence of a non trivial amount of condensate on a lowest energy wave function @xmath486 described by the addendum @xmath489 . as explained before ( see also @xcite ) , the careful investigation of the finite volume behavior of the kms states given in plays a relevant role because of the boundary effects which can not be avoided in the case of non amenable graphs as in our situation . due to such boundary effects , this analysis becomes very delicate . yet , the emerging situation is surprisingly in accordance with previous results in @xcite . the main aim of the present section is to investigate , technicalities permitting , the infinite volume limits of finite volume gibbs states arising from the grand canonical ensemble . we start by considering standard finite volume approximations of kms states . we focus the attention to any uniformly bounded degree connected network @xmath4 , equipped with the exhaustion @xmath95 made of finite volume regions . fix a common root @xmath490 , @xmath121 and consider the sequence @xmath491 of perron frobenius eigenvectors of the adjacency of the finite regions @xmath308 , all normalized to @xmath40 at @xmath178 . in addition , we also suppose that @xmath492 ( or equivalently the pure hopping hamiltonian @xmath309 ) admits the ids w.r.t . the exhaustion @xmath95 . we reduce the matter to the @xmath105subalgebra @xmath493 , see section 5.2.5 of @xcite . we denote by @xmath494 the identity on @xmath275 . fix a sequence of finite volume chemical potentials @xmath495 , with @xmath496 converging to @xmath289 . the finite volume gibbs states @xmath497 are given , for such a sequence of chemical potential , by @xmath498 the case when @xmath499 is quite standard ( cf . @xcite ) and presents no further complications . thus , we restrict ourselves to the condensation regime @xmath296 . in correspondence of such a sequence @xmath307 of chemical potential , we put @xmath500 the following result takes into account what happens when the graph @xmath4 is recurrent . [ roec ] if @xmath4 is recurrent then for each @xmath501 , @xmath502 we consider the splitting , and @xmath503 given in . it is readily seen that @xmath504 and @xmath505 . by recurrence and the monotone convergence theorem , we get @xmath506 on the other hand , as @xmath507 in the strong operator topology , @xmath508 then , again by the monotone convergence theorem and fatou lemma , @xmath509 then we have just shown that the two point function @xmath510 diverges in the infinite volume limit . the last result together with proposition [ add2 ] , tells us that , even when the critical density is finite ( which certainly happens in presence of hidden spectrum ) , it is impossible to construct locally normal states exhibiting bec when the graph @xmath4 under consideration is recurrent ( compare with the similar results in @xcite ) . in our situation , proposition [ roec ] covers the case @xmath172 , with @xmath462 . now we specialize to the situation of main interest here when the network @xmath4 ( denoted symbolically by @xmath116 ) is @xmath49 and @xmath47 . in this situation , @xmath173 denotes the perron frobenius eigenvector for @xmath146 described in , , and @xmath511 the finite volume perron frobenius eigenvectors of the finite regions @xmath308 . all such perron frobenius eigenvectors are normalized at @xmath40 on the fixed common root @xmath178 . [ faxpx111 ] let @xmath511 be the ( unique ) perron frobenius eigenvector of @xmath512 normalized to 1 at the common root @xmath490 . then @xmath513 . in addition @xmath514 . we start by considering the networks @xmath515 consisting of @xmath177 , perturbed along the subset @xmath516 , see fig . [ figgk ] . the vectors @xmath517 , also described in @xcite , are the perron frobenius weights normalized at @xmath40 on the common fixed root . let @xmath91 be any weak cluster point of @xmath491 which exists by compactness ( cf . proposition 4.1of @xcite ) . it is straightforwardly seen that there exists a collections @xmath518 of smooth universal functions ( on a sufficiently large neighborhood of @xmath143 ) such that for each @xmath519 and @xmath520 , @xmath521 here , @xmath522 is the unique element such that @xmath183 ( cf . lemma 4.2 of @xcite ) , @xmath486 is either @xmath91 or @xmath517 , and , correspondingly , @xmath140 is @xmath143 or @xmath523 . notice that , it was shown in @xcite that @xmath524 this first implies that there exists a unique weak cluster point , say @xmath91 , of the sequence @xmath491 . in addition , as @xmath525 , we get for each @xmath364 , @xmath526 concerning the limits of the norms we can suppose that the @xmath511 act on @xmath385 by extending those to @xmath178 on the complement of the balls @xmath176 , respectively . we have by fatou lemma @xmath527 which completes the proof . the precise estimate of the behavior of the @xmath39norms @xmath528 of the perron frobenius eigenvectors @xmath511 of @xmath512 as @xmath529 is needed in the sequel . unfortunately , even if we have a precise estimate of the behavior of @xmath530 , this does not guarantee the analogous estimate for @xmath528 . however , it is expected that @xmath531 still holds . for the reader convenience , we report a sufficient condition , tested on simpler models , which implies . suppose that @xmath532 we automatically get @xmath533 in @xmath534 . thus , as in the proof of proposition [ 3a3c ] , we compute , @xmath535 because @xmath536 , and @xmath537 as @xmath538 in the strong operator topology . the estimate implies the weaker one @xmath539 needed in proposition [ pprroo ] . now we pass to the transient situation by showing a preliminary result which , in this case , takes into account the case when the parameter @xmath131 describing the amount of the condensate in on the base point subspace is zero . [ faxpx ] for each @xmath364 we have @xmath540 by using the splitting in , we can reduce the matter to the analysis of the resolvents . notice that @xmath541 . thus , by monotone convergence theorem , we get @xmath542 we now recall some notations standardly used in the sequel . put @xmath543 eventually acting on @xmath544 . consider @xmath176 the ball of radius @xmath166 of @xmath177 , @xmath545 is the support of the perturbation inside @xmath308 ( i.e. @xmath546 for @xmath547 , and @xmath175 is isomorphic to the finite set consisting of @xmath548 points for @xmath163 ) . finally , @xmath549 is the selfadjoint projection onto @xmath550 . as @xmath308 is the finite volume perturbation of @xmath551 , @xmath552 . [ korila ] if @xmath553 $ ] , then @xmath554 the proof immediately follows from the same computations in lemma 3.1 of @xcite , by taking into account that the resolvent is differentiable and convex . point to investigate the infinite volume limit for the remaining cases , that is when the underlying perturbed graph is transient , is the following [ resspr ] let @xmath555 ( i.e. @xmath504 ) such that @xmath307 ( i.e. @xmath556 ) . then , @xmath557 where @xmath558 . in addition , if @xmath559 then @xmath560 . by using the splitting , the krein formula and finally lemma 2.4 of @xcite , we easily obtain for @xmath561 the above formula with @xmath562 given by , @xmath563 where the last goes to zero by propositions [ density0 ] and [ 6 ] . by lemma [ korila ] , the second half directly follows as @xmath564 now we specialize the situation to the transient situation for the case when finite volume gibbs states are prepared in order to describe in the infinite volume limit , a mean density @xmath46 . to this end , we chose the sequence of the chemical potential by putting @xmath565 even if it is expected that implies that the resulting limit density describes a mean density @xmath566 , in order to prove that we should show that @xmath567 the last result would follow again by the krein formula for the resolvent . unfortunately , its proof relies by unavoidable boundary effects which seriously affect the related formulas and are not easily manageable to take the infinite volume limit . however , for the infinite volume mean density @xmath45 , if holds true , we get by proposition [ resspr ] , @xmath568 this simply means that , if holds true , the sequence of finite volume gibbs states obtained with the choice of the chemical potentials , describes a fixed mean density @xmath46 in the infinite volume limit . it is expected that the condition is incompatible with local normality as it is shown in the following [ pprroo ] suppose that the estimate holds true . if the sequence of the finite volume chemical potentials satisfies , then for the two point function in , we get @xmath569 we have @xmath570 the proof easily follows collecting together and . the last result together with proposition [ add2 ] , would mean that it is impossible to construct locally normal states exhibiting bec with mean density @xmath46 in all the situations under consideration in the present paper . finally , we pass for the transient cases to the construction of the states by the infinite volume limits of finite volume gibbs states in with a suitable choice of the sequence of chemical potentials . also in this situation , we assume that @xmath571 holds true . in order to obtain itself for each other point @xmath364 . ] in order to recover by the infinite volume limits of , we put @xmath572 let @xmath116 be either @xmath49 with @xmath573 , or @xmath47 where @xmath463 , @xmath160 with @xmath464 given in . suppose that holds true , and is satisfied under the choice . then for each @xmath364 , we get for the two point function in , @xmath574 in addition @xmath560 . the second half directly follows from the assumption , by the second half of proposition [ resspr ] . indeed , we get , @xmath575 concerning the first part , as usual we reduce the matter to the resolvent . as @xmath576 , for each fixed @xmath447 , we get by monotony @xmath577 by proposition [ faxpx111 ] , we get after taking the limits in @xmath166 and in @xmath578 , @xmath579 on the other hand , by using the krein formula , we obtain @xmath580 which , taking into account , leads to @xmath581 collecting together , we get the assertion . we end by noticing that the case without condensate , but at the critical point @xmath296 , is directly covered by proposition [ faxpx ] without any further assumption . the author would like to thank the university of pretoria , and especially a. strh , for the warm hospitality and for financial support during the period march may 2010 , when the present work has been started . silvestrini p. , russo r. , corato v. , ruggiero b. , granata c. , rombetto s. , russo m. , cirillo m. , trombettoni a. , sodano p. _ topology induced critical current enhancement in josephson networks _ , phys . lett . a * 370 * ( 2007 ) , 499503 .

we analyze late - time evolution of the universe in the framework of the self - consistent model , in which the dark matter is influenced by the archimedean - type force proportional to the four - gradient of the dark energy pressure . the dark energy is considered as a fluid with the equation of state of the relaxation type , which takes into account a retardation of the dark energy response to the universe accelerated expansion . the dark matter is guided by the archimedean - type force , which redistributes the total energy of the dark fluid between two its constituents , dark energy and dark matter , in the course of the universe accelerated expansion . we focus on the constraints for the dark energy relaxation time parameter , for the dark energy equation of state parameter , and for the archimedean - type coupling constants , which guarantee the big rip avoidance . in particular , we show that the archimedean - type coupling protects the universe from the big rip scenario with asymptotically infinite negative dark energy pressure , and that the little rip is the fate of the universe with the archimedean - type interaction inside the dark fluid .

1212.4094

the present time accelerated expansion of the universe discovered in the observations of the supernovae ia @xcite revived the discussion about the fate of our universe . a psychologically jeopardy variant of the universe future is that the permanently increasing scale factor @xmath0 and the hubble function @xmath1 can reach infinite values during a finite time interval . the life - time of such universe is finite , and the catastrophe of this type produces cosmic inertial ( tidal ) forces , which destroy the bounds in all physical systems . various aspects of the singular scenaria of the future stage of the universe evolution were discussed during last two decades ( see , e.g. , the review @xcite ) . the specific term big rip ( or doomsday ) entered the scientific lexicon after publication of the paper @xcite ; nowadays this term indicates a new trend in theoretical cosmology ( see , e.g. , @xcite-@xcite ) . it seems to be uncomfortable for physicist to think that the big rip is the fate of our universe , probably , it is a reason that many authors consider the models , in which the big rip can be _ avoided_. in particular , the big rip can be avoided in the @xmath2 , @xmath3-gravity models and their modifications , in the models with chaplygin gas , in the model for the dark energy with various effective time - dependent equations of state ( see , e.g. , @xcite-@xcite ) . as a more optimistic variant of the universe behavior one could consider the case , when the universe life - time is infinite , and the hubble function tends asymptotically to a constant , @xmath4 . the asymptotic regime of this type appears , in particular , in the @xmath5cdm model , which converts at @xmath6 into the de sitter model with @xmath7 ( @xmath5 is the cosmological constant ) . more general case , when @xmath8 , but @xmath9 is not necessary equal to @xmath10 , is indicated as pseudo rip in @xcite . various intermediate scenaria with the infinite universe s life - time , in which the scale factor , the hubble function and ( probably ) its time derivative tend to infinity , belong to the class indicated by the term little rip ( see , e.g. , @xcite-@xcite for references ) . in order to classify the models of three types mentioned above in more detail , we use the terminology ( see , e.g. , @xcite ) based on the asymptotic properties of the scale factor @xmath0 , hubble function @xmath11 , and its time derivative @xmath12 , which are the basic quantities in the isotropic spatially homogeneous cosmological models of the friedmann - lematre - robertson - walker ( flrw)-type with the metric @xmath13 \ , . \label{metric}\ ] ] ( here and below we use the units with @xmath14 ) . in fact this classification is originated from the idea that two bonded particles are influenced by the inertial force proportional to the quantity @xmath15 where @xmath16 is the acceleration parameter . the internal structure of the physical system is assumed to be destroyed , when this inertial force exceeds the internal ( e.g. , intermolecular ) forces . we think that such force can also be indicated as a _ tidal _ force , if we take into account the following motives . let us start with the well - known equation of the world - line deviation ( see , e.g. , @xcite ) @xmath17 where @xmath18 is the velocity four - vector , @xmath19 is the deviation four - vector , @xmath20 is the world - line parameter and @xmath21 is the parameter describing the world - line family . the second term in the right - hand side of ( [ devi1 ] ) describes the contribution of the force of the non - gravitational origin ( its structure for the electromagnetic interactions is described , e.g. , in @xcite ) . the first term in the right - hand side of ( [ devi1 ] ) is interpreted as the tidal ( curvature induced ) force proportional to the convolution of the riemann tensor @xmath22 with the deviation four - vector . taking into account that in the flrw model the velocity four - vector is of the form @xmath23 , we can choose arbitrary direction , say @xmath24 , and calculate the corresponding tidal force component @xmath25 . the result is @xmath26 , thus the interpretation in terms of tidal force is equivalent to the one in terms of inertial force . this force tends to infinity , when @xmath12 or @xmath27 ( or both of them ) tend to infinity . the standard classification uses the following terminology : \1 . power - law inflation , if @xmath28 , @xmath29 ; \2 . @xmath5cdm model @xmath30 , where @xmath5 is the cosmological constant ; \3 . pseudo rip , if @xmath31 ; \4 . little rip , when @xmath32 , e.g. , as @xmath33 ; \5 . the future singularities at finite time @xmath34 , when @xmath35 . the fifth subclass can be specify , for instance , as follows ( see @xcite ) : 5.1 . big rip ( type i ) , if @xmath36 , e.g. , as @xmath37 ; 5.2 . sudden singularity ( type ii ) , if the pressure @xmath38 tends to infinity , but @xmath0 and energy density @xmath39 remain finite ; 5.3 . singularity of the type iii , if @xmath40 and @xmath41 , but @xmath0 is finite ; 5.4 . singularity of the type iv , if only @xmath42 . this paper is considered to be the third ( final ) one describing the fundamentals of the model of the archimedean - type interaction between the dark energy ( de ) and dark matter ( dm ) . in @xcite we established the self - consistent model in which the dark matter is influenced by the force proportional to the four - gradient of the dark energy pressure . in @xcite we classified the corresponding scenaria of the universe evolution with respect to a number of the transition points , in which the acceleration parameter @xmath43 takes zero values , and thus the epochs of decelerated expansion is replaced by the epoch of the accelerated expansion . in @xcite we considered the application of the model to the problem of light propagation with non - minimal coupling in a cosmic dark fluid with an archimedean - type interaction between the de and dm , and described the so - called unlighted cosmological epochs , for which the effective refraction index @xmath44 , the phase and group velocities , @xmath45 and @xmath46 , respectively , were imaginary functions of time . in @xcite we focused on the solutions with asymptotically _ hubble function and finite acceleration parameter , and claimed that the case of _ infinite _ @xmath27 and @xmath47 is the topic of special discussion . now we systematically consider the last case in terms of big rip and little rip solutions . why this consideration might be interesting for readers ? there are at least three motives to complete our investigation . the first motif is the following . the dark energy is considered in @xcite as a fluid with the simplest rheological property : the equation of state is of the relaxation type , i.e. , it takes into account a retardation of the dark energy response , and includes an extra parameter , @xmath48 . the first question arises : is it possible to avoid the big rip by the appropriate choice of the relaxation parameter @xmath48 ? the second motif is connected with the effective constant of the archimedean - type coupling , @xmath49 . when this coupling is absent , the energy density and pressure of the dark matter decrease in the course of the universe expansion , so that the dark matter does not play any role in the asymptotic regime . however , when the pressure of the dark energy , being negative , becomes infinite in the course of accelerated expansion of the universe , the archimedean - type coupling leads to an effective heating of the dark matter , so that the de and dm contributions happen to be of the same order . the second question arises : for what values of the coupling parameter @xmath49 the archimedean - type interaction is able to protect the universe from the big rip singularity ? the third motif is that the pressure of the dark energy , @xmath38 , is the key element in our model , which guides the behavior of both state functions of the dark matter , the energy - density @xmath50 and pressure @xmath51 ( via the archimedean - type force ) , the energy - density @xmath39 of the dark energy ( via the equation of state ) , thus predetermining the behavior of the hubble function @xmath27 and its derivative @xmath12 ( via the einstein equations ) . this means that in the model under consideration one can classify the asymptotic types of the universe behavior using only one element , just the function @xmath52 . this function , @xmath52 , satisfies the key equation , which is the nonlinear differential equation of the second order and includes six guiding parameters : @xmath48 , @xmath53 and @xmath54 coming from the three - parameter equation of state for the dark energy , the effective coupling constant of the archimedean - type interaction @xmath49 , the starting energy , @xmath55 , and temperature , @xmath56 , of the dark matter . thus , ( keeping in mind the initial data @xmath57 and @xmath58 ) we have a possibility to classify all the types of asymptotic behavior of the model using eight parameters . the question arises : in what domains of the 8-dimensional effective space of the guiding parameters the big rip is reliably avoided ? in order to answer these questions we organized this paper as follows . in section ii we remind briefly the master equations for the de and dm evolution in the framework of the archimedean - type model ( subsections iia - iic ) . in subsection iid we describe our classification scheme , which is based on eight parameters and includes three bifurcation points : @xmath59 ( @xmath49 is the archimedean - type coupling constant ) ; @xmath60 ( @xmath48 is the de relaxation parameter ) ; @xmath61 , the phantom - crossing point . in section iii we focus on the analysis of the models indicated as models with asymptotic de domination , for which the contribution of the dm to the universe expansion is asymptotically negligible . in subsection iiia we describe the big rip , pseudo rip , cyclic solutions with @xmath59 and @xmath60 in terms of two guiding parameters @xmath53 , @xmath54 , and the initial value of the de pressure . in subsection iiib we focus on the models with @xmath59 and @xmath62 , thus analyzing the role of the de relaxation parameter in various scenaria of the big rip and little rip formation ( in terms of three guiding parameters @xmath48 , @xmath53 , @xmath54 , and two initial data for the de ) . in section iv we consider the models with @xmath63 and @xmath60 ( in terms of three guiding parameters @xmath49 , @xmath53 , @xmath54 and the initial value of the de pressure ) , and show that the archimedean - type coupling of the dm with the barotropic de avoids the big rip regimes in the universe expansion . in section v we discuss the general case @xmath63 and @xmath62 qualitatively and numerically , and show that the archimedean - type coupling of the dm with the de of the rheological type converts the big rip regimes into the little rip , pseudo rip and cyclic regimes of the universe expansion . let us remind briefly the master equations of the model of the archimedean - type coupling between the dark energy and dark matter . these equations are derived in @xcite and now we use them to analyze the late - time behavior of the universe in the framework of this model . the flrw - type isotropic spatially homogeneous cosmological model with two interacting constituents : the dark energy and dark matter are known to be described by two nontrivial einstein equations @xmath64 \ , , \label{einredu10}\ ] ] @xmath65 here the dot denotes the derivative with respect to time , the functions @xmath66 and @xmath52 are the energy - density and pressure of the dark energy , respectively , @xmath67 and @xmath68 describe the corresponding state functions of the dark matter . the cosmological constant @xmath5 is considered to be incorporated into the de energy density @xmath39 and de pressure @xmath38 . as usual , @xmath69 is the hubble function . the conservation law as the compatibility equation for the set ( [ einredu10])-([einredu1 ] ) has the form @xmath70 the energy density and pressure of the one - component dark matter can be effectively presented , respectively , by the integrals @xmath71 @xmath72 the auxiliary function @xmath73 of the dimensionless variable @xmath74 @xmath75\ } } \ , , \label{ff}\ ] ] and convenient parameters @xmath76 contain the effective archimedean - type coupling constant @xmath49 , an effective number density @xmath77 , the mass @xmath78 and temperature @xmath56 of the leading sort of the dark matter particles . the term @xmath79 } \ , , \label{mcd}\ ] ] is the modified bessel function , @xmath80 is the boltzmann constant . to describe the dark energy fluid we use the linear three - parameter equation of state @xmath81 when @xmath82 and @xmath60 , eq . ( [ simplest0 ] ) introduces the model in which the dark energy relates to the cosmological constant @xmath5 . when @xmath83 and @xmath60 , eq . ( [ simplest0 ] ) gives the well - known linear relation @xmath84 with @xmath85 . the retardation of response in the dark energy evolution is taken into account by inserting the term containing the first derivative of the pressure @xmath86 . an equivalent scheme is widely used in the extended thermodynamics and rheology ( see , e.g. , @xcite ) , in which the extended constitutive equation for the thermodynamically coupled variables @xmath87 and @xmath88 has the form @xmath89 here @xmath20 is a relaxation time , a new coupling parameter of the model . in the cosmological context @xmath20 is generally the function of time , @xmath90 . we assume that @xmath91 , i.e. , this relaxation time can be measured in natural cosmological scale @xcite . our ansatz here is that the dimensionless parameter @xmath48 is constant . when the quantities @xmath66 and @xmath52 depend on time through the scale factor @xmath0 only , i.e. , @xmath92 , @xmath93 , the so - called @xmath94-representation is convenient , which is based on the following relations @xmath95 @xmath96 in these terms the balance equation ( [ balance ] ) and the constitutive equation ( [ simplest0 ] ) give the key equation for the pressure of the dark energy @xmath97 @xmath98 the prime denotes the derivative with respect to @xmath94 . the source @xmath99 is defined as follows : @xmath100^{\prime}}{2 x^4 } \int_0^{\infty}\frac{q^4 dq e^{{-}\lambda _ { * } \sqrt{1{+}q^2}}}{\sqrt{1{+}q^2 f(x ) } } . \label{key2}\ ] ] the quantity @xmath99 vanishes , when the archimedean - type coupling parameter vanishes , i.e. , @xmath101 . we deal with the differential equation of the second - order linear in the derivatives but nonlinear in the unknown function @xmath97 . the function @xmath38 plays the key role in the analysis of the whole model behavior . indeed , when @xmath97 is found from the key equation ( [ key1 ] ) , we can reconstruct the de energy - density @xmath39 using the formula @xmath102 then calculate the dm state functions @xmath103 and @xmath104 using ( [ e(x ) ] ) and ( [ p(x ) ] ) , then find @xmath105 from the einstein equation ( [ einredu1 ] ) , and finally , reconstruct the scale factor as the function of time using ( [ diffa2 ] ) . we are interested to discuss the late - time period of evolution , i.e. , the period , when @xmath106 . we consider both cases : @xmath107 ( future finite time singularity ) and @xmath108 . clearly , there are three possible types of the asymptotic behavior of the dark fluid composed of the dark energy and dark matter . the first type refers to the asymptote @xmath109 . this case includes the submodel with @xmath110 ) . since the archimedean - type force is proportional to the four - gradient of the dark energy pressure @xmath38 , then asymptotically the dark matter decouples from the dark energy , its energy - density @xmath50 decreases as @xmath111 and we obtain the case indicated as an asymptotic de domination . the second type of behavior refers to the case @xmath112 . now the dm becomes frozen , since the function @xmath73 ( [ ff ] ) tends to zero exponentially , the dm decouples from the de , and again we deal with the model of the de domination . the third and the most interesting case , refers to the asymptotic behavior @xmath113 , for which the function @xmath73 tends to infinity , the dm becomes effectively ultrarelativistic and thus plays an active role in the energy redistribution process . the model as a whole contains four effective guiding parameters ( @xmath48 , @xmath53 and @xmath54 describing the de equation of state , and @xmath49 , the effective archimedean - type coupling constant ) and four initial parameters ( @xmath57 and @xmath58 , initial data for the de pressure , and @xmath55 , @xmath56 initial data for the dm ) . the classification scheme includes three bifurcation points : the first , @xmath59 indicates whether the archimedean - type coupling is switched on or not ; the second , @xmath60 indicates whether the de possesses the simplest rheological property or not , the third , @xmath61 , relates to the phantom - crossing point . _ remark on the big rip symptom : _ when the integral in the right - hand side of ( [ diffa2 ] ) converges on the upper limit , the infinite value of the scale factor @xmath0 can be reached at finite time @xmath114 , thus we deal with the big rip . this first model relates to the case , when the dark energy does not possess rheological properties , and there is no archimedean - type coupling between de and dm . the contribution of the dm is asymptotically vanishing , and in this sense we deal with the case of the de domination . the main results for this model are well - known and our goal is to recover them as the limiting case of our model in the corresponding terminology . let us recover , first , the well - known results presenting the big rip solutions , when @xmath60 , @xmath83 and @xmath59 . the key equation ( [ key1 ] ) takes the form @xmath115 and the solution is @xmath116 it is convenient to introduce the new dimensionless parameter @xmath117 since it plays an important role below . for the de energy - density @xmath118 we immediately obtain @xmath119 so that @xmath120 clearly , the real solution for the hubble function exists , when the product @xmath121 is positive . using the remark on the big rip symptom one can conclude that the integral in ( [ diffa2 ] ) converges , when @xmath122 is positive , i.e. @xmath123 . this simplest submodel allows us to verify this fact directly by computing the scale factor analytically . indeed , according to ( [ diffa2 ] ) the scale factor has the form @xmath124^{-2\alpha } \ , . \label{key141}\ ] ] when @xmath123 or in other words , when @xmath125 , one obtains that the parameter @xmath122 is positive . this means that the scale factor tends to infinity at some finite time value @xmath126 , given by @xmath127 when @xmath128 , the solution is real , when the initial value of the de pressure is negative @xmath129 . thus we deal with the well - known big rip solution , which is characterized by the following asymptotes at @xmath130 : @xmath131 , @xmath132 , @xmath133 , @xmath134 and @xmath42 . let us add that when the parameter @xmath53 is negative and belongs to the interval @xmath123 , the parameter @xmath122 is positive ( see ( [ key1119 ] ) ) . there are two special cases . when @xmath135 , the parameter @xmath122 becomes infinite , @xmath136 ; when @xmath137 , the parameter @xmath122 tends to zero also . finally , when @xmath122 is negative , i.e. , @xmath138 or @xmath139 , the solution ( [ key141 ] ) describes power - law expansion of the universe . the little rip solutions do not appear in this submodel . when @xmath140 , the solution of the key equation @xmath141 has the form @xmath142 \ , . \label{key115}\ ] ] the square of the hubble function reads now @xmath143 \ , , \label{key12}\ ] ] where @xmath144 \ , . \label{key13}\ ] ] the signs of the parameters @xmath145 , @xmath54 , @xmath122 and @xmath53 provide important details of the classification . in order to classify the models we , first , distinguish three ( principal ) cases and then show that all other cases can be reduced to these three ones . _ ( 1 ) the model with @xmath146 , @xmath123 , @xmath147 . _ let us consider , first , the case when @xmath123 and @xmath146 . the corresponding initial value @xmath148 should be negative , and the inequality @xmath149 has to be satisfied . the solution for the scale factor has the form @xmath150^{{-}2\alpha } \ , , \label{key14}\ ] ] where @xmath151 there exists a moment @xmath126 , for which @xmath152 . at @xmath130 the plot of @xmath0 has a vertical asymptote . this critical moment of time , @xmath153 , can be found from the equation @xmath154 and has the form @xmath155 again we deal with the big rip solution , characterized by the asymptotes @xmath156 , @xmath157 , @xmath133 , @xmath134 , @xmath42 , but now , in the case @xmath140 , we can indicate the big rip singularity as that of hyperbolic type in contrast to the one of the power - law type , presented in the previous subsection ( with @xmath158 ) . _ remark 1 : _ two submodels with negative @xmath122 , namely , @xmath146 , @xmath159 , @xmath147 and @xmath146 , @xmath139 , @xmath160 , give formally the same scale factor ( [ key14 ] ) . however , the essential difference is that at @xmath161 the hubble function @xmath27 ( [ key12 ] ) tends to the constant @xmath162 , and we deal with the pseudo rip instead of the big rip . _ ( 2 ) the model with @xmath163 , @xmath123 , @xmath147 . _ when @xmath123 and @xmath164 , one obtains from ( [ key12 ] ) , that there is a critical value @xmath165 @xmath166 for which the function @xmath167 @xmath168 \label{key28}\ ] ] takes zero value . let us assume that @xmath169 , i.e. , according to ( [ key27 ] ) @xmath170 and @xmath171 . it is possible when @xmath172 . since the model solutions for @xmath105 obtained from ( [ key28 ] ) can not be prolonged for @xmath173 , @xmath94 starts to decrease after the moment @xmath174 . the corresponding solution for the scale factor has now the form @xmath175^{2\alpha } \ , , \label{key24}\ ] ] where @xmath176 clearly , the plot of this function is symmetric in reference to the moment @xmath177 ; the formula ( [ key24 ] ) gives the maximal value ( [ key27 ] ) @xmath178 at this moment . at @xmath6 the scale factor @xmath0 tends to zero as @xmath179 the hubble function is described by the finite function @xmath180 } \ , , \label{key249}\ ] ] which changes the sign at @xmath174 . this means that the universe expansion turns into a collapse at this moment . the acceleration parameter @xmath181 takes infinite value at @xmath174 , since the hubble function in the denominator vanishes at this moment . clearly , the universe passes two eras with @xmath182 ( accelerated expansion and accelerated collapse ) , and two eras with @xmath183 ( decelerated expansion and decelerated collapse ) . we deal with a closed universe which does not admit the big rip behavior , despite @xmath184 . _ remark 2 : _ two submodels with negative @xmath122 , namely , @xmath163 , @xmath159 , @xmath160 and @xmath163 , @xmath139 , @xmath147 , give formally the same expression for the scale factor ( [ key24 ] ) . since now @xmath122 is negative , the hubble function @xmath27 ( [ key12 ] ) tends again to the constant @xmath162 , if we require @xmath185 to avoid the internal singular point . we deal now with the pseudo rip instead of big rip . _ ( 3 ) the model with @xmath163 , @xmath123 , @xmath160 . _ now the scale factor can be expressed in terms of trigonometric functions @xmath186^{-2\alpha } \ , , \label{key34}\ ] ] the parameter @xmath145 has to satisfy the inequality @xmath187 , thus , the initial value @xmath148 is restricted by @xmath188 . again , there exists a moment @xmath153 @xmath189 for which the plot of the scale factor has the vertical asymptote typical for the big rip with @xmath156 , @xmath157 , @xmath133 , @xmath134 , @xmath42 . it is convenient to indicate such singularity as the big rip singularity of the trigonometric type . _ remark 3 : _ two submodels with negative @xmath122 , namely , @xmath163 , @xmath159 , @xmath160 and @xmath163 , @xmath139 , @xmath147 , give formally the same scale factor ( [ key34 ] ) . since now the parameter @xmath190 becomes imaginary ( see ( [ key15 ] ) ) , at @xmath161 one should replace the trigonometric functions by the hyperbolic one . again , asymptotically we deal with the pseudo rip solution instead of big rip . when we consider this submodel we assume that the archimedean - type coupling is switched of , but the dark energy possesses the simplest rheological properties . since @xmath59 , the asymptotic regime again will be characterized by the dark energy domination . in order to analyze this model let us introduce the new unknown function @xmath191 as follows : @xmath192 and define the initial value @xmath193 as @xmath194 the function @xmath191 satisfies the euler equation @xmath195 the characteristic polynomial of this euler equation has two roots @xmath196 \ , , \label{roots17}\ ] ] which can be real or complex depending on the values of the parameters @xmath48 and @xmath53 . one can distinguish three subcases . _ ( i ) _ @xmath198 . when the discriminant in ( [ roots17 ] ) is positive , one obtains @xmath199 @xmath200}{2\gamma } \left(x^{\gamma } { - } x^{{-}\gamma } \right ) \right\ } \ , , \label{aperiod2}\ ] ] where @xmath201 according to the remark about the big rip symptom we can conclude that the integral in ( [ diffa2 ] ) converges now , when @xmath202 , thus this inequality guarantees the big rip existence . in order to _ illustrate _ the existence of the big rip by exact solutions , let us assume that initial data satisfy the equality @xmath203 \ , . \label{ap1}\ ] ] then the solutions can be simplified as follows @xmath204 and @xmath191 becomes proportional to the required leading order term exactly . now the de state functions are of the form @xmath205 @xmath206 where the new parameter @xmath207 is introduced . in the case of the de domination the square of the hubble function @xmath208 \ , , \label{ap4}\ ] ] formally coincides with ( [ key12 ] ) , if we use the substitutions @xmath209 based on the results of the previous section one can conclude that the big rip takes place , if @xmath202 . there are two domains on the plane of the parameters @xmath210 , where this inequality holds at @xmath198 : first , @xmath211 , @xmath212 , second , @xmath213 . the first domain is a natural extension of the interval @xmath214 , which is obtained if we put @xmath60 into the right - hand side of the inequality ; let us remind that just for the case @xmath214 the big rip appears , when @xmath60 . the second interval for the big rip existence , @xmath213 at @xmath198 , has no analog at @xmath60 . the parameter @xmath54 distinguishes three situations admitting the big rip at @xmath215 : \(1 ) @xmath158 . in both domains the singularity of @xmath0 is of the power - law type . \(2 ) @xmath216 . in the first domain , where @xmath211 , @xmath212 , the singularity is of the hyperbolic type . in the second domain , where @xmath213 , the singularity is of the trigonometric type . \(3 ) @xmath217 . in the first domain , where @xmath211 , @xmath212 , the singularity is of the trigonometric type . in the second domain , where @xmath213 , the singularity is of the hyperbolic type . more detailed constraints for the guiding parameters follow from the inequalities @xmath218 or @xmath219 , respectively . _ ( ii ) _ @xmath220 . when the parameter @xmath48 is negative , we use the following scheme of analysis . we put @xmath221 and obtain that @xmath222 the de state functions take now the form @xmath223 @xmath224 and the square of the hubble function reads @xmath225 \ , . \label{7ap4}\ ] ] again , these functions can be obtained asymptotically at @xmath226 , since now @xmath227 and in this limit one obtains that @xmath228 . the sum @xmath229 should be negative , and the product @xmath230 should be positive , if we try to find the analogs of the big rip . this is possible , when @xmath231 . clearly , we obtain the power - law singularity , when @xmath158 , the hyperbolic singularity , when @xmath216 , and the trigonometric singularity , when @xmath216 . the discriminant can take zero value , only when @xmath198 . for the positive @xmath48 there are three interesting cases : @xmath233 , @xmath234 and @xmath235 . _ ( i ) _ @xmath233 . the corresponding solution for @xmath191 is @xmath236\right\ } \ , , \label{aperiod22}\ ] ] where @xmath237 , if @xmath238 , and @xmath239 , if @xmath240 . according to the mentioned symptom , the big rip is possible , when @xmath241 is negative ; clearly , it is possible only if we deal with the second solution , i.e. , when @xmath233 and @xmath240 . these constraints require that @xmath184 . in order to illustrate the behavior of the scale factor , let us assume that , first , @xmath83 , second , @xmath242 and , third , @xmath243 . then we obtain @xmath244 @xmath245 the square of the hubble function reads @xmath246 and the integral in ( [ diffa2 ] ) @xmath247 \ , , \label{aperi3}\ ] ] happens to be reduced to the error - function @xmath248 \equiv \frac{2}{\sqrt{\pi } } \int_0^{z } du \ e^{{-}u^2 } \ , . \label{aperi4}\ ] ] the error - function @xmath249 $ ] takes finite value @xmath250 { = } 1 $ ] , when the upper limit in the integral goes to the infinity , thus , the value @xmath226 can be reached at the moment @xmath251 . the scale factor has the form @xmath252 \right]^2 \right\ } \ , , \label{aperi5}\ ] ] where the symbol @xmath253 stands for the function inverse to the error - function . thus , when @xmath233 the big rip is possible , if @xmath240 . _ ( ii ) _ @xmath234 . this case corresponds to the positive parameter @xmath241 . this means that @xmath191 ( see ( [ aperiod22 ] ) ) asymptotically vanishes , the hubble function tends to constant , and we deal with the pseudo rip . _ ( iii ) _ @xmath254 . double real roots appear now , when @xmath255 , i.e. , when @xmath256 or @xmath61 . the first root , @xmath256 , corresponds to the solution ( [ aperiod22 ] ) with @xmath257 , and we again deal with the pseudo rip . the second root , @xmath258 , refers to the special case with diverging quantity @xmath259 , and we have to return to the key equation ( [ key1 ] ) . as we emphasized in @xcite , in this special model the de pressure and the de energy density contain the logarithmic terms squared . in particular , for the initial data , which satisfy the equality @xmath260 , these de state functions are @xmath261 @xmath262 the scale factor has the form @xmath263 } \right\ } \ , , \label{sc61}\ ] ] it describes the super - exponential expansion . the hubble function is monotonic @xmath264 } \label{sc67}\ ] ] and increases infinitely . the function @xmath12 @xmath265 } \label{sc679}\ ] ] also tends to infinity at @xmath6 . the acceleration parameter @xmath266}}{\cosh^2{\left[\sqrt{12\pi g \rho_0 } \ ( t{-}t_0 ) \right ] } } \label{sc69}\ ] ] starts with @xmath267 , reaches the maximum @xmath268 at @xmath269 , and tends asymptotically to @xmath270 . this is the explicit example of the little rip . complex roots appear at positive @xmath48 . the solution is quasiperiodic @xmath272 @xmath273}{\beta } \sin{(\beta \log{x } ) } \right\ } \ , , \label{period2}\ ] ] where @xmath274 in this section it is convenient to use the variable @xmath275 and to present the square of the hubble function as follows @xmath276 \right\ } \ , , \label{period3}\ ] ] where the parameters @xmath277 @xmath278 - \xi \beta z(1 ) \label{period5}\ ] ] depend on the initial data @xmath193 and @xmath279 ; for the illustration one can choose them so that @xmath280 . again there are three interesting cases . _ ( i ) @xmath281 for arbitrary @xmath20 . _ for instance , when @xmath282 , @xmath280 and @xmath170 we obtain that @xmath281 . then the hubble function @xmath283 is real and tends to @xmath284 at @xmath285 ; we deal with the pseudo rip . _ ( ii ) @xmath286 . _ to illustrate the situation , for which @xmath287 is nonnegative and has a number of zeros , let us choose , first , that @xmath288 , second , that @xmath289 , @xmath280 , providing @xmath290 the hubble function is periodic with maximal value @xmath291 . the corresponding scale factor @xmath292 \right\ } \label{period7}\ ] ] takes finite values satisfying the inequalities @xmath293 thus , there are neither big rip , nor little rip , despite @xmath123 ( @xmath233 ) . _ ( iii ) @xmath287 changes the sign . _ when @xmath294 , the quasiperiodic function @xmath295 given by ( [ period3 ] ) inevitably reaches zero value at some moment @xmath296 . near this point the function @xmath295 can be presented as @xmath297 depending on the guiding parameters @xmath53 , @xmath48 , @xmath54 , and on the initial data @xmath193 and @xmath279 , the quantity @xmath298 can vanish or be positive . \(1 ) @xmath299 . in the vicinity of the moment @xmath300 one obtains that @xmath301 , and thus @xmath302}$ ] . \(2 ) @xmath303 . the decomposition ( [ period9 ] ) starts with @xmath304 , we obtain that @xmath305}$ ] . in both cases the universe has finite size and its life - time is finite . to summarize the results of the analysis let us indicate the domains on the plane of the parameters @xmath48 and @xmath53 , for which the big rip scenaria can be realized : these domains are displayed on fig.1 . first , let us focus on the case @xmath60 . according to the results of subsection iia , the big rip can be realized on the interval @xmath123 of the vertical line @xmath60 . it is the `` classical '' big rip domain , which is characterized by the condition @xmath306 . when @xmath307 this `` classical '' big rip zone contracts along the line @xmath53 to the interval @xmath308 , and disappears at all when @xmath309 ( see the domain i on fig.1 ) . in this sense , the simplest rheological property of the de , namely , the retardation of the de response to the universe expansion , can avoid the big rip , which was the fate of the universe at @xmath123 , if the relaxation parameter @xmath48 exceeds the critical value @xmath235 . in other words , when the de relaxation time @xmath310 is bigger than @xmath311 , where @xmath312 is the universe expansion scalar , the regime of the big rip can not be supported by the de with such relaxation time . if @xmath48 is negative , the interval @xmath123 again relates to the big rip scenario of the universe expansion ( see the domain ii on fig.1 ) . the choice of the parameter @xmath54 can not avoid the big rip scenario , but it predetermined the type of the big rip : the power - law type , hyperbolic or trigonometric ones . when @xmath138 , there was no classical big rip at @xmath60 . the same fact can be indicated at @xmath198 . however , if the parameter @xmath48 is negative , the big rip scenario is possible for @xmath138 , since instead of retardation the de displays the acceleration of the response to the universe expansion ( see the domain iii on fig.1 ) . when @xmath139 , there was no `` classical '' big rip at @xmath60 . however , the big rip becomes possible at @xmath139 for positive relaxation parameter ( see the domain iv on fig.1 ) . mathematically , this new domain for the big rip scenaria appears , since the characteristic equation is quadratic , and the presence of the second root in ( [ roots17 ] ) extends the possibilities of the big rip scenaria . indeed , when @xmath62 , but @xmath313 , the roots of the mentioned characteristic equation can be estimated as @xmath314 the first root does not depend on @xmath48 and gives us the `` classical '' big rip , when @xmath123 . the second root is positive and thus describes the `` non - classical '' big rip , when either @xmath128 and @xmath198 , or @xmath138 and @xmath220 . this explains the appearance of two new domains iv and v on fig.1 . for small values of @xmath48 the second characteristic number @xmath315 is respectively big ; from the physical point of view , we can speak about instability of the de response to the universe expansion near the bifurcation line @xmath60 . * * [ htmb ] and @xmath53 , for which the universe evolution follows the big rip scenario in the model of the asymptotic dark energy domination . the horizontal stripe @xmath123 is divided into three domains . there are two domains of the big rip in this stripe : the first domain is bounded by the part of parabola @xmath316 ( @xmath317 ) , by the vertical line @xmath60 and horizontal line @xmath258 ; the second domain is situated at @xmath220 . two rectangular sectors @xmath198 , @xmath139 and @xmath220 , @xmath138 also refer to the case of big rip . , title="fig:",width=311,height=321 ] in this model we suggest that the dark energy is described by the equation of state @xmath319 , but now the archimedean - type coupling is switched on . in other words , there are two coupled energy reservoirs in the universe : the dark energy and dark matter , and an effective energy redistribution between them is possible in the course of the universe accelerated expansion . let the moment @xmath321 ( or equivalently , @xmath322 ) be chosen so that the value @xmath148 is large and negative . the term @xmath323 ( see ( [ key2 ] ) ) takes the form @xmath324 where we use the new parameter @xmath325 here and below the symbols with asterisk : @xmath326 , @xmath55 , @xmath49 , relate to the parameters of the leading dm component . the key equation ( [ key1 ] ) transforms into nonlinear differential equation of the first order @xmath327 } { -}\sigma \right ] { = } 3\left[\rho_0 { + } ( 1{+}\sigma ) \pi \right]\ , , \label{ar3}\ ] ] and its structure prompts us the following transformations . let us introduce new variables @xmath328 , @xmath329 and parameter @xmath330 . for the case @xmath331 the transformations have the form : @xmath332 \ , , \label{ar4}\ ] ] @xmath333 when @xmath334 , we replace formally @xmath335 by @xmath336 and @xmath328 by @xmath337 . for the case @xmath338 the mentioned transformations are not appropriate and we consider this submodel in the special section . in terms of new variables the key equation can be transformed into the linear equation for the function @xmath339 : @xmath340 the energy density of the de and dm in these new terms can be written as @xmath341 thus , the square of the hubble function takes the form @xmath342 \ , , \label{ar12}\ ] ] and the rate of the hubble function evolution @xmath12 is described by the formula @xmath343 \ , . \label{ar121}\ ] ] the function @xmath0 can be extracted from the equation @xmath344}{y \sqrt{\frac{\rho_0}{1{+}\sigma } { - } \frac{\sigma}{\nu_{*}}\left(y { + } \frac{1}{4\alpha x } e^y \right ) } } \ , . \label{ar13}\ ] ] here we have to put @xmath339 as the solution of ( [ ar5 ] ) and to use the inverse function @xmath345 = y\left[\left(\frac{a(t)}{a(t_0)a _ { * } } \right)^{4 } \right ] \label{ar14}\ ] ] on the upper limit of the integral . the solution to the equation ( [ ar5 ] ) is @xmath346^{4\alpha } + y^{4\alpha } \int^y_{y(1 ) } \frac{du}{u^{1 + 4\alpha } } e^{u } \ , . \label{ar6}\ ] ] here the constants @xmath347 and @xmath348 are defined using the initial data as follows : @xmath349 \ , . \label{ar70}\ ] ] whatever the parameter @xmath122 is , when @xmath156 ( i.e. , @xmath350 ) the leading order terms in the asymptotic decomposition of the function @xmath339 @xmath351 \label{ar7}\ ] ] come from the generalized integral exponential @xmath352 which appears in the right - hand side of ( [ ar6 ] ) . thus , when @xmath353 , the integrand in ( [ ar13 ] ) behaves at @xmath354 as @xmath355 this asymptotic estimation is valid when @xmath356 , i.e. , in both interesting cases : @xmath123 and @xmath139 . big rip can exist , when the integral in the right - hand side of ( [ ar13 ] ) converges at @xmath354 . in our case the integral diverges at @xmath354 as @xmath357 , and big rip can not be realized in the scenario with @xmath358 and @xmath359 . in the asymptotic limit @xmath354 the formula ( [ ar12 ] ) for the @xmath287 yields @xmath360 \ , . \label{aar12}\ ] ] in other words , when @xmath356 the hubble function infinitely grows as @xmath361 . as for @xmath12 , ( [ ar121 ] ) yields @xmath362 \ , , \label{aar121}\ ] ] i.e. , the rate of the hubble function growth tends to the positive constant @xmath363 . keeping in mind ( [ ar13 ] ) and ( [ ar15 ] ) we can reconstruct now the function @xmath0 using the function @xmath364 for the case @xmath365 . when @xmath354 , the behavior of the function @xmath339 is predetermined by the second term in ( [ ar6 ] ) , since the generalized integral exponent ( [ ar8 ] ) gives the exponential leading order term @xmath366 ( see ( [ ar7 ] ) ) . the inverse function @xmath367 is of the logarithmic type , which can be presented by the following iteration procedure @xmath368 ] } \ , , \label{ar9}\ ] ] zero - order estimations yielding @xmath369 the scale factor @xmath370 is described by the anti - gaussian function , which was obtained in @xcite as the exact solution of master equations of the archimedean - type model . since the corresponding hubble function @xmath371 grows linearly with time , the obtained solution can be classified as the little rip with @xmath372 and @xmath373 . to complete the analysis let us consider two special cases . _ ( i ) special case @xmath374 _ in the special case : @xmath375 , or equivalently , @xmath374 , we should consider the solutions in more details . the key equation @xmath327 } { + } 1 \right ] { = } 3 \rho_0 \label{3ar3}\ ] ] can be now transformed into @xmath376 using the modified replacements @xmath377 @xmath378 \ , , \quad \tilde{y}(1 ) = 0 \ , . \label{3ar44}\ ] ] the solution to ( [ 3ar4 ] ) @xmath379 e^{{-}\frac{4}{3\rho_0 \nu _ { * } } \tilde{y } } { - } \frac{1}{1{+}\frac{3\rho_0 \nu_{*}}{4 } } e^{\tilde{y } } \ , , \label{3ar45}\ ] ] shows that @xmath380 takes zero value at some @xmath381 and then changes the sign . since @xmath382 is a positively defined function , we can conclude that the regime @xmath36 with @xmath383 can not be realized , when @xmath374 . in other words , the phantom - crossing value @xmath61 , which corresponds to @xmath375 is the critical value : when @xmath384 , the little rip regime with @xmath385 and @xmath386 is possible ; when @xmath374 , the scale factor can not reach infinite value . _ ( ii ) special case @xmath256 _ the special case : @xmath387 , or equivalently , @xmath388 , relates to the ultrarelativistic de , when @xmath83 , since @xmath389 . the functions @xmath328 , @xmath287 and @xmath12 take now the form @xmath390 @xmath391 \ , , \label{3332}\ ] ] @xmath392 where the new constant @xmath393 is introduced @xmath394 when @xmath395 , @xmath396 , @xmath397 , thus , we deal with the de sitter asymptote . in terms of the dynamic variables ( [ ar4 ] ) the key equation for the de pressure takes the form @xmath399 { + } y = 0 \ , . \label{5ar1}\ ] ] at @xmath358 it coincides with ( [ ar5 ] ) , as it should be . the function @xmath12 and the square of the hubble function can be presented as follows @xmath400 \ , , \label{es1}\ ] ] @xmath401 \ , . \label{es2}\ ] ] let us , first , analyze these functions in terms of the variable @xmath329 . let us assume that the asymptotic solution of ( [ 5ar1 ] ) at @xmath403 differs insignificantly from the solution to ( [ ar5 ] ) , and let us consider the function @xmath404 \ , , \label{5ar21}\ ] ] in which the parameter @xmath405 is unknown , but should satisfy the condition @xmath406 this means that we consider the solution in the form ( [ ar7 ] ) and restrict ourselves by the term @xmath407 in the parentheses . with mentioned accuracy the function @xmath339 ( [ 5ar21 ] ) satisfies ( [ 5ar1 ] ) with ( [ 5ar2 ] ) , when @xmath408 from the qualitative point of view the behavior of the de with @xmath62 at @xmath131 is analogous to that of the de with @xmath60 and effective parameter @xmath409 equal to @xmath410 it is interesting that @xmath411 , when @xmath61 , i.e. , these two parameters coincide for arbitrary @xmath48 at the phantom - crossing point @xmath412 . for the solution of the type ( [ 5ar21 ] ) the term @xmath413 behaves as @xmath414 , when @xmath354 . thus , the contributions of the last terms in ( [ es1 ] ) and ( [ es2 ] ) ( proportional to the parameter @xmath48 ) can be neglected at @xmath354 . thus , the main conclusion that the archimedean - type interaction avoids the big rip remains valid at @xmath307 , @xmath402 . the key equation for this case can be obtained from ( [ 5ar1 ] ) by the formal replacement @xmath416 and @xmath417 with positively defined @xmath382 . numerical calculations show that in this case the regime @xmath354 does not exist . to illustrate qualitatively such a behavior one can mention the following : the regime with @xmath354 would require that the leading order terms in the equation ( [ 5ar1 ] ) are linked by @xmath418 , which is in contradiction with the requirement @xmath419 . then two scenaria are available : first , with @xmath420 , second with @xmath421 . in the first scenario there is no big rip , and , as a rule , we deal with the pseudo rip . in the second scenario the quantity @xmath329 with positive initial value @xmath422 would have negative derivative , thus , the equation ( [ es1 ] ) would be transformed into @xmath423 \ , , \label{es4}\ ] ] showing explicitly that the derivative of the hubble function is negative . this means that at some moment @xmath424 ( finite or infinite ) the hubble function will take zero value providing the scale factor @xmath0 to reach its maximal value @xmath425 . in other words , in this scenario the big rip can nor appear in contrast to the model with @xmath59 . when @xmath48 is negative , the procedure similar to the one used in the case @xmath307 , @xmath402 gives the asymptotic formulas analogous to ( [ aar12 ] ) and ( [ aar121 ] ) @xmath427 @xmath428 which do not contain the parameter @xmath48 in the leading order term . however , now we consider the parameter @xmath335 to be positive , and can conclude that we deal with the little rip , when @xmath429 , and the solution with finite @xmath430 , when @xmath431 . in any case the archimedean - type interaction avoids the big rip . as in the general case described in @xcite , asymptotic model admits the solution of the anti - gaussian type , when @xmath432 . indeed , the corresponding key equation @xmath433 } \label{ag1}\ ] ] is satisfied with the solution @xmath434 if the guiding parameters are coupled by the equality @xmath435 since @xmath436 is positive , one requires that @xmath437 , i.e. , this equality can not be satisfied when @xmath60 . when the right - hand side of ( [ ag3 ] ) is positive , the logarithmic solution can appear as a result of some `` fine tuning '' : one can satisfy ( [ ag3 ] ) , e.g. , by varying the initial temperature @xmath56 of the dm , which enters the parameter @xmath438 , the argument of the function @xmath439 ( see ( [ ar2 ] ) ) . the asymptotic de energy - density is now @xmath440 where the relation @xmath441 \label{toy011}\ ] ] is used . the sum @xmath442 remains constant @xmath443 the dm energy density @xmath103 is now constant , i.e. , @xmath444 and the hubble function @xmath105 can be found from the equation @xmath445 \ , . \label{toy4}\ ] ] the corresponding scale factor @xmath0 can be written in the form @xmath446 where the parameters with asterisks are defined as follows @xmath447 \right\ } \,,\ ] ] @xmath448 } \ , . \label{toy6}\ ] ] the acceleration parameter @xmath449 for the anti - gaussian expansion is positive and exceeds the unity : @xmath450 clearly , this solution illustrates one of the versions of the little rip . in order to confirm the qualitative conclusions made for the model with @xmath62 , @xmath359 , we analyzed numerically the models for the parameters @xmath48 and @xmath53 belonging to the domains i , ii , iii , iv displayed on fig.1 . these calculations are illustrated by fig.2 and fig.3 . these figures contain four panels : the first one displays the hubble function @xmath105 , the second panel demonstrates the behavior of the scale factor @xmath0 , the third panel describes the rate of the growth of the hubble function @xmath12 , the fourth panel displays the function @xmath118 . * * [ htmb ] , @xmath451 , @xmath452 ( i.e. , for the case @xmath123 , @xmath220 and @xmath453 , which refers to the big rip scenario at @xmath59 according to fig.1 ) . clearly , when @xmath63 , i.e. , the archimedean - type coupling is switched on , the model solution is of the little rip type with @xmath454 , @xmath134 , @xmath373 , @xmath41 , @xmath131 . , title="fig:",width=311,height=363 ] * * [ htmb ] , @xmath455 , @xmath452 ( i.e. , for the case @xmath139 , @xmath198 and @xmath456 , which refers to the big rip scenario at @xmath59 according to fig.1 ) . when @xmath63 the model solution is of the little rip type with @xmath454 , @xmath134 , @xmath373 , @xmath41 , @xmath131 . , when the dark matter is effectively coupled to the dark energy by the archimedean - type interaction , the late - time evolution of the universe with negative de pressure is protected from the big rip singularity , and the little rip becomes a typical fate of the universe . this is the main conclusion of the paper . more detailed discussion includes the following three items . \1 . in the presented model of archimedean - type coupling between the dark matter and dark energy the de pressure , @xmath38 , is the key element of modeling . all the model solutions can be divided into three classes with respect to asymptotic behavior of the state function @xmath457 ( we consider both cases : @xmath107 ( future finite time singularity ) and @xmath108 ) . the first class is characterized by finite @xmath458 ; the second class relates to @xmath459 ; we deal with the solutions of the third class , when @xmath460 . in @xcite we focused on the solutions of the first class , and have shown qualitatively that the solutions of the second and third classes exist for some values of the effective guiding model parameters @xmath48 , @xmath53 , @xmath54 and @xmath49 . three parameters @xmath48 , @xmath53 and @xmath54 describe the equation of state of the dark energy ( see ( [ simplest0 ] ) ) , @xmath49 is the effective archimedean - type coupling constant . this paper is devoted to the analysis of the solutions of the second and third classes . when @xmath458 is finite or @xmath459 , the dark matter is asymptotically decoupled from the dark energy and its energy density becomes negligible in comparison with the de energy density . we indicated this asymptotic situation as the de domination ( the coupling constant @xmath49 is not equal to zero , but it becomes a _ hidden _ parameter of the model ) . in this case only the appropriate choice of the constitutive parameters @xmath48 , @xmath53 and @xmath54 can protect the universe from the big rip singularity . to summarize the analytical results for the de domination epochs we prepared fig.1 . when @xmath60 , the big rip can be realized on the interval @xmath123 ( or in other words @xmath461 ) . it can be indicated as the `` classical '' big rip domain . when @xmath307 this `` classical '' big rip zone contracts along the line @xmath53 to the interval @xmath308 , and disappears at all when @xmath309 . in this sense , the retardation of the de response to the universe expansion , can avoid the big rip , if the relaxation parameter @xmath48 exceeds the critical value @xmath235 . in other words , when the de relaxation time @xmath310 is bigger than @xmath311 , where @xmath312 is the universe expansion scalar , the regime of the big rip can not be supported . if @xmath48 is negative , the interval @xmath123 again refers to the big rip scenario of the universe expansion . the choice of the parameter @xmath54 can not provide the avoidance of the big rip scenario , but it predetermined the type of the big rip : the power - law type , hyperbolic or trigonometric ones . when @xmath138 , there was no `` classical '' big rip at @xmath60 . the same fact can be indicated at @xmath198 . however , if the parameter @xmath48 is negative , the big rip scenario is possible for @xmath138 , since , instead of retardation , the de displays the acceleration of the response to the universe expansion . when @xmath139 ( or , in other words , @xmath462 ) , there was no `` classical '' big rip at @xmath60 , however , the big rip becomes possible at @xmath139 for positive relaxation time parameter , @xmath198 . the most interesting case studied in the paper concerns the model with asymptotic behavior @xmath113 , for which the archimedean - type coupling leads to the effective heating of the dark matter component of the dark fluid . due to the force proportional to the four - gradient of the de pressure the dm becomes effectively ultrarelativistic and thus plays an active role in the energy redistribution processes inside the dark fluid . the reviving of this second player in the late - time scenario of the universe evolution change essentially the character of expansion : at @xmath63 the big rip scenaria happen to be _ avoided _ and instead of them the little rip scenaria become typical for the universe late - time evolution . this avoidance is typical both for the cases @xmath60 and @xmath62 . we would like to emphasize that the so - called anti - gaussian type solutions for the scale factor @xmath0 appearing as some specific exact solution in @xcite , become typical asymptotic solutions in the case under discussion ( see , e.g. , ( [ ar17 ] ) and ( [ toy5 ] ) ) . these anti - gaussian type solutions correspond to the little rip scenario @xcite , since @xmath463 , @xmath464 , @xmath465 and @xmath466 . in other words , in this model the life - time of the universe is infinite , the hubble function and scale factor tend to infinity without vertical asymptotes . discussing the mechanism of the big rip avoidance in the presence of the archimedean - type interaction , we would like to attract the attention to the following feature . when the coupling constant @xmath49 vanishes , the de pressure behaves as the power - law function of the scale factor . in terms of dimensionless variables @xmath329 and @xmath328 ( see ( [ ar4 ] ) ) this power - law function is predetermined by the first term in ( [ ar6 ] ) . when @xmath359 , and the dm contribution to the total energy - density gains the same order as the de energy - density , the de pressure behaves according to the super - logarithmic law ( [ ar9 ] ) , which is typical for the little rip . super - logarithmic law for @xmath329 means that the de pressure grows more slowly than in the case of @xmath59 . in other words , since the de transfers energy to the dm due to the archimedean - type interaction , the rate of the de pressure growth decreases , thus protecting the universe from the big rip singularity . the model under discussion displays that there are three _ critical _ values of the guiding parameters . the first one is the value @xmath61 , which corresponds to the well - known phantom - crossing value @xmath467 for the de equation of state parameter @xmath468 . the symptom of criticality is that the term @xmath469 systematically appears in the key equations and expressions ( see , e.g. , ( [ key1119 ] ) and ( [ key0167 ] ) ) . the manifestation of the criticality is that the solutions with @xmath470 differ principally from the ones with @xmath471 , so that we considered the last case as the special one . in this sense , one can see some analogy between this model and the critical behavior of the magnetic and electric fields affected by the gravitational wave : we obtained principally different solutions for the case , when the refractive index tends to one , @xmath472 , and for the case , when this parameter is equal to one identically , @xmath473 , since the term @xmath474 appeared in the denominators of the sought - for functions @xcite . the second critical value is @xmath59 . again , the parameter @xmath49 appears in the denominators in the functions @xmath0 and @xmath1 ( see , e.g. , ( [ ar17 ] ) , ( [ ar199 ] ) , as the symptom of criticality , and the solutions with @xmath59 ( we mean the solutions of the big rip type ) differ principally from the solutions with @xmath63 ( the solutions of the little rip type ) . the third critical value is connected with the parameter @xmath48 . the value @xmath60 can be considered as the critical one , since crossing the vertical line @xmath358 on fig.1 , we see the structural rearrangement of the domains corresponding to the big rip type solutions . in addition , according to ( [ rootsqq1 ] ) , this parameter appears in the denominator , indicating that the situation with @xmath313 ( there are two roots of the characteristic equation ) differs essentially from the situation with @xmath475 ( there exists only one root ) . the authors are grateful to prof . s.d . odintsov for valuable comments . this work was partially supported by the federal targeted programme `` scientific and scientific - pedagogical personnel of the innovative russia '' ( grants nos 16.740.11 . 0185 and 14.740.11 . 0407 ) , and by the russian foundation for basic research ( grants nos . 11 - 02 - 01162 and 11 - 05 - 97518 - p - center - a ) .

we present a study of the deuterium abundance along the extended sight line toward 9 with the _ far ultraviolet spectroscopic explorer _ ( _ fuse _ ) . 9 is a o9.5iii star located in the galactic disk at a distance of kpc away from the sun . both in terms of distance and column densities , 9 has the longest and densest sight line observed in the galactic disk for which a deuterium abundance has been measured from ultraviolet absorption lines so far . because many interstellar clouds are probed along this sight line , possible variations in the properties of individual clouds should be averaged out . this would yield a deuterium abundance which is characteristic of the interstellar medium on scales larger than the local bubble . the spectra of 9 show numerous blended interstellar and stellar features . we have measured interstellar column densities of neutral atoms , ions , and molecules by simultaneously fitting the interstellar absorption lines detected in the different channels . as far as possible , saturated lines were excluded from the fits in order to minimize possible systematic errors . _ iue _ ( _ international ultraviolet explorer _ ) archival data are also used to measure neutral hydrogen . we report d / o and d / h ( s ) . our new results confirm that the gas - phase deuterium abundance in the distant interstellar medium is significantly lower than the one measured within the local bubble . we supplement our study with a revision of the oxygen abundance toward , a moderately distant ( pc ) sdob star , located pc below the galactic plane . excluding saturated lines from the fits of the spectra is critical ; this led us to derive an column density about two times larger than the one previously reported for . the corresponding updated d / o ratio on this sight line is d / o ( s ) , which is lower than the one measured within the local bubble . the dataset available now outside the local bubble , which is based primarily on measurements , shows a contrast between the constancy of d / o and the variability of d / h . as oxygen is considered to be a good proxy for hydrogen within the interstellar medium , this discrepancy is puzzling .

astro-ph0508611

deuterium , which is produced during primordial nucleosynthesis and then destroyed by astration , is a key element in cosmology . whereas measurements in low - metallicity qso absorption systems probe d / h at look - back times of @xmath6 gyrs , the present epoch deuterium abundance can be measured in the interstellar medium ( ism ) . the ( _ far ultraviolet spectroscopic explorer _ ) mission has brought significant progress on the measurement . d / h likely has a single value in the local bubble pc - size low - density cavity in which the solar system is embedded ( see , i.e. , sfeir et al . @xcite ) ] , in the range @xmath7 ( moos et al . @xcite ; hbrard & moos @xcite ) . however , it now appears that this local abundance should not be considered as a canonical value characteristic of the milky way at the present epoch , as it was usually believed before studies . indeed , numerous distant sight lines , i.e. sight lines probing the interstellar medium beyond the local bubble , have shown deuterium abundances in disagreement with the local bubble one ( see , e.g , laurent et al . @xcite , jenkins et al . @xcite , lemoine et al . @xcite , sonneborn et al . @xcite , hoopes et al . @xcite ) . hbrard & moos ( @xcite ) reported a trend in the deuterium abundance based on d / o , d / n , and previously reported d / h measurements : the deuterium abundance is lower than in the local bubble for the most distant sight lines exhibiting the highest hydrogen column densities . more recent work by wood et al . ( @xcite ) confirms this low value . these results suggest that the deuterium abundance might be locally abnormally higher than the present epoch value . while hbrard & moos ( @xcite ) suggested that the present epoch ratio would be significantly lower than @xmath8 , linsky et al . ( @xcite ) recently proposed a ratio higher than @xmath9 assuming that deuterium could be significantly depleted onto dust grains ( see jura @xcite ; wood et al . @xcite ; draine @xcite ) . both values may challenge deuterium evolution models , the baryonic density of the universe inferred from primordial d / h measurements , and our understanding of the physics of the interstellar medium . regardless of the actual scenario , it is now clear that local measurements are not enough to assess , and that deuterium studies in the distant interstellar medium are mandatory . only a few are available now . here we report the measurement of the deuterium abundance in the interstellar medium toward 9 ( [ hd90087 ] ) . this sight line is studied thanks to new observations of this target . _ iue _ ( _ international ultraviolet explorer _ ) archival spectra are also used to constrain the neutral hydrogen column density from the fits of the line wings . both in terms of distance and column densities , 9 provides the farthest galactic line of sight for which deuterium abundance has been measured from far ultraviolet absorption lines to date . hence , many interstellar clouds are probed along this sight line . this will tend to average out the peculiarities of individual galactic regions , such as the local bubble , in order to reach measurements that are characteristic of the interstellar medium on large scales . we supplement our study with a reconsideration of the column density measurement toward from spectra , allowing an updated interstellar d / o measurement on this sight line ( [ feige ] ) . 9 is a bright o - type star located in the fourth quadrant of the galactic plane . it was assigned different spectral types and luminosity classes in the past , ranging from o9 to b3 and from ii to v. the most recent classification is the one given by mathys ( @xcite ) : o9.5iii . savage , meade , & sembach ( @xcite ) reported the spectroscopic parallax distance @xmath10 pc , with an accuracy estimated to be about 30 % ( i.e. about 800 pc here ) . this value is in good agreement with the one reported by diplas & savage ( @xcite ) , @xmath11 pc . table [ table_star_parameters ] summarizes relevant sight line and atmospheric parameters for 9 . this target is well beyond the local bubble . the column densities that we report below for 9 represent 50 - 100 times the column densities usually measured within the local bubble ( e.g. , moos et al . the interstellar matter located within the local bubble is a negligible fraction of the total column of material probed along this sight line , which is dominated by the distant interstellar medium . the target was first observed with using the low - resolution ( lwrs ) aperture in april 2000 as a part of a wide study of the hot gas in the milky way through the doublet observations ( bowen et al . several interstellar absorption lines were detected against the stellar continuum on this 3.9 ksec snapshot observation . this indicated that the target is well - suited for distant deuterium abundance measurement . so , an additional 14 ksec observation was performed in june 2003 through the medium - resolution ( mdrs ) aperture , allowing higher signal - to - noise ratio to be reached and reducing interference from geocoronal emission from the terrestrial atmosphere . the analysis presented below uses both the lwrs and mdrs data . the two observations were obtained in histogram mode . the observing log is summarized in table [ table_obslog ] . details of the instrument may be found in moos et al . ( @xcite ) and sahnow et al . ( @xcite ) . note that the lwrs spectra of 9 were first reported in the atlas of galactic ob spectra presented by pellerin et al . ( @xcite ) . the one - dimensional spectra were extracted from the two - dimensional detector images and calibrated using version 2.4.1 of the calfuse pipeline . the data from each channel and segment ( sic1a , sic2b , lif1a , lif2b , etc . ) were co - added separately for each of the two apertures , after wavelength shift corrections of the individual calibrated exposures . we did not sum them because the line spread function ( lsf ) and the distortions in the wavelength scale are different for the lwrs and mdrs apertures , and even between segments for the same aperture . we fit these datasets simultaneously in the analysis reported below . for the lwrs spectra , we used the 5 available exposures . in the case of mdrs , the individual spectra present flux variations due to the motions of the target in the aperture , whose spread is larger that the size of the aperture . we did not include the exposures which have too little flux to be useful , which led us to exclude one or two exposures from the sum of some of the segments . exceptions are the sic2 segments , for which we used only 20 of the 32 available exposures ; the 12 remaining ones present too little flux , or even no flux at all in certain cases ( because the target was out of the aperture during the entire exposure ) . because of their good photometric quality , we use the lwrs spectra to set the absolute flux level of the mdrs spectra . the spectral resolution in the final spectra ranges between @xmath12 and @xmath13 , depending on detector segment and wavelength . the final lwrs spectra are plotted in fig . [ fig_spectre_all ] . a sample of the sic1b mdrs spectrum is shown in fig . [ fig_spectre_lines ] with the , , and lines identified . in that figure , the flux has been multiplied by a factor 1.4 in order to match the lwrs one . we measured the column densities @xmath14 for several species on this line of sight by fitting voigt profiles to interstellar absorption lines . we used the profile fitting method presented in detail by hbrard et al . ( @xcite ) , which is based on the procedure , developed by martin lemoine and the french team ( lemoine et al . we split each spectrum into a series of small sub - spectra centered on absorption lines , and fitted them simultaneously with voigt profiles using minimization ( see fig . [ fig_fit ] ) . each fit includes typically one hundred or more spectral windows , and around 300 transitions of neutral atoms , ion , and molecules . the different species are assumed to be in six different clouds : the first one includes only , the second one , , , , and , the third one and its two excited levels and , the fourth one all the rotational levels of h@xmath15 , the fifth one hd , and the last one all the rotational levels of co. note that thanks to the redundancy of spectral coverage , a given transition might be observed in one , two , three , or four different segments ( see fig . [ fig_spectre_all ] ) . in addition , as lwrs and mdrs data are available for a given spectral feature , we can have up to eight independent spectra . these different observations allow for some instrumental artifacts to be identified and possibly averaged out . the laboratory wavelengths and oscillator strengths ( @xmath16-values ) used in the analysis are from abgrall et al . ( @xcite , @xcite ) and e. roueff ( private communication ) for the molecules , and from morton ( @xcite ) for atoms and ions . morton ( @xcite ) was highly used in the past in the literature for similar studies . following the publication of that atomic data compilation , some important @xmath16-value revisions occurred ( beideck et al . @xcite ; verner el al . @xcite ; tripp et al . @xcite ; savage & sembach @xcite ; howk et al . @xcite ; sofia et al . @xcite ) , but for the most part , differences between morton ( 1991 ) and the updated atomic tables from morton ( @xcite ) are small . we point out , however , the case of transitions , for which some oscillator strength values have been significantly updated in morton ( @xcite ) . such differences ( up to a few tens of percent ) would have significant effect on the column density measurements . our analysis takes into account the updated @xmath16-values compiled by morton ( @xcite ) , including those of . several parameters are free to vary during the fitting procedure , including the column densities of all the species , the radial velocities of the six interstellar clouds assumed in the fit , their temperatures and turbulent velocities , and the shapes of the stellar continua , which are modeled by low order polynomials . produces solutions that are coherent between all the fitted lines , assuming a given cloud has a single radial velocity , temperature and turbulence . some instrumental parameters are also free to vary , including the flux background , the spectral shifts between the different spectral windows , and the widths of the gaussian used to convolve with the voigt profiles . the velocity shifts fitted for the spectral windows are plotted in fig . [ fig_shift ] . they allow the inaccuracies of the wavelength calibration to be corrected . typically , the velocity corrections are consistent within a few within a given segment . they are not a smooth function of wavelength , and shifts as high as 10 to 20 might occur over only a few ngstrms in extreme cases ( fig . [ fig_shift_ex ] ) . larger velocity corrections might occur from one segment to another . the velocity shifts are relative and we did not attempt to measure absolute ( e.g. heliocentric or lsr ) velocities . we note that the wavelength distortion problem is generally more severe in data obtained in histogram mode . since most observations employ the time - tag mode , distortion of the wavelength scale is usually not as bad as the examples shown in fig . [ fig_shift_ex ] ( which were recorded in histogram mode ) . the widths of the gaussian fitted for the spectral windows are shown in fig . [ fig_fwhm ] . the averages for the apertures lwrs and mdrs are respectively @xmath17 and @xmath18 pixels ( full widths at half maximum ) . we also attempted to fit all spectral windows with a unique lsf width . the best was obtained for a 11.4-pixel width . however , significant variations occur from one segment to the other , and as a function of the wavelength for a given segment . these averaged widths are in agreement with those reported in hbrard et al . ( @xcite ) . note that during the fits , all the spectra were binned to three pixel samples in order to reduce computing time . considering the width of the lsf , this binning does not degrade the spectral resolution . when possible , the saturated lines located on the flat part of the curve of growth were excluded from the fit . indeed , such saturated lines can introduce systematic effects on column density measurement ( hbrard et al . @xcite ; hbrard & moos @xcite ; friedman et al . @xcite ) , mainly because the actual widths of the lines , the velocity structure of the sight line , or the of the instrument are unknown but expected to be complex . a particular difficulty with distant lines of sight such as 9 is the rarity of unsaturated lines for some species , especially and , which is caused by high column densities . this is significantly different from local sight lines , for which numerous unsaturated lines are available . another relevant difference between the long line of sight toward 9 and those of targets located in the local bubble is the number of detected absorption lines . the overview of the 9spectra plotted in fig . [ fig_spectre_all ] can be compared with the fig . 1 in hbrard et al . ( @xcite ) , which shows the corresponding spectra for a target located 53 pc away . the 9 spectra are particularly crowded . this is due to stellar lines , which are more numerous here than in the case of the targets used to probe the more local interstellar medium ( mainly white dwarfs ) . in the case of 9 , however , the high rotational velocity of this star smooths the stellar continuum . an even bigger effect on the absorption line number arises from interstellar lines , as the column densities here are typically tens or hundreds of times larger than those measured in the local bubble . in particular , many h@xmath15 lines are detected in the 9 spectra . this makes line blending a critical problem , which can be resolved with simultaneous fits of all the lines together . only one interstellar component , at a given radial velocity , was assumed for the sight line for a given species . this assumption is unlikely to be true but it will have no effect on the column densities measured from unsaturated lines ( see tests of this assumption in hbrard et al . thus , we report total column densities , integrated along each line of sight . the simultaneous fit of numerous lines allows the reduction of statistical and systematic errors , especially those errors due to continua placement , lsf uncertainties , line blending , flux and wavelength calibrations , and atomic data uncertainties . the error bars were obtained using the method presented by hbrard et al . ( @xcite ) . we also estimated the error bars by varying the assumptions for the fits . this included the lsf ( normally free to vary , but tests were made with fixed lsfs ) , the velocity structure ( several species are assumed to be in the same component , but fits with one component per species were performed ) and also the aperture ( comparing ( a ) mdrs and lwrs together , ( b ) only lwrs , and ( c ) only mdrs ) . we discuss below the different species . the measured column densities and error bars are reported in table [ table_columns ] and examples of fits are given in fig . [ fig_fit ] . the curves for , , and are plotted in fig . [ fig_chi2_don ] . all the rotational levels of h@xmath15 were included in a single interstellar component . the fits used a total of about @xmath19 h@xmath15 absorption lines from rotational levels @xmath20 to @xmath21 available in one to eight independent segment / aperture configurations . the @xmath22 h@xmath15 lines present damping wings and the @xmath23 that we used are unsaturated . this allows accurate column density measurements to be performed for these five levels . only saturated lines located on the flat part of the curve of growth are available for the @xmath24 h@xmath15 transitions ; this precludes accurate measurements of their column densities . the excitation diagram corresponding to the derived h@xmath15 column densities is shown in fig . [ fig_excitation ] . h@xmath15 is not thermalized for levels @xmath25 . the column densities of the levels @xmath20 and @xmath26 give the temperature @xmath27 k , which can be considered as the kinetic temperature of the h@xmath15 molecular gas . a temperature of 78 k would correspond to @xmath28 . thus , the single - component doppler parameter @xmath29 that we found for this molecular component is not purely thermal but contains additional contributions from the gas turbulent motions and/or additional velocity components . note that since the @xmath22 levels are damped and the @xmath30 to @xmath31 levels are optically thin , the doppler parameter we derived is mainly characteristic of the @xmath24 levels . on another hand , since we did not detect any significant velocity shift between the different rotational levels ( all are @xmath32 ) , this suggests that these levels arise from the same dominant cloud component(s ) on average . the only hd lines detected in the spectra are those corresponding to the level @xmath20 ; no hd lines are detected for @xmath33 . our global fits include the seven hd lines used by lacour et al . ( @xcite ) , and also four extra ones at 1007.28 (see fig . [ fig_fit ] , left - bottom panel ) , 925.78 , 967.53 , and 1042.84 . the other hd ( @xmath20 ) lines are blended with strong lines ( as @xmath341001.89 and @xmath341078.81 , which are lost in saturated h@xmath15 lines ) and are not detected . among those eleven hd lines , the five strongest appear to be saturated or nearly so . we then performed extra fits including only the six weakest hd lines , namely @xmath341054.29 , @xmath34959.82 , @xmath34967.53 , @xmath34925.78 , @xmath341066.27 , and @xmath341105.83 ( in decreasing order of @xmath16-values ) . excluding those five saturated lines increased the hd column density estimate by 0.1 dex , from 14.42 to 14.52 . we adopt the latter value . interestingly , the radial velocity of the hd lines was found to be systematically redshifted by @xmath35 with respect to the velocity of the h@xmath15 lines . according to the error bars , the velocity shift is significant ; it is mainly constrained by the 13 spectral windows that include close hd and h@xmath15 lines . by comparison , no velocity shifts were detected between the different rotation levels of h@xmath15 ( [ h2 ] ) , between the different atomic or ionic species ( [ ni ] ) , nor between and ( [ hi ] ) . thus , the shift between h@xmath15 and hd appears to be an actual effect , and not a systematic effect due to an instrumental artifact . in addition , we found a doppler parameter for the hd lines of @xmath36 , about half the doppler parameter estimated for h@xmath15 ( mainly constrained by @xmath24 ) . these two results therefore suggest that hd and h@xmath15 rotational lines do not trace exactly the same gas conditions . consequently , the hd and h@xmath15 lines were not assumed to be located in the same component during the fits . this might jeopardize the interpretation of the hd/2h@xmath15 ratio computed from the measured column densities of the two species . note that a similar radial velocity shift ( 4.6 ) was also detected between h@xmath15 and hd on the sight line to pg0038 + 199 ( williger et al . as numerous different h@xmath15 and hd transitions are used in these two analyses , this suggests that hd tabulated wavelengths are unlikely to be the cause of the velocity shifts . a broad study of h@xmath15 and hd radial velocities toward numerous targets is however mandatory to progress on this issue . numerous unsaturated lines are detected in the spectra of 9 . the fits include eight of them , namely @xmath34955.26 , @xmath34959.49 , @xmath34951.29 , @xmath34955.53 , @xmath34960.20 , @xmath341159.82 , @xmath341160.94 , and @xmath34955.44 ( in decreasing order of @xmath16-values ) . some of these lines are detected in several segments and/or through the two available apertures . thus , the final fits include 18 independent unsaturated lines . these lines were fitted simultaneously with all the other ones , assuming that they are in the same cloud as , , , and . indeed , these species are supposed to co - exist in neutral regions ( and ionization potentials are respectively 7.9 ev and 10.5 ev , i.e. below the hydrogen ionization potential ) . we checked this hypothesis by performing fits with these five species in five different components , and we did not detect any significant radial velocity shifts between them . the final error bar reported in table [ table_columns ] for @xmath14 ( ) is small due to the numerous unsaturated lines available . many lines are available in the bandpass . however , in the range @xmath37@xmath38 , most of those lines are saturated , which precludes accurate column density measurement ( see [ feige ] and friedman et al . @xcite ) . in this column density range , only one unsaturated line is available in the bandpass , namely @xmath39 . this weak line is blended with three h@xmath15 lines , including two lines ( levels @xmath40 and @xmath41 ) with equivalent widths larger than the equivalent width of @xmath42 . the third one ( level @xmath21 ) is negligible . the h@xmath15 contributions to this blend can be calculated thanks to the @xmath43 h@xmath15 lines included in the fits , many of which are unblended . thus , the shapes of the three @xmath42 h@xmath15 lines are well - constrained . @xmath42 is detected in the segments sic1b and sic2a , through the two available apertures . our fits include these four different observations of this line , which reduces potential instrumental effects . the fits of @xmath42 observed through mdrs on sic1b and sic2a are plotted in fig . [ fig_fit ] . the h@xmath15 ( @xmath41 ) line seems slightly under - fitted , which might be due to a weak unknown feature . however , this extra absorption appears to be weak enough and far enough from @xmath42 to have no significant effect on @xmath44@xmath45 . measuring a column density from only one transition makes the result highly sensitive to possible errors caused by an erroneous @xmath16-value or uncontrolled blends from unknown lines . however , up to now , no clues suggest such problems . no strong inconsistencies were found between @xmath42 and the stronger lines available in the bandpass in the case of bd@xmath46 + 28@xmath474211 ( hbrard & moos @xcite ) . slight disagreements might be difficult to see , however , as @xmath42 is weak on this sight line . we also measured in the sight line of hd195965 from data , using @xmath42 with the same method as those we use for 9 ( see fig . [ fig_hd195 ] ) . we obtained @xmath48@xmath49 , in agreement with the value obtained by hoopes et al . ( @xcite ) using @xmath50 , @xmath48@xmath51 . thus , there are apparently no inconsistencies between @xmath42 and @xmath50(the studies of oxygen abundance by meyer et al . ( @xcite ) , andr et al . ( @xcite ) , or cartledge et al . ( @xcite ) are based on @xmath50 ) . the two comparisons we present above suggest that @xmath42presents no strong oscillator strength inconsistencies nor significant uncontrolled blends ( see also [ hd900987_others ] ) . as in the case of discussed in the previous subsection , only a few unsaturated lines are available in this high column density regime . strong blending ( especially for h@xmath15 lines ) and saturation prohibit the use of most of the lines . the lines of @xmath52 , @xmath53 , @xmath54 , and at larger wavelengths in the lyman series are all saturated , so we did not include them in the fits . @xmath55 is not detected because it is located at @xmath56 from a strong h@xmath15 ( @xmath26 ) line with damping wings , so it can not be used to constrain @xmath44@xmath45 . we could not obtain a good fit to @xmath57 . this line is blended with two strong lines ( and h@xmath15 ( @xmath58 ) ) but apparently also with some unknown feature(s ) . our fit therefore includes only two transitions , namely @xmath59 and @xmath60 . the first one is available only on segment sic1b , while the second one is available on both sic1b and sic2a . as we use simultaneously the data from the mdrs and lwrs apertures , our fits include a total of six lines . a weak h@xmath15 ( @xmath41 ) line is located @xmath61 from @xmath59 . it is weak and far enough to have no significant effects . @xmath59 is probably the highest detectable in the lyman series . indeed , in the high column density regime necessary to produce significant absorption for lyman lines this high , interstellar absorption lines are too numerous and crowded below 916 , and therefore almost no flux is detected there ( see figs . [ fig_spectre_all ] and [ fig_spectre_lines ] ) . the @xmath60 line is blended with h@xmath15 ( @xmath30 ) , located @xmath62 blueward and of similar equivalent width . the fits include nine other h@xmath15 ( @xmath30 ) lines , observed on different segments and through the two apertures , allowing the blend to be deconvolved . @xmath60 is slightly saturated . we tested the results by including only @xmath59 , and the ensuing @xmath44@xmath45 was in agreement with the fit to @xmath63 . the choice of the continuum used for fitting the absorption lines has an effect on the measured column density . here the continua are fitted by polynomials for each spectral window . as the parameters of the polynomials are free to vary , the uncertainty due to the continuum should be included in the final error bar . we checked that by varying the continua , i.e. by using different polynomial degrees or changing the wavelength intervals on which they are fitted . our result is @xmath37@xmath64 ( s ) . to date , this is the largest published column density ever measured . as @xmath59 would saturate for @xmath44@xmath45 values 2 or 3 times larger , it would be difficult to accurately measure larger column densities for other sight lines . note , that this limit also provides a selection bias against sight lines with higher d / h or d / o ratios in this high column density regime . interesting lower limits on @xmath44@xmath45 might however be estimated from saturated lines . the column density of hd represents only @xmath65 % ( s ) of @xmath44@xmath45 . most of the deuterium detected on the line of sight of 9 is in atomic form . therefore , given the size of the errors , deuterium in molecular form can be neglected for the computation of the total deuterium column density . interstellar absorption lines of co , , , , and its two excited levels and are also detected in the spectra of 9 . we report accurate column density measurements for co , , and . the measurements for the remaining species are more uncertain . we discuss here briefly all these species . numerous unsaturated lines are available in the spectra of 9 . we used the following ones : @xmath66 , @xmath67 , @xmath68 , @xmath69 , and @xmath70 ( in decreasing order of @xmath16-values ) . we also made fits including these lines and @xmath71 and @xmath72 , whose @xmath16-values are larger and which might be slightly saturated . no significant difference was found . for , we used only @xmath73 , since @xmath74 is saturated . as , the lines are at the same radial velocity as , , and . lebouteiller et al . ( @xcite ) show that might be considered as a good proxy for . the / ratio we measure toward 9 is in good agreement with the average ratio they report . this is another argument to support the reliability of our measured @xmath44@xmath45 , based on @xmath42 ( see [ hd900987_oi ] ) . the co a - x bands at @xmath75 and @xmath76 were also detected toward 9 . since the band structure can not be resolved with the data , we included @xmath77 levels up to @xmath78 in our initial fits . our best fit indicated that the column densities of the @xmath20 and @xmath79 rotational levels are similar while the contribution from the @xmath25 levels is negligible ( column densities ten times lower or less ) . the co lines are detected at the same radial velocity as h@xmath15 . in both regions , there is one @xmath20 co transition , and two @xmath26 co transitions on each side of the first one . the lines are broad enough to allow the two @xmath77-levels to be measured , despite their not being resolved . the excitation diagram leads to a temperature @xmath80 of about 4 k , indicating that the rotational levels are sub - thermally populated . the derived co / h@xmath15 ratio is @xmath81 ( see table [ table_ratios ] ) , a value about 3 orders of magnitude smaller than that typically found in purely molecular clouds . the latter results are consistent with rotational population distributions and co abundances typically found in diffuse molecular clouds where photo - processes still largely dominate the gas chemical pathways ( sonnentrucker et al . @xcite ) . for carbon , the strongest line , at 945 , is near to saturation . numerous weaker lines are detected in the range @xmath82 , together with and transitions . however , it is not possible to obtain a fit in agreement for the different transitions . this is probably due to inaccurate @xmath16-values . using a large sample of interstellar lines recorded at high resolution , jenkins & tripp ( @xcite ) have shown that many @xmath16-values require significant revisions , especially @xmath16-values of weak transitions . unfortunately , they did not analyze transitions in the wavelength range . our reported carbon column density measurements therefore remain inaccurate . finally , @xmath14 ( ) is obtained only from saturated lines ( @xmath83 and @xmath84 ) on the flat part of the curve of growth ; it is thus uncertain . to determine the total column density in the interstellar medium toward 9 , we fitted the damped profile recorded in an _ iue _ spectrum of the star . 9 was observed with _ iue _ once in 1982 ( table [ table_obslog ] ) . we retrieved the iuesips ( _ iue spectral image processing system _ ) version of the data and extracted the spectrum using standard procedures . fig . [ fig_hi_iue ] shows the portion of the spectrum in the vicinity of the line . the strongly damped profile is readily apparent along with the nearby pcygni profile ( which causes the emission feature at @xmath85 and the adjacent trough at @xmath86 ) and a variety of stellar and interstellar absorption lines . we measured @xmath44@xmath45 using the procedure described in jenkins et al . ( @xcite ) and sonneborn et al . ( @xcite ) . briefly , we used minimized in order to determine the column that provided the best fit to the lorentzian wings of the profile . the following five parameters were freely varied using powell s method ( press et al . ( @xcite ) : ( 1 ) @xmath44@xmath45 , ( 2)-(4 ) three coefficients of a low - order polynomial fitted to the continuum over a broad range on each side of , and ( 5 ) a correction for the flux zero level . 9 is an early - type star , and any stellar line is unlikely to significantly affect the interstellar profile , especially since the interstellar column toward this star turns out to be large and the profile is accordingly quite broad . to evaluate upper and lower confidence limits on @xmath44@xmath45 based on changes in , we then increased ( or decreased ) the column density while allowing the other parameters to vary . by allowing the polynomial coefficients to vary throughout the entire process , the fitting procedure can adjust the height and curvature of the continuum and thereby provide some assessment of the continuum placement uncertainty . with _ iue _ data , background subtraction near is difficult due to order crowding , and this introduces significant uncertainty in the flux zero point . for this reason , it is important to include the flux zero point as a free parameter in the fitting process . our final best fit and the corresponding continuum placement are shown in fig . [ fig_hi_iue ] along with the profiles corresponding to s upper and lower limits on @xmath44@xmath45 . from these fits we obtain @xmath37@xmath87 ( suncertainties ) . this result is in good agreement with the values obtained from the fits of , , and observed with . we obtained @xmath48@xmath88 from the fits of these three lines ( fig . [ fig_hi_fuse ] ) , using all the available data . however , if presents damping wings , the stellar continuum at this wavelength is a critical issue . it is probably reached by strong features due to the stellar wind ( see , e.g , bianchi & garcia @xcite ) . no robust models of the stellar continuum are available up to now . the polynomials that we used give an estimation of the stellar continua shape , but they are probably too naive . thus , we adopt the measurement for our @xmath44@xmath45 result . on this line of sight , a minority of the material is in molecular form . the hydrogen molecular fraction @xmath89h@xmath90h@xmath91h@xmath92@xmath93 = 10.1\pm2.6 $ ] % . thus , molecular species can be neglected when the abundances are computed from column densities ratios . the implied errors on abundance would be of the order of 10 % , probably even less , as the molecular species are unlikely to significantly coexist in the same clouds with the atomic and ionic species . by comparison with the final reported error bars , the effect of the molecules on the abundance is negligible . for example , we obtain /@xmath94 and ( + hd)/(+2h@xmath95 . on the contrary to h@xmath15 and hd ( [ hd ] ) , we did not detect any significant velocity shifts between the and lines ( @xmath32 ) on the spectra , which is reassuring as the ratio of the column densities of these two species are directly used to compute the d / h ratio . some targets have been reported with significant shifts between the and lines . wood et al . ( @xcite ) , for example , reported a shift of 5 between the radial velocities of and on the sight line of jl9 . the abundances on the line of sight to 9 are reported in table [ table_ratios ] . the d / h , d / o , and d / n ratios are all significantly lower than the ratios measured within the local bubble . this is in agreement with the picture reported by hbrard & moos ( @xcite ) , which suggests that the deuterium abundance is significantly lower in the distant interstellar medium than locally . the n / h ratio is in good agreement with the typical interstellar n / h ratio reported by meyer et al . ( @xcite ) . the o / h ratio is rather large when compared with the typical o / h ratios in the distant interstellar medium ( meyer et al . ( @xcite ; andr et al . @xcite ; cartledge et al . @xcite ) , although compatible within the error bars . other targets have been found to show similar o / h ratios : hd91824 , hd122879 , hd12323 ( cartledge et al . @xcite ) , hd195965 ( hoopes et al . @xcite ) , and jl9 ( wood et al . it can be argued that the low d / o observed toward 9 is due to the high o / h ratio rather than to a true low deuterium abundance . however , the d / h and d / n ratios are low as well . in addition , if we assume the column density is overestimated by a factor @xmath96 ( in order to agree with the above o / h studies ) , this would still imply a low d / o , below @xmath97 . the deuterium abundance seems thus to be actually low on this line of sight . it is interesting to note that 9 shows simultaneously a low deuterium abundance together with a rather high oxygen abundance . hd195965 ( hoopes et al . @xcite ) and jl9 ( wood et al . @xcite ) also show a similar effect . this might be a hint of an astration signature . the hd/2h@xmath15 ratio is about 5 times lower than the d / h ratio . this is characteristic of interstellar clouds with moderate molecular fraction ( we obtain @xmath89h@xmath98 % ) . indeed , compared to h@xmath15 , self - shielding of hd becomes significant deeper in the molecular clouds since deuterium is much less abundant than hydrogen ( ferlet et al . @xcite ; lacour et al . thus , the transition between and hd takes place deeper in the molecular clouds than the transition between and h@xmath15 . our hd/2h@xmath15 ratio is similar to those reported by lacour et al . ( @xcite ) . an extensive study of the sight line was performed by friedman et al . ( @xcite ) . all the information about this target and its observations are in that initial paper . briefly , is a sdob star . most of the interstellar material probed along its sight line is probably located outside the local bubble . the parallax and photometric distances , @xmath99 pc and @xmath100 pc respectively , are in good agreement . its high galactic latitude ( -5907 ) locates at @xmath101 pc from the galactic plane . the spectra of were obtained in 2000 ( table [ table_obslog ] ) . friedman et al . ( @xcite ) measured @xmath44@xmath45 and @xmath44@xmath45 from these spectra , and used _ iue _ archival data to derive @xmath44@xmath45 using the same technique we applied to 9 ( [ hi ] ) . subsequently , hbrard & moos ( @xcite ) used the spectra of to measure @xmath44@xmath45 on this sight line . the column density measurement reported by friedman et al . ( @xcite ) was derived using only a curve - of - growth analysis . however , the weak line at @xmath39(see [ hd900987_oi ] ) was not included in this early analysis , and all the lines fitted in the curve - of - growth are likely to be significantly saturated . as saturated lines might lead to erroneous column density evaluations ( friedman et al . @xcite ) , we revisit here the @xmath44@xmath45 measurement toward . we used exactly the same dataset previously employed by friedman et al . ( @xcite ) and hbrard & moos ( @xcite ) . as for 9 ( [ overview ] ) , we measured the column densities on the sight line toward using the profile fitting method presented by hbrard et al . ( @xcite ) , which is based on the procedure ( lemoine et al . , , , and h@xmath15 lines were included in the fits . we focus here on . we note that we obtain results in agreement with friedman et al . ( @xcite ) and hbrard & moos ( @xcite ) for @xmath44@xmath45 and @xmath44@xmath45 , respectively . as for 9 , we found that @xmath39 is the only unsaturated line available in the bandpass . however , this line is less blended in the case of : @xmath39 is blended with only one h@xmath15 ( @xmath40 ) absorption ( see fig . [ fig_oi_feige ] ) . the shape of this saturated h@xmath15 line is constrained thanks to the other h@xmath15 transitions available in the spectra . the two other h@xmath15 transitions located around 974.07 ( corresponding to the level @xmath41 and @xmath21 ) are negligible for this sight line . our fits include the two different available observations of @xmath42(segments sic1b and sic2a ) . we checked the results of our fits by performing a curve - of - growth analysis , using the same procedure as friedman et al . ( @xcite ) . we normalized the spectrum by a fitted continuum and a synthetic h@xmath15 model constrained by the other h@xmath15 transitions . then , as we did for the other , stronger lines , we measured the equivalent width of @xmath42 . we obtained @xmath102m and @xmath103m for the segments sic1b and sic2a , respectively . the resulting curve of growth is plotted in fig . [ fig_oi_cog_feige ] . the column densities measured using these two methods ( profile fitting and curve of growth ) are different by only 0.03dex . this difference is negligible with respect to the uncertainties ( @xmath104dex , s ) , so there are apparently no systematic effects due to the method . our final value , which reflects the combined effort of these two independent analyses , is @xmath37@xmath105 . this is a factor @xmath106 larger than the one reported by friedman et al . ( @xcite ) , namely @xmath37@xmath107 ( s ) . [ fig_oi_feige ] shows the fit of @xmath42 using these two @xmath44@xmath45 values , which clearly suggests a higher value than the one reported by friedman et al . this illustrates that measuring column densities from saturated lines might lead to erroneous results and underestimation of error bars . our past experience with different targets together with simulations of data ( moos et al . @xcite ; hbrard et al . @xcite ; friedman et al . @xcite ) has shown that measurements from saturated lines seem to preferentially underestimate @xmath44@xmath45 . the case of is in agreement with this tendency . we note that the value of @xmath108 for @xmath42 is at least 10 times smaller than for the other lines in the bandpass . such large differences might explain significant systematic effects due to saturation in the range @xmath37@xmath38 when the @xmath42transition is not used in the analysis . in addition , one can note that the equivalent width of the @xmath109 line seems to indicate a lower column density ( see fig . [ fig_oi_cog_feige ] ) . the reason is unclear , but it can be due to structure along the sight line , some components being saturated before the other ones . a slight inaccuracy in @xmath16-value for the @xmath109 transition and/or blending issue implying continuum uncertainties in that spectral area might also be present . such problems with @xmath109 might increase the error for @xmath44@xmath45 even more , if @xmath42 is not used . targets in that column density range ( @xmath48@xmath38 ) and with poor signal to noise should be considered with caution . indeed , if the signal to noise is not high enough to allow @xmath42 to be detected , @xmath44@xmath45 measurements might be significantly underestimated ( see the dashed line in fig . [ fig_oi_cog_feige ] ) . since it is mainly based on one transition , our final result on @xmath14 ( ) toward is highly sensitive to possible errors caused by uncontrolled blends from unknown lines ( numerous stellar lines due to metals are detected in the spectrum of ) or an erroneous @xmath16-value . up to now , studies of other targets have not suggested such problems ( see [ hd900987_oi ] and friedman et al . @xcite ) . using our new column density measurement together with the @xmath44@xmath45 reported by friedman et al . ( @xcite ) , we obtain d / o @xmath110 ( s ) . this ratio is lower than d / o measured in the local bubble , @xmath111 ( s , hbrard & moos @xcite ) . the lower d / o value is in agreement with the fact that the deuterium abundance might be lower for the most distant lines of sight and the highest column densities , as reported by hbrard & moos ( @xcite ) . friedman et al . ( @xcite ) , however , reported d / h @xmath112 ( s ) on this sight line . this ratio is _ larger _ than the local bubble value , which is in the range @xmath7 ( see [ introduction ] ) . this ratio is also about three times larger than the d / h typical from the distant interstellar medium reported by hbrard & moos ( @xcite ) . the apparent contradiction between d / h and d / o is surprising . thus , this sight line appears atypical . according to our new evaluation of @xmath44@xmath45 and the @xmath44@xmath45 from friedman et al . ( @xcite ) , the oxygen abundance toward is o / h @xmath113 ( s ) . this ratio is about two times larger than the typical o / h ratios reported by meyer et al . ( @xcite ) , andr et al . ( @xcite ) , or cartledge et al . ( @xcite ) . similarly , the @xmath44@xmath45 measurement from hbrard & moos ( @xcite ) and the @xmath44@xmath45 measurement from friedman et al . ( @xcite ) imply n / h @xmath114 ( s ) for . this ratio is about three times larger than the typical n / h ratios reported by meyer et al . ( @xcite ) . the fact that d / h , o / h , n / h are all approximately two - three times larger than the values usually measured in the distant interstellar medium is suspicious . it suggests that the column density might be affected by some systematic effects , and the actual @xmath44@xmath45 being two - three times larger . however , the iue spectrum of studied by friedman et al . ( @xcite ) appears not consistent with such a high column density . new observations would be very useful for that target but due to the recent failure of the spectrograph stis onboard hst , such observations wo nt be possible at short term . whatever the solution is , appears as an abnormal line of sight , with abnormally high d / h , o / h , and n / h ratios . finally , the ti / h ratio appears also particularly high toward ( prochaska et al . this point is important , as is the target that mainly drives the correlation between and recently reported by prochaska et al . ( @xcite ) . such a correlation might be interpreted as a support for a significant depletion of deuterium onto dust grains ( linsky et al . @xcite ) , as ti is a refractory element . however , the systematic effect that we suspect on @xmath44@xmath45 might also be the cause for the high ti / h ratio measured toward . in this section we compare our new d / h and d / o measurements with those obtained toward other lines of sight . d / o is considered to be a good proxy for d / h and , in addition , it may be less sensitive to systematic effects ( timmes et al . @xcite ; hbrard & moos @xcite ) . [ fig_dso_dsh ] ( left panel ) shows d / o as a function of the column density . this plot includes 27 sight lines : the 24 included in hbrard & moos ( @xcite ) , but with the revised @xmath14 ( ) for ( [ feige ] ) , plus jl9 ( wood et al . @xcite ) , pg0038 + 199 ( williger et al . @xcite ) , and hd90087 ( [ hd90087 ] ) . d / o measurements apparently show a bimodal picture : there is a high , homogeneous d / o ratio for low column densities , and a low , homogeneous d / o ratio for high column densities . the weighted mean for the 14 lines of sight within the local bubble is @xmath115 ( hbrard & moos @xcite ) . the weighted mean for the 7 lines of sight with the highest @xmath14 ( ) is about two times lower : d / o@xmath116 , with a of 5.9 for 6 degrees of freedom ( reduced of 1.0 ) . we do not see any variations in the distant d / o ratio . these 7 sight lines are feige110 ( friedman et al . @xcite and [ feige ] ) , hd195965 and hd191877 ( hoopes et al . @xcite ) , lss1274 ( hbrard & moos @xcite ; wood et al . @xcite ) , jl9 ( wood et al . @xcite ) , pg0038 + 199 ( williger et al . @xcite ) , and hd90087 ( [ hd90087 ] ) ; they all present @xmath37@xmath117 . the transition between the two d / o ratios , @xmath118 and @xmath119 , is located around @xmath37@xmath120 . this simple , bimodal picture of d / o is rather different from what d / h shows . indeed , a larger dispersion is seen through d / h measurements . on fig . [ fig_dso_dsh ] ( right panel ) are plotted d / h ratios as a function of @xmath37@xmath45 for 43 lines of sight . this includes the 39 targets reported in the table 4 from wood et al . ( @xcite ) , to which we added sirius ( hbrard et al . @xcite ) , lan23 ( oliveira et al . @xcite ) , pg0038 + 199 ( williger et al . @xcite ) , and hd90087 ( [ hd90087 ] ) . note that we adopt d / h @xmath121 for lan23 , according the and column densities reported by oliveira et al . ( @xcite ) . the @xmath44@xmath45 toward this target is derived from low euve flux , which makes the measurement uncertain , especially for such high column density . the @xmath122-axes of the two plots of fig . [ fig_dso_dsh ] are identical , and their @xmath123-axes are scaled to the same size ( assuming o / h@xmath124 ; meyer @xcite ) ; this allows the two panels to be easily compared . within the local bubble , at low column densities , both panels show a homogeneous deuterium abundance ( note the slight but significant disagreement between the two averaged values ; see hbrard & moos @xcite ) . however , the two panels present different pictures at larger column densities : whereas d / o shows a decrease toward a homogeneous value , d / h shows dispersions . considering as above the same 7 lines of sight with @xmath48@xmath117 , the for the weighted mean of this d / h distribution is 34.4 , whereas it is 5.9 in the case of d / o ( for 6 degrees of freedom in both cases ) . d / h seems more dispersed than d / o in the distant interstellar medium . since oxygen is considered as a good proxy for hydrogen , these two different pictures are unexpected . this discrepancy might call into question the homogeneity of the o / h ratio within the interstellar medium , but numerous studies show that is a good tracer of in the nearby galactic disk ( meyer et al . @xcite ; andr et al . @xcite ; cartledge et al . @xcite ; oliveira et al . andr et al . ( @xcite ) and cartledge et al . ( @xcite ) suggested however o / h can vary somewhat as a function of distance and/or average density . on another hand , it can be argued that some column densities are under - evaluated , as it may be possible for ( [ feige ] ) , but no systematics allowing such under - evaluation have been yet clearly identified . fig . 13 in williger et al . ( @xcite ) shows explicitly how poor the pg0038 + 199 line is fitted if @xmath44@xmath45 is forced to be high enough to recover the expected d / h and o / h values . column densities might also be suspected to be erroneous for distant lines of sight , as it is mainly measured from only one transition , namely @xmath39 . this transition appears however to be reliable ( [ hd900987_oi ] ) , and there are several targets that show low d / o ratios whereas their @xmath44@xmath45 measurements are based on @xmath50 ( @xmath125ori@xmath46a , @xmath126cas , @xmath127ori , hd195965 ; see hbrard & moos @xcite ) . finally , if deuterium is significantly depleted onto dust grains , as proposed by wood et al . ( @xcite ) and linsky et al . ( @xcite ) , one would have to argue that oxygen is depleted at the same rate , in order explain the observed constancy of d / o and variability of d / h . such a tight correlation between deuterium and oxygen depletion seems unlikely , however . indeed , oxygen tends to be associated with silicates , whereas draine ( @xcite ) proposed that pahs are causing the deuterium depletion . up to now , the discrepancy between the constancy of d / o and the variability of d / h remains unexplained . several processes might be implied , including the above ones , but the complete explanation is still to be found . it should also be recalled that distant targets with deuterium measurements remain sparse . actually , only three lines of sight exhibit significantly high d / h ratios , namely pg0038 + 199 , feige110 , and @xmath128vel . these targets are important as they are those that are assumed to be characteristic of the actual value if deuterium is significantly depleted onto dust grains . for the two first ones , both d / h and d / o measurements are available . for pg0038 + 199 , d / h@xmath129 and d / o@xmath130 ( williger et al . @xcite ) , and for feige110 , d / h@xmath131 ( friedman et al . @xcite ) and d / o@xmath132 ( [ feige ] ) . thus , both present `` low '' d / o ratios , but `` high '' d / h . no direct @xmath37@xmath45 measurements are available for the last target , @xmath128vel , for which sonneborn et al . ( @xcite ) reported d / h@xmath133 . the only available estimation is reported by knauth et al . ( @xcite ) : @xmath44@xmath134 . this is based on and observations by fitzpatrick & spitzer ( @xcite ) . it leads to a low d / o ratio ( @xmath135 ) , but the systematic uncertainties in this @xmath44@xmath45 indirect estimation are too large to allow any firm conclusion to be drawn . indeed , using updated solar abundances and a lower depletion factor might produce d / o up to @xmath136 times larger . more d / h and d / o measurements toward distant lines of sight are mandatory to compare the enigmatic behavior of these ratios . we have presented a new interstellar deuterium abundance study of the sight line toward 9 , and the revision of a result for , both based on spectra . our new measurement of @xmath44@xmath45 toward shows that column density measurements based on saturated lines might be erroneous , even up to a factor of two , and that the error bars might be underestimated due to systematic effects . this sight line exhibits a low d / o ratio , in agreement with those obtained toward other distant targets . 9 is the farthest galactic target for which the deuterium abundance has been measured from ultraviolet absorption lines so far , both in terms of distance and column densities . the low deuterium abundance measured on this line of sight is in agreement with the trend reported by hbrard & moos ( @xcite ) and reinforced by wood et al . ( @xcite ) : on large scales , the deuterium abundance is lower than the one measured within the local bubble . the question however remains open to know if this low deuterium abundance is the value representative of the present epoch , or the signature that a significant amount of deuterium has been depleted onto dust grains . finally , we have shown that d / h presents more dispersion than d / o does . as oxygen is considered to be a good proxy for hydrogen , this discrepancy is puzzling . this might call into question the homogeneity of the o / h ratio , or some or column density measurements . the complete explanation is still to be found . this work is based on data obtained by the nasa - cnes - csa mission operated by the johns hopkins university . financial support has been provided by nasa contract nas5 - 32985 . we thank d. massa and a. w. fullerton for their help on the stellar profile of 9 , as well as c. m. oliveira and a. vidal - madjar for useful comments on this paper . most of this work was performed during the stay of g.h . at the johns hopkins university in 2004 . was also supported by cnes . t.m.t . was supported in part by nasa grant nng04gg73 g . this work used the profile fitting procedure owens.f developed by m. lemoine and the french team . abgrall , h. , roueff , e. , launay , f. , roncin , j. y. , & subtil , j. l. 1993 , , 101 , 323 abgrall , h. , roueff , e. , launay , f. , roncin , j. y. , & subtil , j. l. 1993 , , 101 , 273 andr , m. k. , oliveira , c. m. , howk , j. c. , et al . 2003 , , 591 , 1000 beideck , d. j. , schectman , r. m. , federman , s. r. , & ellis , d. g. 1994 , , 428 , 393 bianchi , l. , & garcia , m. 2002 , , 581 , 610 bowen , d. v. , jenkins , e. b. , tripp , t. m. , sembach , k. r. , & savage , b. d. 2004 , in `` astrophysics in the far ultraviolet : five years of discovery with fuse '' , eds . g. sonneborn , h. w. moos , & b .- andersson ( san francisco : asp ) , in press ( astro - ph/0410008 ) cartledge , s. i. b. , lauroesch , j. t. , meyer , d. m. , & sofia , u. j. 2004 , , 613 , 1037 diplas , a. & savage , b. d. 1994 , , 93 , 211 draine , b. t. 2004 , in _ origin and evolution of the elements _ , ed . a. mcwilliams & m. rauch ( cambridge : cambridge univ . press ) , p. 320 ferlet , r. , et al . 2000 , , 538 , l69 fitzpatrick , e.l . , & spitzer , l. , jr . 1994 , , 427 , 232 friedman , s. d. et al . 2002 , , 140 , 37 friedman , s. d. et al . 2005 , , submitted hbrard , g. , mallouris , c. , ferlet , r. , koester , d. , lemoine , m. , vidal - madjar , a. , & york , d. 1999 , , 350 , 643 hbrard , g. et al . 2002 , , 140 , 103 h ' ebrard , g. & moos , h. w. 2003 , , 599 , 297 hog , e. , kuzmin , a. , bastian , u. , fabricius , c. , kuimov , k. , lindegren , l. , makarov , v. v. , & roeser , s. 1998 , , 335 , l65 hoopes , c. g. , sembach , k. r. , h ' ebrard , g. , moos , h. w. , & knauth , d.c . 2003 , , 586 , 1094 howk , j. c. , sembach , k. r. , roth , k. c. , & kruk , j. w. 2000 , , 544 , 867 jenkins , e. b. , tripp , t. m. , woniak , p. a. , sofia , u. j. , & sonneborn , g. 1999 , , 520 , 182 jenkins , e. b. , & tripp , t. m. 2001 , , 137 , 297 jura , m. a. 1982 , in advances in ultraviolet astronomy , ed . y. knauth , d. c. , andersson , b .- g . , mccandliss , s. r. , & moos , h. w. 2003 , , 596 , l51 lacour , s. , andr , m. k. , sonnentrucker , p. , le petit , f. , dsert , j .- m . , ferlet , r. , roueff , e. , welty , d. e. , york , d. g. 2005 , , 430 , 967 laurent , c. , vidal - madjar , a. , & york , d. g. 1979 , , 229 , 923 lebouteiller , v. , kuassivi , & ferlet , r. 2005 , , in press ( astro - ph/0507404 ) lemoine , m. et al . 1999 , new astronomy , 4 , 231 lemoine , m. et al . 2002 , , 140 , 67 linsky , j. l. et al . 2005 , submitted to mathys , g. 1988 , , 76 , 427 meyer , d. m. , cardelli , j. a. , & sofia , u. j. 1997 , , 490 , l103 meyer , d. m. , jura , m. , & cardelli , j. a. 1998 , , 493 , 222 meyer , d. m. 2001 , xviith iap colloquium , paris , edited by r. ferlet et al . , p. 135 moos , h. w. et al . 2000 , , 538 , l1 moos , h. w. et al . 2002 , , 140 , 3 morton , d. c. 1991 , , 77 , 119 morton , d. c. 2003 , , 149 , 205 oliveira , c. m. , h ' ebrard , g. , howk , j. c. , kruk , j. w. , chayer , p. , moos , h. w. 2003 , , 587 , 235 oliveira , c. m. , dupuis , j. , chayer , p. , moos , h. m. 2005a , , 625 , 232 pellerin , a. , et al . 2002 , , 143 , 159 press , w. , teukolsky , s. , vetterling , w. , and flannery , b. 1995 . numerical recipes in c. cambridge university press , second edition prochaska , j. x. , tripp , t. m. , & howk , j. c. 2005 , , 620 , l39 sahnow , d. j. et al . 2000 , , 538 , l7 savage , b. d. , & sembach , k. r. 1996 , ara&a , 34 , 279 savage , b. d. , meade , m. r. , & sembach , k. r. 2001 , , 136 , 631 sfeir , d. m. , lallement , r. , crifo , f. , & welsh , b. y. 1999 , , 346 , 785 sofia , u. j. , fabian , d. , & howk , j. c. 2000 , , 531 , 384 sonneborn , g. , tripp , t. m. , ferlet , r. , jenkins , e. b. , sofia , u. j. , vidal - madjar , a. , & woniak , p. r. 2000 , , 545 , 277 sonnentrucker , p. , friedman , s. d. , welty , d. e. , york , d. g. , & snow , t. p. 2003 , , 596 , 350 timmes , f. x. , truran , j. w. , lauroesch , j. t. , & york , d. g. 1997 , , 476 , 464 tripp , t. m. , lu , l. , & savage , b. d. 1996 , , 102 , 239 verner , d. a. , barthel , p. d. , & tytler , d. 1994 , , 108 , 287 welsh , b. y. , sasseen , t. , craig , n. , jelinsky , s. , & albert , c. e. 1997 , , 112 , 507 williger , g. m. , oliveira , c. , brard , g. , dupuis , j. , dreizler , s. , moos 2005 , , 625 , 210 wood , b. e. , linsky , j. l. , hbrard , g. , williger , g. m. , moos , h. w. , blair , w. p. 2004 , , 609 , 838

we have analyzed five x - ray outbursts of the neutron - star soft x - ray transient and investigated the timing properties of the source in correlation with its spectral states as defined by different positions in the color - color and hardness - intensity diagrams . the hard color and the source count rate serve as the distinguishing parameters giving rise to three spectral states : a low - intensity hard state , an intermediate state and a high - intensity soft state . these states are respectively identified with the extreme island , island and banana states that characterize the atoll sources . the large amount of data analyzed allowed us to perform for the first time a detailed timing analysis of the extreme island state . differences in the aperiodic variability between the rise and the decay of the x - ray outbursts are found in this state : at the same place in the color - color diagram , during the rise the source exhibits more power at low frequencies ( 1 hz ) , whereas during the decay the source is more variable at high frequencies ( 100 hz ) . the very - low frequency noise that characterizes the banana - state power spectra below 1 hz can not be described in terms of a single power law but a two - component model is required . in two outbursts a new 6 - 10 hz qpo has been discovered and tentatively identified with the normal / flaring branch - like oscillation observed only at the highest inferred mass accretion rates . we have compared the spectral and timing properties of with those of other atoll and z sources . our results argue against a unification scheme for these two types of neutron - star x - ray binaries . # 1*#1 *

astro-ph0311281

belongs to the general group of systems known as low - mass x - ray binaries ( lmxb ) . these systems consist of a neutron star or a black hole orbiting around a late - type star ( later than a ) . the x rays are the result of the accretion of material from the companion onto the compact star . mass transfer is thought to occur via roche lobe overflow , and hence proceeds via an accretion disk . in , the compact star is a neutron star and the mass - donating companion is a v=21.6 k7v star , located at an estimated distance of 2.5 kpc ( chevalier et al . an orbital period of 18.95 hours has been suggested ( chevalier & ilovaisky 1991 ; welsh , robinson & young 2000 ) . the high - energy radiation is characterized by 1.52 month long transient x - ray outbursts during which the x - ray luminosity can increase by more than three orders of magnitude . they are thought to be due to thermal instabilities in the accretion disk ( e.g. van paradijs 1996 ) . their recurrence time and duration vary but typical values are @xmath2 200 days and 4060 days , respectively ( @xmath3imon 2002 ) . the source also displays type i bursts ( zhang , yu & zhang 1998 ) that last a few tens of seconds and are interpreted as runaway thermonuclear burning of matter on the surface of the neutron star . various schemes have been proposed to categorize the lmxbs : x - ray spectral behavior as a function of intensity and the requirement of a low - energy blackbody component in the x - ray spectra ( parsignault & grindlay 1978 ; naylor & podsiadlowski 1993 ) , cluster analysis of a large number of source characteristics ( ponman 1982 ) , detailed x - ray spectral fits ( white & mason 1985 ) , x - ray hardness - intensity and color - color diagrams ( schulz , hasinger & trmper 1989 ) , aperiodic variability at very low frequencies ( reig , papadakis & kylafis 2003 ) . most relevant for the purpose of this paper is the classification scheme in terms of the rapid aperiodic variability and the patterns that these sources display in color - color diagrams ( hasinger & van der klis 1989 ) . in this scheme lmxbs are divided into two different subclasses , z and atoll sources . the three spectral branches that make up the z shape are called horizontal branch , normal branch and flaring branch and the two structures that occur in atoll sources are known as the island and the banana ( van der klis 1989 ) . at the lowest count rates an extension of the island state is sometimes seen with a harder spectrum and stronger band - limited noise than the `` canonical '' island state ( van straaten , van der klis & mndez 2003 and references therein ) . the term _ extreme island state _ has been used to designate such state ( prins & van der klis 1997 ; reig et al . the spectral and timing properties of atoll sources in this state are reminiscent of those seen in the low / hard state of black hole systems ( van der klis 1994a , 1994b ; berger & van der klis 1998 ; olive et al . 1998 ; belloni , psaltis & van der klis 2002 ) . both z and atoll sources move through these patterns continuously , that is , without jumping from one branch to another , although the source motion along the spectral tracks is much slower in atoll sources when they are in the island state than in the banana state and in z sources in all states . the classification of aquila x1 in this scheme was investigated by reig et al . ( 2000 ) see also cui et al . ( 1998 ) , who studied the correlated x - ray timing and spectral variations of and presented evidence for its classification as an atoll source , exhibiting all classic atoll source states . this scheme has been revisited by muno , remillard & chakrabarty ( 2002 ) and gierliski & done ( 2002 ) . they reported that three transient atoll sources which display a wide dynamic range in intensity ( @xmath4 ; 4u 160852 , 4u 170544 and trace out three - branch patterns in the color - color diagram similar to those of z sources , and suggested this may be a general feature . their study was based on the spectral properties only . however , as pointed out by hasinger & van der klis ( 1989 ) it is difficult to make a clear distinction between z and atoll sources on the basis of the motion through the color - color diagram alone . analysis of the rapid aperiodic variability to study the noise components in different regions of the color - color diagram and the actual time scales of the motion of the source through the diagram are important to identify source type and state . in this work we have studied the spectral and timing properties of in order to investigate whether atoll and z sources can be unified into one classification scheme and provide new insights into the poorly known extreme island states of atoll sources . we find that although the color - color diagram shows a branch whose topology is similar to the normal branch of z sources , neither the timing properties nor the motion in the color - color diagram of agree with those of z sources . the data were retrieved from the _ rxte _ archive and comprise all observations of available from 1997 february to 2002 may . data taken during satellite slews and earth occultation were removed . likewise , all type i bursts were excluded from our analysis . the observations contain five x - ray outbursts . although the duration and maximum intensity differ , the profile of the outbursts is very similar and is characterized by a fast rise and a slower decay . during intensity maximum the light curve is complex and multi - peaked . the fourth outburst showed two minor outbursts ( the peak intensity was one order of magnitude lower than the main outburst ) 70 days and 110 days after the main peak , respectively . the long - term light curve of the observations is shown in fig . [ lightc ] . background subtracted light curves corresponding to the energy ranges 2.03.5 kev , 3.56.0 kev , 6.09.7 kev and 9.716.0 kev were used to define the soft and hard colors as sc=3.56.0/2.03.5 and hc=9.716.0/6.09.7 , respectively . the color - color diagram ( cd ) of was then constructed by plotting the hard color as function of the soft color ( fig . [ ccd ] ) . during the time spanned by the observations the response of the detectors varied due to ageing . in addition , gain changes are applied occasionally , making the channel boundaries for a given energy range change with time , and also slightly affecting the effective areas of the detectors . each gain change is the start of a new `` gain epoch '' . we reduced the effect of the color shifts that results from these gain changes by linearly interpolating between the count rates in the two energy channels straddling the energy boundaries in each epoch . each data point of was normalized with the closest corresponding crab point within the same gain epoch in order to mitigate the response change effects on the colors . the final cd was obtained by averaging the five ( one for each detector ) normalized cds and rebinning into 256-s data points . we excluded data for which the resulting relative errors were larger than 5% . the total number of data points excluded from the analysis was @xmath0 10% . in order to recover the true values of the colors of the soft color should be multiplied by 2.35 and the hard color by 0.56 ( quoted values are averages of the crab colors during the observations and for all five pcus ) . the variation of the crab colors computed as the root - mean - square , i.e. ( the standard deviation over the mean color ) was 5.3% and 1.7% for the soft and hard colors , respectively see also fig . 1 in van straaten et al . ( 2003 ) . can be found in two main states : a low - intensity hard state and a high - intensity soft state . in addition , a short - lived intermediate or transition state is found displaying values of the hard color in between the two main states . the aperiodic variability as a function of spectral hardness , i.e. position in the cd of was previously studied by reig et al . ( 2000 ) . with only two of the five outbursts analysed in that work , the cd consisted of a soft branch , the banana branch , and two isolated groups of points which were called extreme island and island states . the large amount of data now analyzed reveals a more structured cd , in which the extreme island state appears as a more stable and longer - lived branch than the island state . the high / soft state corresponds to the classical banana state and the island state represents the transition between the two main states . outbursts 1 , 4 and 5 ( o1 , o4 and o5 ) contain points in the two main states while outbursts 2 and 3 ( o2 and o3 ) provide points to the banana state only . observations of o2 and o3 began and finished when the source was still at a relatively high level of emission ( @xmath5 c / s ) and did not experience large amplitude changes ( i@xmath6/i@xmath7 @xmath8 5 ) . in contrast , outburst 1 , 4 and 5 extended over a larger dynamic range in intensity ( i@xmath6/i@xmath7 @xmath9 50 ) . none of the points of the two minor outbursts which followed o4 ( which will be termed here as o4@xmath10 ) contributed to the banana state . outbursts 1 , 4 , 4@xmath10 and 5 include points in the extreme island state . while o4 and o5 include points both during the rise and during the decay , o1 gives points during the decay only . the island state occurred only during the decay of o1 and o5 . all five outbursts ( except for o4@xmath11 ) have points in the banana state but only o3 provides banana state points during the rise . figure [ cdmotion ] displays the motion of in the cd for outbursts 4 and 5 . data points are @xmath21-day averages . figure [ coltime ] shows the evolution of the colors with time for those outbursts that include spectral transitions ( o1 , o4 , o5 ) . the 216 kev intensity just before the transitions is also given . all quoted intensities in this section are background subtracted and correspond to 5 pcus in the energy range 216 kev . there is a strong correlation between the position of the source in the cd and its intensity . at the onset of the outbursts the data points distribute in the softest part of the extreme island state . the count rate is @xmath0 100 c / s . as the intensity increases the source moves toward the right along the extreme island branch . the hard color ( hc ) slightly decreases on average . in about 6 days the count rate increases by one order of magnitude and the soft color ( sc ) increases by @xmath12 + 0.3 . then the source seems to jump to the banana state , recovering the initial values of the soft color . the hard color decreases by @xmath13 . the 216 kev intensity is @xmath2 3 or 4 times that of the extreme island state prior to the spectral transition , namely 40006000 c / s . no intermediate points between the two states are observed during the rise of the outbursts . this can be attributed to the fast rise and the lack of good sampling of the data . indeed , the observational gaps between the last point of the extreme island branch and the first one of the banana branch were 2.6 and 3.1 days for o4 and o5 , respectively . as the intensity continues to increase the source becomes harder , i.e. , it moves to the right along the banana , with approximately constant hard color . the peak of the outburst is characterized by flaring activity with erratic changes in count rate . the count rate at the peak is 50100 times the minimum detected count rate . despite this irregular behavior the correlation between the intensity and the colors is maintained in the sense that the lower intensity points in the flares display a lower soft color as it is illustrated in fig . [ flare2ccd ] . in other words , during the flare maxima the source lies at the right end of the banana branch , and in between flares it moves to the left . thus the flaring variability in the light curve translates into the cd in a back and forth motion which approximately extends over the right half part of the banana state . this motion is quite fast . as an example , in one of the flares in o5 the soft color changed by 0.08 in about 2.4 hr . in another flare by 0.06 in 0.55 hr . this flaring behavior is seen in all outbursts but o2 shows it exclusively ( fig . [ flare2ccd ] ) . such behavior stops once the source gets to half way the decay of the outburst , at which point the source shifts to lower soft colors ( to the left along the banana branch ) . as the outburst declines the source moves back along the banana branch , abandoning it when the count rate becomes lower than @xmath2 800 c / s . the transition to the hard state occurs at much lower soft color than when it entered the banana , although not necesarily at the lowest soft color . the banana state covers values of the soft color between 0.87 ( o5 ) and 1.16 ( o3 ) . the transition to the hard state takes place at sc @xmath12 0.920.93 ( fig . [ cdmotion ] ) . the count rate at which the transition between the two main states takes place is higher when the source is in the rise of the outburst . hard to soft transitions occur when the count rate is well above @xmath2 1000 c / s . soft to hard transitions when the count rate is well below @xmath2 1000 c / s as it can be seen in fig . [ coltime ] . rather than jumping directly to the hard branch the source remains for a short time ( @xmath14 days ) in an intermediate state , the island state . note that such a short island state episode would usually have been missed due to data gaps in the rise . the count rate in the island state is @xmath2 200400 c / s . at even lower count rates the source finds itself back in the extreme island branch , moving toward the left as the intensity decreases , to eventually become too weak to measure the colors sufficiently well . the source re - enters this state during the decay when the count rate goes below @xmath15 c / s . the re - entry point occurs at a lower soft color ( sc @xmath12 1.051.10 ) then when the source left the extreme island state ( sc @xmath12 1.3 ) . as the speed of motion along the extreme island state is approximately constant , the time during which the source can be found in this state is shorter during the decay of the outburst than during its rise . a difference in the position of the source in the cd depending on whether the source intensity increases or decreases was already recognized by reig et al . ( 2000 ) . the main result that should be stressed is that transitions between states do not occur at the same point of the spectral branches . during the rise of the x - ray outburst the source occupies the hardest parts of the extreme island state before the spectral transition to the banana state . during the decay it tends to occupy the softest part of the banana branch before moving into the island state . however , the points of departure and arrival are different . in the cd , this translates into some sort of rectangular track ( fig . [ cdmotion ] ) along which moves clockwise as the count rate first increases and then decreases . in addition to fig . [ coltime ] , tables [ dur ] and [ timetran ] also illustrate the time scale for the motion through the diagram . table [ dur ] and table [ timetran ] give upper limits on the duration of the source in each state and the time scales of the spectral transitions , respectively . note that the observational gap for the spectral transitions from the extreme island state to the banana state is approximately two times longer than viceversa . given the speed of motion of the soft color along the extreme island branch ( roughly 0.05 day@xmath16 ) a three - day gap might imply a change in soft color of about 0.15 . thus , it is possible that the actual entry point into the banana branch is at lower values of the soft color . in this case the motion of in the cd would be similar to that of 4u 1705 - 44 ( barret & olive 2002 ) , in which both the transition from ( to ) the extreme island state to ( from ) the banana state occur at the same point of the banana branch . the motion in the cd then would then resemble an inverted triangle rather than a rectangle . in order to investigate the aperiodic variability of we obtained power spectra by dividing the 260 kev pca data of each observation into 256-s segments and calculated the fourier power spectrum of each segment up to a nyquist frequency of 2048 hz . the high - frequency end ( 15002048 hz ) of the power spectra was used to determine the underlying poisson noise ( approximated by a constant power level ) , which was subtracted before performing the spectral fitting . our aim is to investigate the aperiodic variability in correlation with the spectral states , which in turn , correlate with the source count rate . thus we divided the color - color plane into various regions and obtained a mean power spectrum for each region . the mean power spectrum was the result of averaging all the power spectra corresponding to the points enclosed in each color region . see table [ reg ] to find the average values of the colors , boundaries and intensity of each region . prior to this , we investigated whether power spectra from the same cd region taken at different outbursts were consistent with having the same properties . we found that while this was true for the banana state , some differences existed in the extreme island state depending on whether the source was in the rise or the decay of the x - ray outburst . the extreme island state is populated with points from the rising part of o4 and o5 , the decaying part of o1 , o4 and o5 and from the extended extreme island state of o4@xmath11 . consequently , the extreme island state was first separated into points pertaining to the rise , to the decay or to o4@xmath11 . then each one of these groups was further divided into three subgroups according to the value of the soft color . note that in the case of the rise and decay regions this division implies a division in count rate as well as in hard color . on average , the extreme island sates eisrise1 and eisdecay2 , ( see table [ reg ] ) have lower intensity and are softer than eisrise2 and eisdecay1 , respectively . in turn , eisrise2 contains softer points than eisrise3 . the island state points ( intermediate state ) defined another region . the timing properties of the banana state in atoll sources have been seen to vary smoothly as the soft color , or equivalently , the count rate increases , that is , from the lower ( left ) to the upper ( right ) banana branch . therefore , the banana branch was split into five regions ( bs1 to bs5 ) in increasing order of the soft color . each region approximately contains the same amount of data . there is no unanimity in the literature about the names of the power spectral noise components nor about the mathematical functions used to fit those components . in this work we have followed the terminology of belloni et al . ( 2002 ) and fitted the power spectra using lorentzian functions only . also , for displaying our power spectra we have used the power times frequency representation @xmath17 vs @xmath18 , in which the product of the power density in the rms normalization and the fourier frequency is plotted as a function of the fourier frequency ( belloni et al . 1997 ) . as fit parameters we use the frequency @xmath19 at which the lorentzian attains its maximum value in such representation , and the @xmath20 value , which gives a measure of the coherence of the lorentzian . in terms of the half width at half maximum or @xmath21 , these quanties are related by @xmath22 and @xmath23 , where @xmath24 is the central frequency of the lorentzian . between three and five lorentzians were needed in order to get a good fit . if during the fitting procedure a lorentzian component resulted with negative @xmath20 value then we set it to zero . it should be noted that the naming scheme of belloni et al . ( 2002 ) was applied , in the case of low - luminosity neutron stars , to systems displaying the island state only . thus , a complete description of some power spectra , especially those in the banana state , in terms of the noise components given in that work is not always possible and extra or new components are needed in order to obtain acceptable fits . the extensions to the belloni et al . naming scheme we use are those of van straaten et al . ( 2003 ) . in the terminology of belloni et al . ( 2002 ) , the band - limited noise of atoll sources in the ( extreme ) island state consists of three lorentzians namely , @xmath25 that accounts for the power spectrum at low frequencies ( @xmath26 hz ) , @xmath27 and @xmath28 that fit its high - frequency end . @xmath27 may show up as a broad bump at frequencies 1025 hz ( normally in the island state ) or as a narrow qpo at frequencies @xmath1 100 hz ( normally in the banana state ) . in fact , at frequencies above @xmath1 100 hz @xmath27 and @xmath28 represent the khz qpos seen in low - mass x - ray binaries . however , note that this identification of @xmath27 with both a broad peak and the lower khz qpo is only tentative and the broad bump at 10 - 25 hz might alternatively represent a separate component ( van straaten et al . the intermediate frequency range ( 1100 hz ) is described by one broad lorentzian or `` hump '' , @xmath29 and one narrow lorentzian @xmath30 referred to as the low - frequency qpo . not detected by belloni et al . ( 2002 ) but also present in some atoll sources ( van straaten et al . 2002 ) is the so - called hectoherz qpo , @xmath31 , a broad feature with characteristic frequency at @xmath2 100 hz . table [ fitres ] lists the results of the fits and fig . [ pds ] depicts the power spectra and fit functions for different regions of the cd ( see table [ reg ] to identify the position in the cd ) . the extreme island state power spectrum contains , in order of increasing frequency , the low - frequency band - limited noise , @xmath25 , the broad peak low - frequency noise @xmath29 , the broad version of @xmath27 and one more component which could be either the hhz qpo or a khz qpo . based on the rms / frequency correlations of the noise components seen in other atoll sources , van straaten et al . ( 2002 , 2003 ) favored the upper khz interpretation . if @xmath27 is also interpreted as the lower khz qpo then the intervals eisrise2 , eisrise3 and iso4@xmath32 ( see fig.[pds ] ) would represent the first detection of the two khz qpo in . alternatively , the fourth component could be the hhz qpo . note that the narrow @xmath21160 hz peak in eiso4@xmath32 is not statistically significant . an f - test reveals that the probability that the improvement of the fit ( in terms of a lower @xmath33 ) by the addition of an extra lorentzian happen by chance is larger than 30% . the characteristic frequencies of the noise components increase and their rms values decrease slightly as the count rate increases , i.e. , the source moves toward the right in the extreme island state . figure [ corr ] displays the characteristic frequency of the @xmath29 and @xmath27 components as a function of the band - limited noise component @xmath25 for , the atoll sources 4u 160852 ( van straaten et al . 2003 ) , 4u 172834 and 4u 0614 + 09 ( van straaten et al . 2002 ) and the z source gx51 ( jonker et al . 2002 ) . in the case of gx51 te horizontal branch oscillation is plotted instead of @xmath29 . note also that the broad - band noise in the z source was not fitted with a lorentzian but with a cutoff power law , @xmath34 . by differentiating @xmath35 and equating the resulting function to zero the maximum frequency of the z band - limited noise , @xmath36 , was derived . figure [ corr ] seems to confirm that _ i ) _ the 120 hz broad feature @xmath29 observed in atoll sources at low count rates ( island states ) has the same origin as the horizontal branch oscillations ( hbo ) observed in z sources ( psaltis , belloni & van der klis 1999 , van straaten et al . 2003 ) , _ ii ) _ the broad bump at frequencies 1025 hz is associated with the lower khz qpo . the extreme island state data in sample the lowest characteristic frequencies so far detected . one interesting result from the timing analysis of the extreme island state is the differences in the aperiodic variability of the source during the rise and decay of the outbursts . figure [ island ] shows three power spectra of and an enlargement of the cd displaying the extreme island state . all three power spectra are associated with the same value of the colors , i.e , they occupy the same position in the cd ( marked in fig . [ island ] ) . however , they correspond to different parts of the x - ray outburst : rise ( diamonds ) , decay ( stars ) and when the source found itself in the extended island state of o4@xmath11 ( crosses ) . during the rise the power is concentrated at low frequencies whereas during the decay , the source exhibits more power at high frequencies during the extended island state of outburst o4@xmath11 there is roughly equal power at all frequencies . the differences in the characteristic frequencies of the noise components are much less pronounced and consistent with one onother whenever the same component appears . the low - frequency end ( @xmath0 1 hz ) of the banana state power spectra is dominated by the very - low frequency noise ( vlfn ) . one single power law is normally used to fit this component . however , the power spectrum of below 1 hz can not be described by one component only . in order to fit the vlfn we used two zero - centered lorentzians that were called low vlfn ( @xmath37 ) and high vlfn ( @xmath38 ) ( schnerr et al . 2003 ) . the characteristic frequencies of these components are found to be independent of the position of the source in the cd : @xmath39 hz and @xmath20.03 hz , respectively . however , while the fractional rms variability of lower frequency vlfn roughly remains constant at @xmath23.4% , the higher frequency vlfn becomes stronger increasing from @xmath20.9% in the lower banana to @xmath22% in the upper banana as the count rate increases . above 1 hz the banana state power spectra contain one broad lorentzian , describing the band - limited noise and one qpo ( two in bs1 ) . the characteristic frequency and width of the broad band - limited noise component decrease as the count rate increases ( i.e. as the source moves toward the right ) . this component has also been seen in 4u 160852 ( van straaten et al . 2003 ) , 4u 0614 + 09 and 4u 172834 ( van straaten et al . 2002 ) . however , the lack of data precludes a more detailed comparison . only 4u 172834 , with just two measurements , seems to have the same behavior as i.e. , the frequency of this noise component decreases along the banana . following van straaten et al . ( 2003 ) we will refer to this component as @xmath40 . next to , sometimes on top of , the band - limited noise in the banana state there is a relatively narrow peak , whose characteristic frequency and width do not correlate with any other parameter . this narrow qpo has a frequency in between that of the hhz lorentzian and @xmath29 . van straaten et al . ( 2003 ) see also di salvo et al . ( 2001 ) and van straaten et al . ( 2002 ) argued that it is the band - limited noise component @xmath25 which transforms into a narrow qpo above certain frequency ( @xmath41 hz in 4u 160852 ) . the appearance of @xmath40 and this transformation of @xmath25 occurs coincidentally . in , although the narrow version of @xmath25 appears systematically in all banana - state power spectra it is always below @xmath42 . the detection of a second khz qpo in is only marginal ( van der klis 2000 ) . based on the values of the fitting parameters and the relation between the qpo frequency and the x - ray colors , mndez , van der klis & ford ( 2001 ) concluded that whenever a khz qpo is detected in it is always the lower khz qpo . the fractional rms and characteristic frequency of the second qpo in bs1 are consistent with being the lower khz qpo ( fig . 1 of mndez , van der klis & ford 2001 ) . also , the characteristic frequencies of the @xmath27 and @xmath25 in bs1 agree with the correlation of fig . [ corr ] if an extrapolation to higher frequencies of the points is done . the analysis of the transition state is hampered by poor statistics given the relatively low number of points . one single lorentzian with peak at @xmath43 hz and q value of 0.15 accounts for the entire 0.01100 hz frequency band . this noise component is identified as @xmath25 . a second narrower lorentzian with @xmath44 hz and @xmath45 fits the noise at higher frequencies . its fractional rms , @xmath46% , is too high to agree with the lower khz qpo . in fact , the value of the frequency and rms are similar to what it is seen in 4u 160852 and 4u 172834 for the upper khz qpo ( fig . 1 of mndez , van der klis & ford 2001 ) . nevertheless , given the relatively noisy spectrum at high frequencies such identification should be taken with care . a qpo with frequency 714 hz has been found in at least two atoll sources , xte j1806246 ( wijnands & van der klis 1999 ; revnivtsev , borozdin & emelyanov 1999 ) and 4u 182030 ( wijnands et al . 1999 ) . these qpos are detected at the highest inferred mass accretion rates only , are short - lived ( a few hundreds of seconds ) and are localized in a very narrow region of the cd , namely the tip of the upper banana . they are reminiscent of those observed in z sources in the normal and flaring branches , when the source is accreting near the eddington limit . thus the detection of similar qpo in atoll sources at significantly lower accretion rates questions the current models that explain nbo in terms of near - eddington accretion ( fortner , lamb & miller 1989 ) . we searched for nbo - like qpo in and tentatively found two of them at @xmath47 significance , at frequencies [email protected] hz in o3 and [email protected] hz in o5 ( fig . [ qpo ] ) . they have similar rms amplitude ( @xmath49% ) and q - values ( @xmath50 ) . the average values of the colors and intensity associated with these qpos are sc=1.11 , hc=0.52 , @xmath51=5045 count s@xmath16 and sc=1.06 , hc=0.51 , @xmath51=6870 count s@xmath16 , respectively . for the sake of comparison with other sources we also give the value of the x - ray luminosity : @xmath52 erg s@xmath16 for outburst 3 and @xmath53 erg s@xmath16 for outburst 5 . as a confirmation that they occur at the highest count rates , typical luminosities in the softest parts of the extreme island and banana states are @xmath54 and @xmath55 erg s@xmath16 , respectively . the values of the luminosity are for a distance of 2.5 kpc and the energy range 216 kev . in the other three outbursts the detection is only marginal . the low statistical significance of these qpos ( @xmath56 ) can be ascribed to their extremely short life . we have analysed all the available rxte data of between 1997 february and 2002 may and investigated its spectral and timing properties in correlation . this work has been motivated by recent reports that show that if large data sets are used , the atoll sources displaying the largest ampitude variations in intensity show three - branch patterns in the color - color diagram ( cd ) , reminiscent of z sources ( muno , remillard & chakrabarty 2002 ; gierliski & done 2002 ) . consequently , they conclude that the distinction between atoll and z sources might be no longer justified . the extended island state , island state and banana state of atoll sources would then correspond to the horizontal branch ( hb ) , normal branch ( nb ) and flaring branch ( fb ) , respectively . the results of these studies were based on the spectral properties ( i.e. color - color diagram ) only . in order to be able to distinguish between the different types of low - mass x - ray binaries the rapid aperiodic variability associated with the various spectral states needs to be addressed as well ( hasinger & van der klis 1989 ) , and that is what we have attempted in this paper . we first make a comparison of the timing properties of with those of other atoll sources and then discuss whether ot not it can be considered as a z source . 4u 160852 and are the atoll sources that display the largest dynamic range in x - ray intensity ( @xmath57 ) . thus a comparison between these two sources is of particular interest . 4u 160852 has been extensively studied by van straaten et al . ( 2003 ) . by comparing the shape of the power spectra ( see fig . 7 in van straaten et al . 2003 and our fig . [ pds ] ) and the results of the power spectral fits ( their table 2 and 3 and our table [ fitres ] ) one can find distinct similarities between the two sources : the power spectra of intervals a , b and c in 4u 160852 resemble those of the extreme island state of , while interval d of 4u 160852 is reminiscent of the island state in . likewise , the shape and number of lorentzians appearing in the power spectra of the extreme island state of are similar to those of the three low - luminosity x - ray bursters studied by belloni et al . ( 2002 ) . it is worth noting the relevant role of the hard color in connection with the timing properties of the extreme island state . in the characteristic frequencies of the timing features in the extreme island state increase from left to right this is opposite to what it is seen in 4u 1608 - 52 , where frequencies increase from right to left . however , in both sources the increase in frequencies occurs in correlation with an overall decrease in hard color . we therefore suggest that in the extreme island state it is the hard color , rather than the intensity , which is the main determining factor of the timing properties . we note , that black hole candidates are likewise characterized by power spectral states whose occurrence seems to primarily depend on spectral hardness and which are relatively insensitive to intensity , and which are associated with two - dimensional motion through the color - intensity plane rather than motion along a one - dimensional track ( homan et al . it may be that in the island and extreme island states , the states in which atoll sources are most similar to black holes ( e.g. , van der klis 1994a ; belloni , psaltis & van der klis 2002 ) , possibly because in these states the inner disk radius is furthest from the neutron star surface , this 2-d picture better describes neutron star behavior than the 1-d description appropriate to the banana state and to z sources . leaving aside the khz qpos , the overall shape of the banana - state power spectra of present similarities with those of other atoll sources , namely , a strong red - noise component , the very - low frequency noise ( vlfn ) below @xmath2 1 hz and a broad and a narrow components describing the band - limited noise ( bln ) in the frequency range 1100 hz . compared to 4u 0614 + 09 , 4u 1608 - 52 and 4u 172834 , exbibits the strongest vlfn but the weakest bln . while the vlfn is normally described by either a power law with index between 1.5 - 2.0 or a zero centered lorentzian in the other atoll sources , it requires the use of two model components in , similar to the case of gx 13 + 1 ( schnerr et al . 2003 ) . the broad band - limited noise component @xmath40 in and 4u 1608 - 52 , or equivalently the zero centred lorentzian in 4u 172834 and 4u 0614 + 09 have similar rms values . in contrast , @xmath25 , which is clearly detected in these other atoll sources , is not significant in in the banana state . gierliski & done ( 2002 ) suggested that the name atoll source is no longer appropiate as these type of low - mass x - ray binaries also display a three - branch pattern in the color - color diagram when large data sets of observations are used , reminiscent of the z pattern for which z sources are named . this effect is more pronounced in systems exhibiting large flux variations ( muno et al . 2002 ) , like . the cd of ( fig . [ ccd ] ) does indeed show a branch ( the island state ) that can be interpreted to connect the two main branches like in a z shape . in addition , the correlation between the characteristic frequency of the 120 hz broad lorentzian @xmath29 and the band - limited noise component ( fig . [ corr ] ) at low count rates , suggesting a common origin with the horizontal branch oscillations and lfn in z sources and the discovery of 714 hz qpo at the highest x - ray flux ( fig . [ qpo ] ) , reminiscent of the z - source normal / flaring branch oscillations , strengthen the similarities between and z sources . however , there are a number of systematic differences that clearly puts outside the group of z sources . first , there is the motion in the cd . in the extreme island state as the count rate increases the soft color increases and the hard color decreases . then the source makes a rapid transition to the banana state and proceeds to harder colors ( right ) as the count rate increases further . however , after the peak of the outburst , when the intensity decreases the source does not follow exactly the reverse path . as shown in fig . [ cdmotion ] , enters the banana state ( extreme island state ) at a higher ( lower ) value of the soft color than when it leaves it . the source moves clokcwise following a rectangle track . it should be pointed out that the entrance into the banana state at such high soft color is not caused by the time resolution used ( @xmath2 1 day ) in fig . [ cdmotion ] . a 256-s bin shows the same effect . however , we can not rule out an observational origin since the time gap between the transition from the extreme island state to the banana state is @xmath12 5 days for outburst 5 and @xmath12 2.6 for outburst 4 ( table [ timetran ] ) . if this is the case and the source entered the banana state from the left then the pattern that it would trace out in the cd would be that of an inverted triangle , much like 4u 170544 ( barret & olive 2002 ) . in contrast , the re - entrance into the extreme island state at a low soft color is probably not cause by the lack of data given the relatively short gap between the island state and the extreme island state ( just 1.6 days in outburst 5 ) . in either case this behavior is not that of z sources . in z sources , the nb is not a fast transition between the other two branches , and the source follows the same path whether it is going up or down along the nb . in this respect , is like the other two atoll sources that display high - amplitude flux variations . more detailed looks into the motion of 4u 160852 and 4u 170544 have revealed different topologies from the classical z - shaped track . the motion of 4u 160852 resembles the greek letter @xmath58 ( van straaten et al . 2003 ) and 4u 170544 describes an inverted triangle ( barret & olive 2002 ) . as noted above , these various topologies may simply be different aspects of 2-d motion through the color - intensity plane in which it is hard color that determines the timing properties . second , the amplitude of the x - ray luminosity change over the z track is also different . in the x - ray luminosity throughout the cd , i.e. , from the left of the banana state to the right of the extreme island state is about three orders of magnitude . in z sources , luminosity changes are typically less than a factor of 2 ( di salvo et al . 2000 , 2001 , 2002 ) . third , another difference is the velocity and time spent by the source in each spectral state . jonker et al . ( 2002 ) calculated that the z source gx 51 spent most of the time in the normal branch . in contrast , spends the smallest percentage of the the total observing time in the transition state ( the analogue to the normal branch would be a z source ) . wijnands et al . ( 1997 ) found that cyg x2 moves through the z most slowly on the hb , faster on the nb and fastest on the fb . during a typical x - ray outburst , i.e. , when is x - ray active , it spends most of its time in the banana branch ( the analogue to the flaring branch ) . finally , the properties of the aperiodic variability of also differ from those in z sources , especially at very low count rates . we can establish the following differences : * the very existence of the extreme island state . although the characteristic frequencies in increase along the extreme island state as the count rate increases as z sources do in the hb , the latter never reach the low frequencies that are seen in atoll sources in the extreme island state , even when at the left end of the hb . likewise , the rms amplitude of the noise components in the extreme island state is significantly higher than in a z source hb . the total rms amplitude of those components in amounts to @xmath59% , while typically in z sources , it does not go above @xmath60% . the extreme island - state power spectra are typical of atoll sources . * the peaked noise components , such as @xmath29 and the upper khz qpo are narrow qpos on the hb in z sources , with typical values of the q parameter above 3 for the upper khz qpo , whereas they are considerably broader peaked noise components in the extreme island state in ( q @xmath8 0.5 ) . * in , the normal / flaring branch - like oscillation does not occur during the `` normal / flaring branch '' , i.e. the island or the lower banana branches but in the upper banana branch , just like in other atoll sources . * the strength of the khz qpos decreases as the z source moves along the z track from hb to nb . typically , the khz qpos become undetectable by the time the source reaches the middle of the nb . in contrast , presents khz qpos in the island and banana states . * the time scales of the aperiodic variability tend to decrease ( frequencies increase ) as the z source moves along the fb , which is the opposite behavior to what is is observed in in the banana branch . we have investigated the timing and spectral properties of the soft x - ray transient during five x - ray outbursts covering a period of more than five years . three spectral states show up in the color - color diagram of : a low / hard state , identified with an extreme island state , a high / soft state corresponding to the classic banana state and an intermediate state associated with the island state of atoll sources . we have found that the hard color plays a crucial role in determining the timing properties of the extreme island state . although the overall distribution of these states in the color - color diagram may resemble a z , neither the motion in the cd , nor the typical time scales through the different branches of the cd , nor the range of x - ray luminosities , nor the timing properties at very low count rates ( the existence of the extreme island state ) are compatible with the classical z - source behavior . pr is a researcher of the programme _ ramn y cajal _ funded by the university of valencia and the spanish ministery of science and technology . pr also wants to thank the university of crete for providing in part the resources needed to carry out this work . this research has made use of data obtained through the high energy astrophysics science archive research center online service , provided by the nasa / goddard space flight center . barret , d , & olive , j.f . , 2002 , apj , 576 , 391 belloni , t. , van der klis , m. , lewin , w. h. g. , van paradijs , j. , dotani , t. , mitsuda , k. , & miyamoto , s. , 1997 , a&a , 322 , 857 belloni , t. , psaltis , d. , van der klis , m. , 2002 , apj , 572 , 392 berger , m. , & van der klis , m.,1998 , a&a , 340 , 143 chevalier , c. & ilovaisky , s. a. 1991 , a&a , 251 , l11 chevalier , c. , ilovaisky , s.a . , leisy , p. , patat , f. , 1999 , a&a , 347 , l51 cui , w. , barret , d. , zhang , s.n . , chen , w. , boirin , l. , swank , j. , 1998 , apj , 502 , l49 di salvo , t. , stella , l. , robba , n. r. , van der klis , m. , burderi , l. , israel , g. l. , homan , j. , campana , s. , frontera , f. , parmar , a. n. , 2000 , apj , 544 , l119 di salvo , t. , mndez , m. , van der klis , m. , ford , e. , robba , n. r. , 2001 , apj , 546 , 1107 di salvo , t. , farinelli , r. , burderi , l. , frontera , f. , kuulkers , e. , masetti , n. , robba , n. r. , stella , l. , van der klis , m. , 2002 , a&a , 386 , 535 fortner , b. , lamb , f. k. , miller , g. s. , 1989 , nature , 342 , 775 gierlinski , m. , done , c. , 2002 , mnras , 331 , l47 hasinger , g. & van der klis , m. 1989 , a&a , 225 , 79 homan , j. , wijnands , r , van der klis , m. belloni , t. , van paradijs , j. , klein - wolt , m. fender , r. mndez , m. , 2001 , apjs , 132 , 377 jonker , p. g. , van der klis , m. , homan , j. , mndez , m. , lewin , w.h.g . , wijnands , r. , zhang , w. , 2002 , mnras , 333 , 665 mndez , m. van der klis , m. , ford , e.c . , 2001 , apj , 561 , 1016 muno , m.p . , remillard , r.a . , chakrabarty , d. , 2002 , apj , 568 , l35 naylor , t. , podsiadlowski , ph , 1993 , mnras , 262 , 929 nowak , m.a . , 2000 , mnras , 318 , 361 olive , j. f. , barret , d. , boirin , l. , grindlay , j. e. , swank , j. h. , & smale , a. p. 1998 , a&a , 333 , 942 parsignault , d.r . , grindlay , j.e . , 1978 , apj , 225 , 970 ponman , t , 1982 , mnras , 200 , 351 prins , s. , van der klis , m. , 1997 , a&a , 319 , 498 psaltis , d. , belloni , t. , van der klis , m. , 1999 , apj , 520 , 262 reig , p. , mndez , m. , van der klis , m. , ford , e. c. , 2000 , apj , 530 , 916 reig , p. , papadakis , i. , kylafis , n. , 2003 , a&a , 398 , 1103 revnivtsev , m. , borozdin , k. , emelyanov , a. , 1999 , a&a , 344 , l25 schnerr , r.s . , reerink , t. , van der klis , m. , homan , j. , mndez , m. , fender , r.p . , & kuulkers , e. , astro - ph/0305161 schulz , n. s. , hasinger , g. & trmper , j. 1989 , a&a , 225 , 48 @xmath3imon , v. , 2002 , a&a , 381 , 151 van der klis , m. , 1989 , ara&a , 27 , 517 van der klis , m , 1994a , a&a , 283 , 469 van der klis , m , 1994b , apjs , 92 , 511 van der klis , m. , 2000 , ara&a , 38 , 717 van paradijs , j. , 1996 , apj , 464 , l139 van straaten , s. , van der klis , m. , di salvo , t. , belloni , t. , 2002 , apj , 568 , 912 van straaten , s. , van der klis , m. , mndez , m. , 2003 , apj , in press , astro - ph/0307041 welsh , w.f . , robinson , e.l . , young , p. , 2000 , aj , 120 , 943 white , n.e . , mason , k.o . , 1985 , ssrv , 40 , 167 wijnands , r. , van der klis , m. , 1997 , apj , 482 , l65 wijnands , r. , van der klis , m. , 1999 , apj , 514 , 939 wijnands , r. , van der klis , m. , rijkhorst , e. 1999 , apj , 512 , l39 zhang , s. n. , yu , w. & zhang , w. , 1998 , apj , 494 , l71 lcccc states & duration & i@xmath61 & i@xmath62 & i@xmath63 + & days & c / s & c / s & c / s + + eis & /0.14 & /56 & /63 & /60 + bs & 15.1 & 583 & 3057 & 2164 + is & 0.10 & 200 & 220 & 209 + + eis & & & & + bs & 30.96 & 1213 & 2694 & 1778 + is & & & & + + eis & & & & + bs & 38.95 & 1227 & 6175 & 4656 + is & & & & + + eis & 6.02/2.03 & 110/70 & 1295/156 & 650/127 + bs & 18.05 & 368 & 4835 & 2330 + is & & & & + + eis & 63 & 77 & 530 & 340 + bs & & & & + is & & & & + + eis & 6.54/1.33 & 87/128 & 1444/255 & 542/203 + bs & 41.53 & 810 & 8555 & 4925 + is & 0.01 & 458 & 463 & 460 + lccccccccc states & @xmath65&@xmath66&@xmath67 & @xmath68&mean & mean & mean count & number of&@xmath33(dof ) + & & & & & @xmath69 & @xmath70 & rate@xmath71 ( c / s ) & power spectra & + eisrise1 & ( 0.89,1.08)&(1.12,1.08)&(0.89,1.25)&(1.11,1.26)&1.01 & 1.14 & 185 & 54 & 122/92 + eisrise2 & ( 1.14,1.06)&(1.21,1.07)&(1.14,1.15)&(1.26,1.14)&1.18 & 1.09 & 925 & 34 & 105/84 + eisrise3 & ( 1.19,1.04)&(1.34,1.05)&(1.23,1.10)&(1.29,1.11)&1.25 & 1.07 & 1063 & 23 & 95/87 + eisdecay1 & ( 1.01,1.02)&(1.11,1.03)&(1.00,1.12)&(1.11,1.23)&1.05 & 1.07 & 183 & 37 & 87/92 + eisdecay2 & ( 0.92,1.11)&(1.11,1.14)&(0.88,1.43)&(1.06,1.45)&0.98 & 1.24 & 76 & 40 & 106/92 + eiso4@xmath72&(0.96,1.04)&(1.24,1.06)&(0.96,1.26)&(1.10,1.26)&1.05 & 1.15 & 188 & 152 & 106/87 + eiso4@xmath73&(1.08,1.04)&(1.07,1.06)&(0.96,1.26)&(1.10,1.26)&1.14 & 1.14 & 356 & 367 & 114/84 + is & ( 0.97,0.71)&(1.11,0.89)&(0.82,0.97)&(0.95,1.05)&0.98 & 0.90 & 255 & 28 & 99/91 + bs1 & ( 0.83,0.47)&(0.93,0.41)&(0.89,0.59)&(0.95,0.57)&0.91 & 0.50 & 1275 & 313 & 193/87 + bs2 & ( 0.93,0.41)&(0.99,0.40)&(0.95,0.57)&(1.01,0.57)&0.97 & 0.49 & 2086 & 368 & 139/89 + bs3 & ( 0.99,0.40)&(1.04,0.41)&(1.01,0.57)&(1.02,0.57)&1.02 & 0.49 & 3507 & 368 & 100/88 + bs4 & ( 1.04,0.41)&(1.10,0.42)&(1.02,0.57)&(1.05,0.61)&1.07 & 0.50 & 5248 & 381 & 89/88 + bs5 & ( 1.01,0.42)&(1.18,0.50)&(1.05,0.61)&(1.09,0.68)&1.12 & 0.51 & 5315 & 250 & 99/88 + lccccccc state@xmath71 & @xmath37&@xmath38 & @xmath25 & @xmath40 & @xmath29 & @xmath27 & @xmath31,@xmath28 + + eisrise1 & - & - & [email protected] & - & [email protected] & [email protected] & - + eisrise2 & - & - & [email protected] & - & [email protected] & [email protected] & 233@xmath4823 + eisrise3 & - & - & [email protected] & - & [email protected] & 32@xmath482 & 273@xmath4824 + eisdecay1 & - & - & [email protected] & - & 17@xmath483 & - & 1160@xmath74 + eisdecay2 & - & - & [email protected] & - & - & - & 880@xmath48250 + eiso4@xmath72 & - & - & [email protected]&- & [email protected] & [email protected] & 180@xmath74 + eiso4@xmath73 & - & - & [email protected]&- & [email protected] & [email protected] & 223@xmath4817 + is & - & - & 20@xmath482 & - & - & - & 700@xmath4840 + bs1 & [email protected] & [email protected] & 62@xmath485 & 42@xmath482 & - & 861@xmath486 & - + bs2 & [email protected] & [email protected] & 73@xmath485 & 45@xmath485 & - & - & - + bs3 & [email protected] & [email protected] & 76@xmath483 & 24@xmath483 & - & - & - + bs4 & [email protected] & [email protected] & 73@xmath484 & 16@xmath481 & - & - & - + bs5 & [email protected] & [email protected] & 65@xmath486 & 13@xmath481 & - & - & - + + eisrise1 & - & - & [email protected] & - & [email protected] & [email protected] & - + eisrise2 & - & - & [email protected] & - & [email protected] & [email protected] & [email protected] + eisrise3 & - & - & [email protected] & - & [email protected] & [email protected] & [email protected] + eisdecay1 & - & - & [email protected] & - & [email protected] & - & 19@xmath482 + eisdecay2 & - & - & [email protected] & - & - & - & [email protected] + eiso4@xmath72 & - & - & [email protected] & - & [email protected] & [email protected] & [email protected] + eiso4@xmath73 & - & - & [email protected] & - & [email protected] & [email protected] & [email protected] + is & - & - & [email protected] & - & - & - & 17@xmath482 + bs1 & [email protected] & [email protected] & @xmath81.3 & [email protected] & - & [email protected] & - + bs2 & [email protected] & [email protected] & @xmath81.2 & [email protected] & - & - & - + bs3 & [email protected] & [email protected] & @xmath81.4 & [email protected] & - & - & - + bs4 & [email protected] & [email protected] & @xmath81.1 & [email protected] & - & - & - + bs5 & [email protected] & [email protected] & @xmath81.0 & [email protected] & - & - & - + + eisrise1 & - & - & 0@xmath74 & - & 0@xmath74 & 0@xmath74 & - + eisrise2 & - & - & 0@xmath74 & - & [email protected] & [email protected] & [email protected] + eisrise3 & - & - & 0@xmath74 & - & [email protected] & [email protected] & [email protected] + eisdecay1 & - & - & 0@xmath74 & - & 0@xmath74 & - & 0@xmath74 + eisdecay2 & - & - & 0@xmath74 & - & - & - & 0@xmath74 + eiso4@xmath72 & - & - & [email protected] & -&[email protected]&[email protected] & 0@xmath74 + eiso4@xmath73 & - & - & [email protected] & -&[email protected]&[email protected] & [email protected] + is & - & - & [email protected] & -&- & - & [email protected] + bs1 & 0@xmath74 & 0@xmath74 & 0@xmath74 & 5@xmath74 & - & 7@xmath481&- + bs2 & 0@xmath74 & 0@xmath74 & [email protected] & 3.5@xmath74 & - & - & - + bs3 & 0@xmath74 & 0@xmath74 & [email protected] & [email protected] & - & - & - + bs4 & 0@xmath74 & 0@xmath74 & [email protected] & [email protected] & - & - & - + bs5 & 0@xmath74 & 0@xmath74 & [email protected] & [email protected] & - & - & - + +

we study the critical behavior and the ground - state entanglement of a large class of supersymmetric spin chains with a general ( not necessarily monotonic ) dispersion relation . we show that this class includes several relevant models , with both short- and long - range interactions of a simple form . we determine the low temperature behavior of the free energy per spin , and deduce that the models considered have a critical phase in the same universality class as a-dimensional conformal field theory ( cft ) with central charge equal to the number of connected components of the fermi sea . we also study the rnyi entanglement entropy of the ground state , deriving its asymptotic behavior as the block size tends to infinity . in particular , we show that this entropy exhibits the logarithmic growth characteristic of -dimensional cfts and one - dimensional ( fermionic ) critical lattice models , with a central charge consistent with the low - temperature behavior of the free energy . our results confirm the widely believed conjecture that the critical behavior of fermionic lattice models is completely determined by the topology of their fermi surface .

1610.03989

integrable spin chains often provide a fertile ground for studying key theoretical concepts in a simple framework that captures the essential features of the problems under consideration . an important example of this assertion is the analysis of the entanglement of a quantum system , which can be considered as one of the fundamental characteristics of the quantum realm @xcite . one of the most common ways of measuring the degree of entanglement of a state of a quantum system @xmath2 is via the bipartite entropy of a subsystem @xmath3 @xcite . this entropy is defined by @xmath4 , where @xmath5 is the reduced density matrix of the subsystem @xmath3 , @xmath6 is the density matrix representing the state of the whole system , and @xmath7 is an appropriate entropy functional ( von neumann , rnyi , etc . ) . the small class of models for which the entanglement entropy can be evaluated in closed form ( at least in the thermodynamic limit ) includes certain integrable spin chains , like the lipkin meshkov glick model @xcite , its @xmath8 generalization @xcite and the nearest - neighbors heisenberg @xmath9 and @xmath10 models @xcite . as is well known , the latter two models are critical ( gapless ) for a certain range of values of the applied magnetic field , their corresponding virasoro algebras having central charge respectively equal to @xmath11 and @xmath12 . in both cases , the bipartite rnyi entropy of a block of @xmath13 consecutive spins when the whole chain is in its ground state scales as @xmath14 in the critical phase , where @xmath15 is the rnyi parameter ( @xmath16 for the von neumann entropy ) and @xmath17 is the central charge . this behavior is consistent with the scaling of the rnyi entanglement entropy of a ( @xmath18)-dimensional conformal field theory ( cft ) @xcite . in fact , the logarithmic scaling of the ground state entanglement entropy is a characteristic feature of critical ( fermionic ) one - dimensional lattice models with short - range interactions ( see , e.g. , ref . @xcite ) . in a previous paper @xcite , we showed that the above results also apply to a large class of supersymmetric spin chains with general ( not necessarily short - range ) interactions , which turn out to be equivalent to a suitable free fermion model . the critical character of these chains ( for appropriate values of the chemical potential @xmath19 ) was ascertained via the analysis of the low - temperature behavior of the free energy per spin . indeed , we proved that when the dispersion relation @xmath20 of the corresponding free fermion model is monotonic in the interval @xmath21 $ ] , for @xmath22 the free energy per spin is approximately given ( in natural units @xmath23 ) by @xmath24 where @xmath25 is the fermi velocity ( or effective speed of `` sound '' ) and @xmath26 . this is precisely the expected behavior of the free energy for any critical model ( @xmath17 being the central charge of its virasoro algebra ) , since at low temperatures the free energy of a quantum system is determined by its lowest energy levels , and the free energy per spin of a ( @xmath18)-dimensional cft with central charge @xmath17 satisfies for sufficiently small @xmath27 @xcite . we also studied the ground - state rnyi entanglement entropy of the above mentioned supersymmetric spin chains , showing that in the thermodynamic limit @xmath28 it again behaves as that of a @xmath1-dimensional cft with central charge @xmath26 . the aim of this paper is to extend the results of ref . @xcite by suppressing the requirement that the dispersion relation be monotonic in @xmath21 $ ] . as shown in section [ sec.disp ] , this makes it possible to treat a host of naturally arising models , like supersymmetric spin chains with near and next - to - near interactions , or with long - range rational interactions , whose dispersion relation is not always monotonic . in fact , the entanglement entropy of fairly arbitrary energy eigenstates of one - dimensional free fermionic systems ( in particular , of the ground state of such systems with a non - monotonic dispersion relation ) has been previously studied in the literature ; see , e.g. , refs . @xcite . in general , the entanglement entropy of the ground state of these models grows logarithmically with the size @xmath13 of the subsystem , with a constant prefactor determined by the number of boundary points of the fermi `` surface '' in @xmath29 . this logarithmic scaling is a manifestation of the so - called `` area law '' , which is believed to hold for critical fermionic systems in an arbitrary number of dimensions @xcite . we shall show that the @xmath0 supersymmetric chains studied in this paper do indeed satisfy the area law . more precisely , by analyzing the low - temperature behavior of the free energy we shall first show that the models under consideration are critical for @xmath30 , where @xmath31 and @xmath32 respectively denote the minimum and maximum values of the dispersion relation . ( as explained in section [ sec.crit ] , strictly speaking this is only true if the roots of the equation @xmath33 are all simple . ) from the latter analysis it also follows that the central charge of these models is equal to the number @xmath34 of disjoint intervals that make up the fermi sea . we shall next study the ground state rnyi entanglement entropy , showing that in the thermodynamic limit @xmath28 it behaves as @xmath35 . we shall explicitly compute the ( non - universal ) constant @xmath36 , and prove that the prefactor @xmath37 is equal to @xmath38 . this is in agreement with the value of the central charge deduced from the low - temperature analysis of the free energy , and once again confirms the conjecture that the entanglement properties of critical fermion models are entirely determined by the topology of their fermi surface @xcite . we shall end this section with a few words on the paper s organization . in section [ sec.model ] we recall the definition of the supersymmetric chains under consideration and review their main properties . section [ sec.disp ] is devoted to the analysis of the models dispersion relation and the construction of simple examples of supersymmetric chains , featuring both short- and long - range interactions , with a non - monotonic dispersion relation . in section [ sec.crit ] we derive the asymptotic behavior of the models free energy per spin at low temperature , showing that they are critical in an appropriate range of the chemical potential , and determine the central charge of the corresponding virasoro algebra . the asymptotic behavior of the entanglement entropy of the models ground state is determined in section [ sec.gsee ] using a particular case of the fisher hartwig conjecture for toeplitz matrices @xcite rigorously proved by bttcher and silbermann @xcite . we briefly state our conclusions and outline several future developments suggested by the present work in section [ sec.conc ] . the paper ends with three appendices in which we present a review of the application of the fisher hartwig conjecture in the present context , as well as the proofs of several technical results used throughout section [ sec.gsee ] . the type of models we shall study in this work is the class of @xmath0 supersymmetric spin chains with translationally invariant interactions introduced in ref . @xcite . in the latter models each site is occupied either by a scalar boson or a spinless fermion , whose creation operators we shall respectively denote by @xmath39 and @xmath40 , the subindex @xmath41 indicating the site on which these operators act . thus the hilbert space is the @xmath42-dimensional subspace @xmath43 of the infinite - dimensional fock space defined by the constraints @xmath44 the hamiltonian of the models under consideration is given by to @xmath45 unless otherwise stated . ] @xmath46 where the operator @xmath47 is the total fermion number @xmath48 so that the real parameter @xmath19 has the natural interpretation of the fermions chemical potential . the real - valued function @xmath49 giving the strength of the interaction between two particles @xmath50 sites apart is assumed to satisfy the constraint @xmath51 but is otherwise arbitrary was implicitly assumed to be nonnegative for all @xmath52 in ref . @xcite . here we shall drop the latter requirement , which is not essential in what follows . ] . in other words , the chain is _ closed _ , i.e , translationally invariant . finally , @xmath53 is the @xmath0 spin permutation operator , defined by @xcite @xmath54 equivalently , let @xmath55 ( with @xmath56 ) be a state of the canonical spin basis , where @xmath57 and @xmath58 respectively denote the state with one boson or one fermion . the action of @xmath53 on the latter state is then given by @xmath59 where @xmath60 if @xmath61 while for @xmath62 @xmath63 equals the number of fermions in the state @xmath64 occupying the sites @xmath65 . the operator @xmath53 is clearly invariant under the supersymmetry transformation @xmath66 ( @xmath67 ) , and on @xmath43 we have @xmath68 , where @xmath69 is the total boson number . hence the hamiltonian is indeed supersymmetric - invariant , up to a constant term and the usual relabeling @xmath70 . the fundamental feature of the @xmath0 supersymmetric chain , explained in detail in refs . @xcite , is that it can be mapped into a free - fermion model by interpreting the boson state @xmath57 as the fermion vacuum . more precisely , consider the operators @xmath71 which can be regarded as a new set of fermion creation operators as they obviously satisfy the canonical anticommutation relations ( car ) on @xmath43 . it was shown by haldane @xcite that on @xmath43 the @xmath0 permutation operator @xmath53 can be simply expressed as @xmath72 substituting into eq . we readily obtain @xmath73 where @xmath74 we thus see that the spin chain is indeed equivalent to a free - fermion model with hopping amplitude @xmath75 and chemical potential @xmath76 . since the hamiltonian is translationally invariant on account of eq . , it can be diagonalized by the discrete fourier transform @xmath77 indeed , the operators @xmath78 obviously satisfy the car , and can therefore be considered as a new set of fermionic operators . moreover , a straightforward calculation shows that @xcite @xmath79 where @xmath80h_n(j)\,.\ ] ] likewise , the system s total momentum operator @xmath81 is given by @xcite @xmath82 with @xmath83 thus the operator @xmath84 creates a ( non - localized ) fermion with well - defined energy @xmath85 and momentum @xmath86 . it follows from eq . that the spectrum of @xmath87 is the set of numbers of the form @xmath88 with @xmath89 , whose corresponding eigenstates are given by @xmath90 an essential requirement making it possible to study the chain or , equivalently , its fermionic counterpart in the thermodynamic limit is the existence of a smooth function @xmath20 _ independent of @xmath45 _ such that when @xmath91 we have @xmath92 when this is the case , we shall refer to @xmath20 as the model s _ dispersion relation_. from the latter equation and the identity @xmath93 it follows that the dispersion relation is always symmetric about @xmath94 , namely @xmath95 likewise , @xmath96 implies that @xmath97 . it is also customary to extend @xmath20 to the whole real line as a @xmath98-periodic function , in which case eq . entails that @xmath99 . for instance , for the @xmath0 haldane shastry chain @xcite , whose interaction strength is given by @xmath100 it was shown in ref . @xcite that @xmath101 hence in this case holds with @xmath102 and no error term . in fact , it can be shown that eq . also holds ( again with no error term ) for a suitable dispersion relation @xmath103 in the more general chain with elliptic interactions studied in ref . @xcite . we shall next present a few relevant examples of models of the form for which the dispersion relation is guaranteed to exist . to this end , it is convenient to rewrite eq . to take into account conditions , namely @xmath104h_n(j)\\ + 2[1-\pi(n)]\,\pi(l)\,h_n(n/2)\,,\end{gathered}\ ] ] where @xmath105 denotes the parity of the integer @xmath50 and @xmath106 is the integer part of @xmath107 . clearly , the values of @xmath108 with @xmath109 appearing in the latter equation are no longer restricted by eq . . for this reason , from now on we shall implicitly restrict the domain of @xmath110 to the range @xmath109 , since for @xmath111 we simply have @xmath112 . in this vein , we shall say ( with a slight abuse of language ) that the interaction is _ independent of _ @xmath45 if there is a fixed function @xmath113 such that @xmath114 for @xmath109 . if this is the case we shall simply write @xmath115 , again implicitly assuming that we are restricting ourselves to the range @xmath109 . an important class of models of the form for which the dispersion relation @xmath20 is guaranteed to exist are those whose interaction strength @xmath110 is short - ranged and independent of @xmath45 . by this we mean that there is a positive integer @xmath116 ( the range of the interaction ) such that @xmath117 for @xmath118 , and @xmath119 with @xmath120 independent of @xmath45 and @xmath121 . obviously , in this case we have @xmath122\,.\ ] ] in fact , the same is true if we drop but assume instead that the limit @xmath123 exists for all @xmath124 . on the other hand , the short range of the interaction @xmath110 is by no means a necessary condition for the existence of the dispersion relation @xmath20 . indeed , suppose for simplicity that @xmath115 is independent of @xmath45 , and that the series @xmath125 is absolutely convergent . then clearly holds with @xmath126\,.\ ] ] for instance , for the power law interaction @xmath127 with @xmath128 the previous series can be summed in closed form in terms of the polylogarithm function @xcite @xmath129 namely ( taking , for simplicity , @xmath130 ) @xmath131 , \label{cepli}\end{gathered}\ ] ] where @xmath132 denotes riemann s zeta function ( cf . [ fig.epp ] ) . chain with power - law interaction @xmath133 for several values of the exponent @xmath134 between @xmath135 and @xmath136 . ] from the integral representation @xmath137 where @xmath138 is euler s gamma function , we obtain the equivalent expression @xmath139 using the latter formula in the identity @xmath140 and reversing the order of integration we arrive at the somewhat simpler expression @xmath141 remarkably , for @xmath142 eq . reduces to eq . ( see , e.g. , @xcite ) . thus the @xmath0 chain with rational interaction @xmath143 has the same dispersion relation as the haldane shastry chain . this is of course not entirely unexpected , since for _ fixed _ @xmath144 we have @xmath145 . note , however , that for @xmath146 both interactions , although negligibly small as @xmath91 , differ by a factor@xmath147 ( cf . fig . [ fig.hs-rat ] ) . hs chain ( red ) with the simple inverse - square law @xmath148 ( blue ) for @xmath149 . inset : same plot for the range @xmath150 . ] of course , although holds for a wide range of interesting interactions , it is not universally true . for instance , it is not satisfied by the @xmath45-independent interaction @xmath151 , since @xmath152 converges for @xmath153 while the series @xmath154 is divergent . in a previous paper @xcite we analyzed the critical behavior of supersymmetric spin chains of the type whose dispersion relation is monotonic in the interval @xmath21 $ ] . these models include the @xmath0 haldane shastry chain ( cf . ) and , more generally , its elliptic generalization introduced in ref . @xcite . as is apparent from fig . [ fig.epp ] ( and can be analytically checked differentiating eq . ) , the chain with power - law interactions also exhibits this property . however , this behavior is not universal , and there are in fact simple examples of supersymmetric chains of the form with a non - monotonic dispersion relation . indeed , consider to begin with the chain with nearest and next - to - nearest interactions , whose hamiltonian ( up to an irrelevant multiplicative constant ) is given by @xmath155 with @xmath156 , @xmath157 and @xmath158 . note that when @xmath159 the fermionic version of the latter model can be mapped to the ( closed ) heisenberg @xmath9 chain by a wigner jordan transformation @xcite . from eq . with @xmath160 and @xmath161 we easily obtain @xmath162 since @xmath163 the dispersion relation will have a critical point in @xmath164 if and only if @xmath165 ( more precisely , a maximum for @xmath166 and a minimum for @xmath167 ) . thus in this case the dispersion relation is not monotonic in @xmath21 $ ] provided that @xmath165 . the same is clearly true for chains of the form with interactions of finite range @xmath168 , for suitable values of the interaction strengths . it is also easy to construct simple examples of chains of the form with long - range interactions with a non - monotonic dispersion relation . take , for instance , @xmath169 whose dispersion relation is given by @xmath170\,.\ ] ] if @xmath171 , differentiating the latter equation we obtain @xmath172\,,\ ] ] with @xmath173 it can be shown ( cf . [ fig.ep23 ] ) that the function @xmath174 increases monotonically over the interval @xmath164 , with @xmath175 and @xmath176 . ] , so that @xmath177 changes sign once ( from positive to negative ) in @xmath164 if and only if @xmath178 . we conclude that the dispersion relation of the chain with interaction is not monotonic on @xmath21 $ ] provided that @xmath179 . in particular , for @xmath180 the dispersion relation is not monotonic in @xmath21 $ ] even if the interaction strength is positive for @xmath181 ( see fig . [ fig.ep23 ] for a plot of @xmath20 when @xmath182 ) . . inset : plot of the function @xmath174 in eq . . in this section we shall study the critical properties of the spin chain when its dispersion relation @xmath20 is not necessarily monotonic over the interval @xmath21 $ ] . to this end , we shall examine the low temperature behavior of the helmholtz free energy per spin @xmath183 which for this model is given by ( cf . @xcite ) @xmath184\,{\mathrm{d}}p\,.\ ] ] in the previous expressions @xmath185 denotes the partition function of the chain with @xmath45 spins , and @xmath186 ( in natural units @xmath23 ) . as remarked in the introduction , at low temperatures the free energy of a critical model should satisfy eq . . moreover , it was shown in ref . @xcite that when @xmath20 is monotonic and nonnegative in the interval @xmath21 $ ] the model is critical when the chemical potential @xmath19 lies in the interval @xmath187 , with central charge @xmath26 , and noncritical for @xmath19 outside the closed interval @xmath188 $ ] . we shall next extend this result to the more general case in which @xmath20 is not necessarily monotonic ( nor nonnegative ) in @xmath21 $ ] . to begin with , it is immediate to show that the model is not critical when @xmath19 lies outside the interval @xmath189 $ ] . indeed , suppose first that @xmath190 , so that @xmath191 , @xmath192 and @xmath193 in contradiction with the asymptotic behavior characteristic of a critical model . similarly , when @xmath194 we have @xmath195{\mathrm{d}}p\ ] ] and @xmath196\,{\mathrm{d}}p\\&<t{\mathrm e}^{-{\beta}(\mu-{{{\mathcal e}}_{\text{max}}})}\,,\end{aligned}\ ] ] again in disagreement with eq . . this conclusion is also borne out by the fact that when @xmath194 or @xmath190 the spectrum is clearly gapped , with energy gap respectively equal to @xmath197 or @xmath198 . let us now consider the more interesting case in which @xmath199 , in which the spectrum is clearly gapless . we shall suppose that the equation @xmath33 has @xmath200 roots @xmath201 in the interval @xmath164 , which we will assume to be _ we start by expressing the free energy as @xmath202 where @xmath203{\mathrm{d}}p\ ] ] and the last integral is extended to the subset of the interval @xmath21 $ ] defined by the inequality @xmath204 . clearly , as @xmath205 the main contribution to the integral in eq . comes from an increasingly small neighborhood of the `` turning points '' @xmath206 , near which @xmath207 is small . to exploit this fact , we choose @xmath208 small enough that @xmath209\cap[p_j-{\delta}p , p_j+{\delta}p]=\emptyset$ ] for @xmath210 and @xmath211 on @xmath212 $ ] this is certainly possible , since by hypothesis @xmath213 for all @xmath214 . obviously , @xmath215 depends only on the dispersion relation @xmath20 , and is therefore independent of @xmath27 . calling @xmath216-\cup_{i=0}^m[p_i-{\delta}p , p_i+{\delta}p]$ ] we have @xmath217{\mathrm{d}}p=\int_a\log\bigl[1+{\mathrm e}^{-{\beta}|{{\mathcal e}}(p)-\mu|}\bigr]{\mathrm{d}}p\\ + \sum_{i=0}^m\int_{p_i-{\delta}p}^{p_i+{\delta}p}\log\bigl[1+{\mathrm e}^{-{\beta}|{{\mathcal e}}(p)-\mu|}\bigr]{\mathrm{d}}p\,.\end{gathered}\ ] ] the first integral can be easily estimated . indeed , let @xmath218 be the minimum value of @xmath207 on the compact set @xmath219 , which is clearly positive since @xmath220 on @xmath219 , and denote by @xmath221 the length of @xmath3 . we then have @xmath222{\mathrm{d}}p{\leqslant}{\mathrm e}^{-a{\beta}}|a|{\leqslant}\pi { \mathrm e}^{-a{\beta}}\,,\ ] ] with @xmath218 ( and @xmath221 ) obviously independent of @xmath27 . consider next the integral @xmath223{\mathrm{d}}p\,.\ ] ] to analyze its low temperature behavior , we perform the change of variable @xmath224 separately in each of the intervals @xmath225 $ ] and @xmath226 $ ] . since @xmath213 , this change of variable is one to one and @xmath227 in both of the latter intervals , and we have @xmath228 the asymptotic behavior of these integrals as @xmath205 can be easily determined taking into account that by construction @xmath177 does not vanish on both intervals @xmath225 $ ] and @xmath226 $ ] , and therefore @xmath229 as @xmath230 implies that @xmath231 . since the integral @xmath232 is convergent we have @xmath233 where we have set @xmath234 moreover , if @xmath235 is independent of @xmath236 we have @xmath237 using this inequality and the integral @xmath238 in eq . we thus have @xmath239 from eqs . we finally obtain the asymptotic estimate @xmath240 this is the low temperature behavior of the free energy of a @xmath1-dimensional cft with @xmath34 free bosons with fermi velocities @xmath241 . thus in this case the model is critical , with central charge @xmath242 . the situation is markedly different if any of the roots of the equation @xmath33 is not simple . indeed , assume that @xmath243 is a root of order @xmath244 of the latter equation , so that we can write @xmath245\ ] ] with @xmath246 we now choose @xmath208 such that @xmath209\cap[p_j-{\delta}p , p_j+{\delta}p]=\emptyset$ ] for @xmath210 and @xmath211 on @xmath247 $ ] for all @xmath214 . proceeding as before we again arrive at eqs . - and obtain the estimate for the first integral in eq . . in order to analyze the low temperature behavior of the integral @xmath248 , we again perform the change of variable in each of the intervals @xmath249 $ ] and @xmath250 $ ] , thus obtaining eq . with @xmath251 . in each of the latter intervals we now have @xmath252\ ] ] and @xmath253\,,\ ] ] so that @xmath254\,.\ ] ] substituting into eq . and proceeding as before we easily obtain @xmath255 ( see ref . @xcite for more details on the evaluation of the last integral ) . thus at low temperatures the contribution of @xmath243 to the free energy , given by @xmath256 dominates over the @xmath257 contribution coming from the simple roots @xmath206 . moreover , since the coefficient of @xmath258 in eq . is always negative , this term can not be compensated by similar terms in eq . coming from other multiple roots . we thus conclude that when @xmath199 , but the equation @xmath33 has at least one multiple root , the model can not be critical . a similar analysis shows that this is also the case when @xmath259 or @xmath260 is smooth on @xmath21 $ ] , so that @xmath261 vanishes at the points where @xmath103 attains its maximum and minimum values over the latter interval . ] . this shows that the model is critical if and only if @xmath30 and all the roots of the equation @xmath33 are simple . when that is the case , the central charge of the model is equal to the number of connected components of its fermi sea ( or , equivalently , half the number of connected components of its fermi `` surface '' ) . thus , the universality class of the model depends exclusively on the topology of its fermi sea , which confirms the general assertion in ref . as mentioned in the introduction , one of the hallmarks of a critical fermionic lattice model in one dimension with short - range interactions is the logarithmic growth of its ground state bipartite entanglement entropy with the length @xmath13 of the block of spins considered . more precisely , let @xmath262 denote the rnyi entropy of the block when the whole chain is in its ground state @xmath263 , where @xmath264 . the expected behavior of @xmath265 in this type of models is then @xmath266 where @xmath17 is the central charge of the corresponding virasoro algebra and @xmath36 is a non - universal constant ( independent of @xmath13 ) . we showed in a previous paper @xcite that the latter formula is also valid for the supersymmetric chains when their dispersion relation is monotonic ( and nonnegative ) in the interval @xmath21 $ ] , even in the case of long - range interactions . in this section we shall extend this result to a general model of the type , whose dispersion relation need not be monotonic ( or nonnegative ) in @xmath21 $ ] . to this end , recall first of all that the ground state entanglement entropy @xmath265 can be expressed in terms of the eigenvalues of the ground state correlation matrix @xmath267 , with matrix elements @xmath268 indeed , it was shown in ref . @xcite that @xmath269 where @xmath270\ ] ] and @xmath271 $ ] are the eigenvalues of the matrix @xmath272 . the asymptotic behavior of @xmath265 can be determined following the method developed by jin and korepin @xcite for the @xmath9 model . to this end , for @xmath273 we define the complex - valued function @xmath274\,,\ ] ] where @xmath275}z$ ] and @xmath276 . this function has a logarithmic branch cut on the set @xmath277 and no other singularities on a sufficiently small open subset ( independent of @xmath278 ) containing the interval @xmath279 $ ] then for @xmath280 $ ] @xmath281 is real and satisfies @xmath282 . on the other hand , if @xmath283 then for @xmath284 with @xmath285 and @xmath280 $ ] integrating the derivative of @xmath286 along the segment from @xmath52 to @xmath287 we obtain @xmath288 thus @xmath289 for @xmath280 $ ] provided that @xmath290 . ] . by cauchy s theorem and eq . , if @xmath291 is the path sketched in fig . [ fig.path ] we therefore have @xmath292 where @xmath293 as explained in appendix [ app.fh ] , the latter integral can then be approximated using a proved case of the fisher hartwig conjecture to estimate the logarithmic derivative of @xmath294 . in eq . . ] in order to derive the asymptotic behavior of @xmath294 , we first need to determine the symbol of the toeplitz matrix @xmath295 ( see again appendix [ app.fh ] for the definition of the symbol and its calculation in two simple cases ) . we shall compute this symbol for a general model of the type , whose dispersion relation is not assumed to be monotonic over @xmath21 $ ] . more precisely , we shall only suppose that the equation @xmath33 has @xmath200 _ simple _ roots @xmath201 in the interval @xmath164 . from the symmetry of @xmath103 around @xmath94 ( cf . it then follows that the remaining roots of the equation @xmath33 in the interval @xmath296 are @xmath297 . in general , the system s ground state @xmath263 is determined by the conditions for all @xmath298 , so that the ground state is unique . ] @xmath299 \hat a_k{|\psi\rangle}&=0\,,\qquad { \varepsilon}_n(k)>\mu\ , , \end{aligned } \right.\ ] ] so that @xmath300 it immediately follows from eq . that the matrix elements of the correlation matrix @xmath267 are given by @xmath301 where the sum ranges over the set @xmath302 of integers in the range @xmath303 $ ] satisfying the condition @xmath304 . in the thermodynamic limit @xmath91 the latter formula becomes @xmath305 where the integral is extended to the subset of the interval @xmath296 defined by the inequality @xmath204 . in fact , by the @xmath98-periodicity of the integrand we can replace the interval @xmath296 by any interval of length @xmath98 , which we shall take as @xmath306 $ ] . let us suppose , for definiteness , that @xmath307 ( the case @xmath308 is dealt with similarly ) . from the simple nature of the roots @xmath309 , @xmath310 , it then follows that the subintervals of the interval @xmath311 on which @xmath312 is negative are @xmath313 with @xmath314 , and @xmath315 with @xmath316 . by eq . , the symbol of the toeplitz matrix @xmath317 is given by @xmath318 thus @xmath319 is piecewise constant and alternates between the two values @xmath320 and @xmath321 . the discontinuities of this symbol at the points @xmath322 ( with @xmath323 ) suggest the ansatz @xmath324 for suitable @xmath325 and @xmath326 . to verify this ansatz , we note that for @xmath327 ( with @xmath323 ) we have @xmath328 whereas for @xmath329 ( with @xmath330 ) @xmath331 and thus in either case @xmath332 comparing the latter formula with eq . we arrive at the system @xmath333 these equations easily imply that @xmath334 is an integer multiple of @xmath98 for @xmath335 . we shall take @xmath336 , so that calling @xmath337 we have @xmath338 from the equations with @xmath339 and @xmath340 we then obtain @xmath341 so that we can take @xmath342 finally , from the equation with @xmath340 we have @xmath343{\mathrm e}^{-2{\mathrm i}{\beta}\sum_{k=0}^m(-1)^kp_k}\\ = \big[{\lambda}+(-1)^{m}\big]\bigg(\frac{{\lambda}+1}{{\lambda}-1}\bigg)^{-\sum_{k=0}^m(-1)^k\frac{p_k}\pi}\,,\end{gathered}\ ] ] which can also be written as @xmath344 where @xmath345 and @xmath346 denotes the parity of @xmath347 . it is easy to check that with this choice of @xmath325 and @xmath326 eqs . are all satisfied . since eq . coincides with the first eq . , as explained in appendix [ app.fh ] , the condition @xmath348 is satisfied , so that we can apply the fisher hartwig conjecture to estimate @xmath349 . to this end ( using the notation in appendix [ app.fh ] ) , note first of all that @xmath350 and , by eq . , @xmath351 moreover , from eq . it easily follows that @xmath352^{2(m+1)}\ ] ] and @xmath353^{4(-1)^{i+j}{\beta}^2}\\ & \times \prod_{0{\leqslant}j < i{\leqslant}m}\big[2\sin\bigl(\tfrac{p_i+p_j}2\bigr)\big]^{-4(-1)^{i+j}{\beta}^2}\,.\end{aligned}\ ] ] equation and the fisher hartwig conjecture thus yield the asymptotic formula @xmath354^{-2{\beta}^2 } ( { \lambda}+1)^l\bigg(\frac{{\lambda}+1}{{\lambda}-1}\bigg)^{\!\!-lp}\\\label{dlgen } \times \big[g(1+{\beta})g(1-{\beta})\big]^{2(m+1)}\big(1+o(1)\big)\,,\end{gathered}\ ] ] with @xmath355^{(-1)^{i+j}}\end{gathered}\ ] ] independent of @xmath13 and @xmath356 . we shall next use the approximate formula and eq . to derive an asymptotic formula for the rnyi entanglement entropy of the ground state of a general model of the form in the limit @xmath28 . first of all , from eq . we easily obtain @xmath357\ , , \label{dlapprox}\end{gathered}\ ] ] with @xmath358\\= 1+{\gamma}_e+\sum_{n=1}^\infty\frac{{\beta}^2/n}{n^2-{\beta}^2}\end{gathered}\ ] ] ( cf . . in fact , the dominant term ( proportional to @xmath13 ) in the previous expression does not contribute to eq . , since by cauchy s residue theorem we have @xmath359 thus eqs . - yield @xmath360{\mathrm{d}}{\lambda}\,.\end{gathered}\ ] ] moreover , it is straightforward to verify that the integral along the circular arcs of @xmath291 vanishes identically , since each of these arcs is mapped to the opposite of the other by the transformation @xmath361 , and the integrand changes sign under the latter mapping ( cf . eqs . , and ) . we thus obtain @xmath362{\mathrm{d}}{\lambda}\,.\end{gathered}\ ] ] in order to evaluate these integrals , we note that along the segments @xmath363 with @xmath364 we have @xmath365 so that @xmath366 on the other hand , @xmath367 is negative for sufficiently small @xmath368 , while @xmath369 tends to @xmath370 as @xmath371 and has the same sign as @xmath372 , so that @xmath373}w=\mp\pi\,.\ ] ] we thus have @xmath374 \equiv -{\mathrm i}b(x)\mp\frac12\,,\ ] ] with @xmath375 from eq . it immediately follows that @xmath376i_1({\alpha})\\+(m+1)i_2({\alpha})\,,\end{gathered}\ ] ] with @xmath377 { \mathrm{d}}x\ , . \label{i2}\end{aligned}\ ] ] the value of the integral @xmath378 can be deduced from ref . @xcite ( cf . also @xcite ) , namely @xmath379 ( see appendix [ app.int ] for an elementary derivation of the latter formula ) . we thus obtain @xmath380 where @xmath381 { \mathrm{d}}x\ , . \label{calfinal}\end{gathered}\ ] ] comparing with eq . , we see that the ground - state rnyi entanglement entropy of the model behaves as that of a critical system with central charge @xmath242 , as expected . moreover , the constant term @xmath36 is given in this case by @xmath382 where the first term is model dependent ( it depends on @xmath19 and @xmath20 through the momenta @xmath206 ) , while @xmath383 is a universal constant ( independent of @xmath13 and @xmath206 ) characteristic of the class of models under consideration . it is shown in appendix [ app.cal ] that @xmath383 in eq . can be expressed as @xmath384{\mathrm{d}}x\,,\ ] ] where @xmath385 is the digamma function . in particular , eq . implies that @xmath383 coincides with the function @xmath386 defined in eq . ( 64 ) of ref . since for @xmath387 we have @xmath388 , eq . yields the formula derived in ref . @xcite for the rnyi entanglement entropy of the model when its dispersion relation is monotonic over the interval @xmath21 $ ] ( which , as explained in the latter reference , includes the @xmath9 model studied in ref . in fact , using the ideas of ref . @xcite eq . can be written in the simpler form @xmath389 ( see appendix [ app.cal ] for details ) . from the previous expression it is straightforward to evaluate @xmath383 numerically for any specific value of the rnyi parameter @xmath390 [ fig.cal ] . it can be numerically verified that @xmath383 vanishes for @xmath391 , and attains its maximum value @xmath392 for @xmath393 ( cf . [ fig.cal ] ) . in eq . ] it also follows from eq . that @xmath394 as @xmath395 , and that when @xmath396 @xmath383 tends to a finite ( nonzero ) limit , given by @xmath397 taking the @xmath398 limit in the previous formulas we obtain the following asymptotic expression for the von neumann entropy @xmath399 : @xmath400 where @xmath401 and the universal constant @xmath402 is given by @xmath403 note , in particular , that the latter equation agrees with the formula for the analogous constant @xmath404 in ref . @xcite . the formula - ( or its counterpart for the von neumann entropy ) provides an excellent approximation to the ground - state rnyi entanglement entropy of the supersymmetric chain for even moderately large values of @xmath13 . as an example , in fig . [ fig.error ] we have represented the relative error @xmath405 , where @xmath406 is the approximation to the von neumann entropy @xmath7 , for the finite - range chain in the case @xmath407 and @xmath408 . the value of @xmath7 has been numerically computed diagonalizing the correlation matrix @xmath267 and using the exact formula ( with @xmath16 ) . as explained in section [ sec.disp ] , for the value of @xmath409 considered the dispersion relation has exactly one maximum in the interval @xmath164 , and hence is not monotonic . in particular , for @xmath410 the fermi sea consists of two disjoint intervals , as is also apparent from the inset in fig . [ fig.error ] . as can be seen from the latter figure , the relative error decreases ( though not monotonically ) from @xmath411 to @xmath412 when @xmath13 ranges from @xmath413 to @xmath414 . of the approximation @xmath406 in the rhs of eq . to the von neumann ground - state entanglement entropy of the chain with @xmath407 and @xmath408 as a function of the block length @xmath13 . inset : dispersion relation of the chain with the latter values of @xmath409 and @xmath19 . ( the interval @xmath415 is in this case is @xmath416 , @xmath417 , @xmath418 and @xmath419 . ) ] in this paper we have analyzed the critical behavior of a large class of supersymmetric spin chains whose dispersion relation @xmath20 is not assumed to be monotonic in the interval @xmath21 $ ] . we have examined the conditions under which the dispersion relation is well - defined ( i.e. , is a continuous function ) in the thermodynamic limit , providing several simple examples of models of this type , with both short- and long - range interactions , whose dispersion relation is not monotonic . the main conclusion of our work is that the criticality properties of the supersymmetric chains are determined exclusively by the topology ( the number of points ) of their fermi `` surface '' . more precisely , through the analysis of the free energy per spin in the critical ( gapless ) phase , we have shown that these models are equivalent to a system of @xmath34 free bosons with fermi velocities @xmath420 , where @xmath421 are the points of the fermi surface in the interval @xmath21 $ ] . in particular , the central charge is equal to the number @xmath34 of connected components ( intervals ) of the fermi sea . this result is corroborated by the asymptotic behavior of the ground - state rnyi entanglement entropy @xmath265 as the block size @xmath13 tends to infinity , which has been derived applying a proved case of the fisher hartwig conjecture . indeed , we have shown that @xmath422 , where @xmath36 is a nonuniversal constant ( independent of @xmath13 ) which we have computed in closed form in terms of the momenta @xmath421 . in particular , for large @xmath13 the entanglement entropy exhibits the logarithmic growth characteristic of @xmath1-dimensional conformal field theories with central charge @xmath242 . this behavior , which is typical of critical ( fermionic ) one - dimensional lattice models with short - range interactions ( see , e.g. , @xcite ) , was recently established by the authors for supersymmetric spin chains of the type considered here under the assumption that the dispersion relation is monotonic in @xmath21 $ ] . the present work opens up several possible lines for future research . in the first place , one could consider a generalization of our results on the ground - state entanglement entropy to more general situations ( for instance , considering excited states , as in ref . @xcite ) , in which the fermi sea is not necessarily a finite union of disjoint intervals but exhibits a more complicated topological structure . another interesting generalization of the present work is the analysis of the entanglement of a subset consisting of the union of two or more disjoint blocks . in fact , the entanglement entropy of this type of subsystems has already been discussed in ref . @xcite , giving rise to an unproved conjecture on the asymptotic behavior of the determinant of a block toeplitz matrix . the authors would like to thank p. tempesta for several enlightening discussions . this work was partially supported by spain s mineco under research grant no . fis2015 - 63966-p . jac would also like to acknowledge the financial support of the universidad complutense de madrid through a 2015 predoctoral scholarship . in this appendix we shall briefly review the fisher hartwig conjecture on the asymptotic behavior of the determinant of a toeplitz matrix when its order tends to infinity . ( recall that a matrix @xmath27 is toeplitz if its matrix elements @xmath423 depend only on @xmath424 . ) if @xmath425 is a ( complex - valued ) function defined on the unit circle @xmath426 , we define its fourier coefficients @xmath427 ( @xmath428 ) by @xmath429 note that the last integral can in fact be extended to any interval of length @xmath98 , by the @xmath98-periodicity of the integrand . for any @xmath430 , the function @xmath431 defines a toeplitz matrix @xmath432 of order @xmath13 through the relation @xmath433 we shall say that the function @xmath17 is the _ symbol _ of the toeplitz matrix @xmath432 . the fisher hartwig conjecture applies to matrices @xmath432 whose symbol satisfies certain requirements that we shall next describe . more precisely @xcite , @xmath17 should be of the form @xmath434^{{\alpha}_r}\ , , \label{cz}\ ] ] where @xmath435 , @xmath436 is a nonvanishing smooth function with zero winding number and @xmath437 note that @xmath438 has in general ( i.e. , unless @xmath236 is an integer ) a single jump discontinuity at @xmath439 . if @xmath17 satisfies eq . , we denote by @xmath440 ( @xmath428 ) the @xmath63-th fourier coefficient of @xmath441 ( which is well defined and smooth , from the smoothness of @xmath325 and the assumption on its winding number ) , and define @xmath442 it is immediate to show that @xmath443 ( resp . @xmath444 ) can be analytically prolonged to the interior ( resp . exterior ) of the unit circle . it also follows from the definition of @xmath445 that on the unit circle we have the _ wiener hopf decomposition _ @xmath446 let us further set @xmath447=\exp\bigg(\sum_{n=1}^{\infty}n\,l_nl_{-n}\bigg)\ ] ] and @xmath448\prod_{r=1}^rb_+\bigl({\mathrm e}^{{\mathrm i}{\theta}_r}\bigr)^{{\beta}_r-{\alpha}_r}b_-\bigl({\mathrm e}^{{\mathrm i}{\theta}_r}\bigr)^{-{\alpha}_r-{\beta}_r}\\ & \hphantom{={}}\times\prod_{1{\leqslant}s\ne r{\leqslant}r}\big(1-{\mathrm e}^{{\mathrm i}({\theta}_s-{\theta}_r)}\big)^{({\alpha}_r+{\beta}_r)({\beta}_s-{\alpha}_s)}\\ & \hphantom{={}}\times\prod_{r=1}^r\frac{g(1+{\alpha}_r+{\beta}_r)g(1+{\alpha}_r-{\beta}_r)}{g(1 + 2{\alpha}_r)}\,,\end{aligned}\ ] ] where the barnes @xmath449-function is the entire function defined by @xmath450\ ] ] and @xmath451 is the euler mascheroni constant . the fisher hartwig conjecture states @xcite that if @xmath432 is the toeplitz matrix with symbol then when @xmath28 we have @xmath452 with @xmath453 the above conjecture was actually proved by bttcher and silbermann @xcite in the case @xmath454 which , as we shall see below , is the relevant one for our purposes . in fact , as explained in section [ sec.gsee ] , we shall only need to consider the case in which @xmath455 for all @xmath116 and @xmath325 is a constant ( i.e. , independent of @xmath456 ) . the fisher hartwig conjecture simplifies considerably in this case , since @xmath457=1\,,\quad{\mathrm e}^{l_0}=b\,,\ ] ] and therefore @xmath458 with @xmath459 and @xmath460 note also that the product @xmath461 reduces to @xmath462\ , . \label{gsimp}\ ] ] we shall be mainly interested in the case in which @xmath463 , where @xmath356 is a spectral parameter and @xmath267 is the correlation matrix of a block of @xmath13 spins of the @xmath0 supersymmetric spin chain . as a first example , let us express in the form - the symbol of the matrix @xmath432 when the chain s dispersion relation is monotonic in the interval @xmath21 $ ] . to begin with , in this case we have @xmath464 where @xmath465 $ ] is the fermi momentum @xcite . thus the symbol of the toeplitz matrix @xmath267 is @xmath466 and that of @xmath432 is therefore given by @xmath467 note that @xmath17 has two jump discontinuities on the unit circle at the points @xmath468 . we shall next show that @xmath469 for suitable @xmath236 and @xmath470 . indeed , first of all we have @xmath471 on the other hand , if @xmath472 then @xmath473 while for @xmath474 we have @xmath475 hence @xmath476 in order for eq . to hold we must therefore have @xmath477 from which we easily get @xmath478 although these equations admit an infinite number of solutions @xmath479 provided that @xmath480 , it will prove convenient for our purposes to take @xmath481 where @xmath275}z$ ] and @xmath276 . note , in particular , that @xmath325 is a nonvanishing constant . it is also important to observe that @xmath482\implies|{\operatorname{re}}{\beta}|=\frac1{2\pi}\arg_{(-\pi,\pi]}\bigg(\frac{{\lambda}+1}{{\lambda}-1}\bigg)<\frac12\,,\ ] ] since by definition @xmath483}z{\leqslant}\pi$ ] and @xmath484}\bigg(\frac{{\lambda}+1}{{\lambda}-1}\bigg)=\pi&\iff \frac{{\lambda}+1}{{\lambda}-1}\in(-\infty,0)\\&\iff{\lambda}\in(-1,1)\,.\end{aligned}\ ] ] thus the fisher hartwig conjecture can be applied provided that @xmath356 lies outside the closed interval @xmath279 $ ] , with @xmath485 , @xmath486 and @xmath487 by eqs . and , when @xmath28 the characteristic polynomial @xmath488 is given by @xmath489 with @xmath236 defined by eq . this is precisely the formula used by jin and korepin @xcite for the determination of the asymptotic behavior of the ground state entanglement entropy of the @xmath9 model . as a second example , we shall consider a simple case in which the dispersion relation @xmath103 is not monotonic in @xmath21 $ ] . more precisely , suppose that @xmath103 is nonnegative and has a single maximum in the open interval @xmath164 ( cf . [ fig.disp ] ) . with a single maximum in @xmath164 . ] for @xmath490 , the equation @xmath33 has one root @xmath491 in the interval @xmath164 , and the determinant @xmath294 is approximately given by eq . . we shall next determine the asymptotic behavior of @xmath294 when @xmath492 , where @xmath32 is the maximum value of @xmath20 , and thus the equation @xmath33 has two roots @xmath493 in the interval @xmath164 . to begin with , from eq . it follows that in this case @xmath494 thus the symbol of @xmath317 is given by @xmath495 in other words , @xmath496 alternatively takes on the two values @xmath320 and @xmath321 on each of the four intervals @xmath497 on which @xmath312 has constant sign , starting with @xmath320 . since the symbol has four discontinuities at the points @xmath468 and @xmath498 , we shall try to express it as @xmath499 for suitably chosen @xmath325 , @xmath500 . in fact , we only need compute @xmath501 , which is straightforward : @xmath502 combining the previous equations with eq . and comparing with eq . we immediately arrive at the system @xmath503 from the first equation it follows that @xmath504 must be an integer . choosing the simplest solution @xmath505 and dividing the last equation by the first one we again obtain eq . for @xmath236 . finally , from the last equation it follows that @xmath506 note that @xmath236 is still given by eq . , so the condition @xmath348 , necessary for the validity of the fisher hartwig conjecture , also applies in this case if @xmath507 $ ] . since now @xmath508 , @xmath509 and @xmath510 by the fisher hartwig conjecture , the determinant @xmath294 is given in this case by @xmath511^{\!\!-2{\beta}^2}\notag\\ & \hphantom{={}\bigg(}\times({\lambda}+1)^l\bigg(\frac{{\lambda}+1}{{\lambda}-1}\bigg)^{\frac l\pi(p_1-p_0-\pi)}\notag\\ & \hphantom{={}\bigg(}\times g(1+{\beta})^4g(1-{\beta})^4\big(1+o(1)\big)\ , . \label{dlp0p1}\end{aligned}\ ] ] in this appendix we shall provide an elementary derivation of the integral @xmath378 in eq . , which appears in the asymptotic expression of the rnyi entanglement entropy of the model . to begin with , we have @xmath512 with @xmath513\,\frac{{\mathrm{d}}x}{1-x^2}\,,\ ] ] or equivalently ( performing the change of variables @xmath514 ) @xmath515\frac{{\mathrm{d}}t}t\,.\ ] ] integrating by parts we obtain the equivalent expression @xmath516 on the other hand , differentiating eq . with respect to @xmath15 we easily get @xmath517 and hence @xcite @xmath518 solving this linear differential equation with the initial condition @xmath519 we finally obtain @xmath520 which immediately yields eq . . in this appendix we derive eq . for the constant @xmath383 in eq . , and show that it can be more simply expressed by means of the integral . to this end , we use the elementary identity @xmath521 which can be immediately derived from the well - known infinite product for the gamma function . our starting point is the definition of @xmath522 , which can be written as @xmath523 { \mathrm{d}}x\,,\ ] ] with @xmath524 and @xmath525 from the relation @xmath526 and the functional identity @xmath527 satisfied by the digamma function it immediately follows that @xmath528-z{\gamma}_e-\frac12\,,\ ] ] and therefore @xmath529\,.\ ] ] substituting into eq . and using eqs . and we easily obtain eq . . in order to prove eq . , we first make the change of variable @xmath530 in eq . , which yields @xmath531\\&\kern9em\times\big[\psi\bigl(\tfrac12+{\mathrm i}\,w\bigr ) + \psi\bigl(\tfrac12-{\mathrm i}\,w\bigr)\big]\,{\mathrm{d}}w\\ & = \frac2{{\mathrm i}\pi}\int_0^\infty\log\biggl[\frac{2\cosh(\pi{\alpha}w)}{\big(2\cosh(\pi w)\big)^{\alpha}}\biggr]\frac{{\mathrm{d}}}{{\mathrm{d}}w}\log\bigg [ \frac{{\gamma}\bigl(\tfrac12-{\mathrm i}\,w\bigr)}{{\gamma}\bigl(\tfrac12+{\mathrm i}\,w\bigr)}\bigg]\,{\mathrm{d}}w\,.\end{aligned}\ ] ] we next integrate by parts , taking into account that by stirling s formula we have @xmath532 = o(w\log w)\ ] ] while @xmath533=o\bigl({\mathrm e}^{-2\pi\min(1,{\alpha})w}\bigr)\,,\ ] ] so that the boundary term vanishes . we thus obtain @xmath534\log\bigg [ \frac{{\gamma}\bigl(\tfrac12-{\mathrm i}\,w\bigr)}{{\gamma}\bigl(\tfrac12+{\mathrm i}\,w\bigr)}\bigg]{\mathrm{d}}w\,.\end{gathered}\ ] ] on the other hand , from gauss s integral representation of the digamma function @xcite @xmath535 it easily follows that @xmath536\frac{{\mathrm e}^{-t}}t\,{\mathrm{d}}t\ ] ] and hence @xmath532 = { \mathrm i}\int_0^\infty\bigg[{\operatorname{csch}}(t/2)\sin(wt)-2w{\mathrm e}^{-t}\bigg]\frac{{\mathrm{d}}t}t\,.\ ] ] substituting into the last formula for @xmath383 and using the elementary identity @xmath537 we obtain @xmath538 with @xmath539{\mathrm{d}}w.\ ] ] the latter integral can be evaluated in closed form by elementary means . indeed , @xmath540 while @xmath541 and therefore @xmath542 we thus obtain @xmath543 from which eq . easily follows . 32ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop _ _ , ed . ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) in @noop _ _ , , vol . , ( ) pp . @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) , , , and , eds . , @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , )

we present results from multiwavelength observations of the galaxy ngc5005 . we use new neutral hydrogen ( hi ) observations from the very large array to examine the neutral gas morphology and kinematics . we find an hi disc with a well - behaved flat rotation curve in the radial range 20 - 140 . ionized gas observations from the sparsepak integral field unit on the wiyn 3.5 m telescope provide kinematics for the central 70 . we use both the sparsepak and hi velocity fields to derive a rotation curve for ngc5005 . deep 3.6 m observations from the _ spitzer space telescope _ probe the faint extended stellar population of ngc5005 . the images reveal a large stellar disk with a high surface brightness component that transitions to a low surface brightness component at a radius nearly 1.6 times farther than the extent of the gas disk detected in hi . the 3.6 m image is also decomposed into bulge and disk components to account for the stellar light distribution . optical broadband b and r and narrowband h from the wiyn 0.9 m telescope complement the 3.6 m data by providing information about the dominant stellar population and current star formation activity . the neutral and ionized gas rotation curve is used along with the stellar bulge and disk light profiles to decompose the mass distributions in ngc5005 and determine a dark matter halo model . the maximum stellar disk contribution to the total rotation curve is only about 70% , suggesting that dark matter makes a significant contribution to the dynamics at all radii .

1503.05981

understanding the role and significance of dark matter in the evolution of baryonic components ( i.e. , conversion of the gaseous disk into stars ) is a critical aspect for realistic models of galaxy evolution . observational studies aimed at addressing the influence of dark matter in galaxy formation and evolution often use mass decompostion to constrain the distribution of dark matter . many of these projects utilize neutral hydrogen ( hi ) synthesis observations to derive the kinematics for rotation curve decompostion ( i.e. , the hi nearby galaxy survey ( things ) ; @xcite , westerbork survey of hi in spiral galaxies ( whisp ) ; @xcite ) . in this paper , we take advantage of the well - behaved outer disk kinematics in the galaxy ngc5005 to perform a mass decomposition analysis . we use radio synthesis observations of hi and co to probe both the gas surface density and dynamics as the gas follows the gravitational potential of the galaxy . the stellar contribution to the overall mass budget of the galaxy is determined from deep near - infrared ( nir ) observations . finally , we use optical broadband b and r and narrowband h@xmath1 observations to supplement the nir and radio synthesis data to provide information about the dominant stellar populations and star formation activity . large projects like things and whisp have provided a foundation for numerous other studies investigating gas kinematics and morphology ( e.g. , @xcite , @xcite , @xcite ) and dark matter haloes ( e.g. , @xcite , @xcite ) . however , an examination of the statistical significance of connections between the dark matter and baryonic properties of these galaxies may be biased due to the representative nature of the samples . in an effort to address fundamental questions regarding the growth and distribution of stellar disks in dark matter haloes in a statistical manner , we have undertaken a project correlating structural properties and star formation activity with the dark matter properties of the host galaxy . the project uses a statistical sample of 45 nearby galaxies and builds on existing data obtained as part of the extended disk galaxy exploration science ( edges ) survey ( @xcite ) , which includes deep _ spitzer space telescope _ imaging observations . the statistical sub - sample consists of all galaxies in the edges survey that are optimally suited for rotation curve decomposition , based primarily on the limited range of inclination angles that yield both accurate rotation curves and surface density profiles . our combination of optical , nir , and radio synthesis observations will enable a comprehensive statistical analysis of the evolution of galactic disks as a function of baryonic mass , environment , and dark matter mass . in this paper , we focus our efforts on a single galaxy from the sample , ngc5005 . this sab(rs)bc type galaxy has been studied extensively in the literature ( e.g. , @xcite , @xcite , @xcite , @xcite ) . it is known to host a low - ionization nuclear emission - line region ( liner ) in its center ( @xcite ) and has been classified as a low - luminosity active galactic nucleus ( llagn ) with a polycyclic aromatic hydrocarbon ( pah)-dominated mid - infrared spectrum by @xcite . the morphology of ngc5005 as viewed in the infrared and optical features a strong bulge with boxy isophotes ( @xcite ) and a complicated dust morphology with no coherent structure ( @xcite ) . there is also evidence of a fairly weak , round bar which is nearly aligned with the major axis of the galaxy . despite the weakness of the bar , @xcite found that it could still be responsible for creating shocks visible in the central molecular gas velocity field . the molecular gas dynamics in ngc5005 were studied in more detail by @xcite . their observations reveal co emission from the nucleus , a ring at about 3 kpc , and a stream of gas northwest of the nucleus . the s - shape in the nuclear velocity field in @xcite is a signature of non - circular gas motions . @xcite cite a stellar bar of length @xmath25 kpc as being accountable for infalling gas seen as the northwest stream and for creating the 3 kpc ring which may form at the ultraharmonic resonance of the bar or alternatively constitute a pair of spiral arms that originate at the bar s ends . a position - velocity cut along the major axis of ngc5005 reveals a steep rise in the rotational velocity of the co emission ( their figure 5 ) . the rotational velocity rises to approximately 300 km s@xmath3 within the central 10 . ngc5005 lies in a relatively populated environment that is near the ursa major cluster . the distance to ngc5005 is uncertain ; distance measurements range from 13.7 mpc to 34.6 mpc . in this paper , we adopt the type ia supernova distance of 16.5 mpc ( @xcite ) . ngc5005 is thought to be part of a physical galaxy pair with the nearby spiral galaxy ngc5033 ( @xcite ) . the two galaxies are separated by a projected distance of approximately 41 ( @xmath2200 kpc at 16.5 mpc ) and are at similar redshifts . there are no clear signs of any interaction between the galaxy pair . there are two dwarf galaxies , sdss j131115.77 + 365911.4 and sdss j131051.05 + 365623.4 , that are within 10 of ngc5005 and also lie at similar redshifts . farther out , there are many additional dwarf galaxies within one degree of ngc5005 which are part of the ngc5033 group . the goal of the present work is to investigate the dark matter and baryons in ngc5005 . ngc5005 is also a proof of concept for the larger study . the observational data are presented in section [ obsdata ] and we focus on the kinematics in section [ kin ] . section [ dmdecomp ] describes the rotation curve decomposition analysis and the determination of the dark matter distribution in ngc5005 . section [ conclusion ] provides a brief summary of the results . multiwavelength observations have been acquired to probe the stellar and gas content in ngc5005 . deep nir images taken at 3.6@xmath0 m from the _ spitzer space telescope _ trace the extended stellar population , while moderate depth optical broadband b and r and narrowband h@xmath1 provide information about the dominant stellar population and star formation activity . radio synthesis observations of neutral hydrogen from the very large array ( vla ) are used to probe both the gas surface density and kinematics . spectroscopic integral field unit ( ifu ) observations from sparsepak on the wiyn 3.5 m telescope provide ionized gas kinematics in the central region of ngc5005 . figure [ n5005 ] shows the new vla hi observations along with optical and nir imaging data . table [ intprops ] provides a summary of integrated properties derived from this multifrequency dataset . we describe the data acquisition and processing below . lc distance & 16.5 mpc + morphological type & sab(rs)bc + m@xmath4 & [email protected] + m@xmath6 & [email protected] + m@xmath7 & [email protected] + m@xmath4 & [email protected] + ( b - r)@xmath8 & [email protected] + ( b-3.6)@xmath8 & [email protected] + ( r-3.6)@xmath8 & [email protected] + @xmath9(h@xmath1 flux ) & [email protected] erg s@xmath3 @xmath10 + equivalent width & [email protected] + total sfr & [email protected] m@xmath11 year@xmath3 + total stellar mass & ( [email protected])@xmath1210@xmath13 m@xmath11 + hi flux & [email protected] jy km s@xmath3 + total hi mass & ( [email protected])@xmath1210@xmath14 m@xmath11 + co flux & 1278@xmath5484 jy km s@xmath3 + total h@xmath15 mass & ( [email protected])@xmath1210@xmath14 m@xmath11 + total dynamical mass & ( [email protected])@xmath1210@xmath16 m@xmath11 + radio synthesis observations of ngc5005 using the 21 cm line of hi were obtained with the vla in c configuration on 2013 august 4 and 5 for a total on - source time of 407 minutes . the data had an initial frequency resolution of 7.8125 khz channel@xmath3 ( @xmath21.65 km s@xmath3 channel@xmath3 ) . the standard flux calibrator 3c286 was observed at the beginning and end of each observing block , and the phase calibrator j1227 + 3635 was observed approximately every 40 minutes , so that the data may be flux and phase calibrated . in order to flag uniformly , the data were loaded into aips and the inner 75% of the bandpass was combined to generate a `` channel zero '' dataset . the channel zero data was flagged and phase calibrated . the calibration solutions were applied to the line data , which was then bandpass calibrated with 3c286 . after calibrations , the data were continuum subtracted in the @xmath17 plane and then combined . the combined , continuum subtracted datasets were imaged using the aips task imagr . three cubes were created with different robust weighting parameters for varying spatial resolution and a channel averaging of 3 for a resulting velocity resolution of @xmath25 km s@xmath3 channel@xmath3 ( for simplicity , these cubes are designated as low , medium , and high ; table [ cubeprops ] ) . only the cubes with the highest sensitivity ( low ) and highest resolution without severe loss of sensitivity ( high ) are shown in figure [ n5005 ] . a lower velocity resolution cube ( 15 km s@xmath3 channel@xmath3 ) was additionally created to probe the lowest column density of hi gas . this velocity resolution was chosen to match the channel width to the expected line width from thermal gas motions . this cube was also smoothed spatially post - imaging to increase the sensitivity to low column density gas . in addition to the new vla c configuration data , there are existing archival vla d configuration data taken on 2000 august 11 for a total on - source time of 80.3 minutes ( project code aw536 ) . these archival data were retrieved from the archive and processed in aips . properties of the imaging of this data set are presented in table [ cubeprops ] . the archival d configuration data was not combined with the new c configuration data due to the lower velocity resolution and the smaller frequency coverage of the archival data . the galaxy signal nearly fills the bandpass in the d configuration observations . despite this , the integrated hi flux and hi mass measured from the archival data is consistent with the new c configuration values within the measurement errors . both this d configuration data and the spatially smoothed , 15 km s@xmath3 channel@xmath3 velocity resolution cube were used to confirm the lack of detection of low column density gas in the higher resolution cubes . lccccc low & 5.0 & 5 & [email protected] & 64.2 & 0.57 + medium & 5.0 & 0.5 & [email protected] & 67.0 & 0.61 + high & 5.0 & -0.5 & [email protected] & 72.6 & 0.72 + smooth & 14.9 & 5 & [email protected] & 64.2 & 0.53 + d config & 10.4 & 5 & [email protected] & -45.2 & 0.96 + the total intensity map and velocity field of the final combined c configuration data imaged at the low spatial and 5 km s@xmath3 channel@xmath3 velocity resolution are presented in figures [ n5005 ] ( a ) , ( b ) , and ( d ) . the velocity field will be discussed in greater detail in section [ hikin ] . figure [ n5005 ] ( c ) shows the same data in contours imaged at high spatial resolution . the first contour in figures [ n5005 ] ( a ) , ( b ) , and ( c ) corresponds to a column density of 10@xmath18 atoms @xmath10 . there are two notable features of the hi emission in ngc5005 visible in figures [ n5005 ] ( a ) and ( b ) . first , there is a slight hole or depression in the hi column density at the center . this is expected to occur in galaxies where there is a phase change from atomic to molecular gas in the central kiloparsec . bar - driven transport also tends to concentrate the molecular gas in the centers of galaxies . ( @xcite ; see also @xcite ) . archival co data allows us to trace the molecular gas component in the nucleus to fill in the missing gas surface density information ( section [ codata ] ) . second , the gaseous disk does not extend nearly as far as the stellar disk . indeed , ngc5005 is known as an hi - deficient galaxy with a measured mass in hi that is almost half of its expected value ( @xcite ) . furthermore , we did not find low column density gas in either the lower velocity resolution , spatially smoothed map or in the archival d configuration data beyond what is shown in figure [ n5005 ] ( a ) . the higher resolution hi contours displayed in figure [ n5005 ] ( c ) show that the neutral gas appears to be spatially coincident with the ionized gas seen in h@xmath1 , as is expected for star forming regions in galaxies . global properties of ngc5005 measured from this hi dataset are presented in table [ intprops ] . the field of view of the hi observations includes not only ngc5005 , but also nearby dwarf galaxies sdss j131051.05 + 365623.4 , sdss j131115.77 + 365911.4 , sdss j131105.57 + 371036.1 , and sdss j131126.81 + 371842.3 . only sdss j131051.05 + 365623.4 is detected in the hi images ( figure [ bigfov ] ) with a measured flux of 0.21 jy km s@xmath3 and line width at 50% of the peaks of 33 km s@xmath3 ; this yields a corresponding hi mass of 1.33@xmath1210@xmath19 m@xmath11 at the adopted distance of 16.5 mpc . outside of the primary beam , we also detect ngc5002 and ngc5033 . no other known or suspected companions were detected in the hi observations . the central hole in hi column density makes it difficult to extract trustworthy total gas surface density measurements from the vla data alone . therefore , archival co data from the bima survey of nearby galaxies ( bima song ) ( @xcite ) was acquired to supplement the hi observations of ngc5005 . observations for the bima song survey were carried out from 1997 november through 1999 december using the berkeley - illinois - maryland association ( bima ) millimeter interferometer in hat creek , ca . the archival co data cube for ngc5005 has a velocity resolution of about 10 km s@xmath3 and a spatial resolution of 6.2 @xmath12 6.0 . the integrated co intensity is mapped in red contours in figure [ n5005 ] ( a ) . a more detailed map can be found in @xcite . this map allows the central molecular gas surface density to be determined . note that the spatial extent of the molecular gas is confined within the central @xmath240 or @xmath23.2 kpc at 16.5 mpc . as is typical for barred galaxies with co emission , the molecular gas is concentrated in the center ( @xcite ) . in the present study , we focus on using the archival co observations to improve the accuracy of the gas surface density measurements only . a complete analysis of the complex molecular gas dynamics in ngc5005 can be found in @xcite . we obtained an h@xmath1 velocity field for the central region of ngc5005 using the sparsepak ifu ( @xcite ) on the wiyn 3.5 m telescope . the sparsepak ifu is composed of eighty - two 5 diameter fibers arranged in a fixed 70 @xmath12 70 rectangle . observations were acquired on the night of 26 april 2014 during reasonably clear , but non - photometric weather with good seeing . we used the sta1 ccd binned 4 @xmath12 3 with the [email protected] grating in eighth order , centered at 6686 to cover a wavelength range from 6480 to 6890 . this resulted in a velocity resolution of 13.9 km s@xmath3 pixel@xmath3 , which is sufficient to derive an accurate velocity field for ngc5005 . for simplicity , the sparsepak array was aligned on the sky with a position angle of 0 . we used a three pointing dither pattern to spatially fill in gaps between fibers . we took three exposures at 900 s for each dither pointing in order to be able to detect diffuse ionized gas , not just star forming regions . additional observations of blank sky were also taken to remove sky line contamination more accurately , as ngc5005 is much more extended than the sparsepak field - of - view ( see figure [ spectra ] ) . the sparsepak data was processed using standard tasks in the hydra package within iraf . the data were bias - subtracted and flattened , and the iraf task dohydra was used to fit and extract apertures from the ifu data . the spectra were wavelength calibrated using a wavelength solution created from thar lamp observations . the three exposures for each dither pointing were cleaned of cosmic rays . the individual images were sky subtracted using the separate sky pointing scaled to the strength of the 6577 sky line , which is close to the redshifted h@xmath1 and [ nii ] @xmath216584 lines in some locations . the cleaned , sky subtracted images were averaged together to increase the signal - to - noise ratio and then flux calibrated using observations of spectrophotometric standards from @xcite . we illustrate representative spectra from four sparsepak fibers in the central region of ngc5005 in figure [ spectra ] to demonstrate the types of spectral emission found in ngc5005 . the locations of these fibers start in the central nucleus of ngc5005 and move out towards the northwest just off the minor axis ( see the rotated h@xmath1 image on the right ) . while our primary analysis relies on cross - correlation of the entire spectrum , we also fit gaussian profiles to emission lines of h@xmath1 , [ nii ] , and [ sii ] in all the spectra to estimate the full width half maximum ( fwhm ) values and the emission line ratios . as seen previously ( @xcite ) the strength of the [ nii ] @xmath216584 line emission relative to h@xmath1 in the central fiber ( + 000 + 000 ) implies non - thermal emission processes ( @xcite ) . the broadness of the nuclear emission lines from the center of ngc5005 are suggestive that the nucleus is accretion powered ( @xcite ) . we find an approximate fwhm of 900@xmath5200 km s@xmath3 based on single gaussian fits to [ nii ] and [ sii ] lines within the central 5 fiber . the second spectrum in figure [ spectra ] ( -005 + 003 ) was extracted from the adjacent fiber to the northwest of the central fiber . this spectrum displays double - peaked emission lines that suggests either infalling or outflowing gas . indeed , this fiber is placed near where @xcite detected infalling molecular gas in their co observations of ngc5005 . the third fiber s spectrum ( -010 + 006 ) still shows broad emission lines with non - thermal line strengths , but less so in both aspects than the central fiber . this spectra is representative of diffuse ionized gas emission in ngc5005 , which we measure to have typical fwhms of about 140@xmath510 km s@xmath3 and [ nii ] @xmath216584/h@xmath1 line ratios of [email protected] , which is consistent with non - thermal or shock excited gas . the fourth fiber s spectrum ( -015 + 008 ) shown in the bottom panel of figure [ spectra ] is placed on the ring of h@xmath1 emission surrounding the nucleus . this spectrum displays narrower line widths and stronger h@xmath1 to [ nii ] @xmath216584 line emission strengths . we find typical fwhms of approximately 80@xmath55 km s@xmath3 and [ nii ] @xmath216584/h@xmath1 ratios of [email protected] in similar spectra , which are indicative of primarily thermal emission processes most likely due to recent star formation activity ( @xcite ) . inspection of spectra extracted from a few other fibers revealed additional locations with double - peaked emission lines , similar to the -005 + 003 fiber spectrum . these spectra were located approximately east - west on the sky along where evidence from @xcite suggests a stellar bar exists . we note , however , that double - peaked emission lines may also occur due to smearing of the rotation curve within the 5 sparsepak fibers . the spectra along the major axis would be most susceptible to this beam smearing as the velocity gradient is highest here . since the bar is oriented close to the major axis of ngc5005 , it would be difficult to verify that the double - peaked emission lines are signatures of infalling gas along the bar . alternatively , the beam smearing effect would be minimized for velocities along the minor axis , so it is reasonable to assume that the double - peaked emission lines seen in the -005 + 003 spectrum in figure [ spectra ] are real and are not merely due to observational effects . the flux calibrated , sky subtracted spectra were cross - correlated with a template emission line spectrum to extract the luminosity weighted mean recessional velocity at every position using the iraf task fxcor . we used the spectrum of an hii region from within ngc5005 as our reference template for the cross - correlation analysis due to its strong narrow emission lines . spectra with a tonry and davis ratio ( tdr ; @xcite ) less than 5 were not included in deriving the velocity field . spectra with broad or double - peaked emission line profiles resulted in broad cross - correlation peaks that were nonetheless of high significance and had a well - defined centroid . the velocity values from fxcor were then placed into a grid that mapped the sparsepak fiber locations . to fill in the missing spacings , the velocity field was then interpolated using the average value from the nearest eight pixels . analysis of the velocity field will be discussed further in section [ spakkin ] . imaging observations at 3.6@xmath0 m were obtained using the infrared array camera ( irac ) on the _ spitzer space telescope _ on 2012 july 12 and 19 . observations were conducted in a grid - like mapping pattern of 10 @xmath12 9 dithered pointings , where each pointing was offset by @xmath2100 to cover a total area of 18 @xmath12 33 . observations were taken in two sets separated by several days to enable asteroid removal from the mosaicked images . the resulting map has a total exposure time of 1800 s pixel@xmath3 . this exposure time provides a sensitivity of a [email protected] m@xmath11 pc@xmath22 over a field of view 5 times the size of the optical radius r@xmath23 . after standard processing , the individual pointings were combined into maps using the mosaicker and point source extractor ( mopex ) software . images were drizzled to resample the data to 0.75 pixel@xmath3 ( @xcite ) . the maps were fit for a first order sky subtraction to remove any gradient in the sky level . figure [ bigfov ] shows the full extent of the faint stellar disk component of ngc5005 detected in the final mosaicked and sky subtracted 3.6@xmath0 m image . the image on the left has been smoothed to bring out the lowest surface brightness features . the immense sensitivity over such a large field of view enables a complete study of the faint extended stellar component of ngc5005 . ngc5005 has a large high surface brightness disk , which is surrounded by an even more extended low surface brightness component . this faint stellar population extends about twice as far as the break or transition radius between the high and low surface brightness components and nearly three times farther than the gas disk detected in hi . this is an unusual result , as most spiral and irregular galaxies with hi have a diameter in hi that is larger than the optical diameter ( @xcite ) . there is also a feature that appears to be a faint stellar stream extending to the northeast towards an apparent companion galaxy sdss j131105.57 + 371036.1 . to measure the stellar mass associated with the apparent stellar stream , we used the smoothed and masked image to create a hand - drawn polygonal aperture that traces the area associated with the diffuse light ( figure [ bigfov ] ) . we measured a lower limit for the flux in the stream of 0.15 @xmath5 0.13 mjy based on a conservatively masked image . however , this value significantly underestimates the total flux associated with the stream due to the significant masking required to remove the large area associated with the bright foreground star located in the middle of the stream . thus , to obtain a more accurate estimate of the total mass associated with the stream , we replaced the masked pixels with a value interpolated from neighboring pixels and remeasured the total flux to be 0.35 mjy . assuming the best - fitting disk mass - to - light ratio of 0.5 ( section [ dmdecomp ] ) and a distance to ngc5005 of 16.5 mpc , this corresponds to a total stellar mass for the stream of 1.8 @xmath12 10@xmath24 m@xmath25 , which is an order of magnitude less massive than notable streams identified around other galaxies , such as m63 ( @xcite , @xcite ) and m83 ( @xcite ) . the apparent companion galaxy is not detected in the hi observations and does not have a redshift measurement available in the literature . however , if sdss j131105.57 + 371036.1 is at the same distance as ngc5005 and we again assume a mass - to - light ratio of 0.5 , it would have a total stellar mass of 2.3 @xmath12 10@xmath24 m@xmath25 ( table [ otherprops ] ) . given the slightly larger mass of the apparent companion galaxy , interactions between the apparent dwarf companion and ngc5005 can not be immediately ruled out as an origin for the potential stellar stream . lcccccrccccc ngc 5005 & 13:10:56.2 , 37:03:33 & 948 & - & - & [email protected] & 65.55 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] + sdss j131115.77 + 365911.4 & 13:11:15.7 , 36:59:14 & 956 & 5.85 & 28.1 & [email protected] & -22.24 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] + sdss j131051.05 + 365623.4 & 13:10:51.0 , 36:56:23 & 1026 & 7.24 & 34.7 & [email protected] & 32.21 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] + sdss j131105.57 + 371036.1 & 13:11:06.2 , 37:10:40 & - & 7.29 & 35.0 & [email protected] & -62.08 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] + sdss j131058.75 + 364943.8 & 13:10:58.6 , 36:49:42 & - & 13.72 & 65.9 & [email protected] & -49.14 & [email protected] & [email protected] & - & [email protected] & [email protected] + sdss j131058.04 + 364812.4 & 13:10:58.1 , 36:48:12 & - & 15.35 & 73.7 & [email protected] & -8.15 & [email protected] & [email protected] & - & [email protected] & [email protected] + sdss j131126.81 + 371842.3 & 13:11:26.9 , 37:18:42 & 960 & 16.33 & 78.4 & [email protected] & -7.90 & [email protected] & [email protected] & - & [email protected] & [email protected] + kkh 081 & 13:11:11.6 , 36:40:46 & 1032 & 22.94 & 110.1 & [email protected] & 17.85 & [email protected] & [email protected] & - & [email protected] & [email protected] + ngc 5002 & 13:10:37.9 , 36:38:03 & 1091 & 25.78 & 123.7 & [email protected] & -6.64 & [email protected] & [email protected] & - & [email protected] & [email protected] + the 3.6@xmath0 m image of ngc5005 was decomposed into bulge and disk components to represent the stellar mass distributions in the rotation curve decomposition analysis . note that the emission at 3.6@xmath0 m is predominantly stellar , and not greatly affected by pah emission at 3.3@xmath0 m ( @xcite ) . we used diskfit ( @xcite ) to derive a photometric model that fit bulge and disk components to the 3.6@xmath0 m image itself . we did not include a bar component in the photometric model , despite evidence for its existence ( @xcite , @xcite ) , since we only fit stellar bulge and disk components in the mass decomposition ( section [ dmdecomp ] ) . our models are not sensitive to the exact inner structure of the galaxy , so the lack of a bar component will have a negligible effect on the derived dark matter profile . radial surface brightness profiles of the resulting bulge and disk models are shown in figure [ phot ] ( a ) . we additionally extracted a surface brightness profile from the 3.6@xmath0 m image using concentric , con - eccentric ellipses to provide a consistency check . the diskfit model and the ellipse photometry yielded similar total surface brightness profiles . ngc5005 s surface brightness profile displays a relatively small bulge component with a complex disk component featuring a break , or change in slope of the exponential disk where the transition from high to low surface brightness stellar disk occurs . the break radius was estimated by fitting two exponentials to the radial ranges of 80 - 180 ( inner ) and 280 - 443 ( outer ) . the radial ranges were chosen so as to avoid locations where the surface brightness profile displays a change in slope . specifically , the inner limit of 80 avoids the bulge component and the dip in surface brightness around 40 . scale lengths of these inner and outer exponential fits are given in table [ radprops ] . the break radius is defined as the radius at which the two different exponential fits intersect , which is 216.3 ( or 17.3 kpc at 16.5 mpc ) . this break radius is indicated on the bottom axis of the panels in figure [ phot ] , along with the optically defined r@xmath23 ( see section [ optdata ] ) and radial extent of the hi disk . the size of the neutral gas disk is only about 60% of the size of the high surface brightness stellar disk , as measured by r@xmath26 . figure [ phot ] ( b ) shows the radial trends in b-3.6 and r-3.6 colors . the colors display nearly identical trends with radius featuring a relatively complex distribution in the inner @xmath280 most likely due to star formation in spiral arms . lc inner scale length ( 3.6@xmath0 m ) & 31.6 ( 2.5 kpc at 16.5 mpc ) + outer scale length ( 3.6@xmath0 m ) & 124.6 ( 10.0 kpc at 16.5 mpc ) + outer break radius ( 3.6@xmath0 m ) & 216.3 ( 17.3 kpc at 16.5 mpc ) + d@xmath23 & 336.4 ( 26.9 kpc at 16.5 mpc ) + hi size at 10@xmath18 atoms @xmath10 & 276 ( 22.1 kpc at 16.5 mpc ) + concentration ( c=5@xmath27 ) & 4.3 + @xmath9(equivalent width ) gradient & -0.11 @xmath28 arcmin@xmath3 + color gradient ( b - r ) & -0.13 color arcmin@xmath3 + aperture photometry on the 3.6@xmath0 m image was carried out using the galphot package for iraf . concentric , con - eccentric ellipses were used to measure the integrated magnitude at r@xmath23 ( section [ optdata ] ) given in table [ intprops ] . no foreground extinction correction was used for the nir since the expected value is negligible . the total stellar mass presented in table [ intprops ] was determined from the total stellar luminosity at 3.6@xmath0 m assuming a mass - to - light ratio ( m / l ) of 0.5 ( see section [ dmdecomp ] ) . the luminosity was calculated from the absolute magnitude at 3.6@xmath0 m , which includes a slight correction ( 0.04 mag ) based on an extrapolation of the 3.6@xmath0 m surface brightness . the ellipse photometry was also used to estimate the concentration of light in ngc5005 using the definition from @xcite ( see table [ radprops ] ) . our calculated value of 4.3 is towards the higher end of concentrations and is in agreement with the value of 4.16 measured in the r - band for ngc5005 by @xcite . @xcite performed a study of nir surface brightness distributions of virgo cluster galaxies and discovered a bimodality of concentration with peaks at c = 3 and 4.5 in high surface brightness disk galaxies . the concentration value determined here for ngc5005 is consistent with the second peak in the bimodal distribution . optical broadband b and r and narrowband h@xmath1 images provide insight into changes in the age or metallicity of the dominant stellar population in ngc5005 . observations were taken with the wiyn 0.9 m telescope at kitt peak national observatory in february 2013 with the s2 kb imager and april 2014 with the half degree imager ( hdi ) . ngc5002 , which was detected outside the primary beam in the vla data , was also observed with hdi . during both observing runs , narrowband imaging was done using a filter with fwhm of about 60 centered at a wavelength of 6580 . a filter with similar fwhm and slightly offset central wavelength was used for continuum subtraction . the h@xmath1 images were taken with total exposure times of 2 @xmath12 20 minutes in the 6580 filter , and 20 minutes in the narrowband continuum filter . broadband imaging was done with three exposures of 900 and 300 s with the b and r filters , respectively . an additional short r - band exposure was acquired to replace saturated pixels in the core of ngc5005 . the optical images were reduced and analyzed with the iraf package . the image reduction included bias subtraction and flat fielding . both the s2 kb and hdi images were flat - fielded using twilight flats for the broadband filters and dome flats for the narrowband . prior to combination , sky values were measured as the mode of galaxy - free regions of the images . the images were then scaled , aligned , and averaged together to make a final combined image . the wcs was updated using star lists from the usno catalog . the s2 kb and hdi images were processed independently in order to provide internal consistency checks for the derived structural parameters . since the hdi images have slightly higher quality and a larger field of view than the s2 kb images , the subsequent analysis will focus on results from the hdi observations . in addition to ngc5005 , our hdi observations include eight known or suspected companion galaxies , including ngc5002 ( table [ otherprops ] ) . velocities presented for the known companions in table [ otherprops ] are from the nasa / ipac extragalactic database ( ned ) . the three unconfirmed companions are diffuse , irregular galaxies that do not appear to be larger background galaxies based on visual inspection of the broadband images . these galaxies are identified using the sloan digital sky survey ( sdss ) designation that most closely matched the photometric center of the galaxies . no corresponding hi or h@xmath1 emission was detected for the unconfirmed companion galaxies that look like they might be associated with ngc5005 . the absolute b - band magnitudes presented in table [ otherprops ] are calculated assuming a distance of 16.5 mpc . aperture photometry of ngc5005 using concentric , con - eccentric ellipses was carried out to derive surface brightness magnitudes for both the b- and r - band hdi images . table [ intprops ] gives measured integrated magnitudes derived from the aperture photometry . the integrated apparent magnitudes tabulated here are measured at the radius at which the b - band surface brightness equals 25 mag arcsec@xmath22 ( r@xmath23 ) and are not corrected for foreground extinction . the absolute b - band magnitude has been corrected for extinction ( @xcite ) and includes a correction ( 0.10 mag ) based on an extrapolation of the b - band surface brightness profile . integrated colors are also measured at r@xmath23 and have been extinction corrected . inspection of the h@xmath1 image indicates that there are no visible star formation regions beyond r@xmath23 . aperture photometry was performed on the h@xmath1 image to measure an h@xmath1 flux and derive an equivalent width . the current total star formation rate ( sfr ) is derived from the h@xmath1 flux ( measured at r@xmath23 ) using the calibration given in @xcite . the calculated sfr of [email protected] m@xmath11 yr@xmath3 is much lower than the median value of about 1.5 m@xmath11 yr@xmath3 found by @xcite for galaxies of the same hubble t - type as ngc5005 . however , this sfr is most likely an underestimate due to internal extinction , which we can not accurately account for without a spatially resolved total infrared map due to the strong active galactic nucleus ( agn ) contamination . the equivalent width ( ew ) of the h@xmath1 emission line is used as a tracer of the specific star formation rate . it is calculated by dividing the h@xmath1 flux by the continuum flux density measured from the h@xmath1 narrowband continuum filter . it , therefore , serves as an indicator of the strength of the current sfr relative to the past average sfr . a larger ew value would indicate a larger current sfr relative to the continuum , or past average star formation . ngc5005 has a relatively small ew value of [email protected] , which is consistent with the little h@xmath1 emission seen in the narrowband image . this ew is also more of an upper limit as it is contaminated by strong [ nii ] emission from the agn in ngc5005 . despite this , the derived ew value places ngc5005 in the same regime as other sbc morphological type galaxies in the m@xmath4-ew and ( v@xmath29)-ew planes as found in @xcite . in addition to integrated properties from the aperture photometry , radial profiles were also created to examine the surface brightness values in each elliptical annulus . figure [ phot ] shows the radial changes in the surface colors b-3.6 , r-3.6 ( _ b _ ) , and b - r ( _ c _ ) ( see also section [ 3.6data ] ) , as well as the radial trend in ew ( _ d _ ) . the radial trend in b - r is typical of large spiral galaxies , going from redder in the bulge - dominated nuclear region to bluer in the disk where more of the current star formation is occurring ( e.g. , @xcite ) . after the initial gradient from red to blue , the b - r color remains roughly constant through the disk of ngc5005 . there does not appear to be any change in b - r color corresponding to either the extent of the hi gas , r@xmath23 , or the change in the outer disk exponential slope ( r@xmath26 ) . the radial trend in ew is much less regular than the b - r color trend . the higher values of ew correspond to spiral arms where there are more hii regions from recent star formation activity . table [ radprops ] gives estimates of the ew gradient and b - r color gradient from a linear fit to these profiles . for visual reference , the characteristic radii described above are overplotted on both the h@xmath1 and r - band images in figure [ ellipses ] . the images are displayed with the same scale to exemplify the concentration of the ionized gas emission ( _ left _ ) and extent of the stellar disk ( _ right _ ) . as is expected , the morphology of the ionized gas emission in the central @xmath240 directly matches the morphology of molecular gas emission mapped in co by @xcite . as can be seen in the zoom - in image , this morphology features a central nuclear disk ( first solid ellipse ) with a ring of emission at about 30 - 40 ( second solid ellipse ; referred to as the 3 kpc ring in @xcite ) . the two central ellipses have semi - major axes of 10 and 40 , respectively , assuming an inclination angle of 53 ( from @xcite ) and position angles of 75 . the h@xmath1 emission continues beyond this central region , tracing the spiral arms which appear to emanate from either end of the ring along the semi - major axis . the ionized gas emission is contained within the hi disk ( first dotted ellipse ) . the r - band image on the right shows a high surface brightness disk component that extends beyond r@xmath23 ( middle dashed - dotted ellipse ) followed by a low surface brightness component that begins slightly within the break radius ( outer - most dashed ellipse ) . the hi disk ( inner - most dotted ellipse ) lies well inside the high surface brightness disk component . in this section , we return to the hi synthesis observations to explore the neutral gas kinematics . in addition , we use integral field unit ( ifu ) spectroscopy to extract ionized gas kinematics in the central region of ngc5005 . we use both ionized and neutral gas kinematics to derive the circular rotational velocities of the gas as it follows the overall gravitational potential of the galaxy . starting with the velocity field , we follow well established techniques to determine the bulk motions of the gas from which we extracted a rotation curve . this process is described in more detail below . beginning with the hi , we used standard analysis tools distributed as part of the gipsy software package ( @xcite ) for analysis of hi synthesis data cubes to extract an intensity - weighted mean velocity field for ngc5005 . to create this first moment map , individual channels from the low resolution datacube at 5 km s@xmath3 channel@xmath3 were smoothed by a factor of 3 , and clipped at the 3@xmath30 level , then interactively blotted to identify signal . figure [ n5005 ] ( d ) shows the final velocity field with isovelocity contours . the straight isovelocity contours seen in figure [ n5005 ] ( d ) indicate that the rotational velocity of the gas remains fairly constant with radius with no obvious signs of a warp at large radii . kinematic warps due to changes in position and/or inclination angle of the disk often result in a velocity field with a characteristic s - shaped distortion seen at larger radii ( e.g. , m83 ; @xcite ) . the low signal - to - noise of the hi observations within the central @xmath220 ( 1.6 kpc at 16.5 mpc ) makes it difficult to trace the hi gas motions . however , as introduced previously , @xcite found an s - shaped distortion in the co velocity field . this central kinematic distortion is most likely due to the presence of a bar causing the gas to be in non - circular orbits . the filled circles in the bottom panel of figure [ rotcur ] show the rotation curve for ngc5005 derived by fitting a series of concentric tilted rings to the hi velocity field ( gipsy task rotcur ) . in this analysis the galaxy is described using a set of concentric rings , each with its own center coordinates ( @xmath31 , @xmath32 ) , systemic velocity ( @xmath33 , expansion velocity , inclination angle ( _ i _ ) , position angle ( p.a . ) , and rotation velocity ( @xmath34 ) . the radii and widths of the rings are defined such that the rotation curve is sampled at a rate of two points per synthesized beam width , which in this case is every 10 . the expansion velocity is fixed to zero . initial estimates for the center coordinates , inclination angle , and position angle were first obtained by fitting ellipses to the hi intensity distribution out to 3@xmath30 ( gipsy task ellfit ) . the systemic velocity was measured from the spectral profile derived from the integrated intensity hi map . these initial values were fed into rotcur and iteratively fit to the receding and approaching sides together . once the center position ( @xmath35 , @xmath36 ) was fixed , rotcur was run again to determine @xmath37 with _ i _ and p.a . as free parameters . additional runs of rotcur were used with ( @xmath35 , @xmath36 ) and @xmath37 fixed to determine _ i _ and p.a . for a combination of both sides . the rings are allowed to vary in position and inclination angle as a function of radius to account for warps ( _ top panels _ ) . however , point - to - point variations in _ i _ can be caused by effects such as streaming motions along spiral arms for which the fitting routine tries to compensate by changing ( @xcite ) . a smoothly varying or constant distribution in _ i _ and p.a . was adopted to derive the underlying bulk rotation and prevent spurious second - order effects in the rotation curve . because variations in position and inclination angle were small , constant position and inclination angles were used for this analysis ( _ dashed lines _ ) . smoothly varying fits to both the inclination and position angle were also tried and determined to be unnecessary as this only changed the derived circular rotational velocities by a few km s@xmath3 . a final run of rotcur was done with all parameters fixed to derive the final azimuthally averaged rotation curve ( _ bottom panel , filled circles _ ) . uncertainties on the rotational velocities were conservatively estimated to be a combination of non - circular thermal gas motions and inclination errors . the uncertainties on the rotational velocities for the first two radii are dominated by thermal gas motions and beam smearing . a velocity dispersion , or second moment map , was created from the high resolution hi data cube . uncertainties due to thermal gas motions were estimated to be about 10% of the average velocity dispersion values , which is the accuracy to which we believe the line centroid can be estimated . the velocity dispersion for ngc5005 in the radial range 20 to 40 is quite large ; average values range between 40 - 50 km s@xmath3 . this is why the first two rotation curve points have noticeably larger error bars than the rotational velocities farther out where the dispersion drops to more typical values on the order of 15 km s@xmath3 . outside of r @xmath38 40 , the uncertainties are dominated by inclination angle errors . due to the relatively high inclination angle ( @xmath268 ) of ngc5005 , an inclination angle error of @xmath510 in the inner parts and @xmath55 in the outer parts produces corresponding rotational velocity errors of @xmath3910% and @xmath405% , respectively . circular rotational velocities and their errors in the central 20 were not derived from the neutral gas data due to the low signal - to - noise in the hi data . the rotation curve for ngc5005 is remarkably flat ; it varies by less than 10 km s@xmath3 over a range of radii from 20 to 140 ( 1.6 kpc to 11.2 kpc at 16.5 mpc ) . the last measured rotational velocity is [email protected] km s@xmath3 which yields a dynamical mass of ( [email protected])@xmath1210@xmath16 m@xmath11 at 140 ( 11.2 kpc at 16.5 mpc ) . if we assume that the rotation curve remains flat and extrapolate out to the break radius , this results in a dynamical mass of ( [email protected])@xmath1210@xmath16 m@xmath11 measured at 216.3 ( 17.3 kpc at 16.5 mpc ) . the hi rotation curve for ngc5005 was additionally derived using a code for modeling asymmetries in disk galaxies ( diskfit : @xcite ) . diskfit uses a physically motivated model rather than a parameterization of concentric rings . it is capable of fitting models with both rotation and lopsided or bisymmetric non - circular motions , as would be expected due to the presence of a bar ( @xcite ) . rotation curves for ngc5005 were derived using three different models : rotation only , lopsided flow , and bisymmetric flow . due to the low signal - to - noise ratio of the hi at radii smaller than 20 , only the rotation dominated parts of the velocity field were fit . rotation curves derived from all three diskfit models , therefore , were in good agreement with a rotation only model and the tilted ring fits from rotcur . derived best fitting kinematic parameters from gipsy s rotcur and diskfit s rotation only model are presented in table [ kinprops ] . given the low signal - to - noise ratio of the hi observations in the inner 20 of ngc5005 , we turn to ionized gas to provide kinematic constraints in this inner region . figure [ spakvel ] shows the ionized gas emission map as traced by the [ nii]@xmath216584 line flux ( _ left _ ) and luminosity weighted mean velocity field determined from cross - correlation of the entire spectrum ( _ right _ ) . we used the [ nii]@xmath216584 emission line to demonstrate the distribution of ionized gas in the central 70 of ngc5005 , as the h@xmath1 flux is often affected by underlying stellar absorption in the diffuse gas regions . regions where the [ nii]@xmath216584 line flux is strongest represent areas with higher signal - to - noise ratios with typical tdr values around 40 from the cross - correlation results . the lower signal - to - noise ratio areas still have reliable cross - correlation results with tdr values closer to 20 . we used rotcur in gipsy to derive a rotation curve from the ionized gas velocity field following the same steps for the hi data outlined above . the rotation curve derived from sparsepak ionized gas observations is shown in figure [ rotcur ] ( _ open stars _ ) . the ionized gas velocity field was sampled every 5 out to a radius of 35 . we allowed the position angle , inclination angle , and the systemic velocity to be fit by rotcur and not constrained to the values used for the hi derived rotation curve . constant values of position and inclination angle were adopted ( _ dash - dot lines _ ) . the ionized gas data fills in the central kinematics that are unconstrained by the hi observations and extends up towards the hi derived rotation velocity at 40 . the inner two hi rotation curve data points display higher rotational velocities than the ionized gas , but this is most likely due to the effects of beam smearing . while the ionized gas morphology agrees well with the co morphology found by @xcite , we do not observe as large of a velocity in the nuclear disk ( r @xmath41 10 ) . this difference could arise because the ionized gas rotational velocities were derived from the centroid of the emission line detected in a 5 fiber and , therefore , do not include the high velocity dispersion of gas associated with the agn . we do , however , detect the same central s - shaped distortion in the ionized gas velocity field , as was detected by @xcite in the co. given the similar morphologies , it is plausible that the ionized gas is subjected to the same bar - driven dynamics as the molecular gas discussed extensively in @xcite . lcc kinematic center ( ra , dec j2000 ) & 13:10:56.28 , + 37:03:31.7 & 13:10:39.05 , + 37:05:35.4 + systemic velocity & 948 km s@xmath3 & 944 km s@xmath3 + adopted position angle & 65.6 & 65.2 + adopted inclination angle & 68.1 & 67.8 + rotation curve decomposition analysis is a powerful tool to probe the dark matter content in galaxies ( e.g. , @xcite ) . this well established analysis method combines the derived galaxy kinematics from the observed rotation curve with estimates of the baryonic mass distribution in order to derive an estimate of the dark matter distribution as a function of radius ( e.g. , @xcite ) . @xcite found that both widely used dark matter models ( nfw and isothermal ) fit equally well for galaxies with m@xmath42 @xmath41 -19 . in this analysis , we fit only the spherical pseudo - isothermal dark matter halo model to the observations as the data are not sensitive to the central dark matter distribution . one of the strengths of the present analysis is the use of nir data to derive stellar bulge and disk distributions . in particular , the _ spitzer _ 3.6@xmath0 m images are relatively insensitive to radial changes in mass - to - light ( m / l ) ratio that might be introduced by either internal dust obscuration or changes in the dominant stellar population . indeed , the m / l ratio at 3.6@xmath0 m is nearly independent of star formation history ( e.g. , @xcite , @xcite ) . the combination of nir , hi , and co observations allow many of the uncertainties associated with the spatial distribution of both major baryonic components ( gas and stars ) to be minimized . the contributions from gas and stars to the overall rotation curve for ngc5005 were calculated from the gas mass and stellar light surface density profiles . the gas mass surface density was measured from the hi and co integrated intensity maps . the gas distributions were derived separately and then added together to create a total gas mass surface density distribution . the contribution to the total circular rotational velocity was calculated from the total gas distribution using the gipsy task rotmod assuming an infinitely thin gas disk . following @xcite , the total gas mass was multiplied by a factor 1.3 to account for primordial helium . figure [ gasmsd ] shows the gas mass surface density distributions for the hi and h@xmath15 ( as traced by co ) separately ( _ filled circles _ and _ stars _ , respectively ) , as well as the resulting total gas distribution when added together ( _ dashed line _ ) . the molecular gas is the dominant contribution to the total gas mass surface density for radii within 60 ( 4.8 kpc at 16.5 mpc ) . the bump in the molecular gas distribution between 30 - 40 corresponds to the ring of emission observed in the narrowband h@xmath1 image ( section [ optdata ] and figure [ ellipses ] ) . model rotation curves for the stellar distributions were derived separately for the bulge and disk based on the decomposed light surface density profiles discussed in section [ 3.6data ] . the contribution to the total circular rotational velocity for these stellar components were calculated with rotmod . a vertical sech - squared distribution with a vertical scale height of 0.93 kpc estimated using the fiducial relation derived in @xcite was used for the stellar disk model . a spherical distribution was assumed for the bulge model rotational velocities . figure [ rotfit ] presents the rotation curve decomposition for ngc5005 . circular rotation velocities for the total gas , stellar bulge , and stellar disk were summed in quadrature and subtracted from the observed rotation curve . a spherical pseudo - isothermal dark matter halo model was fit to the residuals . the bulge and disk m / l ratios were left as free parameters to be fit during the decomposition . they were assumed to be constant with radius . for this galaxy , leaving the stellar disk and bulge m / l ratios in addition to the characteristic radius ( r@xmath43 ) and velocity ( v@xmath44 ) of the dark matter halo model as free parameters creates a poorly constrained problem without a unique solution . therefore , the best fitting r@xmath43 , v@xmath44 , m / l@xmath45 , and m / l@xmath46 were achieved using interactive fitting . the best fit achieved in this analysis resulted in a disk m / l@xmath46 = 0.5 and a bulge m / l@xmath45 = 0.4 . this disk m / l is consistent with @xcite who suggest that the m / l ratio at 3.6@xmath0 m should be about 0.5 based on stellar population models and measurements of the large magellanic cloud . the bulge and disk m / l ratios result in a best - fitting dark matter halo model with a characteristic radius of [email protected] kpc and a velocity of [email protected] km s@xmath3 . the observed rotation curve could not be adequately described at large radii using the baryonic contributions alone ( crosses in figure [ rotfit ] ) . the dip in velocity in the observed ionized gas rotation curve between 12 - 24 ( 0.96- 1.92 kpc at 16.5 mpc ) is an interesting feature which demonstrates the connection between the galaxy morphology and its kinematics . it appears to occur in the transition region between the nuclear disk and ring structure , as seen in the h@xmath1 image ( figure [ ellipses ] _ left _ ) and detected in co emission ( @xcite ) . this region corresponds to a decrease in the total gas contribution to the overall rotation curve , which is dominated by the molecular gas at these inner radii ( figure [ gasmsd ] ) , as well as to where the stellar contribution to the overall rotation curve transitions from being bulge- to disk - dominated . it is also visible as a prominent dip in the equivalent width profile seen in figure [ phot ] ( d ) . the mass decomposition reveals that ngc5005 is baryon dominated at the inner - most radii . ngc5005 follows the same trend as most other massive , luminous galaxies where its mass is baryon dominated in the core and gradually becomes more dark matter dominated with increasing radius ( e.g. , @xcite ) . furthermore , galaxies of similar dynamical mass from @xcite , such as ngc6946 and ngc7331 , display rotation curve decompositions similar to that of ngc5005 where the dark matter halo contribution to the total rotation curve does not overtake the stellar disk contribution to the total rotation curve until close to or beyond the last measured point . the fractional contribution of the maximum stellar disk to the rotation curve at 2.2 times the disk scale length is only 0.71 in ngc5005 . this implies that the stellar disk can only explain about 70% of the observed rotation curve at the point where the stellar disk s contribution should be greatest . this is slightly less than average values found for other samples of galaxies ( @xcite and references therein ) , which were between 0.8 and 0.9 . we have acquired multiwavelength observations to examine the gas and stellar content in ngc5005 . we have utilized new vla c configuration observations to explore the neutral hydrogen gas morphology and kinematics , as well as sparsepak ifu observations to derive kinematics of the ionized gas in the central region of ngc5005 . deep 3.6@xmath0 m images from _ spitzer _ allow us to trace the extended stellar population . finally , we have used optical broadband b and r and narrowband h@xmath1 observations to detect properties of the dominant stellar population and star formation activity . the main results from this multiwavelength study are summarized below . 1 . the most striking morphological feature of ngc5005 is its extended stellar population . the neutral and molecular gas components are both confined well within the high surface brightness stellar disk . the low surface brightness stellar component appears to extend nearly 3 times beyond the gas disk detected in hi . the central 40 of ngc5005 has a lack of detectable hi , but displays both molecular and ionized gas emission . spectra of the ionized gas in the central region of ngc5005 reveal signatures of both thermal and non - thermal emission processes . the nuclear agn region displays broad emission lines with non - thermal line intensity ratios . double - peaked emission lines are seen to the northwest of the nucleus , approximately where @xcite found dynamical evidence for infalling gas potentially due to an inconspicuous stellar bar . the ionized gas emission outside of the nuclear disk appears to be dominated by thermal processes from recent star formation activity . 3 . inspection of the optical broadband images reveals numerous known and suspected companions of ngc5005 . these include sdss j131051.05 + 365623.4 which was detected in the vla observations and sdss j131105.57 + 371036.1 which appears to be possibly connected to ngc5005 by a faint stellar stream . surface photometry of ngc5005 shows a nir surface brightness distribution with a central bulge component and an inner and outer exponential disk . the radial profile of optical color displays a distribution that is typical for large spirals like ngc5005 , and is consistent with ngc5005 having little current star formation activity . the unusually low star formation rate and equivalent width measurements from the h@xmath1 imaging are expected given the gas poor nature of ngc5005 . a rotation curve for ngc5005 was derived from both the neutral and ionized gas kinematics . the ionized gas rotation curve in the central region of ngc5005 exhibits structure which can be connected directly to the ionized gas morphology ( which follows the molecular gas ) , and also appears to correspond to the stellar distribution as it transitions from bulge to disk dominated . the hi rotation curve at large radii is remarkably flat and exhibits no signs of warps . the gas kinematics and surface densities were used in conjunction with the stellar bulge and disk light profiles to decompose the observed rotation curve of ngc5005 and derive a model dark matter distribution . the results of the mass decomposition demonstrate that ngc5005 is baryon dominated in its central region and that the prominent stellar disk only maximally contributes about 70% to the total rotation curve . the authors acknowledge observational and technical support from the national radio astronomy observatory ( nrao ) . we acknowledge use of the wiyn 0.9 m telescope operated by wiyn inc . on behalf of a consortium of ten partner universities and organizations , and the wiyn 3.5 m telescope . wiyn is a joint partnership of the university of wisconsin at madison , indiana university , yale university , and the national optical astronomical observatory . this work is based on observations made with the spitzer space telescope , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa . emily e. richards acknowledges support from the provost s travel award for women in science , a professional development fund supported through the office of the provost at indiana university bloomington . this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .

this work is devoted to the development of a novel theoretical approach , named _ hybrid approach _ , to handle a localized bottleneck in a symmetrically coupled two - channel totally asymmetric simple exclusion process with langmuir kinetics . the hybrid approach is combined with singular perturbation technique to get steady - state phase diagrams and density profiles . we have thoroughly examined the role played by the strength of bottleneck , binding constant and lane - changing rate in the system dynamics . the appearances of bottleneck - induced shock , a bottleneck phase and meissner phase are explained . further , the critical values of bottleneck rate are identified , which signify the changes in the topology of phase diagram . it is also found that increase in lane - changing rate as well as unequal attachment , detachment rates weaken the bottleneck effect . our theoretical arguments are in good agreement with extensively performed monte carlo simulations . hybrid approach , bottleneck , monte carlo simulations , langmuir kinetics

1507.03072

in protein synthesis , the genetic information is deciphered into proteins by molecular machines called ribosomes that attach themselves at the start end of mrna , move along the chain in a unidirectional manner and finally detach at the stop end @xcite . each translation step requires the binding of a freely diffusing transfer - rna ( trna ) molecule , carrying the amino acid specific to each codon @xcite . the important factor affecting the ribosome translation rate is relative concentrations of trna , which may vary from codon to codon . the codons with lower concentrations of trna reduce the protein synthesis rate and thus plays the role of an inhomogeneity in a homogeneous system @xcite . apart from this , inhomogeneities also occur naturally in many other transport systems such as vehicular traffic @xcite , blood flow @xcite and flow of data in a von neumann architecture @xcite . in traffic flow , the ongoing construction on roads , a slow moving vehicle or an accident can lead to slow down the flow rate on highways and can lead to congestion . further , the separation of the cpu and the memory in computers creates von neumann bottleneck , which limits the performance of the computer via limited bandwidth between the cpu and the memory . totally asymmetric simple exclusion process ( tasep ) @xcite is well known to be a paradigmatic model for studying stochastic transport in many - particle systems . both single - channel and multi - channel taseps have been well explored theoretically as well as using monte carlo simulations @xcite . further , one has to take into account the fact that the proteins as molecular motors can also attach from the bulk reservoir or detach from it ( langmuir kinetics ( lk ) ) @xcite . in contrast to tasep , where total number of particles remain conserved , the additional attachment - detachment dynamics ( lk dynamics ) violate the particle conservation and leads to many interesting phenomena @xcite . in literature , the consequences of coupling of the two different dynamics : tasep and lk have been well analyzed in single - channel @xcite as well as two - channel homogeneous systems @xcite . the idea of slowing down of particles at certain defect positions can be incorporated in the form of a set of inhomogeneous lattice sites ( bottleneck ) , either as a single unit or randomly distributed over the whole lattice in tasep . this type of disorder is known as site - wise disorder . another type of disorder studied in literature is particle - wise , where a slow moving particle itself acts as an inhomogeneity in the system . the present work focuses on site - wise disorder which is suitable to model the inhomogeneities present in transport of molecular motors . although a lot of work has been done on homogeneous taseps , the effect of disorder on the steady - state dynamics of such systems is not well understood . several studies have been performed on single - channel inhomogeneous taseps with @xcite as well as without lk @xcite . while investigating the role of a bottleneck in a closed tasep @xcite , it was found that even the presence of a single bottleneck site can produce shock profile and a plateau in the fundamental current - density relation . later , kolomeisky @xcite examined even richer case of single - channel open system and explored the consequences of the inhomogeneity with both faster and slower transition rates . he divided the system into two homogeneous taseps coupled at the single bottleneck site ( defect mean - field theory ( dmft ) ) and proved analytically that a fast site has no effect on the phase diagram ; whereas a slow site leads to shifting of the phase boundaries only . he also tested the theoretical results with monte carlo simulations and found a good agreement in low density ( ld ) and high density ( hd ) phases ; while a little deviation in maximal current ( mc ) phase . another analytical approach namely finite - segment mean - field theory ( fsmft ) was introduced by chou and laktaos @xcite to study clusters of slow codons in protein synthesis . they found that ribosome density profiles near neighboring clusters of slow codons suppress the proteins synthesis . dong et al . @xcite generalized the dmft to study the effect of two bottlenecks on the protein production rate and also performed extensive monte carlo simulations to conclude that the location as well as spacing between the bottlenecks affect the production rate , pointing out an important phenomenon namely _ edge effect_. the investigation into edge effects was further carried out by greulich and scadshneider @xcite to generate the phase diagrams of inhomogeneous tasep using interacting subsystem approximation ( isa ) . they could successfully explain the interactions of defects with the boundaries of the single - channel system . the more complex case of inhomogeneous tasep in the presence of langmuir kinetics was studied by qiu et al . they calculated phase diagrams and density profiles by adopting the concept of dmft and also studied the effect of slow hopping rate and detachment rate on the phase diagram . pierobon et al . @xcite provided a detailed study on the role of a bottleneck in a tasep with lk , using an effective mean - field theory and monte carlo simulations . they introduced the concept of carrying capacity to identify various novel phases called bottleneck phases . importantly , all of the above studies focused on single - channel inhomogeneous systems . parallel to the inhomogeneous single - channel systems , particles in multi - channel transport systems @xcite may also confront a bottleneck , present in either one or in more than one channels . the importance of studying the multi - channel system lies in the fact that it can act as a framework for extending the analysis to networks . due to the complexity in dynamics generating from the interactions between different channels , it is difficult to examine the effects of inhomogeneity in a multi - channel open system . up to our knowledge , the only contribution in this direction has been made by wang et al . @xcite , which explored the effect of a local inhomogeneity in one of the lanes of a two - lane tasep with lk under a symmetric lane changing rule . they extended the dmft to a two - channel system by incorporating the concept of effective injection and removal rates at the inhomogeneous lattice site . despite a good agreement between the solution of the mean - field equations and monte carlo simulations in ref . @xcite , this approach fails to produce the steady - state phase diagrams of the two - channel system . moreover , it is not feasible to analyze the role played by various parameters such as lane - changing rate , attachment - detachment rate and strength of bottleneck on the steady - state dynamics . one can infer from here that though dmft is capable to provide analytic solutions for single - channel inhomogeneous taseps with @xcite as well as without lk @xcite , it lacks some important ingredients to generate a complete picture of the dynamics of the corresponding two - channel system as discussed in section [ sec:3 ] . this motivates us to develop a new approach to handle the bottleneck in a two - channel tasep with lk , which not only overcomes the existing limitations , but also demonstrates the unexplored dynamics of two - channel inhomogeneous systems . the objective of the proposed study is two - fold , to develop a general theoretical approach , which is capable to produce the phase diagrams and to study the effect of various system parameters on the steady - state phases . in this paper , we attempt to provide a complete picture of the dynamics of two - channel symmetrically coupled tasep with lk in the presence of a single localized bottleneck in one of the two channels by adopting a new and simplified approach , called the _ hybrid approach_. we have also validated the theoretical results with monte carlo simulations . the paper is organized as follows . in section [ sec:2 ] , we define the model under examination and the governing dynamical rules . we briefly discuss the limitations of the earlier approaches in sec . [ sec:3 ] . the theoretical hybrid approach and monte carlo simulations are covered in section [ sec:4 ] and section [ sec:5 ] . a thorough analysis of the stationary properties of the model is discussed in section [ sec:6 ] . in the concluding section [ sec:7 ] , we summarize the results and future prospectives of our work . we define our model in a two - channel @xmath0 lattice , where @xmath1 is the length of a channel . the two channels are denoted by @xmath2 and @xmath3 , in which particles are distributed under hard - core exclusion principle ( see fig . [ fig:1 ] ) . we adopt random - sequential update rules for the dynamical evolution of the system . for each time step , a lattice site @xmath4 ; @xmath5 ; @xmath6 is randomly chosen . the state of the system is characterized by a set of occupation numbers @xmath7 ( @xmath5 ; @xmath6 ) , each of which is either zero ( vacant site ) or one ( occupied site ) . at entrance ( @xmath8 ) , a particle can enter the lattice with a rate @xmath9 provided @xmath10 ; and at exit ( @xmath11 ) , a particle can leave the lattice with a rate @xmath12 when @xmath13 . in the bulk , if @xmath14 , then the particle at the site @xmath4 firstly tries to detach itself from the system with a rate @xmath15 ( detachment rate ) and if it fails then it moves forward to site @xmath16 with a rate @xmath17 provided @xmath18 ; otherwise it attempts to shift to other lane with a rate @xmath19 , only if the target site is vacant . on the other hand , if @xmath20 ; @xmath21 , a particle can attach to the site @xmath4 with a rate @xmath22 ( attachment rate ) . here , horizontal transition rate @xmath17 is inhomogeneous and is given by the following binary distribution @xmath23 it is clear from eq . that the bottleneck connects the sites at @xmath24 and @xmath25 in lane @xmath2 ( fig . [ fig:1 ] ) . here , @xmath26 denotes the transition rate of a particle on passing through the bottleneck and will be called as _ bottleneck rate _ throughout this paper . we wish to analyze the effects of a localized bottleneck in the bulk with no boundary interactions of the open system . this adds up to the assumption viz . , @xmath27 . site . crossed arrows show forbidden transitions.,width=434,height=170 ] the only step to handle a single - site bottleneck in a two - channel tasep with lk was put forth by wang et al . they generalized the concept of dmft from a single - channel inhomogeneous tasep with lk to the corresponding two - channel system . although inhomogeneity is present in only one lane , yet they divided both the lanes into two sub - lanes separated at the position of bottleneck site and introduced effective exit and entrance rates at bottleneck site , respectively in each lane . using mean - field approximation with some simplified assumptions , they computed the nonlinear system of equations in terms of effective rates and solved it numerically . the steady - state density profiles obtained using dmft are found to be in good agreement with monte carlo simulations . but the incompetency of this approach lies in the fact that the nonlinear system may not have a unique feasible solution always for all choices of parameters , due to which the complete information about the steady - state dynamics is missing . moreover , the phase diagrams can not be obtained and it is not feasible to quantify the role played by different parameters such as lane changing rate , attachment - detachment rates and bottleneck rate etc . since this approach uses the concept of dividing each lane ( inhomogeneous as well as homogeneous ) into subsystems , its generalization either to multi - channel system or to include multiple defect sites would be a cumbersome job to handle . as discussed in the previous section that dmft is incapable of explaining certain important aspects , we introduce a novel _ hybrid approach _ to deal with the bottleneck in a multi - channel tasep . instead of dividing both the lanes into two sublanes each , we divide only the inhomogeneous lane ( lane a ) into two subsystems viz . , @xmath28 and @xmath29 , connected through sites @xmath24 and @xmath25 . firstly , we compute the temporal evolution of occupation probabilities @xmath30 in the bulk ( @xmath31 , @xmath32 ) from the following set of master equations : @xmath33 where @xmath34 denotes the statistical average and last term on right - hand side takes a negative and a positive sign for lane @xmath2 and @xmath3 , respectively . at the boundaries , the particle densities evolve according to @xmath35 factorizing the correlations using mean - field approximation , we get @xmath36 . to find the continuum limit of the master equations , we define lattice constant @xmath37 , rescale the time and other kinetic rates as @xmath38 , @xmath39 and @xmath40 @xcite . writing @xmath41 , we replace binary discrete variables @xmath30 with continuous variables @xmath42 $ ] and retain the terms up to second - order in taylor s series expansion to obtain @xmath43 since lane b is free from any inhomogeneity , we can drop the subscript @xmath44 and compute the following time evolution equation from master equation for average density in lane b , denoted by @xmath45 . @xmath46 where @xmath47 is the binding constant and an important parameter in analysing the system dynamics @xcite . we divide lane a into two sublattices viz . , @xmath48 and @xmath49 , connected through sites @xmath24 and @xmath25 . the two homogeneous subsystems connected by the bottleneck in lane a will be referred to as left and right subsystem . since , each of the two subsystems is individually a homogeneous tasep with lk , we can find the continuum limit of the mean - field approximate equation for each subsystem in lane a by proceeding in a similar fashion as for lane b. keeping the master equations at @xmath50 and @xmath51 sites intact after rescaling of aforementioned variables , we obtain the following _ hybrid _ system of equations for time evolution of average density in lane a , denoted by @xmath52 . * hybrid system * + _ continuum part : _ for @xmath53 , we have @xmath54 _ discrete part : _ @xmath55 in the continuum limit , the boundary equations and reduce to @xmath56 and @xmath57 . along with these boundary conditions , the system of equations , and is consistent for which , a unique feasible solution always exists . with this favour , our new approach overcomes the limitation of dmft , which was unable to guarantee a unique feasible solution to the nonlinear system of equations @xcite . moreover , no approximation , other than mean - field , has been used to generate the hybrid system , which is again an advantage over the methodology of dmft @xcite for a two - channel inhomogeneous tasep with lk . it can be easily seen that the hybrid approach only segments the inhomogeneous lane into two parts , without disturbing the homogeneous lane . this idea makes it applicable to more general systems such as multi - channel or networks and those with multiple bottleneck sites . we introduce a new space variable @xmath58 to merge the equations , and into a single system , which further reduces to the following system in steady - state . @xmath59 where @xmath60 , @xmath61 and @xmath62 . since it is difficult to solve the system analytically due to its hybrid nature , we propose to find the steady - state solution of the hybrid system using the singular perturbation technique @xcite . recently , this technique has been successfully applied for studying two - channel homogeneous tasep with lk @xcite . singular perturbation technique typically involves obtaining solutions of the differential equations describing the boundary layer region and the bulk solution , separately and then matching both the solutions to get the global solution . the bulk part of the solution , known as outer solution , is found in the limit @xmath63 . we propose a generalized methodology to find the outer solution of the hybrid system . instead of solving the system explicitly , we find the steady - state outer solution by capturing the long time solution of , and using the scheme discussed below @xcite . we discretize the continuum part of model equations using finite - difference and replace the time derivative in discrete equations at sites @xmath50 and @xmath51 with forward - difference formula . for @xmath64 , @xmath65 with @xmath66 and @xmath31 with @xmath67 , the following scheme is applied to model equations , and with the boundary conditions @xmath56 and @xmath57 . @xmath68\\&+ \delta t{'}\big[\omega_d ( k-(k+1)\rho_{i , j}^n ) \mp\omega \big((\rho_{i , a}^n)^2(1-\rho_{i , b}^n)- ( \rho_{i , b}^n)^2(1-\rho_{i , a}^n)\big)\big],\\ \rho_{m , a}^{n+1 } = & \rho_{m , a}^n+ \delta t{'}\big[\frac{q}{\epsilon}\rho_{m-1,a}^n(1-\rho_{m , a}^n)-\frac{1}{\epsilon } \rho_{m , a}^n(1-\rho_{m+1,a}^n)+\omega_d ( k-(k+1)\\&\rho_{m , a}^n)-\omega \rho_{m , a}^n \rho_{m+1,a}^n(1-\rho_{m , b}^n)+\omega \rho_{m , b}^n \rho_{m+1,b}^n(1-\rho_{m , a}^n ) \big],\\ \rho_{m+1,a}^{n+1 } = & \rho_{m+1,a}^n+ \delta t{'}\big[\frac{q}{\epsilon}\rho_{m , a}^n(1-\rho_{m+1,a}^n)- \frac{1}{\epsilon } \rho_{m+1,a}^n(1-\rho_{m+2,a}^n)+\omega_d ( k-(k+1)\\&\rho_{m+1,a}^n ) -\omega \rho_{m+1,a}^n \rho_{m+2,a}^n(1-\rho_{m+1,b}^n)+\omega \rho_{m+1,b}^n \rho_{m+2,b}^n(1-\rho_{m+1,a}^n ) \big].\label{eq:12 } \end{split}\end{aligned}\ ] ] the solution is captured in the limit @xmath69 to ensure the occurrence of a steady - state . clearly , the above system is consistent and gives a unique steady - state outer solution . it is easy to check the consistency of the above system of equations by analyzing the system in the limit @xmath70 . to satisfy the boundary conditions at both the ends , the density profiles incurs a crossover narrow regime in the form of either a boundary layer or a shock this solution is known as inner solution and is found by rescaling the space variable in the neighbourhood of @xmath71 as @xmath72 , where @xmath71 is the position of boundary layer . this rescaling eliminates the non - conservative source terms , which are formed by lane - changing transitions and attachment - detachment dynamics , in the system . in terms of @xmath73 , the inner solution @xmath74 is given by @xmath75 here , the integration constant @xmath76 is computed from the matching condition of outer and inner solution . for example , let the boundary layer appears at right boundary ( @xmath77 ) in lane @xmath78 , the matching condition requires @xmath79 here , @xmath80 is value of the outer solution in lane-@xmath78 at @xmath77 . clearly , @xmath80 is a function of system parameters @xmath81 , @xmath82 , and @xmath83 . solving eq . with @xmath84 ; we get @xmath85 where @xmath86 . the solution given by eq . represents a right boundary layer ( rbl ) in lane @xmath78 with positive slope ( @xmath87 ) . when @xmath88 , the inner solution fails to satisfy the right boundary condition @xmath89 and deconfines from the boundary to enter the bulk of lane-@xmath90 in the form of a shock . thus @xmath91 acts as a bulk phase transition line . apart from deconfinement , the inner solution also undergoes change in its slope across the line @xmath92 . the right boundary layer with negative slope is given by @xmath93 where @xmath94 . here , the change in the slope of boundary layer describes a surface transition , which does not affect bulk density profile . analyzing , in a similar fashion , we can find mathematically all the bulk as well as surface transition lines in the @xmath95 plane and generate the phase diagram for different parameters . for a more detailed solution methodology of singular perturbation technique , we refer @xcite . to testify the outcomes of our theoretical mean - field approach , we simulate the two - channel model using monte carlo simulations ( mcs ) @xcite . we adopt random sequential update dynamical rules , which is the realization of the usual master equation in continuous time . moreover , random - sequential type update rule is appropriate to model the dynamics of intracellular protein transport @xcite . we run the monte carlo simulations for @xmath96 time steps and first @xmath97 steps are ignored to ensure the occurrence of steady - state . the densities in both the lanes have been computed by taking time averages over an interval of @xmath98 . since the size of a real system is normally not very large , it is reasonable to simulate the system for a lattice size up to @xmath99 . note that the dynamical rules involve many kinds of particle hoppings such as horizontal hoppings , vertical transitions and attachment - detachment dynamics , it is computationally expensive to calculate steady - state densities . as we proceed in the next section , it is found that the results of theoretical mean - field approach agree to a good extent with monte carlo simulations . in this section , we provide a detailed discussion of the steady - state properties of the proposed model , with emphasis on the resulting phase diagrams and analyze the effects of bottleneck strength , lane - changing rate and binding constant . for simplicity , we assume that the bottleneck in lane a is located exactly at the middle i.e. @xmath100 . the steady - state phase diagrams as well as density profiles are obtained from hybrid mean - field approach and validated using extensive monte carlo simulations . when there is no bottleneck , i.e. @xmath101 , the present model reduces to the homogeneous two - channel tasep with lk . under symmetric coupling conditions , the two - channel homogeneous tasep with lk is analogous to its single - channel counterpart , which has been studied well in literature @xcite . for @xmath102 , we call the defect to be _ fast defect _ as the bottleneck rate is more than the usual horizontal particle hopping rate in the system . clearly , a single fast defect does not affect the steady - state dynamics as the rate - limiting hoppings occurring with rate @xmath103 dominate the whole lattice @xcite . thus , it is appropriate to focus on the situation where the bottleneck acts as a slow defect i.e. @xmath104 . it is well known that the steady - state dynamics of exclusion process are significantly affected by the binding constant . so , we proceed by segmenting the analysis into two subcases viz . ( i ) @xmath105 , and ( ii ) @xmath106 . the other parameters are kept fixed to @xmath107 and @xmath99 . we examine the steady - state dynamics for different strengths of bottleneck by gradually decreasing the value of @xmath26 . no significant qualitative changes are found in the phase diagram till @xmath108 , below which one notices the effect of bottleneck more clearly . [ fig:2 ] shows the phase diagram of the system with @xmath109 and @xmath110 . specifically , fig . [ fig:2](a ) shows the complete characterization of various phases in terms of the boundary layers . the solid and dashed lines denote the bulk and surface transitions , respectively . as an illustration , the notation ( tanh - r , tanh - r ) signifies that there is a right boundary layer in lane a as well as lane b of the kind tanh - r , given by eq . and @xmath111 denotes the shock phase . for a detailed understanding of the bulk and surface transitions , we refer the readers to references @xcite . as discussed in previous section that a surface transition is responsible for the change in slope of the boundary layer only and does not affect the bulk density , henceforth , our focus will be only on bulk transitions to avoid any complexities in the phase diagrams . [ fig:2](b ) represents the bulk phase transitions only corresponding to fig . [ fig:2](a ) . , @xmath107 , @xmath110 and @xmath99 , ( a ) phase diagram showing bulk as well as surface transitions and classification according to the boundary layers , nbl denotes no boundary layer , ( b ) phase diagram showing qualitative nature of phases with no surface transitions . notations:- m : meissner phase , 1 : ( @xmath112-ld / mc / hd , ld / hd / ld / mc / hd ) , 2 : ( ld / mc / hd - ld / mc / hd , ld / mc / hd / ld / mc / hd ) , 3 : ( ld / mc / hd - s , ld / mc / hd / ld / hd ) , 4 : ( @xmath112-s , ld / hd / ld / hd ) , 5 : ( @xmath112-ld / mc , ld / hd / ld / mc ) , 6 : ( ld / mc / hd - ld / mc , ld / mc / hd / ld / mc ) , 7:(mc / hd - ld / mc , mc / hd / ld / mc ) , 8 : ( mc / hd - ld / mc / hd , mc / hd / ld / mc / hd ) and 9 : ( mc / hd - s , mc / hd / ld / hd ) . solid and dashed lines denote bulk and surface transitions , respectively.,title="fig:",width=255,height=219 ] , @xmath107 , @xmath110 and @xmath99 , ( a ) phase diagram showing bulk as well as surface transitions and classification according to the boundary layers , nbl denotes no boundary layer , ( b ) phase diagram showing qualitative nature of phases with no surface transitions . notations:- m : meissner phase , 1 : ( @xmath112-ld / mc / hd , ld / hd / ld / mc / hd ) , 2 : ( ld / mc / hd - ld / mc / hd , ld / mc / hd / ld / mc / hd ) , 3 : ( ld / mc / hd - s , ld / mc / hd / ld / hd ) , 4 : ( @xmath112-s , ld / hd / ld / hd ) , 5 : ( @xmath112-ld / mc , ld / hd / ld / mc ) , 6 : ( ld / mc / hd - ld / mc , ld / mc / hd / ld / mc ) , 7:(mc / hd - ld / mc , mc / hd / ld / mc ) , 8 : ( mc / hd - ld / mc / hd , mc / hd / ld / mc / hd ) and 9 : ( mc / hd - s , mc / hd / ld / hd ) . solid and dashed lines denote bulk and surface transitions , respectively.,title="fig:",width=255,height=219 ] to understand the dynamics in a better way , we introduce an appropriate notation @xmath113 denoting the kind of stationary phase in the two - channel system . here , @xmath114 denotes the type of stationary phase in left ( right ) subsystem of lane a ; while @xmath115 denotes the stationary phase in lane b. for example , the phase ( ld - s , s ) denotes that the region left to the bottleneck is low density ( ld ) with a shock present in the right subsystem in lane a and a shock is also present in lane b. , @xmath107 , @xmath110 and @xmath99 . ( a ) ( ld - ld , ld / ld ) phase for @xmath116 , ( b ) ( ld - s , ld / s ) phase for @xmath117 , ( c)(hd - hd , hd / hd ) for @xmath118 , ( d ) ( s - hd , s / hd ) phase for @xmath119 , ( e ) ( mc / hd - ld / mc , mc / hd / ld / mc ) phase for @xmath120 and ( f ) ( @xmath112-s , ld / hd / ld / hd ) phase for @xmath121.,title="fig:",width=198,height=179 ] , @xmath107 , @xmath110 and @xmath99 . ( a ) ( ld - ld , ld / ld ) phase for @xmath116 , ( b ) ( ld - s , ld / s ) phase for @xmath117 , ( c)(hd - hd , hd / hd ) for @xmath118 , ( d ) ( s - hd , s / hd ) phase for @xmath119 , ( e ) ( mc / hd - ld / mc , mc / hd / ld / mc ) phase for @xmath120 and ( f ) ( @xmath112-s , ld / hd / ld / hd ) phase for @xmath121.,title="fig:",width=198,height=179 ] + , @xmath107 , @xmath110 and @xmath99 . ( a ) ( ld - ld , ld / ld ) phase for @xmath116 , ( b ) ( ld - s , ld / s ) phase for @xmath117 , ( c)(hd - hd , hd / hd ) for @xmath118 , ( d ) ( s - hd , s / hd ) phase for @xmath119 , ( e ) ( mc / hd - ld / mc , mc / hd / ld / mc ) phase for @xmath120 and ( f ) ( @xmath112-s , ld / hd / ld / hd ) phase for @xmath121.,title="fig:",width=198,height=179 ] , @xmath107 , @xmath110 and @xmath99 . ( a ) ( ld - ld , ld / ld ) phase for @xmath116 , ( b ) ( ld - s , ld / s ) phase for @xmath117 , ( c)(hd - hd , hd / hd ) for @xmath118 , ( d ) ( s - hd , s / hd ) phase for @xmath119 , ( e ) ( mc / hd - ld / mc , mc / hd / ld / mc ) phase for @xmath120 and ( f ) ( @xmath112-s , ld / hd / ld / hd ) phase for @xmath121.,title="fig:",width=198,height=179 ] + , @xmath107 , @xmath110 and @xmath99 . ( a ) ( ld - ld , ld / ld ) phase for @xmath116 , ( b ) ( ld - s , ld / s ) phase for @xmath117 , ( c)(hd - hd , hd / hd ) for @xmath118 , ( d ) ( s - hd , s / hd ) phase for @xmath119 , ( e ) ( mc / hd - ld / mc , mc / hd / ld / mc ) phase for @xmath120 and ( f ) ( @xmath112-s , ld / hd / ld / hd ) phase for @xmath121.,title="fig:",width=198,height=179 ] , @xmath107 , @xmath110 and @xmath99 . ( a ) ( ld - ld , ld / ld ) phase for @xmath116 , ( b ) ( ld - s , ld / s ) phase for @xmath117 , ( c)(hd - hd , hd / hd ) for @xmath118 , ( d ) ( s - hd , s / hd ) phase for @xmath119 , ( e ) ( mc / hd - ld / mc , mc / hd / ld / mc ) phase for @xmath120 and ( f ) ( @xmath112-s , ld / hd / ld / hd ) phase for @xmath121.,title="fig:",width=198,height=179 ] we list important features of the steady - state dynamics under the current choice of parameters . 1 . _ meissner phase and bottleneck affected region : _ in the upper right quadrant ( @xmath122 , shaded blue ) , the bulk densities in each lane becomes independent of entrance as well as exit rates . this effect is similar to the behavior of the meissner phase found in superconducting materials @xcite . therefore , we also adopt the name meissner phase ( denoted by m ) to identify boundary independent behavior of two - channel system . it is important to mention that meissner phase is actually a part of certain existing phase , denoted by phase @xmath123 in fig . [ fig:2](b ) . meissner phase is also reported in the literature for a homogeneous single - channel tasep with lk @xcite . note that the density profiles get significantly influenced by bottleneck for a specific range of @xmath124 . the corresponding region in phase - plane is marked with red dots and referred as bottleneck phase throughout this paper . similar results also exist for single - channel inhomogeneous tasep with lk , in which bottleneck phase is identified and analyzed in terms of carrying capacity @xcite . in contrast , for low entrance and exit rates one recovers the tasep / lk density profiles perturbed by a local spike or a dip , which will be discussed as we proceed in this paper . 2 . _ phase transitions and mixed phases : _ to gain an intuitive understanding of the steady - state dynamics of the system , it is important to analyze the phase transitions . across the bulk transition line between ( ld - ld , ld)(fig . [ fig:3](a ) ) and ( ld - s , s ) ( fig . [ fig:3](b ) ) phases , one finds the formation of a shock through deconfinement of the right boundary layer @xcite . it is evident from fig . [ fig:2](b ) that @xmath125 and @xmath126 for @xmath127 are the lines of deconfinement of shock from right boundary layer ( ld phase ) and left boundary layer ( hd phase ) , respectively . interestingly , these are also the corresponding lines of deconfinement in analogous single - channel homogeneous system @xcite . from here , one can infer that the phase diagram under a weak bottleneck has almost similar topological structure as for the one with no bottleneck ( @xmath101 ) @xcite though some of the phases differ in their qualitative nature . actually , a weak bottleneck is incapable of affecting the steady - state dynamics of a symmetrically coupled two - channel tasep with lk system globally . the only effects can be seen in the vicinity of the bottleneck . this point is evident from the density profiles in fig . [ fig:3 ] , which will be discussed in detail subsequently . + as one can see from fig . [ fig:3](b ) , a shock is present in both the lanes in ( ld - s , s ) phase . on decreasing @xmath12 , the shock in both the lanes moves leftwards and reaches exactly at @xmath128 on the line @xmath129 ; @xmath130 . this line corresponds to the phase transition between ( ld - s , s ) and ( s - hd , s ) phases . a further reduction in magnitude of @xmath12 moves the shock upstream to the bottleneck giving rise to ( s - hd , s ) phase . the bulk phase transition from ( hd - hd , hd)(fig . [ fig:3](c))to ( s - hd , s)(fig . [ fig:3](d ) ) can also be understood on the similar lines . rest of the phase transitions in fig . [ fig:2](b ) are due to gradual changes in the magnitude of densities with respect to entrance and exit rates . + note that lane b has exactly the same density as that in lane a everywhere , except in the vicinity of the bottleneck ( see fig . [ fig:3 ] ) . interestingly , this slight difference in density near the bottleneck is able to produce peculiar kinds of density profiles , giving rise to a number of coexistence ( or mixed ) phases such as ( ld / mc / hd - s , ld / mc / hd / ld / hd ) and ( mc / hd - ld / mc , mc / hd / ld / mc ) etc . an illustration of a mixed phase is shown in fig . [ fig:3](e ) . formation of upward and downward spikes : _ consider the ld phase in left subsystem of lane a , where the particles have enough space to move along the lattice and the presence of a weak bottleneck does not affect the particle motion significantly . here , the density profile in lane a incurs a local perturbation in the form of an upward spike as shown in fig . [ fig:3](a ) . due to comparatively slower particle hopping rate at the bottleneck , the particles coming from the left feels a little congestion before entering the bottleneck . due to this , the density in lane a rises slightly just upstream to the bottleneck , leading to formation of an upward spike in the density profile . importantly , lane b remains unaffected in this phase . similarly , one can understand the formation of a downward spike shown in fig . [ fig:3](b ) ) in hd phase . this can be attributed to particle - hole symmetry . when a particle comes out of the bottleneck in a highly packed situation , it experiences a sudden increase in its hopping rate . this leads to formation of a lower density region just downstream to the bottleneck . similar findings have been reported in literature @xcite . 4 . _ transition to bottleneck - induced shock : _ now we discuss an important feature of our system dynamics i.e. the formation of a bottleneck - induced shock . with ld in lane a , an increase in entrance rate @xmath9 increases the number of particles in left subsystem of lane a , which ultimately increases height of the existing upward spike . at a particular @xmath131 , this local density perturbation converts into a shock , which travels in the left subsystem in lane a for further increase in @xmath9 . since this shock emerges due to bottleneck only , we call this shock as _ bottleneck - induced shock _ and denote it by @xmath112 . the formation of bottleneck - induced shock is also observed for a single - channel inhomogeneous tasep with lk @xcite . the presence of bottleneck - induced shock leads to the existence of three new phases ( @xmath112-s , ld / hd / ld / hd ) , ( @xmath112-ld / mc / hd , ld / hd / ld / mc ) and ( @xmath112-ld / mc / hd , ld / hd / ld/ mc / hd ) in the phase - plane as shown in fig . [ fig:2](b ) . the existence of mixed phases in lane b indicates the effect of bottleneck in homogeneous lane as well . interestingly , there exist two shocks in lane a in ( @xmath112-s , ld / hd / ld / hd ) phase ( see fig . [ fig:3](f ) ) , the left one being bottleneck - induced ; whereas the right one is formed due to deconfinement of right boundary layer . we propose a terminology to identify the strength of the bottleneck as shown in table [ tab:1 ] . we also provide the number of steady - state phases observed under each case , the details of which follow in the text . cccc & & + & & & + no bottleneck & @xmath132 & 4 & 7 + weak & @xmath133 & 2 & 13 + moderate & @xmath134 & 2 & 9 + strong & & & 14 + blockage & @xmath135 & 3 & 14 + _ weak bottleneck : _ so far , we have discussed the steady - state dynamics for a weak bottleneck and pointed out important features such as appearance of meissner phase , bottleneck phase and bottleneck - induced shock etc . clearly , the number of steady - state phases under a weak bottleneck is more than in homogenous case , which is due to the appearance of mixed phases . now , we will gradually change the strength of bottleneck from weak to moderate and then to strong and see whether the same trend continues or not ? _ moderate strength : _ in the range of moderate strength of bottleneck , we have chosen @xmath136 to generate the phase diagram . it is found that for a moderate bottleneck , the phase diagram ( see fig . [ fig:4](a ) ) is qualitatively as well as quantitatively different from the one with weaker bottleneck ( see fig . [ fig:2](b ) ) . an important distinguishing feature of the phase diagram is the appearance of bottleneck - induced shock in lane b as well . this can be seen from fig . [ fig:4](a ) , which identifies ( @xmath112-s,@xmath112/s ) as a new phase . while the nature of bottleneck - induced shock in lane a has already been discussed , one needs to investigate the occurrence of two shocks in lane b. clearly one located in the right is due to the deconfinement of right boundary layer ; while other is because of the indirect impact of bottleneck in lane a , whose origin can be understood as follows . actually , the presence of a bottleneck - induced shock in lane a produces a local high - density region there , making it difficult to accommodate incoming particles from lane b. as a result , the density in the left subsystem of lane b rises and converts into a shock with an increase in entrance rate @xmath9 . the shock produced in lane b , due to the effect of bottleneck in lane a , is also called as bottleneck - induced shock . , @xmath110 and @xmath99 with ( a)@xmath136 , 1:(@xmath112-s , s ) , and ( b)density profile in ( @xmath112-s,@xmath112/s ) phase with @xmath137 , @xmath138 and @xmath136 , and ( c ) phase diagram for @xmath139 , i : ( @xmath112-s,@xmath112/s ) , ii : ( @xmath112-ld,@xmath112/s ) , iii : ( hd - ld,@xmath112/s ) and iv : ( hd - s,@xmath112/s ) . increase in height of upward spike ( shown in magnified inset ) and conversion of spike into bottleneck - induced shock with a decrease in @xmath26,title="fig:",width=170,height=160 ] , @xmath110 and @xmath99 with ( a)@xmath136 , 1:(@xmath112-s , s ) , and ( b)density profile in ( @xmath112-s,@xmath112/s ) phase with @xmath137 , @xmath138 and @xmath136 , and ( c ) phase diagram for @xmath139 , i : ( @xmath112-s,@xmath112/s ) , ii : ( @xmath112-ld,@xmath112/s ) , iii : ( hd - ld,@xmath112/s ) and iv : ( hd - s,@xmath112/s ) . increase in height of upward spike ( shown in magnified inset ) and conversion of spike into bottleneck - induced shock with a decrease in @xmath26,title="fig:",width=170,height=160 ] , @xmath110 and @xmath99 with ( a)@xmath136 , 1:(@xmath112-s , s ) , and ( b)density profile in ( @xmath112-s,@xmath112/s ) phase with @xmath137 , @xmath138 and @xmath136 , and ( c ) phase diagram for @xmath139 , i : ( @xmath112-s,@xmath112/s ) , ii : ( @xmath112-ld,@xmath112/s ) , iii : ( hd - ld,@xmath112/s ) and iv : ( hd - s,@xmath112/s ) . increase in height of upward spike ( shown in magnified inset ) and conversion of spike into bottleneck - induced shock with a decrease in @xmath26,title="fig:",width=170,height=160 ] on moving from weak to moderate bottleneck , the phase diagram gets simplified resulting into a reduction in number of phases . the coexistence phases ( mc / hd - ld / mc , mc / hd / ld / mc ) , ( mc / hd - s , mc / hd / ld / hd ) and ( @xmath112-ld / mc , ld / hd / ld / mc ) , observed in fig . [ fig:2](b ) , are replaced by ( hd - ld , hd / ld ) , ( hd - s , s ) and ( @xmath112-ld , s ) , respectively . this shows that mc phase , either as a whole or as a part of density profile , does not exist for moderate bottleneck , contrary to the weak bottleneck case . this result is analogous to the one found for single - channel tasep with lk with a single bottleneck @xcite . further , the bottleneck phase ( dotted region ) enlarges , as expected for a stronger bottleneck ; whereas the phases ( ld - ld , ld),(ld - s , s ) , ( s - hd , s ) and ( hd - hd , hd ) shrink . meissner phase remains unaffected with increase in bottleneck strength . _ strong bottleneck : _ fig . [ fig:4](c ) shows the phase diagram for the present system with @xmath139 , a strong bottleneck . it can be seen that topology of the phase diagram is significantly changed in comparison to the one with moderate bottleneck . here , the bottleneck effect is sufficiently strong to increase the density difference between two lanes up to a level that we have a complex phase diagram structure as shown in fig . [ fig:4](c ) . one can see that ld phase does not exist in the left subsystem to the bottleneck in lane a. this is physically justified as the extremely slow hopping rate at bottleneck obstructs the incoming particles , leading to formation of either a shock or a hd region in left subsystem . similar argument and particle - hole symmetry accounts for the non - existence of hd phase downstream to the bottleneck in lane a. it further accounts for the enlargement of ( hd - ld , hd / ld ) phase with a decrease in @xmath26 . note that the bottleneck phase ( dotted region ) further enlarges to maximum extent and covers the whole @xmath95 plane . we have also checked the results for the case of complete blockage at the bottleneck i.e. @xmath140 . the phase diagram for @xmath140 remains essentially similar to the one for @xmath139 . the only change is shifting of the phase boundaries , which leads to expansion of ( hd - ld , hd / ld ) phase ( following the similar trend as explained above ) and shrinkage of rest of the phases . .,width=245,height=198 ] fig . [ fig:5 ] shows the transition of upward spike into bottleneck - induced shock with respect to an increase in bottleneck strength . firstly the height of upward spike in lane a at @xmath128 increases in moderate bottleneck case(clearly seen in the inset ) , then the spike converts into shock for strong bottleneck . a further decrease in value of @xmath26 pulls the shock leftwards . so far , we have investigated the system dynamics under different bottleneck strengths , but for fixed lane - changing rate at @xmath110 . since coupling strength has always been an important parameter in two - channel systems , it is necessary to examine its role for the proposed model . in literature , it is well explored @xcite that higher orders of lane - changing rate does not affect the dynamics of a symmetrically coupled homogeneous two - channel tasep with / without lk because the system decouples into two independent single - channel homogeneous systems . but , in our model , this is not the case as our system is inhomogeneous . so , we thoroughly investigate the effect of coupling strength on the dynamics of the proposed inhomogeneous model under different strengths of bottleneck systematically . our aim is to discuss the important topological changes in the phase diagram induced by higher order of lane - changing rate ( for definition of order , please see @xcite ) . for a weak bottleneck , the effect of lane - changing rate on the phase diagram is insignificant till @xmath141 . only little quantitative differences are seen on density profiles in both the lanes . when @xmath142 , we notice significant changes in the density profiles of both the lanes leading to some qualitative changes in the nature of certain phases . [ fig:6](a ) shows the phase diagram for @xmath109 with @xmath143 . for higher orders of @xmath144 , we have found that the number of stationary phases remain same . interestingly , some of the existing phases get transformed into new phases without a noticeable change in the phase boundaries . for example , the phases ( ld / mc / hd - ld / mc , ld / mc / hd / ld / mc ) , ( mc / hd - ld / mc , mc / hd / ld / mc ) and ( @xmath112-s , ld / hd / ld / hd ) change to ( ld / hd - mc , ld / hd / mc ) , ( mc / hd - mc , mc / hd / mc ) , ( @xmath112-hd,@xmath112/hd ) , respectively on moving from @xmath110 to @xmath143 . the effect of increasing the order of @xmath144 on a moderate bottleneck is similar to the one discussed for weak bottleneck . therefore , we skip the phase diagrams for this case at higher orders of @xmath144 . moving our attention towards strong bottleneck , we analyze the phase diagram for @xmath139 . no structural changes in the phase diagram are seen till @xmath145 , except minor translation of various phase boundaries . as an example , the ( @xmath112-s , s ) phase expands ; while ( hd - ld , hd / ld ) phase shrinks . an important phenomenon occurs at @xmath146 , where the density difference between both the lanes , in the vicinity of the bottleneck , decreases . as a result , the complexity of the phase diagram reduces as shown in fig . [ fig:6](b ) . importantly , four new phases ( ld - ld , ld ) , ( ld - s , s ) , ( s - hd , s ) and ( hd - hd , hd ) appear . the effect of bottleneck , despite its higher strength at @xmath139 , reduces because the bottleneck phase covers comparatively smaller region in @xmath95 plane . further increase in lane - changing rate does not produce any subsequent changes in the phase diagram for @xmath139 . this happens due to the weakening of bottleneck effect by an increase in lane - changing rate , which is discussed as follows . , and @xmath99 with ( a ) @xmath109 , @xmath143 , notations:- 1 : ( @xmath112-ld / mc / hd,@xmath112/ld / mc / hd ) , 2 : ( ld / hd - ld / mc / hd , ld / hd / ld / mc / hd ) , 3 : ( ld / hd - hd , ld / hd ) and 4 : ( @xmath112-hd,@xmath112/hd).(b ) @xmath139 with @xmath147 . , title="fig:",width=255,height=219 ] , and @xmath99 with ( a ) @xmath109 , @xmath143 , notations:- 1 : ( @xmath112-ld / mc / hd,@xmath112/ld / mc / hd ) , 2 : ( ld / hd - ld / mc / hd , ld / hd / ld / mc / hd ) , 3 : ( ld / hd - hd , ld / hd ) and 4 : ( @xmath112-hd,@xmath112/hd).(b ) @xmath139 with @xmath147 . , title="fig:",width=255,height=219 ] _ weakening of bottleneck effect _ it can be seen from figs . [ fig:2 ] , [ fig:4 ] and [ fig:6 ] that the bottleneck affected region ( dotted region ) shrinks at higher orders of lane - changing rate in comparison to the one for lower orders of @xmath144 under all values of @xmath26 . specifically , the weakening of bottleneck effect occurs for lower values of @xmath9 and @xmath12 with an increase in lane - changing rate @xmath144 . this is further justified from fig . [ fig:7 ] , which shows a monotonic reduction in peak s height of upward spike at bottleneck with respect to an increase in @xmath144 . an obvious explanation of this effect is that the particles in lane a , feeling hindered due to the bottleneck , would shift to other lane more frequently . as a result , the congestion in lane a gets slightly reduced as one increase the lane - changing rate . similar kind of effect is also reported in literature @xcite . the insets in fig . [ fig:7 ] shows the similar effect for downward spike as well . , ( b ) @xmath136 and ( c ) @xmath139 . here , @xmath148 , @xmath149 for upward spike in ld phase and @xmath150 , @xmath151 for downward spike in hd phase.,width=255,height=219 ] since the methodology of the proposed hybrid mean - field approach is general and applicable for any value of binding constant @xmath153 , we follow same lines to discuss the case where attachment and detachment rates differ , @xmath154 , i.e. @xmath155 . , @xmath107 and @xmath156 with different @xmath26 : ( a ) @xmath109 , ( b ) @xmath157 , and ( c ) @xmath139 . notations , m : meissner phase ( shaded blue ) and bp : bottleneck phase ( shaded with red dots),title="fig:",width=170,height=160 ] , @xmath107 and @xmath156 with different @xmath26 : ( a ) @xmath109 , ( b ) @xmath157 , and ( c ) @xmath139 . notations , m : meissner phase ( shaded blue ) and bp : bottleneck phase ( shaded with red dots),title="fig:",width=170,height=160 ] , @xmath107 and @xmath156 with different @xmath26 : ( a ) @xmath109 , ( b ) @xmath157 , and ( c ) @xmath139 . notations , m : meissner phase ( shaded blue ) and bp : bottleneck phase ( shaded with red dots),title="fig:",width=170,height=160 ] fig . [ fig:8 ] shows the steady - state phase diagrams with @xmath158 . comparing fig . [ fig:8 ] with figs . [ fig:2 ] and [ fig:4 ] , we notice that the complexity of the phase diagrams reduces tremendously for @xmath158 under all values of @xmath26 . the number of steady - state phases is shown in table [ tab:1 ] . note that due to the increased attachment rate in comparison to detachment rate , the density in both the lanes increases , leading to a complete absence of ld phase in the system . for weaker bottleneck , the phase diagram is much simple consisting of only two stationary phases ( s - hd , s ) and ( hd - hd , hd ) as shown in fig . [ fig:8](a ) . quite surprisingly , no density profile in @xmath95 plane is significantly affected by bottleneck , excluding the presence of a downward spike at @xmath128 . this is in sharp contrast to the parallel case with @xmath105 ( see fig . [ fig:2](b ) ) . consequently , one does not get a bottleneck - induced shock here for any values of @xmath9 and @xmath12 . further decrease in the value of @xmath26 does not change the structure of phase diagram but increases the height of downward spike in hd phase . this trend continues till one reaches @xmath159 , at which the perturbation of downward spike enters the right subsystem in lane a in the form of a shock , called as bottleneck - induced shock . its emergence leads to formation of new steady - state phases such as ( @xmath160 ) and ( hd-@xmath112,hd ) as shown in fig . [ fig:8](b ) . a part of phase - plane ( shaded with red dots ) is strongly influenced by the bottleneck - induced shock called as bottleneck phase , which was absent for @xmath161 ( see fig . [ fig:8](a ) ) . on the other hand , the value of @xmath26 does not influence the meissner phase . a further decrease in value of @xmath26 below @xmath162 leads to significant changes in the steady - state dynamics . the emergence of a new phase ( hd-@xmath112,s ) is seen at another critical value @xmath163 as shown in fig . [ fig:8](c ) . additionally two of the earlier existing phases ( s - hd , s ) and ( hd - hd , hd ) disappear completely ; while the bottleneck phase covers the whole @xmath95 plane reflecting the profound effect of strong bottleneck . we have also investigated the effect of lane - changing rates on the dynamics of the proposed system under case @xmath164 . for @xmath165 , no significant changes are seen except shifting of phase boundaries for any value of @xmath26 . we focus on the situation when @xmath166 . it is found that the steady - state phase diagram for weak bottlenecks remains unchanged and has a structure similar to the one seen in fig . [ fig:8](a ) . but the height of downward spike at @xmath128 decreases with respect to increase in lane - changing rate , analogously to the parallel case @xmath105 . one can infer from here that an increase in lane - changing rate weakens the bottleneck effect with @xmath155 as well . the extent of weakening effect is up to a level that we do not get a bottleneck - induced shock till @xmath167 , which means the global effects of the disturbance caused by bottleneck can be captured only for a very strong bottleneck . as an illustration , we provide the phase diagram for @xmath139 in fig . [ fig:9](a ) , which consists of four new stationary phases . moreover , the bottleneck phase shrinks in its size , which once again indicates the weakening of the bottleneck s effect with respect to an increases in @xmath144 . , @xmath158 , @xmath143 and ( b)@xmath168 , @xmath169 , @xmath170.,title="fig:",width=198,height=179 ] , @xmath158 , @xmath143 and ( b)@xmath168 , @xmath169 , @xmath170.,title="fig:",width=198,height=179 ] we have observed that an increase in lane - changing rate weakens the bottleneck effect and as a result the bottleneck affected region in @xmath95 plane shrinks with respect to an increase in @xmath144 . comparing the two cases @xmath105 and @xmath164 , one can infer that the effect of bottleneck is further reduced when @xmath155 . the above two statements generate interest to know whether there exist some values of @xmath153 and/or @xmath144 , at which we do not get a bottleneck - induced shock even for a strong bottleneck . a thorough analysis of our study has revealed that for @xmath169 and @xmath170 with a very strong bottleneck ( @xmath168 , almost a blockage ) , we get a steady - state situation free of any bottleneck - induced shock . the resulting steady - state phase diagram is shown in fig . [ fig:9](b ) . although we do have a downward spike in the density profiles , yet the non - existence of a bottleneck - induced shock is an important outcome . such dynamics further opens up new challenges to dig deeply into the shock dynamics of the proposed model , which will be undertaken in our future study . in this paper , we attempted to provide a clear picture of the role played by an inhomogeneity in a two - channel tasep with lk under a symmetric coupling environment . the location of bottleneck is fixed at middle site in lane a , while lane b is kept homogeneous . theoretically , we have introduced a new _ hybrid _ approach and also employed singular perturbation technique to get the steady - state density profiles and phase diagrams . we have adopted random - sequential update rules to perform monte carlo simulations and shown a good agreement between theoretical and mcs . the study is segmented into two subcases : @xmath105 and @xmath164 . based on the analysis , one finds very interesting topologies of the phase diagrams under different bottleneck rates . it is concluded that the steady - state phase diagrams for different bottleneck strengths under @xmath105 are comparatively richer in their composition as compared to those under @xmath164 . for lower strengths of bottleneck , the phase diagram comes out to be quite similar to the corresponding homogeneous system and effects on the density profiles are only local in the vicinity of the bottleneck . the origin of upward and downward spike at bottleneck in ld and hd phases , respectively is explained . we also shed light on the emergence of bottleneck - induced shock in the system , which accounts for the global effects of bottleneck . the effect of lane - changing rate on the dynamics is also examined and found that an increase in coupling strength weakens the bottleneck effect for any value of @xmath153 . this effect is supported by the reduction in height of upward and downward spikes with respect to increase in lane - changing rate . another important observation is that unequal attachment - detachment rates are also responsible for weakening of bottleneck effect . here , we focused on a localized bottleneck , fixed at the middle of the lattice far away from the boundaries . the approach developed in this paper might be easily extended to analyze more general inhomogeneous systems such as those with more number of bottlenecks and also to study the influence of the location of the bottleneck in the multi - channel taseps . from the biological significance , one can say that the present study might help in optimizing the protein translation rate in multi - channel inhomogeneous systems . a.k . gupta gratefully acknowledges the financial support from the department of science and technology ( dst ) , government of india . i. dhiman gratefully thanks csir , india for the ph.d . fellowship . 38 natexlab#1#1url # 1`#1`urlprefix stenstrm , c. m. , jin , h. , major , l. l. , tate , w. p. , isaksson , l. a. , 2001 . codon bias at the 3-side of the initiation codon is correlated with translation initiation efficiency in escherichia coli . gene 263 ( 1 ) , 273284 . wang , r. , jiang , r. , liu , m. , liu , j. , wu , q .- s . , 2007 . effects of langmuir kinetics on two - lane totally asymmetric exclusion processes of molecular motor traffic . c 18 ( 09 ) , 14831496 .

the equilibrium structure and exchange constants of camno and lamno have been investigated using total energy unrestricted hartree - fock ( uhf ) and localised orbital configuration interaction ( ci ) calculations on the bulk compounds and mno and mno clusters . the predicted structure and exchange constants for camno are in reasonable agreement with estimates based on its nel temperature . a series of calculations on lamno in the cubic perovskite structure shows that a hamiltonian with independent orbital ordering and exchange terms accounts for the total energies of cubic lamno with various spin and orbital orderings . computed exchange constants depend on orbital ordering . exchange contributions to the total energy vary between -20 and 20 mev per mn ion , differences in orbital ordering energy vary between 3 and 100 mev and a jahn - teller distortion results in an energy reduction of around 300 mev . the lattice constant of the lowest energy cubic perovskite structure ( 3.953 ) is in good agreement with the lattice constant of the high temperature cubic phase of lamno ( 3.947 ) . the total energy of _ pnma _ lamno was minimised by varying lattice parameters and seven internal coordinates and a structure 194 mev per mn ion below that of a structure determined by neutron diffraction was found . this optimised structure is nearly isoenergetic with a cubic perovskite structure with a 5 percent jahn - teller distortion . uhf calculations tend to underestimate exchange constants in lamno , but have the correct sign when compared with values obtained by neutron scattering ; exchange constants obtained from ci calculations are in good agreement with neutron scattering data provided the madelung potential of the cluster is appropriate . cluster ci calculations reveal a strong dependence of exchange constants on mn d e orbital populations in both compounds . ci wave functions are analysed in order to determine which exchange processes are important in exchange coupling in camno and lamno .

cond-mat0111536

camno@xmath0 and lamno@xmath0 are endpoint compounds in the series ca@xmath5la@xmath6mno@xmath0 , which has been thoroughly studied experimentally and theoretically @xcite . they have relatively simple atomic and magnetic structures , their magnetic excitations are well described by a spin wave hamiltonian @xcite and exchange constants , j , are well established by neutron scattering @xcite and from the nel temperature @xcite . exchange coupling in manganites has been extensively studied using model hamiltonian @xcite and _ ab initio _ calculations @xcite . this paper presents results of bulk _ ab initio _ unrestricted hartree - fock ( uhf ) and cluster configuration interaction ( ci ) calculations of exchange constants for both compounds . exchange constants obtained from ci calculations are in excellent agreement with experiment and the localised orbital ci wavefunctions are analysed to determine which quantum fluctuations are most important in exchange coupling . model hamiltonian calculations have attributed the exchange coupling energy to o@xmath7 to mn@xmath8 superexchange @xcite , mn d@xmath9 d@xmath9/mn d@xmath10 d@xmath11 t@xmath12 superexchange @xcite , or both @xcite . results of calculations presented below show that both exchange mechanisms operate and that o superexchange is the more important of the two . this was also found to be the case in the model hamiltonian calculations of meskine _ et al . _ @xcite . ci cluster calculations provide detailed information on exchange couplings between neighbouring mn ions , however a bulk electronic structure technique is required to study orbital ordering in lamno@xmath0 . total energies of lamno@xmath0 with a and g - type anti - ferromagnetic ( a - af and g - af ) and ferromagnetic ( fm ) spin orderings have been computed in several isovolume structures in order to establish whether or not orbital ordering and spin ordering terms in the hamiltonian for lamno@xmath0 are independent . obviously exchange constants will depend on orbital ordering , as the latter determines which empty orbitals are available to participate in exchange coupling . however , it is not known whether the e@xmath4 electron density in lamno@xmath0 for a particular orbital ordering depends on spin ordering . it is shown below that a common orbital ordering energy for any of several orbital orderings can be identified and that this energy is independent of spin ordering , to a high degree . spin and orbital ordering terms in the hamiltonian are therefore independent , although orbital ordering determines the exchange constants . at low temperatures , camno@xmath0 exists in a cubic perovskite structure ( lattice constant 3.73 ) with g - af magnetic ordering @xcite and a nel temperature of 130 k. using the rushbrooke - wood formula @xcite , this nel temperature implies an exchange constant j = 6.6 mev . note that throughout this work the spin hamiltonian is of the form due to domb and sykes @xcite @xmath13 @xmath14 is a spin operator , s is the magnitude of the total spin for an ion and j@xmath15 is the exchange constant for a pair of ions . this form is adopted for the hamiltonian as it is the same as that adopted in modelling spin wave dispersion in neutron scattering studies @xcite , except for a small dzyaloshinsky - moriya term . at low temperatures the space group of lamno@xmath0 is _ pnma _ @xcite . the ground state magnetic structure is a - af and the unit cell contains four formula units consisting of rotated and distorted octahedra . there is one more d electron per mn ion ( @xmath16 camno@xmath0 ) , which occupies an e@xmath4 orbital and induces a jahn - teller distortion in each mno@xmath17 octahedron , resulting in three distinct mn - o bond lengths of 1.91 , 1.97 and 2.18 @xcite . the occupied e@xmath4 orbital is a linear combination of d@xmath18 and d@xmath19 orbitals . the largest orbital component lies along the most elongated mn - o bond . the _ pnma _ structure is shown in fig . [ fig : fig1 ] . @xmath20 , @xmath21 and @xmath22 axes referred to below are indicated on this diagram . the results of a number of neutron and x - ray scattering studies of the structure of lamno@xmath0 @xcite over a range of temperatures are summarised in ref.@xcite . pnma _ structure can be viewed as containing planes of mn@xmath11 ions , each joined to its in - plane neighbours by pairs of short ( 1.91 ) and long ( 2.18 ) mn - o bonds . each mn@xmath11 ion in a particular plane is coupled to mn@xmath11 ions in planes immediately above and below by two mn - o bonds ( 1.97 ) . the symmetry of the _ pnma _ structure is such that there is one in - plane ( nearest neighbour ) exchange constant ( j@xmath23 ) and one out - of - plane constant ( j@xmath24 ) . both j@xmath23 and j@xmath24 have been determined from two independent neutron scattering studies to be -6.7 and 4.8 mev , respectively , @xcite . thus there is fm coupling within planes and af coupling between planes . the first comprehensive attempt to explain atomic and magnetic structures in doped and undoped manganites was made by goodenough @xcite in 1955 . he assumed three classes of exchange interaction between neighbouring mn ions in undoped camno@xmath0 and lamno@xmath0 lattices . a specific ordering of @xmath25e@xmath4 orbitals and relative orientations of pairs of empty e@xmath4 orbitals correponding to goodenough s classification are illustrated in fig . [ fig : fig2 ] . when empty e@xmath4 orbitals are available on a pair of neighbouring mn ions and are oriented towards one another ( fig . [ fig : fig2]b ) then af coupling of the mn ion spins is energetically favoured . this is because electrons from the central o@xmath7 ion of either spin are postulated to delocalise onto both mn ions simultaneously , owing to the favourable exchange interaction ( hund s rule ) between the delocalised electron and the mn ion spin . however , if the mn ion spins are fm aligned , only the electron from the central o@xmath7 ion with the same spin orientation as the mn ions can delocalise onto either mn ion , resulting in a higher energy for that state . thus the empty orbital arrangement shown in fig . [ fig : fig2]b results in an af coupling of mn spins . this is a type i exchange interaction according to goodenough @xcite . when one empty e@xmath4 orbital is suitably oriented for o@xmath7 ion electron delocalisation ( fig . [ fig : fig2]c ) , fm coupling of the mn ion spins is favoured . this is a type ii interaction . finally , when no empty hybrids are available ( fig . [ fig : fig2]d ) , no delocalisation occurs . this is a type iii interaction . this model has been used to explain the relative energies of a - af , g - af and fm magnetic states of camno@xmath0 and lamno@xmath0 with a cubic perovskite structure @xcite . in that work it was found that the relative energies of these magnetic structures could be explained by counting the numbers of each type of interaction in each magnetic state and calculating the relative energy of each type of interaction . for both camno@xmath0 and lamno@xmath0 it was found that the type i af interaction was more energetically favourable than the type ii interaction by 10 mev @xcite . the simplified description of exchange interactions just given assumes that an empty e@xmath4 orbital is either available or not . however , empty e@xmath4 orbitals in lamno@xmath0 are not purely d@xmath18 or d@xmath19 in character @xcite . the mixed character of the empty e@xmath4 orbital permits some exchange coupling even when the empty e@xmath4 orbital is not optimally oriented . this type of reasoning was used by millis @xcite in a calculation of exchange coupling energies in camno@xmath0 and lamno@xmath0 . in that work configurations allowed by the pauli principle in which one or two electrons hop from the central o@xmath7 ion to one or both neighbouring mn ions are considered . configurations which differ by a single electron hop have a single hopping matrix element , t. diagonal elements of the hamiltonian are parametrised using the energy required to excite one electron or a pair of electrons from an o@xmath7 ion to a mn ion . in camno@xmath0 the configuration which is assumed to lead to stabilisation of the af state over the fm state is one in which a pair of electrons on the o@xmath7 ion is excited onto separate mn@xmath9 ions . if this were indeed the origin of exchange coupling in camno@xmath0 then one would expect this configuration to appear in an _ ab initio _ ground state ci wave function , but this is not the case . however , the main idea of this model , that more low energy configurations are available to singlet states than high spin multiplicity states , is in accord with results presented here . ci cluster calculations of exchange constants in la@xmath26cuo@xmath27 @xcite and knif@xmath28 @xcite , which used delocalised molecular orbitals have been reported quite recently . ci cluster calculations described below were carried out in a localised orbital basis . the localised orbital basis provides a means of identifying the exchange coupling mechanism in terms of fluctuations of electrons between localised orbitals . these calculations were performed on mn@xmath1o@xmath2 and mn@xmath1o@xmath3 clusters representing fragments of camno@xmath0 and lamno@xmath0 . details of the calculations , including the method used to generate the localised orbitals , details of a spherical array of point charges surrounding the clusters , _ etc . _ are given in appendix [ app : details ] . the wavefunctions for the clusters contain orbitals which are partitioned into a ( doubly - occupied ) core orbital space , an active space containing the 2p orbitals of the o@xmath7 ion situated between the two mn ions in the cluster as well as singly - occupied mn d orbitals and an external space containing unoccupied mn d orbitals . the core orbital space contains core electrons as well as valence electrons not in the active orbital space . the ions in the clusters treated quantum mechanically consisted of two corner sharing mno@xmath29 octahedra . the localised orbitals in the active space for the mn@xmath1o@xmath2 and mn@xmath1o@xmath3 clusters are shown in figs . [ fig : fig3 ] and [ fig : fig4 ] , respectively . the main electronic configuration for the mn@xmath1o@xmath2 cluster representing camno@xmath0 is one in which each mn ion with a formal 4 + charge contains three t@xmath12 electrons and each oxygen ion exists in a closed shell o@xmath7 configuration . the actual charge on the mn ions is significantly reduced as there is a covalent component to the mn - o bonds , as can clearly be seen in the contour plot of the localised orbital with mainly o 2pz character in the top panel of fig . [ fig : fig3 ] . the actual mn ion charge in camno@xmath0 is + 2.13 , according to a mulliken population analysis of the uhf wave functions obtained for camno@xmath0 . the formal charge on mn ions in lamno@xmath0 is 3 + , however a mulliken population analysis of uhf wave functions for lamno@xmath0 yields a charge of + 2.24 the o ion charges in the two compounds are -1.33 ( camno@xmath0 ) and -1.75 and -1.82 ( lamno@xmath0 ) . exchange constants were calculated by finding the energy difference between the spin singlet and spin septet(nonet ) states of the mn@xmath1o@xmath2 and mn@xmath1o@xmath3 clusters . wave functions were constructed from the localised orbitals shown in figs . [ fig : fig3 ] and [ fig : fig4 ] and doubly occupied core orbitals . a septet state for the mn@xmath1o@xmath2 cluster was constructed from six singly - occupied t@xmath12 orbitals and doubly occupied o 2p orbitals localised on the central o ion in the cluster . the form of this wave function is @xmath30 @xmath31 is the anti - symmetrising operator and the subscripts _ r _ on t@xmath12 orbitals in eq . [ eqn : eqn2 ] indicate that they are centred on the left or right mn ion , respectively . @xmath32 is a product of doubly occupied orbitals in the core orbital space which includes the 2p orbitals on the central o@xmath7 . this is the restricted open shell hartree - fock ( rohf ) wave function for the septet state , constructed using localised molecular orbitals . self - consistent field ( scf ) rohf wave functions can be computed using a number of standard electronic structure packages such as the gamess package @xcite used in this work . the singlet state is constructed from the same set of singly - occupied orbitals with a spin - coupling of the form @xmath33 this is one of five spin eigenfunctions @xcite ( sef ) for six electrons coupled into a singlet state . provided that the spatial orbitals multiplying this sef are ordered such that orbitals localised on each mn ion are grouped together , we expect this sef to dominate the ci wave function , since hund s rule requires spins on each ion to be coupled with the same spin . this is indeed found to be the case in the actual ci wave function for the singlet state of the mn@xmath1o@xmath2 cluster representing camno@xmath0 . the wave function for the singlet state is therefore @xmath34 using conventional rules for evaluating determinantal energies @xcite , the energy difference between the two states is k@xmath35 + k@xmath36 , with the singlet state lying _ above _ the septet state ( assuming that other inter - site exchange integrals are zero because of negligible spatial overlap ) . when the ground state energies of the singlet and septet states of the mn@xmath1o@xmath2 cluster with wave functions in eq . [ eqn : eqn2 ] and [ eqn : eqn4 ] were evaluated , the singlet state was 3.6 mev above the septet state . this implies a value of 1.8 mev for the exchange integrals just mentioned . note that we use the notation k@xmath15 for exchange integrals between specific molecular orbitals while we use the notation j for the ( effective ) exchange coupling energy of two spins on different mn ions . the singlet and septet states of this configuration are analogous to the heitler - london valence bond wave function for the singlet and triplet states of the he atom in a 1s2s configuration . in that case the triplet state is lower than the singlet state by k@xmath37 . in general , ci wave functions with n electrons in the active orbital space consist of linear combinations of spin - adapted functions ( saf s ) @xmath38 @xmath39 where a saf is a product of spatial orbitals and a sef , @xmath40 , for the particular spin state in question . the septet and singlet saf s in eq . [ eqn : eqn2 ] and [ eqn : eqn4 ] are the dominant terms in a more general ci expansion of the septet and singlet wave function of the mn@xmath1o@xmath2 cluster . all saf s which are obtainable by exciting one or two electrons from the dominant saf s to empty orbitals in the active space are included in the expansion . as stated above , the 13 orbitals in the active space in the calculations described here are comprised of 10 orbitals of mainly mn 3d character and 3 of mainly o 2p character localised on the o ion between the two mn ions . these excited electron saf s enter the wave function with a maximum weight of order 10@xmath41 and a corresponding occupancy of order 10@xmath42 and it is these which lower the energy of the singlet state below the septet state when the spins are af coupled . the main excited saf s in the singlet and septet wave functions are those in which : @xmath43 electron is excited from an o 2p orbital to the mn e@xmath4 orbital aligned with the mn - o axis ( o to e@xmath4 ( 1e ) ) ; a @xmath44 of electrons are excited from one o 2p orbital to the @xmath45 mn e@xmath4 orbital ( o to e@xmath4 ( 2e ) ) and an excitation in which a t@xmath12 electron is transferred fron one mn ion to the other ( t@xmath12 exchange ) . obviously the latter exchange process is only allowed in the singlet state as it violates the pauli exclusion principle in the septet state when the t@xmath12 shells are half filled , as in camno@xmath0 . excitations in which a pair of electrons are excited from the o ion to separate mn ions are found to have negligible weight for both spin states . uhf total energy calculations were performed using the crystal98 code @xcite for fm , a- , c- , and g - af spin orderings . the energy of the cubic fm structure with the experimental lattice constant of 3.73 was adopted as the reference energy ( 0 mev ) ; calculations were also performed for each of the spin orderings with a lattice constant of 3.75 . total energies and magnetic moments from these calculations are given in table [ tab : tab1 ] . when these total energy differences are fitted to the hamiltonian in eq . [ eqn : eqn1 ] with nearest ( j@xmath46 ) and second nearest ( j@xmath26 ) neighbour interactions ( _ i.e. _ along [ a,0,0 ] and [ a , a,0 ] , where a is the lattice constant ) , the parameters obtained for a lattice constant of 3.73 are j@xmath46 = 10.7 mev , j@xmath26 = 0.3 mev and for a lattice constant of 3.75 , the parameters are j@xmath46 = 10.1 mev , j@xmath26 = 0.3 mev . it is generally believed that exchange interactions which connect magnetic ions along a linear chain are stronger than those which do not , such as the j@xmath26 interaction here . however , in the cubic perovskite structure , exchange interactions along [ 2a,0,0 ] , _ etc . _ contribute equally to all four spin orderings studied and so can not be extracted from the data presented here . similar values for j@xmath46 have been obtained from model hamiltonian calculations by meskine _ ( j@xmath46 = 9.5 mev ) @xcite . note that the definition used for the exchange energy in that work , the difference between the energy of a pair of ferro- and anti - ferromagnetically coupled mn ions , is @xmath47 the exchange energy defined in eq . [ eqn : eqn1 ] above . hence values for exchange energies from that work have been divided by two in order to compare them to values in the present work . exchange energies obtained from cluster ci calculations depend strongly on mn e@xmath4 and o 2p orbital populations . in turn these populations depend on the madelung potential of a sphere of point charges surrounding the mn@xmath1o@xmath2 cluster . the charges were located on crystal ion sites and mulliken populations of ions in bulk uhf calculations on camno@xmath0 were used as a guide in choosing the magnitudes of these charges . the sphere of point charges had a radius of just over 20 and contained around 3300 charges . the radius was chosen so that the sphere was overall almost charge neutral ; each unit cell of point charges was also neutral . the dependence of exchange energies in camno@xmath0 and lamno@xmath0 on mn ion charge , measured by the mulliken population of that ion in the scf cluster calculation , is shown in fig . [ fig : fig5 ] . the charge on the two mn ions @xmath48 was adjusted by transferring charge from mn point charge sites to ca or la point charge sites @xmath49 . the total charge of the mn and ca ( or la ) point charges was kept constant and the o ion charge was maintained at the uhf mulliken population value . it can be seen that the magnitude of the exchange energy increases as charge is removed from the mn ion , which is reasonable as charge is mainly being transferred to / from the e@xmath4 orbitals which are directly involved in the exchange coupling mechanism - as the e@xmath4 orbital becomes filled , the exchange energy diminishes . the ci calculation value of j@xmath46 = 8.1 mev quoted for camno@xmath0 in table [ tab : tab2 ] is the value obtained for an mn cluster ion charge of + 2.13 , the mn ion charge determined from the uhf calculation . this is to be compared to an estimate of the experimental value of j@xmath46 = 6.6 mev , derived from the nel temperature of camno@xmath0 . the fundamental saf s for the septet and singlet states of the mn@xmath1o@xmath2 cluster were given in eq . [ eqn : eqn2 ] and [ eqn : eqn4 ] . in the fundamental saf wave functions for either spin state , each has a saf coefficient , c@xmath50 , of unity , however when additional saf s are permitted in the wave function ( _ i.e. _ permitting o@xmath7 ion 2p to e@xmath4 excitations , _ etc . _ ) the weights of fundamental saf s are around 0.9950 and additional saf s corresponding to o superexchange and t@xmath12 exchange enter the wave function with saf coefficients of order 0.01 . even for limited active spaces ( as in these calculations ) the number of saf s entering the wavefunction means that a convenient way to analyse the wavefunction is to tabulate the summed occupancies ( _ i.e. _ @xmath51 ) of configurations of a particular type . there are , for example , several saf s in which one electron is excited from an o 2p orbital to an mn e@xmath4 orbital @xcite . the relative magnitudes of these occupancies are a measure of the importance of each type of fluctuation about the fundamental saf configurations . summed occupation numbers for the mn@xmath1o@xmath2 cluster are given in table [ tab : tab3 ] . it can be seen that the fundamental ( or main ) saf has occupancy 0.9926 for the singlet state while it has occupancy 0.9943 in the septet state and therefore that there are larger correlation effects in the singlet state . saf s in which a t@xmath12 electron has hopped from one mn ion to the other have an occupancy of 0.0005 , while these fluctuations are absent from the septet state owing to the pauli exclusion principle , as noted above . however , the main difference in septet and singlet wave functions is in the occupancy of states in which one electron is transferred from an o 2p orbital to an e@xmath4 orbital , the occupancy being 0.0038 for the singlet state and 0.0027 for the septet state . the occupancy of saf s in which a pair of electrons is transferred from o 2p to one mn e@xmath4 orbital is the same for both spin states . the energies of both spin states relative to the energy of the fundamental septet saf are also given in table [ tab : tab3 ] . the septet state with o superexchange fluctuations is 133.4 mev below the fundamental septet saf . this is the correlation energy for that state @xcite . the singlet state with o superexchange and t@xmath12 fluctuations is 149.6 mev below the reference energy and 153.2 mev below the fundamental singlet saf energy . the latter energy is the correlation energy for the singlet state . correlation energies for the mn@xmath1o@xmath2 and mn@xmath1o@xmath3 cluster ci wave functions are illustrated schematically in figure [ fig : fig6 ] . correlation energies are around 50 percent larger in mn@xmath1o@xmath2 than in mn@xmath1o@xmath3 and this is reflected in the larger exchange energy in camno@xmath0 . it is worth noting that when the ci cluster calculation for the exchange energy in camno@xmath0 was performed with no point charge array surrounding the cluster , the exchange energy obtained was 57 mev , well in excess of the experimental value . this emphasises the importance of madelung terms in the crystal hamiltonian in determining exchange energies in strongly correlated materials . total energy calculations were performed on lamno@xmath0 in the ideal perovskite ( cubic ) structure , a tetragonal perovskite structure , a cubic structure with a jahn - teller distortion of the mno@xmath17 octahedra and the _ pnma _ structure with atomic coordinates derived from experiment @xcite and by minimising the total energy by varying lattice parameters and internal coordinates not determined by symmetry . these structures are summarised in table [ tab : tab4 ] . the jahn - teller distortion consisted of elongation or contraction of mn - o bonds parallel to the @xmath52 axes of the unit cell . these are the mn - o bonds which induce fm coupling between mn ions in the _ pnma _ structure . the cubic structure with the lowest energy had a lattice constant of 3.953 ( volume 61.77 @xmath53 per mn ion ) which is comparable to the lattice constant of the cubic phase of lamno@xmath0 ( 3.947 ) which occurs at temperatures above 750k @xcite . all relative energies and lattice volumes will be assumed to be per mn ion hereafter . when this structure is changed by a 5 percent jahn - teller distortion ( table [ tab : tab4 ] ) the energy is lowered by 304 mev and the magnetic ground state of the structure switches from d@xmath18 d@xmath19 fm to d@xmath54 d@xmath55 fm ( see below ) . the total energy of the _ pnma _ structure using coordinates from experiment @xcite ( table [ tab : tab4 ] ) is 200 mev above the jahn - teller distorted structure . the total energy of the _ pnma _ structure was minimised @xcite by varying the lattice parameters and 7 internal coordinates not determined by symmetry of the _ pnma _ space group . the total energy of the energy minimised structure is 6 mev above the jahn - teller distorted structure . the optimised lattice parameters and internal coordinates are given in table [ tab : tab4 ] ; the @xmath20 lattice vector is essentially unchanged while the @xmath21 and @xmath22 lattice vectors increase in magnitude by 1.1 and 1.6 percent , respectively . the lattice volume rises from 60.89 to 62.53 @xmath53 . probably the most important changes which occur on minimising the total energy are : the degree of jahn - teller distortion is reduced ; la - o distances increase significantly . in the lowest energy cubic structure there is one mn - o distance of 1.976 and an la - o distance of 2.795 . on introducing the 5 percent jahn - teller distortion these become mn - o distances of 1.877 , 1.976 and 2.075 and la - o distances of 2.795 and 2.797 . in the experimental _ pnma _ structure @xcite the mn - o distances are 1.903 , 1.957 and 2.185 and the la - o distances are 2.433 , 2.461 and 2.548 . these change to 1.910 , 1.944 and 2.135 and 2.609 , 2.666 and 2.684 in the energy minimised structure . hence lower energies are found for structures with larger la - o distances and a reduced jahn - teller distortion . the combined ionic radii of la@xmath11 and o@xmath7 are 2.76 @xcite . la - o distances in the energy minimised _ pnma _ and cubic structures lie just below the combined ionic radii distance , whereas the la - o distances in the experimental _ pnma _ structure lie well below this distance . the cubic structure with a jahn - teller distortion and the energy minimised _ pnma _ structure are both lower in energy than the lowest energy cubic structure by around 300 mev . this energy lowering by a jahn - teller distortion is half of the lowering assumed by millis @xcite in a calculation of electron - phonon coupling in ca@xmath5la@xmath6mno@xmath0 . the uhf calculations reported here are similar to those reported by su _ they report an energy lowering of 1055 mev when the cubic structure is changed to the experimental _ pnma _ structure with no volume change . this calculation will overestimate the energy difference between such structures as the cubic structure with the _ pnma _ structure equilibrium volume is not the mimimum energy cubic structure . for the cubic perovskite structure it was found that variations of the total energies of different spin and orbital orderings can be fitted very well by a hamiltonian of the form @xmath56 where @xmath57 is an orbital ordering term which depends only on the orbital order . for these calculations the cubic unit cell was doubled along [ 110 ] , [ 101 ] and [ 011 ] directions ( g - af spin and orbital ordering ) and along the [ 001 ] direction ( a - af spin and orbital ordering ) and total energies and charge density difference plots @xcite were computed for d@xmath18 d@xmath18 , d@xmath18 d@xmath19 and d@xmath19 d@xmath19 orbital orderings and fm , a - af and g - af spin orderings . the d@xmath18 d@xmath19 a - af combination is incompatible with the unit cell doublings chosen and was omitted . total energies are given in table [ tab : tab5 ] and charge density difference plots for each of these orbital orderings are shown in fig . [ fig : fig7 ] . for d@xmath18 d@xmath18 and d@xmath19 d@xmath19 orbital ordering , distinct exchange constants in the @xmath58 plane , j@xmath23 , and in the @xmath59 plane , j@xmath24 , are postulated , whereas for d@xmath18 d@xmath19 ordering a single exchange constant , j = j@xmath23 = j@xmath24 , is postulated . exchange constants for each orbital ordering are given in table [ tab : tab6 ] . af exchange constants are obtained when adjacent mn orbital ordering is the same and fm coupling is observed when adjacent mn e@xmath4 orbitals differ . this observation also applies to _ pnma _ structures studied : fm coupling is observed between in - plane mn ions with alternating e@xmath4 orbital orientations ; af coupling is observed when adjacent mn e@xmath4 orbitals have the same orientation . magnitudes of af couplings vary between 5.1 and 14.2 mev and one fm coupling of -6.0 mev is observed in the d@xmath18 d@xmath19 a - af ordering . once exchange constants have been computed for a particular orbital ordering , comparison of structures with the same magnetic structure but different orbital ordering permits differences in orbital ordering energies to be calculated . the actual magnitude of orbital ordering energy , @xmath57 , of course depends on the reference energy chosen . the choice of the fm d@xmath18 d@xmath18 structure as the reference energy structure yields values of -17.4 , -20.0 and -113.4 mev for the d@xmath18 d@xmath18 , d@xmath19 d@xmath19 and d@xmath18 d@xmath19 relative orbital ordering energies . the important result here is that an alternating orbital order ( d@xmath18 d@xmath19 ) is around 90 mev below those with the same orbital order on each site ( d@xmath18 d@xmath18 or d@xmath19 d@xmath19 ) in the cubic perovskite structure . when the values of @xmath57 and exchange constants just mentioned are used to compute the relative energies of the eight spin and orbital orderings considered , the maximum deviation from the relative energies reported in table [ tab : tab4 ] is 0.2 mev , demonstrating the suitability of the hamiltonian in eq . [ eqn : eqn7 ] . the fact that charge density difference plots for different spin ordering and the same orbital ordering are very similar suggests that this should be the case . using the fact that orbital and spin contributions to the hamiltonian are independent , differences in total energy of a particular spin order as a function of lattice distortion may be attributed to differences in orbital ordering energy . [ fig : fig8 ] is a plot of total energy for each orbital ordering with g - af magnetic order as a function of isovolume , tetragonal lattice distortion . these calculations were performed using p4/mmm space group symmetry . it can be seen that d@xmath18 d@xmath19 orbital ordering is the most stable ordering only within a small parameter range about the cubic structure . when the tetragonal distortion is such that the lattice is elongated along the @xmath60 axis , d@xmath19 d@xmath19 ordering is favoured and when it is compressed along this axis , d@xmath18 d@xmath18 ordering is favoured . this may be explained by a simple electrostatic argument - the ordering which is favoured in either case is the one where the occupied e@xmath4 orbitals are oriented along the elongated axis or axes , thereby reducing the coulombic repulsion energy . the greatest stabilisation relative to the cubic lattice is found for an @xmath61/@xmath60 ratio of 0.94 where the energy is 164 mev below that of the cubic g - af reference energy . this stabilisation is still significantly less than the stabilisation of 298 mev which results when the energy minimised _ structure is adopted . relative energies and exchange constants for the jahn - teller distorted structure and both _ pnma _ structures studied are given in tables [ tab : tab7 ] and [ tab : tab8 ] , respectively . charge density difference plots for the jahn - teller distorted and energy minimised _ pnma _ structures are shown in fig . [ fig : fig9 ] . the magnetic ground state of the jahn - teller distorted structure is fm , but this is almost isoenergetic with the a - af structure . this is because the in - plane exchange constant is fm while the out - of - plane exchange constant is fm ( but small ) . the magnetic ground state of the cubic structure is g - af with af coupling between all neighbouring mn ions . the switch to fm coupling between neighbouring mn ions in - plane is due to the jahn - teller distortion in - plane . both _ structures studied have a - af magnetic ground states ( as is the case in nature ) but magnitudes of exchange constants obtained from these calculations are smaller than those obtained from neutron scattering data @xcite ( table [ tab : tab8 ] ) . values of 0.6 and -3.7 mev for j@xmath24 and j@xmath23 may be compared to 0.8 and -3.5 mev obtained in a similar uhf calculation @xcite and 4.8 and -6.7 mev from experiment @xcite . a local spin density approximation ( lsda ) calculation @xcite found values of -3.1 and -9.1 mev for j@xmath24 and j@xmath23 . this calculation did find an a - af ground state for _ pnma _ lamno@xmath0 , however , as second nearest neighbour exchange constants are significant in the lsda calculation and favour an a - af magnetic ground state . cluster ci calculations for lamno@xmath0 were performed using mn@xmath1o@xmath3 clusters with the mn ions in the same configuration as a pair of mn ions in the @xmath52 plane ( fig . [ fig : fig1 ] ) and with the mn ions along a line parallel to the @xmath21 axis . the former cluster corresponds to a pair of mn ions which is expected to be fm coupled while the latter corresponds to a pair of ions which is expected to be af coupled . mn e@xmath4 orbital ordering in the former cluster had the form illustrated schematically in fig . [ fig : fig2]c while the latter had orbital ordering as in fig [ fig : fig2]b . clusters and surrounding point charges with the experimental _ pnma _ structure @xcite and the energy minimised structure were used . exchange constants for lamno@xmath0 derived from these cluster calculations are given in table [ tab : tab8 ] . cluster ci calculations with mn , o and la surrounding point charges of 2.6 , -1.8 and 2.8 ( close to mulliken population values from uhf calculations ) result in exchange constants of 3.3 and -3.6 mev for j@xmath24 and j@xmath23 when the experimental structure is used . these values change to 5.1 and -7.4 mev when the energy minimised structure ( table [ tab : tab4 ] ) is used . the madelung potential has an important role in determining exchange constants in manganites . obviously ions several lattice constants or more distant from the ions in the central cluster may be treated as point charges rather than distributed charges without significantly altering the potential within the central cluster . however point charges adjacent to the central cluster may cause a significantly different potential within the cluster and affect the results of the exchange constant calculation . this question has previously been addressed by other workers @xcite . in order to estimate the effect of terminating the cluster with point charges , cluster ci calculations were performed with the 12 la point charges nearest to the central cluster ions replaced by la@xmath11 pseudopotentials @xcite . this resulted in a small increase in j@xmath24 and no change in j@xmath23 compared to the calculation where only point charges were used . the values obtained for j@xmath24 and j@xmath23 from these calculations were 5.2 and -7.4 mev , which are in good agreement with the experimental values : 4.8 and -6.7 mev . values for the exchange constants derived from the model hamiltonian calculations of meskine _ et al . _ @xcite are also given in table [ tab : tab8 ] . relative energies and saf occupancies for the mn@xmath1o@xmath3 clusters used for calculating exchange constants in lamno@xmath0 in the energy minimised structure are given in table [ tab : tab9 ] . the fundamental saf singlet states are 11.9 mev ( j@xmath24 ) and 17.9 mev ( j@xmath23 ) above the nonet states of the clusters . when additional saf s are permitted in the wave function the singlet(nonet ) states are lowered by 105.4(83.3 ) mev ( j@xmath24 ) and 82.8(79.7 ) mev ( j@xmath23 ) . these are the correlation energies for these states . the singlet state of the cluster used to calculate j@xmath24 is 10.2 mev lower in energy than the nonet state giving a value for j@xmath24 of 5.1 mev while the nonet state of the cluster used to calculate j@xmath23 is 14.7 mev lower in energy than the singlet state giving a value of -7.4 mev for j@xmath23 . from table [ tab : tab9 ] it can be seen that o to e@xmath4 ( 1e ) excitations are the main fluctuations about the fundamental saf state . the weight of the fundamental saf in the singlet states of either cluster is less than in the nonet states , reflecting the greater degree of correlation in the singlet states . in the j@xmath24 calculation the greater correlation energy of the singlet state _ c.f . _ the nonet state is sufficient to make the singlet state the ground state and give an af exchange constant . on the other hand , in the j@xmath23 calculation the singlet correlation energy is just greater than that of the nonet state and , together with the fact that the singlet state of the fundamental saf wave function is 17.9 mev above the nonet state , this results in a nonet ground state and an fm exchange constant . uhf and ci cluster calculations for the exchange constant in camno@xmath0 are in reasonable agreement with estimates for its value based on the rushbrooke - wood formula @xcite and its nel temperature . the calculated exchange constants are larger than the estimate based on experiment . the single af exchange constant is mainly a result of o to e@xmath4 ( 1e ) excitations which lower the energy of the singlet state of a pair of adjacent mn ions below that of the septet state . the magnitude of the exchange constant derived from ci cluster calculations depends strongly on the madelung potential within the cluster and there is agreement between theory and estimates based on experiment only when that potential results in ionic charges in the cluster similar to those in the bulk uhf calculation . lamno@xmath0 is more complex than camno@xmath0 . it is also more ionic than camno@xmath0 with mulliken populations of ions closer to the formal ion charges . a number of orbital and spin ordered states exist within a small energy range , say 300 mev , close to the ground state . the energies of several spin and orbital ordered states of cubic lamno@xmath0 are well described by the hamiltonian in eq . [ eqn : eqn7 ] . in the remainder of this section exchange constants in cubic and jahn - teller distorted lamno@xmath0 are correlated with mn@xmath11 ion orbital ordering and o@xmath7 ion charge density distortions and mn@xmath11 ion interactions are identified as type i , ii or iii according to goodenough s scheme @xcite . finally , the role of correlation and availability of empty orbitals on magnetic ion sites in af and fm coupling is discussed . cubic lamno@xmath0 has a d@xmath18 d@xmath19 fm ground state and a lattice constant of 3.953 . exchange constants depend on orbital ordering and range from -6.0 to 14.2 mev . charge density difference plots ( fig . [ fig : fig7 ] ) show that the charge density on an mn ion is essentially independent of charge densities on neighbouring ions . that density is determined solely by the ion s orbital ordering . however , charge densities on more polarisable o@xmath7 ion sites depend on charge densities at both neighbouring mn ion sites . for d@xmath19 d@xmath19 orbital ordering ( fig . [ fig : fig7 ] , top panels ) , o@xmath7 ions in mn - o bonds in the @xmath58 plane undergo a quadrupolar distortion in which charge is displaced from the mn - o bond axis into directions perpendicular to the bond while o@xmath7 ions in mn - o bonds along the @xmath60 axis are much less severly distorted and the ions tend to elongate along the bond axes . the d@xmath19 character of the ordered mn e@xmath4 orbitals can be seen clearly in the top right panel of fig . [ fig : fig7 ] . the exchange constant for mn@xmath11 ions in the @xmath58 plane with this orbital ordering is j@xmath23 = -0.1 mev while the exchange constant for mn@xmath11 ions along the @xmath60 axis is 14.2 mev . thus a weak exchange coupling is associated with the quadrupolar distortion of charge away from the bond axis while a much stronger coupling is associated with a nearly spherical ion in which charge density tends to concentrate along the bond axis , compared to the spherical o@xmath7 ion . for d@xmath18 d@xmath18 orbital ordering the d@xmath18 character of the ordered mn e@xmath4 orbitals is clearly seen in the middle left panel of fig . [ fig : fig7 ] . there is a relatively weak quadrupolar distortion of the o@xmath7 ions in the @xmath58 plane and a stronger quadrupolar distortion of o@xmath7 ions along the @xmath60 axis . the exchange constant for mn@xmath11 ions in the @xmath58 plane is 5.1 mev while it is 7.2 mev for mn@xmath11 ions along the @xmath60 axis . the d@xmath18 or d@xmath19 character of orbital ordering can be seen in both bottom panels in fig . [ fig : fig7 ] . in the @xmath58 and @xmath59 planes the o@xmath7 ion charge density is polarised in a circulating pattern , even though the o@xmath7 ions are situated midway between the mn@xmath11 ions . charge is polarised towards regions at mn@xmath11 ion sites where there is a reduction in charge density below that of spherical mn@xmath9 ions , as indicated by negative contours in the charge density difference plots . around each mn@xmath11 ion with the d@xmath19 e@xmath4 orbital occupied , charge is deformed towards the d@xmath19 ion in the @xmath58 plane and away from it along the @xmath60 axis , whereas for mn@xmath11 ions with the d@xmath18 e@xmath4 orbital occupied , charge is deformed towards it along the @xmath60 axis and away from it in the @xmath58 plane . thus each mn@xmath11 ion is coupled to each neighbouring mn@xmath11 ion by a polarised o@xmath7 ion and there is one fm exchange constant of -6.0 mev . a simple pattern of orbital ordering is obtained for the jahn - teller distorted structure where the unit cell was doubled in the [ 001 ] direction ( fig . [ fig : fig9 ] ) . this pattern of orbital ordering was obtained without biasing the initial guess wave function in anyway ( appendix a ) . orbital ordering in the @xmath58 plane is an alternating d@xmath54 d@xmath55 pattern which is repeated with period one along the @xmath60 axis . this is the a - type orbital ordering discussed in ref . the fm exchange constant between mn@xmath11 ions in the @xmath58 plane is -8.1 mev and the weak fm exchange constant along the @xmath60 axis is -0.1 mev . there is strong deformation of o@xmath7 ion charge density in the @xmath58 plane towards regions of reduced charge density at the mn@xmath11 ion sites , associated with strong fm exchange coupling ( fig . [ fig : fig9 ] , left panel ) . there is strong quadrupolar distortion of the charge density at o@xmath7 ion sites coupling mn@xmath11 ions along the @xmath60 axis , associated with a weak fm exchange coupling . the same a - type orbital ordering is also found in the _ pnma _ structures studied . there is a similar pattern of circulating charge polarisation towards regions at mn@xmath11 ion sites where the charge density is reduced and there is strong fm exchange coupling between mn@xmath11 ions lying approximately in the @xmath52 plane ( -3.7 and -6.0 mev , uhf calculations table [ tab : tab8 ] ) . the three mn - mn interactions described by goodenough @xcite are now tentatively identified in cubic and jahn - teller distorted lamno@xmath0 using charge densities on the o@xmath7 sites and orbital ordering at the neighbouring mn ion sites . type i interactions are found for : d@xmath18 d@xmath18 orbital ordering in cubic lamno@xmath0 for both j@xmath23 = 5.1 mev and j@xmath24 = 7.2 mev ( fig . [ fig : fig7 ] , middle panels ) ; d@xmath19 d@xmath19 orbital ordering for j@xmath24 = 14.2 mev ( fig . [ fig : fig7 ] , top right panel ) . type ii interactions are found for : d@xmath18 d@xmath19 orbital ordering in cubic lamno@xmath0 ( fig . [ fig : fig7 ] , bottom panels ) for both j@xmath24 and j@xmath23 = -6.0 mev ; d@xmath54 d@xmath55 orbital ordering in jahn - teller distorted lamno@xmath0 ( fig . [ fig : fig9 ] , left panel ( j@xmath23 = -8.1 mev ) . charge densities are characterised by breaking of symmetry of the o@xmath7 ion along the mn - o - mn axis . obviously this can only occur when the orbital orderings on adjacent mn ions differ , however this observation is worth making as such symmetry breaking is characteristic of fm exchange coupling . type iii interactions are found for d@xmath19 d@xmath19 orbital order in cubic lamno@xmath0 ( fig . [ fig : fig7 ] , top left panel ) where j@xmath23 = -0.1 mev ; for j@xmath24 = -0.6 mev in the jahn - teller distorted structure ( fig . [ fig : fig9 ] , right panel ) . in both cases the weak exchange coupling is associated with strong , quadrupolar o@xmath7 charge density deformation . cluster ci calculations provide detailed information on the exchange coupling mechanism . fundamental saf singlet states for clusters representing both camno@xmath0 and lamno@xmath0 lie above the fundamental saf high spin multiplicity states ; this is expected to be the case for a wide range of magnetic ions exchange coupled _ via _ a closed shell anion . the ground state for the pair of magnetic ions is af when the _ difference _ in correlation energies of the singlet and high spin multiplicity states exceeds the singlet / high spin state splitting , otherwise it is fm . correlation energies for singlet states exceed correlation energies of the corresponding high spin multiplicity states in the three cases studied here ( fig . [ fig : fig6 ] ) . this is also likely to be true for a wide range of magnetic ions which are exchange coupled _ via _ a closed shell anion as there are many more singlet saf s than high spin saf s in any particular active space . for example , in the active space used for the lamno@xmath0 cluster ci calculations there are over 18,000 singlet saf s compared to over 1,500 nonet saf s , which simply reflects the fact that there are many more ways to arrange spin - coupled electrons to form singlet states than there are to form nonet states for a specific number of electrons . only a few of either the singlet or nonet saf s appear in the ground state wave functions with a signicant weight , but since there are so many more singlet than nonet saf s , it is not surprising that the singlet state correlation energy is larger . when one empty mn e@xmath4 orbital is available to accept one or two electrons from an o@xmath7 ion , as is the case for j@xmath23 in _ pnma _ lamno@xmath0 , the singlet state correlation energy is only slightly larger than the nonet state correlation energy ( 82.8 versus 79.7 mev ) and the nonet state is the ground state . however , when two empty mn e@xmath4 orbitals are available , as is the case for j@xmath24 in _ pnma _ lamno@xmath0 and j in camno@xmath0 ( fig . [ fig : fig6 ] ) , singlet state correlation energies are significantly larger than the nonet(septet ) state correlation energies ( 105.4 versus 83.3 mev ( lamno@xmath0 ) and 153.2 versus 133.4 mev ( camno@xmath0 ) ) and the ground states are singlets . exchange coupling in camno@xmath0 and lamno@xmath0 is largely due to quantum fluctuations in the ground state in which one electron is excited from an o@xmath7 ion into an mn e@xmath4 orbital . fluctuations in which an electron is exchanged between t@xmath12 orbitals enter the singlet state in mn@xmath1o@xmath2 and mn@xmath1o@xmath3 clusters but are not the main contributors to the exchange interaction . parallel studies of exchange coupling in la@xmath26cuo@xmath62@xcite show that d@xmath18 d@xmath18 exchange interactions dominate the exchange coupling in la@xmath26cuo@xmath62 and o to d@xmath18 excitations have a lesser weight than cu@xmath63cu@xmath11 excitations in the la@xmath26cuo@xmath62 ground state . this difference in exchange coupling mechanism most likely reflects trends in effective hubbard u parameters for mn@xmath8 and cu@xmath64 and o 2p to metal d excitation energies . this work was supported by enterprise ireland under grant number sc/00/267 . m. n. wishes to acknowledge support by the trinity trust . the methods used to generate localised orbitals and point charge arrays and the basis sets and ci computer codes used in this work are described in this section . uhf calculations on the crystalline solid were performed using the crystal98 code @xcite . the basis sets used for both crystal uhf calculations and cluster ci calculations were identical gaussian orbital basis sets designed for use in the solid state . they are slightly modified versions of the basis sets available from the crystal98 website @xcite . outer exponents of the gaussian functions were modified so that the total energy in a uhf calculation on camno@xmath0 was minimised . the original basis sets had been optimised for different mn ionic solids . the basis sets used in all calculations are the all electron basis sets for : mn ( 86 - 411d41 g @xcite with two d orbital exponents , optimised for camno@xmath0 by changing the outer d exponent to 0.259 ) ; o ( 8 - 411 g @xcite with principal quantum number up to n = 4 , optimised for camno@xmath0 by changing the outer sp exponents to 0.4763 and 0.22 ) ; ca ( 86 - 511d3 g @xcite with the outer d orbital exponents optimised for camno@xmath0 to 3.191 , 0.8683 and 0.3191 ) and an la basis set optimised for the la@xmath11 ion @xcite . the la basis set used in this work differs from the cited basis in that the 5d orbital was removed from the basis and the 6sp and 7sp orbitals were replaced by a single sp orbital exponent of 0.3917 . different orbital ordered states in uhf calculations were obtained using a feature in the crystal98 code which increases the diagonal element of the fock matrix corresponding to a particular orbital for the first few iterations of the calculation . this results in that orbital being unoccupied during those scf cycles and allows the wave function to converge to a state which is a local energy minimum with a particular orbital ordering . high spin multiplicity states , such as the septet and nonet states of the clusters used here , are generally well described by a self consistent field ( scf ) restricted open shell hartree - fock ( rohf ) wave function . all electrons on the cluster were treated explicitly - no pseudopotential approximation was used , except in the test calculation with a la@xmath11 pseudopotential described above . ci calculations were performed in localised orbital bases , rather than the canonical molecular orbital bases derived from the scf rohf calculations . localisation of scf rohf molecular orbitals was performed using the foster - boys algorithm @xcite , which generates localised orbitals with maximally separated centroids . doubly occupied o 2p orbitals , singly occupied mn t@xmath12 ( or t@xmath12 and an e@xmath4 orbital for mn@xmath1o@xmath3 ) orbitals and unoccupied mn e@xmath4 orbitals were localised in three separate localisation steps . these must be performed separately in order to preserve invariance of the rohf total energy , since any mixing between orbitals of different occupancy will increase the total energy . in the localised orbital rohf wave functions for either spin state of the mn@xmath1o@xmath2 and mn@xmath1o@xmath3 clusters , each mn d electron occupies a separate orbital . calculations on low spin multiplicity states of the clusters used the same sets of localised orbitals . they demonstrate that the localised orbitals generated for the high spin multiplicity states are very good approximations to the optimal orbitals for open shell low spin multiplicity states and that a high spin multiplicity rohf wave function ought to be an excellent starting point for perturbative calculations on high and low spin multiplicity states in the solid state . in a ci calculation on a cluster of this size it is essential to partition the orbital space into a core space ( with doubly occupied orbitals ) , an active space of orbitals which are occupied or unoccupied in the rohf main configuration and a space of redundant , unoccupied orbitals which are not used in the calculation . the active orbitals in this work were the three o 2p localised orbitals on the central o@xmath7 and a set of t@xmath12 and e@xmath4 orbitals on each mn ion . the direct multi - reference ci module @xcite in the gamess @xcite programme was used for this work . the active space consisted of either the ( single ) fundamental saf orbitals or the fundamental saf plus all possible single or double excitations which can be made from the fundamental saf into empty active space orbitals . calculations were performed for clusters with no surrounding point charges and with point charges in a spherical volume surrounding the cluster . the radius of the sphere was over 20 and included around 3300 charges . the charges were located on the ionic sites of either camno@xmath0 or lamno@xmath0 . mulliken populations derived from uhf crystal calculations were used as guides for point charge magnitudes . for camno@xmath0 , uhf mulliken populations were ca@xmath65mn@xmath66o@xmath67 . however , in the scf rohf cluster calculation , this choice of point charges results in mulliken populations of mn@xmath68o@xmath69mn@xmath68 on the central mn - o - mn chain in the cluster . the mn and ca point charge magnitudes were adjusted to ca@xmath70mn@xmath71o@xmath72 and this resulted in mulliken populations of mn@xmath73o@xmath74mn@xmath73 on the central mn - o - mn chain and populations of -1.64 and -1.67 on the other two o types in the cluster . note that this adjustment leaves each point charge unit cell charge neutral and the point charge sphere radius is adjusted so that the entire cluster has a charge near zero . the major changes which occur on adjusting the point charges are : charge is transferred from the outer o ions in the cluster to the mn ions and central o ion , each gaining about 0.4e ; the af exchange constant changes from 21.0 mev to 8.1 mev , in agreement with other calculation methods and in reasonable agreement with experiment ; the degree of correlation in the wavefunction decreases sharply . when a mn@xmath1o@xmath2 cluster with no external point charges is used , the mulliken populations on the central mn - o - mn chain are mn@xmath75o@xmath76mn@xmath75 and the exchange coupling energy is 57 mev . a similar adjustment of point charge magnitudes was used for the mn@xmath1o@xmath3 cluster ci calculations . the mulliken populations determined from uhf crystal calculations on the experimental _ pnma _ structure were la@xmath77o@xmath78mn@xmath79 . cluster point charges of la@xmath80o@xmath81mn@xmath68 resulted in mulliken populations of mn@xmath82o@xmath83 in the _ pnma _ structure . gamess - uk is a package of _ ab initio _ programmess written by m. f.guest , j. h. lenthe , j. kendrick , k. schoffel , and p. sherwood , with contributions from r. d. amos , r. j. buenker , h. j. j. van dam , m. dupuis , n. c. handy , i. h. hillier , p. j. knowles , v. bonacic - koutecky , w. von niessen , r. j. harrison , a. p. rendell , v. r. saunders , a. j. stone , d. j. tozer , and a. h. de vries . the package is derived from the original gamess code due to m. dupuis , d. spangler and j. wendoloski , nrcc software catalog , vol . 1 , program no . qg01 ( gamess ) , 1980 . fermion eigenstates of a hamiltonian which does not explicitly contain spin may be written as anti - symmetric products of spatial orbitals and spin eigenfunctions ( sefs)@xcite . sef s are eigenfunctions of the total spin operator , @xmath84 , and @xmath85 operator . sef s may be generated in several different ways and the multi - reference doubles ci module @xcite of the gamess @xcite code used in this work uses the yamanouchi - kotani ( yk ) scheme @xcite . there is one sef for six electrons coupled into a septet state while there are five sef s for six electrons coupled into a singlet state . each sef consists of an orthonormal combination of products of the one - electron eigenspinors , @xmath86 or @xmath87 . sums of saf occupancies are slightly less than unity for states other than the fundamental saf state because large numbers of saf s with coefficients smaller than 0.003 , which individually contribute 0.000009 to the total saf occupancy , were omitted from the sums . the correlation energy for a molecule , cluster , _ etc . _ is usually defined to be the energy difference between that of an scf wave function and a correlated wave function . here the correlation energy is defined to be the difference between the fundamental saf energy for each state and the energy when fluctuations within the limited active space described above are taken into account . additional correlation energies , for example those due to fluctuations within an o ion , will be larger . however , it is believed that the correlation effects included in this work are the most important in determining exchange constants . these are plots of the difference in the charge density obtained from the uhf wave functions of the solid and superimposed ( spherical ) densities of mn@xmath9 ions with half - filled t@xmath12 shells and o@xmath7 ions . the plots therefore show the charge density of the mn e@xmath4 electron and the distortion in density of an o@xmath7 ion in the solid state . c d d d d d state & energy ( mev ) & main saf & t@xmath12 exchange & o to e@xmath4 ( 1e ) & o to e@xmath4 ( 2e ) + singlet & + 3.6 & 1.0000 & 0.0000 & 0.0000 & 0.0000 + septet & 0 & 1.0000 & 0.0000 & 0.0000 & 0.0000 + singlet & -149.6 & 0.9926 & 0.0005 & 0.0038 & 0.0017 + septet & -133.4 & 0.9943 & 0.0000 & 0.0027 & 0.0017 + c d d d ion & x & y & z + la & 0.549 & 0.250 & 0.010 + mn & 0.000 & 0.000 & 0.000 + o & -0.014 & 0.250 & -0.070 + o & 0.309 & 0.039 & 0.224 + la & 0.517 & 0.250 & 0.001 + mn & 0.000 & 0.000 & 0.000 + o & -0.002 & 0.250 & -0.027 + o & 0.290 & 0.014 & 0.237 + la & 0.500 & 0.250 & 0.000 + mn & 0.000 & 0.000 & 0.000 + o & 0.000 & 0.250 & 0.000 + o & 0.2625 & 0.000 & 0.2625 + c d d d d d spin and orbital ordering & relative energy ( mev ) & @xmath88(@xmath89 + & & + fm d@xmath18 d@xmath18 & 0.0 & 4.00 + fm d@xmath18 d@xmath19 & -131.5 & 4.00 + fm d@xmath19 d@xmath19 & -6.1 & 3.99 + a - af d@xmath18 d@xmath18 & -14.4 & 4.05 + a - af d@xmath19 d@xmath19 & -34.4 & 3.87 + g - af d@xmath18 d@xmath18 & -34.9 & 3.88 + g - af d@xmath18 d@xmath19 & -95.4 & 3.89 + g - af d@xmath19 d@xmath19 & -34.0 & 3.88 + c d d d d d structure and spin ordering & energy(mev ) & @xmath88(@xmath90 ) + _ pnma _ ( experiment ) fm & 0.0 & 4.00 + _ pnma _ ( experiment ) a - af & -1.2 & 4.00 + _ pnma _ ( experiment ) g - af & 13.9 & 3.96 + _ pnma _ ( optimised ) fm & 0.0 & 4.00 + _ pnma _ ( optimised ) a - af & -2.0 & 3.96 + _ pnma _ ( optimised ) g - af & 21.9 & 3.94 + jahn - teller fm & 0.0 & 4.00 + jahn - teller a - af & 1.1 & 3.98 + jahn - teller g - af & 33.6 & - + c d d & j@xmath24 ( mev ) & j@xmath23 ( mev ) + experiment & 4.8 & -6.7 + uhf(experiment ) & 0.6 & -3.7 + uhf(experiment ) & 0.8 & -3.5 + uhf(optimised ) & 1.0 & -6.0 + uhf(jahn - teller ) & -0.6 & -8.1 + lsda(experiment ) & -3.1 & -9.1 + cluster ci(experiment ) & 3.3 & -3.6 + cluster ci(optimised ) & 5.1 & -7.4 + cluster ci(optimised / la pseudopotential ) & 5.2 & -7.4 + model hamiltonian & 4.8 & -3.7 + c d d d d d state & energy ( mev ) & main saf & t@xmath12 exchange & o to e@xmath4 ( 1e ) & o to e@xmath4 ( 2e ) + singlet & + 11.9 & 1.0000 & 0.0000 & 0.0000 & 0.0000 + nonet & 0.0 & 1.0000 & 0.0000 & 0.0000 & 0.0000 + singlet & -93.5 & 0.9937 & 0.0006 & 0.0037 & 0.0007 + nonet & -83.3 & 0.9954 & 0.0000 & 0.0030 & 0.0007 + singlet & + 17.9 & 1.0000 & 0.0000 & 0.0000 & 0.0000 + nonet & 0.0 & 1.0000 & 0.0000 & 0.0000 & 0.0000 + singlet & -64.9 & 0.9949 & 0.0004 & 0.0025 & 0.0006 + nonet & -79.9 & 0.9946 & 0.0000 & 0.0038 & 0.0008 +

a comprehensive search for variable and transient radio sources has been conducted using 55,000 snapshot images of the _ first _ survey . we present an analysis leading to the discovery of 1,627 variable and transient objects down to mjy levels over a wide range of timescales ( few minutes to years ) . variations observed range from 20% to a factor of 25 . multi - wavelength matching for counterparts reveals the diverse classes of objects exhibiting variability , ranging from nearby stars and pulsars to galaxies and distant quasars . interestingly , more than half of the objects in the sample have either no classified counterparts or no corresponding sources at any other wavelength and require multi - wavelength follow - up observations . we discuss these classes of variables and speculate on the identity of objects that lack multi - wavelength counterparts .

1107.5901

the first celestial radio source detected the sun was discovered as a consequence of its variability @xcite . nonetheless , while variability has continued to be a source of discovery and insight in the radio regime , few dedicated searches for variables and transients have been undertaken , primarily because the observing times required to conduct deep wide - field surveys are large and current instruments tend to have poor figures of merit for variability studies @xcite . indeed , one of the key scientific aims of the next generation of radio telescopes such as lofar , askap , mwa and , eventually , the ska is a study of radio variability . as discussed by @xcite , lofar will survey the sky for pulsars and fast transients at low frequencies ( 30 - 240 mhz ) with an emphasis on timescales of less than a second . cvs , x - ray binaries , grbs , sne , agn , flare stars , exoplanets and many more new phenomena , as well as extrinsic causes such as interstellar and interplanetary scintillation , are expected to contribute to its inventory of radio transients and variables @xcite . the australian ska pathfinder ( askap ; * ? ? ? * ; * ? ? ? * ) is designed to achieve very high survey speeds and noise levels down to @xmath1 @xmath2jy in an hour in the 1 ghz band , and will shed light on grbs , radio supernovae ( rsne ) , intra - day variables ( idvs ) , etc . the murchison widefield array ( mwa ) will operate in the 80 - 300 mhz range carrying out a blind search for variables and transients in addition to targeting explosive events , stellar and planetary phenomena , and compact objects over a range of timescales from nanoseconds to years @xcite . the evla @xcite will offer vastly improved sensitivity along with dynamic scheduling , providing a host of new capabilities for transient and variable searches . finally , a search for radio variables and transients also forms an important part of the scientific objectives of the square kilometer array ( ska ) . the ska is expected to discover a number of classes of variable radio objects such as pulsars and magnetars , grbs that are @xmath3-ray loud and @xmath3-ray quiet in both afterglow and prompt emission , sub - stellar objects such as brown dwarfs and exoplanets , microquasars , and potentially new classes of astrophysical phenomena @xcite . most of the research on radio variability to date has focused on bright radio samples ( e.g. , @xcite , @xmath4 jy at 408 mhz ; @xcite , @xmath4 jy at 5 ghz ; @xcite , @xmath5 jy at 5 ghz ) . at these flux densities , the radio source population is dominated by agn , while at fainter flux densities ( @xmath6 1 mjy ) , it is dominated by star - forming galaxies @xcite . research on the variability of faint radio sources has so far yielded small samples as a consequence of the small areas searched and the long integration times required . @xcite observed the lockman hole region on timescales of 19 days and 17 months down to 0.1 mjy at 1.4 ghz and discovered nine variable sources , providing an upper limit to the areal density of variable radio sources , as well as constraints on the beaming angle of grbs and confusion limits relevant to searches for grb afterglows . archival vla calibrator observations spanning 22 years with 944 independent epochs have been used by @xcite to search for radio transients over a single field with half - power beam widths of 90 and 54 at 5 ghz and 8.4 ghz , respectively . the typical observation s flux density threshold is @xmath7 @xmath2jy . they detect eight transients in single epochs and two transients in two - month averages of the data . two of the transients were identified as rsne , while the absence of optical counterparts for the remainder offers a wide variety of possibilities including orphaned grb afterglows ( ogrba ) , stellar sources , propagation effects , microlensing events , or perhaps mechanisms heretofore unknown . in the galaxy , single radio bursts detected from the parkes multibeam pulsar survey revealed a new population of neutron stars : rotating radio anomalous transients ( rrats ; * ? ? ? galactic center radio transients ( gcrts ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) have diverse light curves with outburst periods and burst durations varying from minutes to months . lacking counterparts at other wavelengths , their hosts , and the physical mechanisms involved , remain a mystery . two large , sensitive , wide - field surveys exist at radio wavelengths . the nrao vla sky survey ( nvss ; * ? ? ? * ) covered the entire sky north of -40declination at 1.4 ghz with a resolution of 45and an rms noise of 0.45 mjy . the faint images of the radio sky at twenty - cm survey ( _ first _ ; * ? ? ? * ) has covered @xmath8 9,000 deg@xmath9 of the sky at 1.4 ghz with a uniform rms of 0.15 mjy and an angular resolution of 54 . a number of authors have used these two surveys in their overlap region to search for radio transients . @xcite tried to constrain the detection rates of orphaned radio afterglows often associated with grbs . @xcite attempted to characterize the sample generated by @xcite . they report the detection of a radio sn in a nearby galaxy and the detection of a source with no optical counterpart , concluding the latter is unlikely to be associated with a grb . they also place tighter constraints on the beaming factor of grbs and a limit on total rate of nearby relativistic explosions , implying that most core collapse sne do not eject unconfined relativistic outflows . these studies have been complicated by the mismatched resolutions and flux density sensitivities of the nvss and the _ first _ surveys . @xcite have used @xmath0 120 deg@xmath9 of the _ first _ survey data near @xmath10taken in 1995 and then repeated in 2002 to study the optical properties of sources that exhibit significant radio variability at 1.4 ghz over this seven - year interval . they find 123 variable objects with flux densities ranging from @xmath0 2 - 1000 mjy . they conclude that there is a higher fraction of quasars in the sample of variables compared to the non - varying sample . more recently , @xcite have raised a very interesting possibility that the mysterious transients of @xcite observed out of the plane of the galaxy , based on their areal number density , duration and energy characteristics , could have a progenitor population consistent with being galactic isolated old neutron stars with mean distances of the order of kpc . x - ray follow - up @xcite of the transients reported by @xcite mostly resulted in non - detections and the x - ray flux upper limits imply consistency of the progenitor population with the above possibility besides being consistent with extreme flare stars at @xmath8 1 kpc or less extreme flares from brown dwarfs at distances of @xmath0 100 pc . here , we use the data from the _ first _ survey to create the largest , unbiased sample of variable radio sources to date down to a sensitivity level of a few mjy . in [ first ] , we describe the attributes of the _ first _ survey relevant to a study of radio variability . in [ dataanalysis ] , we present our approach to extracting a sample of variables from the more than two million individual source observations contained in the database . we then describe ( [ selection ] ) our refinement of the list of variable candidates , the parameterization of their variability , and a summary of their properties . in [ catalog ] , [ crossid ] and [ summary ] , we present the results of our analysis , including the cross - identification of our sample of transients and variables with existing data at optical wavelengths ( and in other spectral regimes ) , and a discussion of the classes of variables identified , as well as the new populations which might exist among the unidentified radio variables . the _ first _ survey , described below , is the most sensitive large - area radio survey in existence , and provides a database that is uniform in angular resolution and flux density sensitivity . until the next generation of radio instruments arrives , it offers the opportunity to produce the largest unbiased survey yet undertaken for radio transients and variables . we outline here the salient characteristics of the survey . the _ first _ survey covered @xmath0 9,055 deg@xmath9 ( 8,444 deg@xmath9 in the north galactic cap and 611 deg@xmath9 in the south galactic cap ) with observations conducted between 1993 and 2004 using the nrao vla in its b - configuration operating at a frequency of 1.4 ghz @xcite . roughly 65,000 three - minute snapshot images were obtained over this period . all of the work published to date , including the _ first _ catalog , has been based on images constructed by co - adding these snapshots with appropriate weightings ( ( for details , see * ? ? ? * ) ) ; this process yielded images with a uniform noise level of @xmath0 0.15 mjy . the final catalog contains 816,331 sources with a source detection threshold of @xmath0 1 mjy , yielding a source density of @xmath0 90 deg@xmath11 . the astrometric accuracy of the cataloged sources is better than 1 . the co - added maps have 18 pixels and an angular resolution of 54 . the fwhm of the vla s primary beam is @xmath0 30 . the catalog also assigns a sidelobe probability to each source . it must be noted that , for this study , we have made use of `` 08jul16 '' version of the catalog available on the _ first _ survey s webpage ( http://sundog.stsci.edu/first/catalogs/readme.html ) . @xcite describe the survey design and objectives in full detail . here , we provide a brief review of the aspects of the survey relevant to a variability search . in order to obtain uniform sensitivity and sky coverage in the most efficient manner possible , a hexagonal grid is optimal . the pointing centers of each snapshot were placed at the vertices of this hexagonal grid , the shape of which changes slowly with declination . the typical spacing between two neighboring snapshots is @xmath6 26 ; thus , all adjacent snapshots overlap . the time between overlapping observations varies : @xmath0 3 minutes between adjacent snapshots along the same declination strip in the east - west direction , @xmath0 1 to many days between adjacent snapshots that lie on two adjacent strips of declination in the north - south direction , and @xmath0 1 to a few years between two adjacent blocks of annual allocations . thus , while the survey was not undertaken to find variables and transients , its design is implicitly conducive for a search for radio variability on timescales of minutes to years . in this project , we use all of the snapshots from the north galactic cap data which covers 8,444 deg@xmath9 . part of the south galactic cap coverage was repeated , and is in the process of being observed at a third epoch ; preliminary variability results have already been reported by @xcite and a comprehensive analysis of this data set will be undertaken separately when the current observations are completed . in addition , we have restricted our analysis to isolated point sources . from light travel - time arguments , extragalactic sources that vary significantly on timescales of minutes to years are necessarily compact . even though compact components of extended sources could vary , we limit our attention herein to point - like ( unresolved ) sources with a sidelobe probability below 0.15 . further , an additional selection criterion was imposed requiring these sources to be relatively isolated , with no neighbors within 18 ; this minimizes the likelihood that differing multi - component fits between observations will produce spurious evidence of variability . figure [ firstflux ] shows the flux density histograms of all sources , point - like sources , and isolated point - like sources . out of a total source count of 758,942 in the northern galactic cap , there are 647,550 sources with sidelobe probabilities less than or equal to 0.15 . of these , 519,537 are isolated sources and 279,407 meet our criterion for being point - like ( major axis @xmath12 597 for @xmath13 4 declination and major axis @xmath12 687 for @xmath14 4 declination ) . we examine the variability characteristics of these 279,407 sources which have been observed between two and seven times each . it recently became clear that the on - line vla recording system suffers from a rare glitch that manifested itself many times in the _ first _ snapshot database . although the exact nature and cause for this glitch have not been completely understood , the result is that a few visibilities from one snapshot can be appended to the adjacent snapshot . this issue has also been pointed out by @xcite . the prototypical example that led to the discovery of this glitch is the source claimed to be a transient based on a comparison with the nvss : vla 172059.9 + 38522 @xcite . the authors speculated that this could be a prototype for a new class of transients . upon investigation , it was discovered that one of the snapshot images containing the location of this source had been affected by the vla glitch . a thorough search of the database found that 190 snapshot images ( 0.29% ) are apparently affected by this problem ; while in many cases the problem can be fixed by deleting a few records from the @xmath15 data set , in other cases subtler anomalies appear . as a consequence , we have simply excluded all these snapshot images from our analysis . additional grid images excised include those found to contain imaging artifacts and those with large calibration errors ( see below ) . using the positions of isolated , point - like sources from the _ first _ catalog , we created a comprehensive inventory of each object location and the corresponding pixel locations in each of the grid images in which it appeared . every object was found to be covered by from two to seven snapshot images as expected from the survey design described above . in each of the snapshot images , at the location of the object , a peak flux density was estimated by fitting a gaussian plus a flat baseline component . the position , width and position angle of the gaussian , already available from the catalog , were held fixed . the rms noise in each image was also measured around the object s location . for every pixel in each image , the `` local '' standard deviation was calculated by considering a box four times the fwhm of the synthesized beam centered on that pixel ( the fwhm is three pixels ) . assuming an object of interest is located at a given pixel , we consider a circle consisting of pixels at radius 5 times the fwhm of the synthesized beam . thus , each of the pixels in this circle has a standard deviation that was effectively calculated using pixels in an annulus that has roughly an inner radius of 3 times and an outer radius of 7 times the fwhm of the synthesized beam . we associate the median of all these standard deviations with the rms noise around the object . the advantage of using a median is it effectively ignores pixel intensities that are not purely noise such as those from another source ( or a sidelobe ) near to the object of interest . in the case of a marginal ( or non- ) detection of the object in a snapshot image , a @xmath16 upper limit was assigned . artefacts in radio interferometric images arise for a number of reasons such as calibration errors , interference , etc . for a large - scale survey such as _ first _ , it is difficult to ensure high data quality at all times during the observations . one not uncommon problem is a striped pattern corrupting an image owing to un - excised interference or other deconvolution problems . after deriving a preliminary list of candidate variables , we plotted their positions on the sky , and found that some objects were spatially clustered . these clusters had different shapes , including linear , elliptical and irregular appearances , and their sizes spanned several degrees on the sky . such strong clustering is highly suggestive of systematic problems with the images . we investigated the snapshot images associated with these clustered objects and found that many , especially toward the southern limit of the survey , have clearly visible striped patterns running across them . the width of these stripes is typically greater than the angular resolution of the images . having restricted ourselves to isolated point - like sources , we were able to make use of the baseline fitting described in [ gaussfit ] . we flagged any image that contained one or more sources with a baseline component greater than 1.75 times the local image noise . a total of 2162 ( 3.96% ) such snapshot images were deleted from subsequent analysis . figure [ sky_positions ] illustrates how this step significantly eliminated the non - physical clustering of variable and transient source candidates . while amplitude and phase calibration errors are minimized in the _ first _ imaging pipeline for individual snapshots , it is important to keep in mind that the survey is spread over very large scales in space and time . the overall calibration scale , then , could well vary over the course of the survey ( see * ? ? ? relevant to our study is the variation of the overall amplitude calibration scale over the tens of thousands of snapshot images . an estimation of the calibration error was performed after a first iteration of steps in [ empiricalpbeam ] but before the identification of any variables or transients . the algorithm used considers sources in each snapshot image and compares their behaviour to that in all the neighbouring snapshots in which they are also found . the null hypothesis , true for a vast majority of the sources , is that sources will exhibit a constant peak flux density across the different snapshots in which they are detected . a best - fit algorithm assigns a multiplicative amplitude calibration correction factor to each snapshot image such that , when it is applied , all the sources in the neighborhood are as close to the null hypothesis behavior as possible in a minimum absolute deviation sense . thus , the amplitude calibration variations are estimated locally . figure [ mwrathist ] shows the histogram of the amplitude correction factors , where a value of unity indicates no correction need be applied . it can be seen that the distribution , especially the inner core comprising the large majority of snapshot images , can be reasonably fit with a gaussian profile with a standard deviation of 4.3% . however , tails can be seen on both sides of the distribution indicating some large fluctuations in the amplitude calibration for a small fraction of the images . we have accounted for the typical calibration uncertainty using a conservative value of 5% in determining the flux density uncertainty for individual snapshots as described in [ errors_contrib ] . we experimented with different thresholds for the largest allowable amplitude calibration offsets . we set the threshold at 20% ; for amplitude offsets below this value , the number of candidate variables eliminated was proportional to the number of images eliminated , whereas snapshots with higher amplitude corrections contained disproportionally larger numbers of candidates . we also performed an extra iteration , computing the amplitude correction factor after removing sources that were found to be candidate variables and recomputing the calibration correction factors . we find evidence that the change in the amplitude correction factors between the two iterations is greater for snapshots that contain putative variables than for ones that do not ( figure [ mwratiter ] ) . however , the final list of variable candidates between one such iteration and the next changes by only @xmath0 2% . thus , we have used only snapshot images that have an amplitude offset less than 20% after one iteration of the amplitude correction factor estimation . for the set of images retained , we apply , on an image by image basis , the overall amplitude calibration correction factors to the peak flux densities , local image noise values , and baseline components for all the sources found in those images . hereafter , these corrected values are treated as the true values . the data were obtained in 2@xmath177 3-mhz channels near 1.4 ghz . while the center frequency of these bands was fixed for most of the survey observations , the interference environment was constantly changing , leading to data editing which affects the effective center frequency of the observing band(s ) . thus , the frequency - dependent primary beam , normally derived from the aips task pbcor , can not be adopted blindly . our selection algorithm makes use of the hypothesis that the non - varying source population is expected to dominate the variable one . hence , with the off - axis angle information available for each source in each snapshot image , the primary beam can be determined empirically . we see from figure [ pbeam ] that the difference between our estimate for the primary beam and the nominal vla primary beam is at most a few percent . as figure [ errors ] demonstrates , the uncertainty arising from the primary beam estimation is a negligible contribution to the overall error budget , independent of source flux density . the peak flux density of each isolated , point - like object measured in each snapshot image ( as described in [ gaussfit ] ) is normalized by the reported catalog flux density ( after taking into account the clean bias as quantified by * ? ? ? * ) . since the snapshot images and the fitted peak flux densities derived therefrom have not been corrected for the primary beam attenuation , adopting the null hypothesis that a source s flux density stays constant means that these normalized values yield a primary beam normalization at the off - axis angle of the measured source when this normalization is performed for all objects , we obtain an empirical primary beam : @xmath18 where cb is the clean bias @xcite . the empirical primary beam values obtained from all the objects in each of their corresponding snapshot images were binned in angular offset from the center of the primary beam . a sixth - degree polynomial was fitted to determine an analytical expression for the empirical primary beam . the error in the primary beam estimate is the uncertainty in the mean estimate for the primary beam in each bin , which is the standard deviation in each bin divided by the square root of the number of sources in that bin . thus , a primary beam error estimate ( used in [ errors_contrib ] ) is assigned to each data point on each light curve depending on which bin of angular offset the data point falls into . figure [ pbeam ] shows the scatter plot of normalized flux densities from all the usable snapshot images in the upper panel ; in the lower panel , the bin - averaged data points from the upper panel are overplotted on the polynomial fit . fits were obtained independently in eight separate declination zones , but no significant variations ( @xmath19 0.9% ) with declination were seen , so a single empirical primary beam was used for the whole survey . we next estimate the sources of uncertainty in the peak flux density obtained for an object detected in a snapshot image . if @xmath20 , where @xmath21 is the reported flux density , @xmath22 is the absolute amplitude calibration ( ideally @xmath23 ) and @xmath24 , the primary beam value at an offset ( @xmath25 ) from the field center , the principal uncertainties that need to be accounted for include the following : 1 . local noise in the grid image in the vicinity of the source ( @xmath26 ) : this will further be amplified because of the primary beam effect , especially for sources far from the field center . uncertainties in primary beam ( @xmath28 ) : we have computed empirically the primary beam and its uncertainty from the data as described in [ empiricalpbeam ] . 3 . uncertainties in the overall flux calibration amplitude ( @xmath30 ) : this uncertainty is @xmath0 5% and was estimated as described in [ mwrat ] . 4 . pointing errors of the vla antenna ( @xmath32 ) : this is not the same as the positional accuracy of the sources in the catalog . the antennas have an rms pointing error of @xmath33 10 . this translates to an error in flux density , @xmath34 . the overall error is obtained by adding all the above errors in quadrature . the dominant contributions come from the map noise for fainter sources and from the amplitude calibration errors for brighter sources . @xmath35 figure [ errors ] shows the contributions of these different sources of uncertainty as a function of angular offset from the center of the image for two sources observed ( without primary beam correction ) to have flux densities of 1 mjy and 10 mjy , respectively . three other potential sources of error have been assessed and found to be negligible compared to the above terms . firstly , we adopt for each source the first catalog position and hold it fixed when fitting for a source s peak flux density in each grid image . any systematic astrometric errors which are a function of position in the primary beam field of view will thus induce an underestimate of source flux density . we have assessed such errors and report the results for the co - added images in @xcite . we have also assessed the errors for the grid images and find similar results : the maximum error is @xmath3603 . this will induce a a maximum flux density error of @xmath37 . a second potential effect involves the ( generally unknown ) source spectral index . since the primary beam shape is a function of observing frequency , the primary beam correction changes from the lower to the upper end of the observing band . while our empirical determination of the primary beam correction assures that the average value is correct for the average source spectral index , a source with an extremely steep or inverted spectrum will have an inappropriate primary beam correction applied . we have evaluated the magnitude of this effect for extreme sources with spectral indices of @xmath38 and @xmath39 compared to a mean source spectral index of @xmath40 . at an off - axis angle of 20the error is less than 1% in all cases . at 30off axis , it is @xmath41 for @xmath42 ; for an extremely inverted ( and very rare ) source with @xmath43 the error does reach 4% , but the other errors rise rapidly this far off - axis and still dominate the total error budget . finally , we address the matter as to whether or not variations in the small clean bias added to each measured flux density could generate spurious variability . clean bias is an at - best partially understood manifestation of the non - linear cleaning algorithm applied to snapshot data ( see * ? ? ? * ; * ? ? ? it is known to vary with source extent and map rms ; in principle , it could also vary with time . for the first catalog , artificial source experiments ( inserting sources of known flux density into a @xmath15 data set and then measuring the flux density recovered in the derived image ) suggested that , for the point - like sources considered here , a constant value of 0.25 mjy provided a satisfactory correction . becker et al . ( 1995 ) did note that for fields with sources brighter than 500 mjy ( and thus higher rms values ) , the bias correction could be as great as 0.50 mjy leading to a 0.25 mjy underestimate of the true flux density . all variable candidates in such fields discovered here are flagged . subsequently , we examined this issue in @xcite using artificial sources as well as an analysis of sources from the deep radio imaging of the spitzer first - look survey @xcite . for 145 sources from different fields in the flux - density range 1 to 3 mjy , the clean bias in first images was shown to be @xmath44 mjy . thus , even for 1 mjy on - axis sources , the size of any clean - bias error is less than that of image noise and the amplitude correction error ; for brighter , and/or off - axis sources , the uncertainty is negligible . owing to the reasons discussed in [ vlaglitch ] , [ hibase ] and [ mwrat ] , the snapshot images that form the basis of our search are a subset of those from which the _ first _ catalog was constructed . the catalog peak flux densities were derived by a weighted average estimate of the inferred fluxes where the weights are related to the square of the primary beam pattern @xcite . we have recalculated the mean value for each source using only those snapshot images actually included in our search and denote it by @xmath45 , where @xmath46 . @xmath45 and @xmath47 will be used interchangeably hereafter . inferred peak flux density points that lie far from @xmath45 are considered outliers . we have selected as candidate variables outliers that deviate from @xmath45 by more than five times their inferred peak flux density uncertainty , defined in [ errors_contrib ] . we have identified several additional effects other than genuine variability that can generate outliers . * * low - quality images * : noisy images and images with interference and other deconvolution errors are significant sources of false detections . such outliers have been largely eliminated using the technique described in [ hibase ] ; a few additional bad fields containing large numbers of spurious sources were also deleted . * * sidelobes * : there were instances in which sidelobes from nearby sources have significant flux density in one grid image but are absent or marginal in others , thus producing an apparent source that looks both genuine and variable . first _ catalog - generation algorithm does calculate a sidelobe probability for each source , and we selected only objects with a sidelobe probability less than 0.15 . nonetheless , some unlabeled sidelobes are still present . all fields containing bright sources were examined carefully to eliminate sidelobes as spurious variables . variable and transient sources suspected of being sidelobes from strong sources with peak flux densities of @xmath19 500 mjy within a radius of 31 have been flagged . * * calibration problems * : defective calibration can cause sources to vary between different snapshot images even after incorporating the grid image amplitude correction factors . to identify and eliminate such sources , we used the catalog to select a few ( usually between two and four ) objects that are neighbors of each outlier ( within a 9 radius ) and are also point - like . these were normalized by their respective mean peak flux densities . each was examined for any deviations from the ideal normalized value of unity . some neighbors exhibited the same variability pattern as the outliers , thus confirming the presence of remaining calibration problems . some others were simply inconsistent with no variation , making the corresponding outliers suspect . we used this procedure on all outliers to select a final list that appeared to be free from calibration problems . about 30 outliers have neighbors with a deviation from the mean of @xmath16 ( @xmath48 would be expected from a normal distribution ) , and 17 have a value @xmath49 ( figure [ nghbrs_behaviour ] ) ; the table includes the deviation for the neighboring sources in sigma for each source . we also checked to see if there is any systematic variation of the fraction of variables and transients with off - axis angle . we plotted the relative frequencies of the off - axis angle for the source in the light curve containing the data point most discrepant from the mean peak flux density . simulataneously , we plotted the same quantities for all the isolated point - like sources in _ first_. in figure [ hist_offset ] , we find no major systematic differences between the outliers and the rest of the isolated point - like sources in _ first_. the slight shift of the distribution of variables toward the field center arises from the fact that the low _ observed _ flux densities for sources far off axis increases their uncertainty , and thus reduces our ability to detect variability . after removal of all identified false positives , we are left with a sample of 1627 variables and transients . we used three indicators to select variable and transient candidates . we define @xmath51 and @xmath50-probability @xmath52 as the probability that a @xmath50-statistic falls above a certain value @xmath51 , with @xmath53 inferred peak flux density of the @xmath54 data point , @xmath55 average inferred peak flux density weighted by the primary beam , @xmath56 uncertainty in @xmath57 , @xmath58 is the number of degrees of freedom , and @xmath59 , is the number of data points in the light curve . since there is only one constraint in the model for the null hypothesis that the peak flux density is constant with time it follows that @xmath60 . the probability threshold chosen is @xmath61 which is equivalent to the probability that the absolute value of a normally distributed variable falls beyond five standard deviations . an alternate indicator of variability is the maximum deviation of any individual data point in a light curve from the null hypothesis that the flux density is constant at a value given by the mean inferred peak flux density . we denote this by @xmath62 defined as @xmath63 . we chose a threshold @xmath64 . in order to preserve information about variations between any pair of data points in a light curve ( and the associated timescale ) , we also estimate the variation of each point from all the others in a light curve . the significance of this difference is given by the ratio obtained by dividing the difference by the uncertainties in the two flux density estimates added in quadrature , @xmath65 where , @xmath66 significance ratio between @xmath54 and @xmath67 data points , @xmath53 inferred peak flux density at the @xmath54 data point , @xmath68 inferred peak flux density at the @xmath67 data point , @xmath56 uncertainty in @xmath57 , and , @xmath69 uncertainty in @xmath70 . we select the pair that has the maximum absolute value defined as @xmath71 . for an outlier selected with this indicator , the threshold is @xmath72 . @xmath73 is an indicator of overall variability of a light curve that can arise from low but sustained variations , large , sudden variations , or both . although @xmath73 provides a valid statistical indicator of variability , information about timescales and the temporal localization of the variability has been ignored . unlike the indicator @xmath73 , @xmath62 and @xmath74 are indicators of large sudden variability and temporally localize the variability in the light curve . a similar estimation of @xmath73 , @xmath62 and @xmath74 are performed for the neighbours of the outliers . the respective thresholds were chosen as those beyond which the population of varying sources and their non - varying neighbours look statistically different as defined by a k - s test . we found that the variability measure given by @xmath75 identified no new variables that were not already found by one of the other two measures , namely , @xmath61 and @xmath64 . however , we include this variability measure in subsequent discussions because it represents a useful quantity in describing the most significant flux density difference between data points in a light curve and provides us with an idea of the timescale of variability . figure [ scatter_plot_best_sigma_mean_fpeak ] shows the distribution of @xmath62 against @xmath45 . note that a few objects have a relatively low significance as per the measure @xmath62 , but are still included as variable when above the threshold of the other indicator of variability @xmath73 . figure [ hist_chisq_prob ] illustrates the distribution of outliers described by the parameter @xmath73 . the fraction of outliers to the right of the dashed line are those selected by the criterion @xmath64 . the number with a left arrow indicates the number of outliers with even smaller probabilities that could not be represented in the plot . the sky positions of the outliers are shown in figure [ sky_positions ] . one notable feature is that the density of outliers below a declination of @xmath76 4is enhanced . this is because the synthesized beam with which the data is convolved increases below a declination of @xmath76 4as is clearly illustrated in figure [ dec_ps_size ] . consequently , more objects meet the criterion for a point - like source , increasing the density of both non - variable and variable objects . table [ onepagesummarytable ] presents a sample page from the list of 1627 highly variable radio sources identified in our study . the complete table is available online in electronic format . column 1 provides the source position derived from the _ first _ catalog . columns 2 and 3 give , respectively , the _ first _ catalog peak flux density and the mean peak flux density derived from the light curves . the range of flux densities for the source derived from the grid images are in column 4 , while column 5 provides the nvss catalog peak flux density . note that , while _ first _ and nvss have very different angular resolutions , our restriction of this study to isolated point sources in the higher - resolution _ first _ survey means resolution effects should not be a factor in most cases . these data are followed by the number of observations available ( col . the next three columns provide the three measures of variability : @xmath77 for the light curve ( col . 7 ) , the maximum deviation of a single data point from the mean flux density in the light curve ( col . 8) , and the difference ( in @xmath78 ) between the most significantly different pair of data points ( col . @xmath77 is the number of standard deviations on both sides of a normal distribution beyond which the probability from this normal distribution matches @xmath73 . the maximum - to - minimum flux density ratio is provided in column 10 . the @xmath79 ( days ) in column 11 denotes the minimum timescale at which two data points differ by at least @xmath80 . in cases where the maximum absolute difference of any pair of data points never reaches a value @xmath80 , the value denotes the timescale at which the maximum absolute difference in column 10 is attained . in a few cases where the date of observation can not be reliably obtained from the snapshot header , the entry is left without a numerical value . column 12 contains a flag ` n ' if the outlier is found to be in the vicinity of a source brighter than 500 mjy within 31 . column 13 provides the distance by which the mean normalized peak flux density of the outlier s neighbors departs from the mean of the distribution of such normalized neighbor peak flux densities from all outliers , and is a proxy for the quality of the field . the higher the absolute value in this column , the less ideal the quality of the field surrounding the outlier and hence , the less reliable the physical variability of the outlier . the type of light curve is denoted in column 14 by v ( variable ) and t ( transient ) . the apparent sdss-@xmath81 band magnitudes are provided in column 15 where a sdss match is available . finally , column 16 gives the best ranked counterpart at other wavelengths . this ranking is explained in [ crossid ] . figure [ histogram_max_min_ratio_all ] shows the distribution of the ratio of maximum to minimum flux density for our 1627 variables which ranges from 25% to a factor of @xmath19 20 ; overplotted is this ratio for the sample of variable sources @xcite with the same flux density distribution as ours . the areal density of their variables is @xmath0 1 deg@xmath11 while we find it to be @xmath0 0.2 deg@xmath11 for our sample . this difference is primarily due to the lower threshold of detection ( @xmath82 ) used by @xcite . sources for which lower limits are used are indicated by shading . the relation between maximum - to - minimum flux density ratio and mean peak flux density is shown as a scatter plot in figure [ scatter_max_min_ratio_mean_fpeak ] . table [ crossid_summary_table ] presents the results of cross - matching our catalog of variable sources with a variety of catalogs and databases . the first column lists the source type , followed by the specific catalogs used for cross identification and the matching radius adopted for each ( col . the rank in column 4 indicates the order in which the matches were applied ; e.g. , we first sought matches with bright star catalogs , then with a pulsar catalog , then with quasar catalogs , etc . this is important in interpreting the number of matches @xmath83 found in column 5 where we list both the number of total matches including those previously identified in a higher - ranked catalog and , in parentheses , the number of unique matches to the specified catalog . for example , the photometric sdss quasar catalog has 68 total matches , but only 67 of them are real ; the @xmath84 is also matched to a pulsar ( which is the true identification ) . column 6 gives the match number as a percentage of all variable sources ( e.g. , a total of 7.4% of all variables are sdss quasars ) . columns 7 and 8 , respectively , give the percentage of all _ first _ sources matching each catalog that are identified as variable ( e.g. , 76% of all _ first _ pulsars are variable , but only 1.2% of sdss spectroscopically identified quasars with _ first _ counterparts are highly variable ) , and the percentage of all _ first _ isolated point sources in the catalog that find identifications in the given catalog . parentheses are used as in column 4 . in all cases , the rate of chance coincidences is low , typically less than one source per category with the exception of sdss galaxies where up to seven chance coincidences ( out of 442 matches ) are possible . in summary , we find 5 bright stars , 13 radio pulsars , 120 quasars , and 489 galaxies among our variable sources , yielding firm identifications for 38% of our objects . an additional 118 objects are identified with unclassified stellar objects in the sdss and gsc ii databases ; a total of 50 objects have x - ray counterparts , although only 5 of these are not also identified optically ( mostly as quasars ) . this leaves 877 objects or 54% of the total without identifications in other wavelength bands ( although 499 of these are found in other radio catalogs ) . we discuss below each class of objects in turn , and compare the properties of the identified objects to the unidentified ones in order to seek clues as to the identities of the latter sources . in figure [ histogram_max_min_ratio_diff_types ] , we show how the maximum - to - minimum flux density ratio differs for the five different classes of variable objects . the pulsars are , as expected the most highly variable , while the sources classified as galaxies have slightly higher variability amplitudes than the quasars . the objects with no counterparts closely mimic the distribution of galaxies . figure [ histogram_radio_fluxes_diff_types ] illustrates the distribution of mean peak flux densities for the various source classes . as expected , the relative fraction of qsos with respect to galaxies increases with increasing mean peak flux density . the pulsars and the stars are much fainter with mean peak flux densities of a few mjy . again , the distribution of unidentified objects appears to follow the trend shown by galaxies and interestingly , it also has a significant tail at the high - mean peak flux density end . while @xmath0 70 known radio pulsars fall within our survey area , most are both faint and have steep radio spectra , leading to only 17 detections in the _ first _ survey catalog ( table [ psr_summary_table ] ) . of these , 9 ( @xmath0 50% ) appear in our catalog of highly variable sources . by slightly relaxing our criterion for a point - source , we find an additional four pulsars that would have been classified as variables by our algorithm ; we add these to the total number of outliers quoted throughout the paper . thus , more than three - quarters of the detected radio pulsars in the survey are found to be highly variable . while the intrinsic radio luminosity of a pulsar is steady , its emission suffers from interstellar scintillation which arises when signals traveling in slightly different directions are alternately scattered into , and out of , our line of sight by the fluctuating electron density distribution of the interstellar medium . the effect is largest when the number of scattering centers is small , so the highest variability is seen for the nearest pulsars ; indeed , 11 of our 13 variable objects have dispersion measures ( @xmath85 ) of @xmath6 20 @xmath86pc ; the other two have @xmath85 values of 27 @xmath86pc and 41 @xmath86pc . of the four pulsars not seen to vary significantly , two have flux densities @xmath36 1.5 mjy , too close to our threshold to detect variability , and a third has a @xmath87 . the typical time scale for scintillation is @xmath0 1 min at 20 cm ( to be compared with our 165 s integration time ) and the decorrelation bandwidth is typically smaller than our 50 mhz bandwidth , so our observational parameters largely smooth over the fluctuations . nonetheless , pulsars are among the most variable objects in our sample , with 8 out of 13 varying by more than a factor of 4 ( see figure [ histogram_max_min_ratio_all ] ) . interestingly , 17 of the 877 variables which lack optical counterparts entirely also vary by this large factor ; this is to be contrasted to the fact that none of the 120 sources identified as quasars are this variable , and only 1% of the sources coincident with galaxies vary by such large factors . all these 17 sources have flux densities less than 4 mjy and most appear only once or twice , falling below our detection threshold in the other observations . an examination of the radio spectrum and polarization of these sources could provide candidates for pole - on , very short period , and/or intermittent pulsars worthy of further study . a total of 732 radio variables have optical counterparts in one or more of the catalogs used in the identification program summarized in table [ crossid_summary_table ] ; the vast majority of these are from a match to the sdss dr7 catalog . figures [ histogram_magnitudes_diff_types ] shows sdss _ i-_band magnitudes for the various source classes , while figure [ scatter_plot_colors_diff_types_1 ] displays sdss color - color diagrams for the identified galaxies and quasars of our sample . we discuss each sample of objects in turn below . most stars are very faint at radio wavelengths . of the more than 800,000 sources in the _ first _ catalog , only 37 match objects in the bright star catalogs we examined to within 14 . to check that the potentially high proper motions of these ( mostly ) nearby stars have not obviated any true matches , we expanded the search radius to 5 ; no additional matches were found when proper - motion - corrected positions were used . @xcite conducted a more exhaustive search of the first 4,760 deg@xmath9 of the _ first _ survey and found a total of 26 radio - detected stars , 5 of which were at flux densities below the survey limit of 1.0 mjy . one of the detections reported in that earlier survey , the flare star ( 08 08 55.47 + 32 49 06.0 ) , is found in our catalog of highly variable objects . it has a light curve showing variability on timescales of days , months and years . in total , we find only 5 stars both bright enough and variable enough to make our sample of variables and transients . all are previously known variables . a total of 53 quasars from the sdss dr5 spectroscopic quasar catalog @xcite are found in our list of variables . in addition , another 67 objects in the dr6 photometric quasar catalog of @xcite are in our list , making a total of 120 quasars in the sample or 7% of all variables . for the whole _ first _ catalog there are 8396 and 10,025 matches from the spectroscopic and photometric catalogs , respectively , accounting for nearly 2.5% of all _ first _ sources . table [ crossid_summary_table ] , however , includes only matches with isolated point - like sources from the _ first _ catalog . the quasars have the highest mean radio flux density ( @xmath88 mjy ) among our identified source classes ( figure [ histogram_radio_fluxes_diff_types ] ) and the lowest mean max - to - min flux density ratio . this is in part a selection effect , in that only in bright sources does modest variability exceed our significance threshold . nonetheless , with our coarse and uneven sampling , we have found only a dozen quasars that vary by more than a factor of 2.5 . the largest identified segment of our sample is coincident with objects classified as galaxies in sdss ( with a few additional galaxies contributed from gsc ii , 2mass , ned , etc . ) . the 489 galaxies comprise roughly 30% of our sample . the galaxies have a broad spread in magnitudes ( figure [ histogram_magnitudes_diff_types ] ) and a lower mean radio flux density than the quasars but a somewhat higher amplitude of variability ( figure [ histogram_max_min_ratio_all ] ) . since our matching radius is only 14 , the radio sources are all coincident , within the uncertainties , with the galaxy nuclei . agn are thus undoubtedly the source of variable radio emission in virtually all of these objects . optical spectroscopy ( and/or hard x - ray imaging ) of this sample could be interesting , in that it could reveal the fraction of buried agn which show no optical evidence of an accreting black hole . we examined by eye sdss images for all 1627 variable source candidates . virtually all of the matched sources had clear counterparts and almost all of the unmatched sources were blank fields as expected . however , there were a handful of cases where there was a coincident optical object that did not appear in the sdss catalog ; these are noted in the table . in addition , there were several cases in which the radio source was superposed on a large ( @xmath1910 ) galaxy and not coincident with the galaxy nucleus . in a few cases , a backgound source was visible through the galaxy light and is the likely radio source counterpart . in several other cases , however , no background source is apparent although it could , of course , simply be obscured by the foreground galaxy s light or dust . a total of 114 radio variables are coincident with unclassified stellar objects in the sdss dr7 catalog ; a few additional such objects are found in the gsc ii catalog in areas lying outside of the sdss coverage . figure [ scatter_plot_colors_diff_types_2 ] shows the sdss color - color diagrams for these stellar objects ; comparison with figure [ scatter_plot_colors_diff_types_1 ] shows a large majority of these objects have colors consistent with quasars , while virtually all of the remainder fall within the galaxy contours in color - color space . the majority of these objects fall below the @xmath89 cutoff for the sdss photometric quasar catalog , but nearly 60 objects lie above this threshold . spectroscopic follow - up of these objects could provide insight into the completeness of the photometric quasar catalog include one source with no spectrum and two with spectra classified as ` star ' . ] . there are no optical ( or x - ray ) counterparts for 877 sources comprising 54% of our sample . figures [ histogram_max_min_ratio_all ] and [ scatter_plot_colors_diff_types_1 ] show that these sources mostly cohabit the parameter - space of galaxies and qsos , although there are a few notable outliers with very high variability amplitudes as noted above . after analyzing @xmath90 snapshot images from the _ first _ survey of the radio sky covering 8444 deg@xmath9 with sensitivity down to 1 mjy for variability and transient phenomena , we have assembled a sample of 1627 sources that are significantly variable . this sample was matched with multi - wavelength catalogs to identify counterparts . we found 13 radio pulsars , 53 sdss spectroscopic qsos , 67 sdss photometric qsos , 489 galaxies , 5 stars , 123 optically detected but unclassified sources and 877 objects lacking optical counterparts . the unidentified sources mostly occupy the parameter - space of galaxies and qsos but there are a few notable outliers . follow - up observations of several of these source classes would likely be fruitful . our study has shown that there is much to be discovered in the dynamic radio sky . exploration by the next - generation instruments sampling different portions of the spectral and temporal domains will prove to be highly productive . we are grateful to jacqueline van gorkom and zoltan haiman for their valuable comments on this paper . rhb s work was supported in part under the auspices of the us department of energy by lawrence livermore national laboratory under contract w-7405-eng-48 . the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc . this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . funding for the sdss and sdss - ii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , the u.s . department of energy , the national aeronautics and space administration , the japanese monbukagakusho , the max planck society , and the higher education funding council for england . the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium for the participating institutions . the participating institutions are the american museum of natural history , astrophysical institute potsdam , university of basel , university of cambridge , case western reserve university , university of chicago , drexel university , fermilab , the institute for advanced study , the japan participation group , johns hopkins university , the joint institute for nuclear astrophysics , the kavli institute for particle astrophysics and cosmology , the korean scientist group , the chinese academy of sciences ( lamost ) , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , ohio state university , university of pittsburgh , university of portsmouth , princeton university , the united states naval observatory , and the university of washington . -probability , @xmath73 . the fraction of outliers to the right of the dashed line are those selected by criteria other than @xmath73 . the number with a left arrow indicates the number of outliers with even smaller probabilities that could not be represented in the plot.[hist_chisq_prob ] ] lrrcrrrrrcccrccl 12 01 05.881 + 20 23 10.85 & 5.91 & 5.85 & 2.56 - 6.19 & 4.28 & 3 & 4.86 & 5.18 & 4.78 & 2.33 & 5 & & 0.09 & v & 21.24 & sdss - g + 12 01 07.863 + 24 53 33.32 & 8.78 & 8.96 & 3.04 - 9.15 & 8.08 & 3 & 6.97 & 7.27 & 6.50 & 3.01 & 2 & & -0.75 & v & & + 12 01 35.629 + 33 12 33.28 & 4.75 & 4.53 & @xmath363.05 - @xmath369.02 & 4.00 & 6 & 5.70 & 5.61 & 5.35 & @xmath191.58 & & & -1.53 & v & & + 12 01 49.193 + 03 39 03.44 & 9.18 & 9.05 & 4.67 - 9.58 & 8.73 & 3 & 4.90 & 5.22 & 4.64 & 1.95 & 1 & & 0.19 & v & 19.73 & sdss - g + 12 02 10.433 + 16 33 28.02 & 1.25 & 1.28 & @xmath361.05 - @xmath363.75 & & 4 & 5.32 & 4.06 & 5.25 & @xmath192.02 & 1 & & 0.23 & v & 21.28 & sdss - g + 12 02 25.829 + 30 37 38.12 & 4.44 & 4.46 & @xmath362.47 - @xmath367.81 & 3.76 & 3 & 6.51 & 6.39 & 6.17 & @xmath191.97 & 3 & & 0.65 & v & 19.39 & sdss - g + 12 02 40.649 + 21 56 20.99 & 8.40 & 7.14 & 6.76 - 11.24 & 9.19 & 3 & 5.27 & 5.27 & 5.28 & 1.66 & 0 & & -3.17 & v & 20.14 & sdss - g + 12 02 49.673 + 31 36 40.33 & 2.51 & 2.70 & @xmath361.67 - @xmath3615.62 & & 6 & 6.10 & 5.48 & 5.70 & @xmath191.83 & 95 & & -0.95 & v & & + 12 02 53.621 + 34 31 29.70 & 4.54 & 4.47 & 1.43 - @xmath3615.83 & 4.61 & 5 & @xmath918.00 & 7.09 & 6.79 & 3.43 & 5 & & 0.68 & v & & + 12 02 55.734 + 04 33 47.25 & 2.23 & 2.21 & 1.39 - 3.10 & & 3 & 5.19 & 3.65 & 4.87 & @xmath192.24 & 388 & n & -0.26 & v & & + 12 03 02.267 + 03 37 53.99 & 2.49 & 2.60 & @xmath361.66 - 3.29 & & 4 & 6.61 & 5.99 & 6.56 & @xmath191.99 & 1 & & -0.57 & v & 19.26 & sdss - qso(p ) + 12 03 19.427 + 04 38 15.57 & 3.64 & 3.58 & 2.24 - @xmath3613.88 & 3.39 & 4 & 6.36 & 5.13 & 5.72 & @xmath191.73 & 388 & n & -0.12 & v & & + 12 03 57.217 -01 38 26.51 & 2.20 & 2.34 & @xmath361.73 - @xmath367.93 & & 4 & 5.51 & 5.28 & 5.61 & @xmath191.70 & 1 & & 1.20 & v & 16.02 & sdss - g + 12 04 28.315 + 03 52 59.25 & 5.38 & 5.42 & @xmath364.02 - 7.42 & & 4 & 5.70 & 4.92 & 5.35 & @xmath191.53 & 1 & n & 0.62 & v & & + 12 04 39.877 + 50 28 20.71 & 2.32 & 2.32 & 1.43 - 3.59 & & 3 & 5.98 & 4.80 & 5.94 & 2.50 & 27 & & -0.10 & v & & + 12 04 50.878 + 29 20 33.40 & 1.90 & 1.66 & @xmath361.40 - 2.96 & & 3 & 5.63 & 4.35 & 5.92 & @xmath192.12 & 11 & & -1.05 & v & 21.79 & sdss - g + 12 04 53.441 -04 59 41.50 & 4.62 & 4.52 & 1.71 - 4.70 & & 3 & 6.30 & 6.49 & 5.94 & 2.76 & 40 & & -1.43 & v & & + 12 05 26.547 + 32 54 33.42 & 3.53 & 3.39 & 1.80 - 4.03 & 4.21 & 4 & 5.11 & 5.32 & 4.98 & 2.24 & 488 & & 0.44 & v & 15.29 & sdss - g + 12 05 51.716 + 21 11 20.45 & 10.99 & 11.12 & 6.29 - 13.44 & 8.62 & 5 & 6.11 & 6.53 & 5.50 & 1.82 & 5 & & 0.23 & v & & + 12 06 06.530 + 61 42 04.63 & 23.40 & 25.19 & 24.94 - 36.24 & 25.18 & 5 & 5.43 & 3.92 & 3.61 & 1.41 & 0 & & 0.18 & v & & + 12 06 08.024 + 36 43 31.29 & 3.11 & 3.01 & @xmath362.06 - @xmath3624.94 & & 6 & 5.28 & 5.33 & 5.40 & @xmath191.62 & 5 & & -0.41 & v & 21.89 & sdss - g + 12 06 18.560 -07 36 45.19 & 6.53 & 7.24 & 4.32 - 8.39 & 9.55 & 2 & 7.32 & 6.93 & 6.31 & 1.94 & 0 & n & -0.25 & v & & + 12 06 26.572 + 00 09 31.41 & 4.44 & 4.35 & 2.60 - 10.80 & 4.47 & 4 & 5.18 & 4.98 & 4.75 & 1.86 & 3 & & 0.93 & v & 18.13 & sdss - g + 12 07 28.795 + 25 30 18.66 & 17.03 & 16.69 & 16.41 - 38.70 & 13.66 & 6 & 5.40 & 4.93 & 5.18 & @xmath191.96 & 0 & & -0.06 & v & 19.62 & sdss - g + 12 07 47.271 -06 08 57.51 & 4.11 & 4.08 & 1.87 - @xmath3613.68 & & 4 & 6.84 & 6.63 & 5.92 & 2.87 & 2 & & -0.96 & v & & + 12 08 06.423 + 36 53 12.45 & 2.84 & 2.75 & 1.65 - 5.35 & 2.66 & 3 & 5.31 & 4.85 & 5.64 & 3.24 & 6 & & 0.10 & v & & + 12 08 12.253 + 33 13 22.10 & 2.37 & 2.61 & 1.50 - 4.22 & 3.29 & 4 & 6.95 & 5.35 & 6.76 & 2.81 & 483 & & 0.35 & v & & + & hipparcos & 1 & 1.4 & 3 ( 3 ) & 0.18 ( 0.18 ) & 27.3 ( 27.3 ) & 0.004 ( 0.004 ) + & tycho & 2 & 1.4 & 4 ( 2 ) & 0.25 ( 0.12 ) & 22.2 ( 22.2 ) & 0.006 ( 0.003 ) + + pulsar & atnf & 3 & 3.0 & 13 ( 13 ) & 0.8 ( 0.8 ) & 76.48 ( 76.48 ) & 0.006 ( 0.006 ) + + & sdss ( s ) & 4 & 1.4 & 53 ( 53 ) & 3.26 ( 3.26 ) & 1.16 ( 1.16 ) & 1.632 ( 1.632 ) + & sdss ( p ) & 5 & 1.4 & 68 ( 67 ) & 4.18 ( 4.12 ) & 1.1 ( 1.09 ) & 2.203 ( 2.239 ) + + & sdss dr7 & 6 & 1.4 & 442 ( 431 ) & 27.17 ( 26.49 ) & 0.55 ( 0.54 ) & 28.92 ( 29.68 ) + & gsc-2 & 7 & 1.4 & 349 ( 56 ) & 21.45 ( 3.44 ) & & + & 2mass & 8 & 1.4 & 121 ( 1 ) & 7.44 ( 0.06 ) & & + & apmuks(bj ) & 13 & 6.0 & 2 ( 1 ) & 0.12 ( 0.06 ) & & + + & sdss dr7 & 6 & 1.4 & 227 ( 114 ) & 13.95 ( 7.01 ) & 0.91 ( 0.75 ) & 8.9 ( 5.7 ) + & gsc-2 & 7 & 1.4 & 111 ( 4 ) & 6.82 ( 0.25 ) & & + & 2mass & 8 & 1.4 & 27 ( 0 ) & 1.66 ( 0.00 ) & & + & chandra & 9 & 3.0 & 7 ( 1 ) & 0.43 ( 0.06 ) & 1.45 ( 1.23 ) & 0.173 ( 0.048 ) + & xmm & 10 & 15.0 & 15 ( 2 ) & 0.92 ( 0.12 ) & 1.52 ( 1.03 ) & 0.353 ( 0.116 ) + & rass - bsc & 11 & 30.0 & 15 ( 0 ) & 0.92 ( 0.00 ) & 2.29 ( 0.00 ) & 0.234 ( 0.011 ) + & rass - fsc & 12 & 60.0 & 13 ( 2 ) & 0.8 ( 0.12 ) & 0.70 ( 0.49 ) & 0.664 ( 0.246 ) + + & gb-87 & 13 & 6.0 & 21 ( 7 ) & 1.29 ( 0.43 ) & & + & vlss & 13 & 6.0 & 5 ( 2 ) & 0.31 ( 0.12 ) & & + & wb-92 & 13 & 6.0 & 13 ( 1 ) & 0.8 ( 0.06 ) & & + & abell-04 & 13 & 6.0 & 1 ( 1 ) & 0.06 ( 0.06 ) & & + & b3 & 13 & 6.0 & 1 ( 1 ) & 0.06 ( 0.06 ) & & + & wn & 13 & 6.0 & 2 ( 1 ) & 0.12 ( 0.06 ) & & + & nvss & 14 & 7.1 & 989 ( 487 ) & 60.79 ( 29.93 ) & 0.75 ( 0.62 ) & 47.46 ( 46.88 ) + & none & 15 & & 377 ( 377 ) & 23.17 ( 23.17 ) & 0.43 & ( 100.0 ) + + * total * & & & & * 1623 + 4 * & * 100.0 * & & + 10 24 38.698 -07 19 19.07 & 0.17 & j1024 - 0719 & 5.21 & 0.005 & 6.49 & 0.53 + 09 22 14.008 + 06 38 22.84 & 0.51 & j0922 + 0638 & 10.33 & 0.431 & 27.27 & 1.20 + 10 22 58.011 + 10 01 52.85 & 0.40 & j1022 + 1001 & 3.69 & 0.016 & 10.25 & 0.40 + 12 39 40.386 + 24 53 49.87 & 1.19 & j1239 + 2453 & 11.53 & 1.382 & 9.24 & 0.86 + 08 26 51.438 + 26 37 22.83 & 1.19 & j0826 + 2637 & 11.14 & 0.531 & 19.45 & 0.36 + 16 52 03.080 + 26 51 39.85 & 0.56 & j1652 + 2651 & 6.27 & 0.916 & 40.80 & 2.93 + 15 18 16.831 + 49 04 34.19 & 0.31 & j1518 + 4904 & 5.03 & 0.041 & 11.61 & 0.70 + 10 12 33.387 + 53 07 02.09 & 0.66 & j1012 + 5307 & 2.20 & 0.005 & 9.02 & 0.52 + 15 09 25.675 + 55 31 32.90 & 0.62 & j1509 + 5531 & 10.02 & 0.740 & 19.61 & 2.13 + 16 07 12.078 -00 32 40.98 & 0.42 & j1607 - 0032 & 3.93 & 0.422 & 10.68 & 0.59 + 10 23 47.622 + 00 38 41.60 & 1.09 & j1023 + 0038 & 3.12 & 0.002 & 14.32 & 0.90 + 09 43 30.092 + 16 31 34.67 & 2.32 & j0943 + 1631 & 1.51 & 1.087 & 20.32 & 1.76 + 16 40 16.699 + 22 24 08.98 & 0.60 & j1640 + 2224 & 1.92 & 0.003 & 18.43 & 1.19 + 09 53 09.287 + 07 55 35.94 & 0.38 & j0953 + 0755 & 83.22 & 0.253 & 2.96 & 0.26 + 15 43 38.837 + 09 29 16.52 & 0.26 & j1543 + 0929 & 6.17 & 0.748 & 35.24 & 7.69 + 07 51 09.148 + 18 07 38.73 & 0.17 & j0751 + 1807 & 1.42 & 0.003 & 30.25 & 0.62 + 11 15 38.456 + 50 30 12.68 & 0.67 & j1115 + 5030 & 1.00 & 1.656 & 9.20 & 0.54 +

dense neutrino gases exhibit collective oscillations where `` self - maintained coherence '' is a characteristic feature , i.e. , neutrinos of different energies oscillate with the same frequency . in a non - isotropic gas , however , the flux term of the neutrino - neutrino interaction has the opposite effect of causing kinematical decoherence of neutrinos propagating in different directions , an effect that is at the origin of the `` multi - angle behavior '' of neutrinos streaming off a supernova core . we cast the equations of motion in a form where the role of the flux term is manifest . we study in detail the symmetric case of equal neutrino and antineutrino densities where the evolution consists of collective pair conversions ( `` bipolar oscillations '' ) . a gas of this sort is unstable in that an infinitesimal anisotropy is enough to trigger a run - away towards flavor equipartition . the `` self - maintained coherence '' of a perfectly isotropic gas gives way to `` self - induced decoherence . ''

hep-ph0701182

neutrino oscillations between two flavors in vacuum are governed by the frequency @xmath0 where @xmath1 is the energy of a given mode . therefore , if the energy spectrum is broad , the energy dependence of the oscillation frequency quickly leads to kinematical decoherence , i.e. , along a neutrino beam the overall flavor content quickly approaches an average value . the situation changes radically when neutrinos themselves provide a significant refractive effect , leading to collective oscillation modes @xcite that can be of practical interest in the early universe @xcite or in core - collapse supernovae @xcite . defining the parameter @xmath2 the neutrino gas is `` dense '' when @xmath3 , i.e. , when a typical neutrino - neutrino interaction energy exceeds the energy corresponding to the vacuum oscillation frequency . when this condition is satisfied , collective effects are important , even if the ordinary matter effect is much larger than that from the neutrino - neutrino interaction @xcite . one characteristic feature of collective oscillations is the phenomenon of `` self - maintained coherence '' @xcite , meaning that all modes oscillate with the same frequency even though the energy spectrum may be broad . it was recently stressed , however , that this is not the complete story @xcite . perhaps the most interesting case for collective effects is provided by neutrinos streaming off a supernova core , a situation that is far from isotropic . the current - current nature of the weak - interaction hamiltonian implies that the interaction energy between particles of momenta @xmath4 and @xmath5 is proportional to @xmath6 where @xmath7 is the velocity . if the medium is isotropic , the @xmath8 term averages to zero , but if there is a net current , test particles moving in different directions experience different refractive effects . therefore , neutrinos moving in a background with a net current will decohere between different directions of motion , even if the energy spectrum is monochromatic . to avoid confusion about terminology we stress that we always mean `` kinematical decoherence '' when we say `` decoherence . '' different modes oscillate differently , leading to `` de - phasing '' and thus to the depolarization of the overall ensemble . if we use the common language of polarization vectors @xmath9 for each mode @xmath4 , then the length of each @xmath9 is conserved , whereas the length of the overall polarization vector @xmath10 shrinks ( kinematical decoherence ) or is conserved ( kinematical coherence ) . the effect of _ dynamical _ decoherence , caused by collisions among the neutrinos or with a thermal background medium , is that each individual polarization vector @xmath9 shrinks , i.e. , neutrinos in individual modes can not be represented as pure states . this effect , relevant for open quantum systems , does not occur in our case where oscillations are the only form of evolution . the multi - angle decoherence effect becomes nontrivial in the most interesting case when the `` background current '' consists of the neutrinos themselves as for neutrinos streaming off a supernova core . numerical examples reveal significant decoherence effects , but on the other hand they also show collective modes very similar to the isotropic ( `` single - angle '' ) case @xcite . the overall behavior is determined by a complicated interplay between the collective evolution and decoherence . rather than trying to understand the supernova case in its full complexity , we here take the opposite approach and study the simplest example that shows decoherence caused by the neutrino - neutrino multi - angle effect . therefore , we consider a dense neutrino gas that is monochromatic , symmetric ( equal @xmath11 and @xmath12 densities ) , and homogeneous , but not isotropic . even this simple model has a rich phenomenology that helps one to develop a better understanding of the full problem . on the other hand , it is simple enough that the most puzzling aspects of its behavior are analytically tractable . a symmetric @xmath13 gas oscillates in the `` bipolar mode '' where pairs of neutrinos of a given flavor coherently oscillate into the other flavor and back with the `` bipolar frequency '' @xcite @xmath14 as an example we consider a dense neutrino gas initially consisting of @xmath15 and assume that the mixing angle @xmath16 with another flavor is small , a situation that could represent 13-oscillations . we characterize the overall flavor content of the ensemble in the usual way by a polarization vector where the positive @xmath17-direction corresponds to the @xmath18-flavor whereas the mass direction is defined by a unit vector that we call @xmath19 . it is tilted against the @xmath17-direction by @xmath20 and we always choose it to lie in the @xmath21-@xmath17-plane with @xmath22 for the normal hierarchy and @xmath23 for the inverted hierarchy . in vacuum , the polarization vectors precess around @xmath19 with frequency @xmath24 . in a dense neutrino gas with @xmath25 , their motion is largely confined to the @xmath21-@xmath17-plane where they perform pendular motions with frequency @xmath26 and where @xmath27 is the `` force direction . '' the maximum excursion is @xmath20 , indicated by the dotted lines in fig . [ fig : trajectory ] where we show the trajectory in the @xmath21-@xmath17-plane . for the normal hierarchy , the pendulum swings between the dotted lines . for the inverted hierarchy ( bottom ) , it performs almost full - circle oscillations as indicated by the diamonds . if the gas is not isotropic , the picture changes completely . taking the `` half - isotropic '' case as a generic example where only one half - space of momentum modes is occupied ( neutrinos streaming off a plane surface ) , we show in fig . [ fig : trajectory ] the trajectory of the average polarization vector . for both hierarchies , its length shrinks to zero so that the final state of the ensemble is an equal mixture of both flavors . in other words , the ensemble quickly decoheres , independently of the mass hierarchy . this behavior is perhaps intuitive because we have assumed a large neutrino current . however , this effect is self - accelerating in that it is triggered even by the smallest nonzero anisotropy . whereas for perfect isotropy the neutrino - neutrino interaction provides for self - maintained coherence between modes of different energy , even an infinitesimal anisotropy has the opposite effect of causing self - induced decoherence between modes with different propagation directions . the average flavor content of the ensemble always approaches an equal mixture of both flavors . this instability is a nonlinear effect caused by the neutrino - neutrino interaction . in other words , the pendular motion of the isotropic ensemble is an unstable limit cycle , whereas the equal flavor mixture is a stable fixed point . our goal is to explain this behavior . as a first step we formulate in sect . [ sec : eom ] the equations of motion for the multi - angle system in terms of the multipole components of the neutrino angular distribution . the special role of the flux ( the dipole term of the angular distribution ) becomes manifest . we begin with the full @xmath28-flavor problem and then consider two flavors in terms of polarization vectors . in sect . [ sec : anisotropy ] we study the role of a large anisotropy in causing decoherence . our main result is developed in sect . [ sec : smallanisotropy ] where we show that a @xmath13 gas is unstable in flavor space and always decoheres if there is an infinitesimally small initial anisotropy . we conclude in sec . [ sec : conclusions ] with a summary and outlook . a statistical ensemble of unmixed neutrinos is represented by the occupation numbers @xmath29 for each momentum mode @xmath4 , where @xmath30 and @xmath31 are the relevant creation and annihilation operators and @xmath32 is the expectation value . the corresponding expression for antineutrinos is @xmath33 , where here and henceforth overbarred quantities refer to antiparticles . for several flavors , the occupation numbers are generalized to `` matrices of densities '' in flavor space @xmath34 that really are matrices of occupation numbers . the reversed order of the flavor indices @xmath35 and @xmath36 in the r.h.s . for antineutrinos ensures that under a flavor transformation , @xmath37 , antineutrinos transform in the same way , @xmath38 . sums and differences of @xmath39 and @xmath40 then transform consistently . for the seemingly intuitive equal order of flavor indices that is almost universally used in the literature , @xmath39 , @xmath40 , @xmath41 and @xmath42 all appear in the equations instead of @xmath39 and @xmath40 alone . analogous matrices @xmath43 can be defined for the charged leptons . we assume that the neutrino ensemble is completely described by these one - particle occupation number matrices , i.e. , that genuine many - body effects play no role @xcite . in this case flavor oscillations of an ensemble of neutrinos and antineutrinos are described by an equation of motion for each mode @xmath44 + \sqrt{2}\,g_{\rm f}\left [ \int\!\frac{d^3{\bf q}}{(2\pi)^3 } \left(\ell_{\bf q}-\bar\ell_{\bf q } + \varrho_{\bf q}-\bar\varrho_{\bf q}\right ) ( 1-{\bf v}_{\bf q}\cdot{\bf v}_{\bf p } ) , \varrho_{\bf p}\right ] , \nonumber\\ { \rm i}\,\partial_t\bar\varrho_{\bf p}&= & -\left[\omega_{\bf p},\bar\varrho_{\bf p}\right ] + \sqrt{2}\,g_{\rm f}\left [ \int\!\frac{d^3{\bf q}}{(2\pi)^3 } \left(\ell_{\bf q}-\bar\ell_{\bf q } + \varrho_{\bf q}-\bar\varrho_{\bf q}\right ) ( 1-{\bf v}_{\bf q}\cdot{\bf v}_{\bf p } ) , \bar\varrho_{\bf p}\right],\end{aligned}\ ] ] where @xmath45 $ ] is a commutator and @xmath46 is the fermi constant . for ultrarelativistic neutrinos , the matrix of vacuum oscillation frequencies , expressed in the mass basis , is @xmath47 with @xmath48 . further , @xmath49 is the velocity of a particle ( neutrino or charged lepton ) with momentum @xmath4 . the total conserved energy of the neutrino ensemble is an important quantity for understanding its evolution @xcite . we find for the energy density @xmath50 ^ 2 -\frac{\sqrt2\,g_{\rm f}}{2}\left [ \int\!\frac{d^3{\bf p}}{(2\pi)^3}\ , ( \varrho_{\bf p}-\bar\varrho_{\bf p}){\bf v}_{\bf p}\right]^2 \biggr\rbrace\,.\end{aligned}\ ] ] this quantity actually represents the energy _ shift _ caused by the neutrino masses and by neutrino interactions . it is straightfoward to show that indeed @xmath51 by taking the time derivative on the r.h.s . of eq . ( [ eq : energydensity ] ) , inserting the equations of motion eq . ( [ eq : eom1 ] ) , and using cyclic permutations of matrices under the trace . for completeness we mention that one can also define the entropy density @xcite @xmath52\,.\end{aligned}\ ] ] this expression is well defined because @xmath39 and @xmath53 are positive semi - definite matrices . in our case where the equation of motion for each @xmath39 is of the form @xmath54 $ ] , the entropy density is conserved . oscillations alone do not lead to a loss of information . kinematical decoherence does not lead to an increase of entropy . since we wish to study the simplest example that shows nontrivial multi - angle effects , we now restrict the neutrino energy distribution to be monochromatic and the geometry to axial symmetry . we then integrate the matrices of densities over all variables except @xmath55 where the angle is relative to the direction of symmetry . therefore , we consider the matrices @xmath56 implying the normalization @xmath57 for convenience we have arbitrarily normalized the @xmath58 matrices to the neutrino density . we further define the matrix representing the particle flux along the direction of symmetry , @xmath59 where @xmath60 is the velocity along the symmetry direction . for relativistic particles @xmath61 . for charged leptons we proceed in the same way , except that we normalize to the electron density @xmath62 . we denote with @xmath63 the matrix of vacuum oscillation frequencies for our fixed neutrino energy @xmath1 . the equations of motion then simplify to @xmath64 + \left[\lambda\left(\ell-\bar\ell)_0 + \mu(\varrho-\bar\varrho\right)_0,\varrho_u\right ] \nonumber\\ & & \kern2.8em{}-\left[\lambda\left(\ell-\bar\ell)_1 + \mu(\varrho-\bar\varrho\right)_1,u\varrho_u\right]\,,\end{aligned}\ ] ] where @xmath65 . the same equation applies for @xmath66 except for a sign change of the vacuum oscillation term . evidently the isotropic part of the medium ( index 0 ) affects all modes in the same way and is ultimately responsible for the phenomenon of self - maintained coherence . on the other hand , the flux term ( index 1 ) involves a factor @xmath67 for every mode @xmath68 . even in the absence of a neutrino - neutrino term , a charged - lepton flux alone causes a trivial multi - angle decoherence effect . the conserved energy of the axially symmetric neutrino ensemble is @xmath69 \nonumber\\ & & { } + \lambda\,{\rm tr}\left [ ( \ell-\bar\ell)_0(\varrho-\bar\varrho)_0 -(\ell-\bar\ell)_1(\varrho-\bar\varrho)_1\right ] \nonumber\\ & & { } + \frac{\mu}{2}\,{\rm tr}\left [ ( \varrho-\bar\varrho)_0 ^ 2 -(\varrho-\bar\varrho)_1 ^ 2\right]\,.\end{aligned}\ ] ] because of our normalization of the @xmath58 matrices the quantity @xmath70 is the energy per @xmath11 . the structure of these equations becomes more transparent if we expand the angular dependence of the matrices in legendre polynomials . ( had we not assumed axial symmetry , spherical harmonics would be the appropriate basis . ) the first few legendre polynomials are @xmath71 we thus define @xmath72 the normalisation @xmath73 implies that the original function is @xmath74 the previously defined overall density @xmath75 and the flux @xmath76 are but the zeroth and first case of eq . ( [ eq : expansion1 ] ) . multiplying both sides of eq . ( [ eq : eom2 ] ) with @xmath77 and integrating over @xmath78 leads to the equation of motion @xmath79 + \mu\left[(\varrho-\bar\varrho)_0,\varrho_n\right ] \nonumber\\ & & { } -\mu\left[(\varrho-\bar\varrho)_1 , { \textstyle\int_{-1}^{+1}}du\,u\varrho_u l_n(u)\right]\,.\end{aligned}\ ] ] here and henceforth we no longer show the ordinary matter term . its structure is similar to the neutrino - neutrino term so that it is easily reinstated . with the expansion eq . ( [ eq : expansion2 ] ) the remaining integral is @xmath80 we use the `` dipole matrix element '' of the legendre polynomials @xmath81 with this result the equations of motion are @xmath82 + \mu\left[(\varrho-\bar\varrho)_0,\varrho_n\right ] \nonumber\\ & & { } -\frac{\mu}{2 } \left[(\varrho-\bar\varrho)_1 , ( a_n\varrho_{n-1}+b_n\varrho_{n+1})\right]\,,\end{aligned}\ ] ] where @xmath83 note that the equation for @xmath75 is consistent even though the quantity @xmath84 is not defined because the coefficient @xmath85 for @xmath86 . the equations of motion eq . ( [ eq : eom3 ] ) together with the corresponding equations for @xmath87 ( sign change for the vacuum term ) form a closed set of equations . if we also include the ordinary matter term they are equivalent to eq . ( [ eq : eom2 ] ) for the momentum - space matrices @xmath58 . one important difference is that @xmath88 is conserved , reflecting the absence of decoherence for individual momentum modes , whereas @xmath89 is not in general conserved , reflecting the effect of kinematical decoherence . we note that in general @xmath90 when the equation of motion is of the form @xmath91 $ ] . this is the case for @xmath58 , but not for the multipole matrices @xmath92 . of course , in a numerical implementation the series @xmath92 has to be truncated at some value @xmath93 , leading to limited angular resolution . this is analogous to the coarse graining of phase space required for the @xmath94 where one needs to use discrete angular bins of nonzero width . the equations of motion for the flux terms are special in that they involve one power of @xmath76 or @xmath95 in each term of the equation . therefore , if initially there is no flux term ( @xmath96 ) , none will develop . in this case the equations for @xmath75 and @xmath97 form a closed set , describing the dynamics of the `` flavor pendulum '' studied in ref . in addition , the higher multipoles @xmath92 with @xmath98 , if initially nonzero , will simply oscillate under the action of the vacuum term and of the density term @xmath99 . the interpretation of the equations of motion in multipole space becomes clearer if we observe that for some function @xmath100 one has to second order @xmath101 taking the sum and difference of these equations provides @xmath102 with eq . ( [ eq : abcoefficients ] ) we may write @xmath103 where we interpret a prime as a derivative with respect to @xmath104 that is now viewed as a continuous variable . if we further observe that in the discrete equation @xmath105 we may write eq . ( [ eq : eom3 ] ) in the continuous form @xmath106 + \mu\left[(\varrho-\bar\varrho)_0-(\varrho-\bar\varrho)_1 , \varrho_n\right ] \nonumber\\ & & \kern2.7em{}-\mu \left[(\varrho-\bar\varrho)_1,\left ( \frac{\varrho^\prime_n}{2n+1}+\frac{\varrho^{\prime\prime}_n}{2 } \right)\right]\,.\end{aligned}\ ] ] therefore , the lowest multipole ( monopole ) is equivalent to the pendulum in flavor space @xcite , the high multipoles are akin to a continuous medium carrying flavor waves . equation ( [ eq : eom4 ] ) is a diffusion equation . the periodic excitation caused by the flavor pendulum then diffuses to higher multipoles ( smaller scales ) . the flavor waves as a function of neutrino direction visible in numerical simulations @xcite are the result of this process . in the equations of motion the difference of the particle and antiparticle densities , @xmath107 , plays a special role in that it is this matrix of net lepton number densities that plays the role of a potential for the other neutrino modes . as noted by several authors , it is sometimes useful to write the equations of motion in terms of the sum and difference of @xmath39 and @xmath40 instead of these matrices themselves . we therefore define the matrix of particle densities ( sum @xmath108 ) and the matrix of lepton number densities ( difference @xmath109 ) by virtue of @xmath110 analogous definitions apply to the case of axial symmetry , @xmath111 and @xmath112 , and their multipole expansion @xmath113 and @xmath114 . the equations of motion for the particle and lepton number matrices are found by adding and subtracting the two lines of eq . ( [ eq : eom1 ] ) and the equivalent equations for axial symmetry or the multipoles . we find explicitly for the multipoles @xmath115 + \mu[{\sf d}_0,{\sf s}_n ] -\mu\left[{\sf d}_1 , \frac{a_n{\sf s}_{n-1}+b_n{\sf s}_{n+1}}{2}\right ] , \nonumber\\ { \rm i}\dot{\sf d}_n&=&[\omega,{\sf s}_n ] + \mu[{\sf d}_0,{\sf d}_n ] -\mu\left[{\sf d}_1 , \frac{a_n{\sf d}_{n-1}\!+\!b_n{\sf d}_{n+1}}{2}\right ] . \nonumber\\\end{aligned}\ ] ] one can easily restore the ordinary matter term because it has the same structure as the neutrino - neutrino term with the substitution @xmath116 and @xmath117 , the latter being the charged - lepton matrix of net lepton - number densities and the corresponding flux , respectively . the commutator structure of the r.h.s . of eq . ( [ eq : eom5 ] ) implies that its trace vanishes . therefore , the particle and lepton numbers for all multipoles , @xmath118 and @xmath119 , are conserved . with the expressions ( [ eq : abcoefficients ] ) for @xmath120 and @xmath121 the lowest - order multipole equation is @xmath122 + \mu[{\sf d}_0,{\sf s}_0 ] -\mu[{\sf d}_1,{\sf s}_{1}]\ , , \nonumber\\ { \rm i}\dot{\sf d}_0&=&[\omega,{\sf s}_0]\,.\end{aligned}\ ] ] one immediate consequence of the second equation is @xmath123 i.e. , in the mass basis a certain combination of flavor - lepton numbers is conserved . the explicit equation for the flux term is @xmath124 + \mu[{\sf d}_0,{\sf s}_1 ] + \frac{\mu}{3}\,[{\sf s}_{0}+2{\sf s}_{2},{\sf d}_1]\ , , \nonumber\\ { \rm i}\dot{\sf d}_1&=&[\omega,{\sf s}_1 ] + \frac{2\mu}{3}\,[{\sf d}_0-{\sf d}_2,{\sf d}_{1}]\,.\end{aligned}\ ] ] as stated earlier , these equations are special in that they involve one power of @xmath125 or @xmath126 in each term . therefore , @xmath125 and @xmath126 vanish at all times if they vanish initially . in this case , the monopole terms ( overall densities ) form a closed set of equations . defining @xmath127 the equations of motion take on the simple form @xmath128\ , , \nonumber\\ { \rm i}\dot{\sf d}&=&[\omega,{\sf q}]\,.\end{aligned}\ ] ] conserved quantities are @xmath129 , @xmath130 , @xmath131 , and @xmath132 . up to a constant , the last quantity is the energy per neutrino . for two flavors , these simplifications are at the origin of the pendulum analogy @xcite where @xmath129 corresponds to its length , @xmath130 its spin , and @xmath131 to the orbital angular momentum around the force direction ( the mass direction in flavor space ) . returning to the general case with arbitrary initial conditions , the energy per neutrino is @xmath133\,.\ ] ] energy conservation @xmath134 is easily verified by using the equations of motion for @xmath135 , @xmath136 and @xmath126 . in the two - flavor case where the @xmath39 are hermitian @xmath137 matrices , a representation in terms of polarization vectors is useful . a hermitian @xmath137 matrix @xmath138 is represented as @xmath139 where @xmath140 is the vector of pauli matrices and @xmath141 the polarization vector . the commutation relations of the pauli matrices imply that an equation of motion of the form @xmath142\ ] ] is represented by @xmath143 moreover , @xmath144 for any two matrices @xmath138 and @xmath145 . as usual , we denote with @xmath9 and @xmath146 the polarization vectors representing @xmath39 and @xmath40 , respectively , and analogous expressions for @xmath147 and @xmath148 . the particle and lepton - number matrices are represented by @xmath149 and @xmath150 , in each case with the subscripts @xmath151 , @xmath68 or @xmath104 for the different variables . finally , we represent the matrix of oscillation frequencies in the form @xmath152 where @xmath153 we have defined the oscillation frequency @xmath24 as a positive number . the `` magnetic field '' is a unit vector that in the interaction basis is explicitly @xmath154 where @xmath16 is the vacuum mixing angle . we have arbitrarily chosen its @xmath155-component to vanish , corresponding to a choice of overall phase of the mixing matrix . in the following , when we say that `` the mixing angle is small , '' we mean that @xmath156 . the sign of @xmath157 is physically irrelevant whereas @xmath22 represents the normal , @xmath23 the inverted hierarchy . the equations of motion for the particle- and lepton - number polarization vectors are explicitly @xmath158 the two lowest - order equations are explicitly @xmath159 and @xmath160 the conserved energy eq . ( [ eq : energy2 ] ) becomes , with the help of eq . ( [ eq : doubletrace ] ) , @xmath161\bigr\rbrace\,.\end{aligned}\ ] ] we have already noted that the individual traces @xmath162 , @xmath163 and @xmath164 are conserved so that is missing in the equivalent expression in our previous paper @xcite . ] @xmath165\ ] ] is the conserved energy per neutrino , up to an irrelevant constant . we use the notation @xmath166 for what is the `` potential energy '' in the pendulum analogy . its minimum is chosen at zero . the `` kinetic energy '' ( neutrino - neutrino interaction energy ) is denoted by @xmath167 where we have introduced @xmath168 for the monopole and dipole contributions . in order to study the impact of kinematical decoherence in the simplest nontrivial case we study a homogeneous ensemble consisting of equal numbers of neutrinos and antineutrinos that are all initially in the electron flavor . before turning to issues of decoherence , we briefly recall the evolution of the isotropic case . to this end we show in fig . [ fig : isotropic ] in the upper panels the @xmath17-component of the overall polarization vector @xmath169 for the inverted and normal hierarchy . the @xmath18 or @xmath170 survival probability is @xmath171 because in our normalization @xmath172 . in the lower panels we show the corresponding energy components . for the normal hierarchy ( left ) the motion is a small - excursion harmonic oscillation with the bipolar frequency @xmath26 . we show one full cycle of oscillation . the potential energy @xmath173 ( solid line in the lower panel ) is measured against the mass direction so that it is back at its maximum after a half - period as behooves an ordinary pendulum . the dashed line shows the `` kinetic pendulum energy '' @xmath174 . the isotropy of the ensemble implies @xmath175 and thus @xmath176 . for the inverted hierarchy ( right ) the motion is that of an inverted oscillator . we show approximately one full cycle that for the chosen mixing angle lasts approximately 3 times the bipolar period @xmath177 . the polarization vector starts in the positive @xmath17 direction , swings almost one full circle and almost arrives back at the vertical direction ( maximum in the middle of the plot ) , then swings back , reaching its original vertical position , and then starts over again . the maximum potential energy is @xmath178 because the energy is normalized to the total energy per @xmath11 and because there are equal numbers of neutrinos and antineutrinos , i.e. , two particles per @xmath11 . reducing the mixing angle has the effect of reducing the oscillation amplitude for the normal hierarchy , whereas it increases logarithmically the duration of the `` plateau phases '' for the inverted hierarchy . for quantitative discussions of the isotropic case in terms of the pendulum analogy see ref . @xcite . next we consider the same example , but assume a large degree of anisotropy where only one half - space of momentum modes is occupied . in fig . [ fig : halfisotropic ] we show , in the upper panels , the evolution of @xmath179 . moreover , we show the length of the overall polarization vector , @xmath180 , as a thin dashed line . both in the normal hierarchy ( left ) and the inverted hierarchy ( right ) , an equal mixture of the two flavors is quickly achieved . in the isotropic case the length of @xmath181 and @xmath182 is conserved and that of @xmath183 is approximately conserved , up to corrections of order @xmath184 . here , the lengths of @xmath185 , @xmath182 and @xmath183 shrink to zero , reflecting kinematical decoherence . of course , @xmath179 can become small or zero without its length shrinking as during the first swing in the inverted hierarchy case . it is the length of the polarization vector , not its @xmath17-component , that is a measure of decoherence . of course , in the normal hierarchy , @xmath183 performs only small - excursion oscillations so that a significant change of its @xmath17-component can be achieved only by a change of its length . in this case @xmath186 and @xmath187 are almost identical . the evolution of @xmath183 was also illustrated in fig . [ fig : trajectory ] where we showed its trajectory in the @xmath21-@xmath17-plane . for our chosen geometry where @xmath19 is in the @xmath21-@xmath17-plane , @xmath188 at all times . for the inverted hierarchy ( bottom ) we indicate the isotropic - gas trajectory with diamonds at time intervals of @xmath189 where the motion starts in the vertical position . in the half - isotropic case , the particle - number polarization vector @xmath183 spirals in to a position close to the origin . the final offset is very small and depends on the magnitude of @xmath190 for which we have chosen @xmath191 . in the normal hierarchy ( top ) , the isotropic - case motion is a small - excursion harmonic oscillation between the dotted lines . in the half - isotropic case , the evolution begins from the vertical position . a full oscillation back to the vertical position is performed before the shrinking of @xmath183 becomes noticeable . in the lower panels of fig . [ fig : halfisotropic ] we show the evolution of the different energy components . the simpler case is the inverted hierarchy ( right ) where the `` potential '' and `` kinetic '' energies begin to oscillate as in the isotropic case of fig . [ fig : isotropic ] . on the time scale of a few bipolar oscillation periods , the two components essentially equipartition , although asymptotically a small offset remains that depends on @xmath190 . moreover , the neutrino - neutrino energy ( `` kinetic energy '' ) now develops a nonvanishing flux term @xmath192 that inevitably is negative . the normal hierarchy ( left ) is initially similar in that @xmath173 and @xmath174 oscillate as for a pendulum , even though this motion is not visible on the scale of the plot . as decoherence sets in , a qualitatively different mode of behavior obtains in that the neutrino - neutrino energy @xmath174 is dominated almost entirely by the negative @xmath193 whereas @xmath194 now is subdominant . all individual polarization vectors of all modes start aligned with the @xmath17-direction , i.e. , almost aligned with the force direction since the mixing angle is small . therefore , the initial @xmath173 is near its minimum . if the overall polarization vector @xmath183 is supposed to shrink , the individual polarization vectors must develop significant deviations from each other and thus the potential energy must increase . energy conservation then dictates that @xmath174 becomes negative . one important conclusion is that the angular dependence of the neutrino - neutrino interaction alone is not enough to cause kinematical decoherence , but its absolute sign is also crucial . if @xmath193 were not negative , energy conservation would prevent significant decoherence for the normal hierarchy . changing this sign in a numerical example indeed reveals the absence of decoherence for the normal hierarchy , but no significant change of behavior for the inverted case . in the real world there is no choice about this sign . it derives from the negative sign of the spatial part in the neutrino current - current interaction , i.e. , it is the negative sign inherent in the lorentz metric . from the equations of motion in the form eq . ( [ eq : eom6 ] ) one would have never guessed that the absolute sign of the term proportional to @xmath195 plays a crucial role . it is also instructive to study the evolution of the higher multipole components @xmath196 . we consider the same example as in the previous section . for a half - isotropic gas the initial relative length of the multipole components is @xmath197 where @xmath198 and all of them are initially oriented in the @xmath17-direction . here , @xmath199 is an integer , i.e. , the even multipoles vanish initially except for @xmath200 . instead of using these initial values , we calculate the evolution for a case where all multipoles vanish initially except for @xmath201 and @xmath202 , i.e. , we use the same density and flux terms as in the half - isotropic case . the evolution is very similar . we show the evolution of the first 21 multipoles in fig . [ fig : multipoleevolution ] for the inverted hierarchy ; the picture is similar for the normal hierarchy . the initial horizontal part of the curves corresponds to their zero initial value so that the offset by 0.1 vertical units between the curves is directly apparent . initially , @xmath200 and @xmath203 are large , but quickly decay away by decoherence , whereas the higher multipoles vanish initially and get excited one after another , but then decrease again . the length of each multipole sports `` wiggles , '' i.e. , the evolution is not monotonic . moreover , they have large relative angles in the @xmath21-@xmath17 plane ( not shown here ) , i.e. , the spread of the inital excitation to larger and larger multipoles is far from simple in detail , but the overall process is as expected . in fig . [ fig : multipoleevolution2 ] we show @xmath204 and @xmath205 as a function of @xmath104 at @xmath206 for the example of fig . [ fig : multipoleevolution ] . at @xmath206 , multipoles @xmath207 are not yet excited . the nearly linear increase of the phases between individual polarization vectors corresponds here to a diffusion of the `` multipole wave '' to larger @xmath104 with approximately constant speed , @xmath208 . qualitatively , this can be understood as follows . the second line in eq . ( [ eq : eom6 ] ) together with the analogue of eq . ( [ eq : diffusion ] ) implies that @xmath209 the last term does not contribute to the growth of @xmath210 and can be ignored . approximate equipartition between potential and kinetic energies in eq . ( [ eq : energy4 ] ) implies @xmath211 , as indeed was observed in fig . [ fig : halfisotropic ] . the second term in eq . ( [ eq : n_growth ] ) then represents a combination of a drift in @xmath212space with velocity @xmath213 and diffusion with diffusion coefficient @xmath214 . up to a factor @xmath215 this gives rise to an exponential factor @xmath216 $ ] . on the other hand , since @xmath217 for @xmath98 , the first term in eq . ( [ eq : n_growth ] ) can give rise to exponential growth whenever @xmath218 . this growth will be non - monotonic , but can roughly be estimated as @xmath219 before @xmath220 saturates . using perturbation theory , we will find a similar growth rate for @xmath221 at early times in sect . [ sec : analytic ] . combining these two factors , one sees that at time @xmath222 the wave front of @xmath223 , and thus also of @xmath196 , should be located at @xmath208 , which , according to figs . [ fig : multipoleevolution ] and [ fig : multipoleevolution2 ] indeed it is within about @xmath224% . the slow decrease of @xmath196 and @xmath220 at late times can be interpreted as due to the @xmath215 factor in the diffusive behavior . the overall behavior of the inverted - hierarchy case is qualitatively easy to understand . to this end we consider an ensemble of polarization vectors @xmath225 , initially oriented in the @xmath17-direction , that rotate around the @xmath155-axis , each with a frequency @xmath68 , so that @xmath226 both @xmath68 and @xmath222 are normalized to be dimensionless . if the frequencies are spread over @xmath227 , then the overall polarization vector evolves as @xmath228-\sin(t)}{t}\ , , \nonumber\\ p^x_0&=&\int_1^{1+\delta u}\!\!du\,\sin(u t ) = \frac{\cos(t)-\cos[(1+\delta u)\,t]}{t}\ , , \nonumber\\ |{\bf p}_0|&=&\left[(p^x_0)^2+(p^z_0)^2\right]^{1/2 } = \frac{2\,|\sin(\delta u\,t/2)|}{t}\,.\end{aligned}\ ] ] for @xmath229 we show in fig . [ fig : toy1 ] the evolution of @xmath230 ( solid line ) and the length @xmath231 ( dashed line ) . in fig . [ fig : toy2 ] we show the trajectory of @xmath181 in the @xmath21-@xmath17-plane . unsurprisingly , this simple model represents qualitatively the features of the inverted - hierarchy case . in particular , the trajectory in the @xmath21-@xmath17-plane is not a true spiral , but there are many crossings of the origin as the length of the overall polarization vector shrinks . its length does not shrink monotonically . its envelope decreases as a power law @xmath232 , not exponentially . this example also shows that it is not trivial to define a useful measure of kinematical decoherence . the @xmath17-component of the polarization vector is not useful because @xmath181 can tilt so that its @xmath17-component vanishes , i.e. , both flavors are present with the same probability , yet the flavor content of the ensemble is perfectly coherent . the length of @xmath181 is a much better measure in analogy to dynamical decoherence of a single mode where the length of @xmath9 is a measure of dynamical decoherence ; unit length would correspond to a pure state . in our case , the length of the individual @xmath9 is conserved , whereas the length of the total @xmath181 shrinks , but not monotonically . it is not evident if there is an `` entropy '' measure that evolves monotonically as a consequence of kinematical decoherence . we note , however , that the origin of kinematical decoherence is the differential evolution between neighboring polarization vectors . in our toy example , the angle between neighboring polarization vectors grows linearly as @xmath233 and thus is a measure of decoherence . motivated by this observation we can define the `` winding number '' @xmath234 in the symmetric case where the evolution of the polarization vectors is essentially restricted to the @xmath21-@xmath17-plane , this quantity tells us the number of windings around the @xmath155-direction of the full ensemble of polarization vectors . in fig . [ fig : windingnumber ] we show the evolution of this quantity for the normal and inverted hierarchy examples of fig . [ fig : halfisotropic ] . @xmath28 counts how often the length of @xmath183 shrinks to zero . closely related to kinematical decoherence are recurrence effects that arise when one uses a limited number of polarization vectors , i.e. , when the full ensemble is represented by a coarse - grained ( binned ) ensemble of polarization vectors or , in multipole space , when the series is truncated at some multipole @xmath93 . binning or truncation are unavoidable in numerical treatments and represent equivalent approximations . if the phase differences evolve linearly between different polarization vectors as in the toy example of the previous section , then the initial state will recur when each polarization vector has turned around the @xmath155-axis by @xmath235 relative to its neighbors . put another way , if we use @xmath236 bins to represent the polarization vectors , we expect recurrence to begin when @xmath237 . we demonstrate this effect in fig . [ fig : recurrence ] for the same example as in fig . [ fig : halfisotropic ] , normal hierarchy , using @xmath238 . indeed , the winding number increases until approximately @xmath239 and then decreases back to almost zero , and so forth . this system is almost periodic on the recurrence time scale , reflecting that nonlinear effects play a subleading role here . since a large flux term is present from the start , kinematical decoherence is very similar to the linear toy example of the previous section where recurrence would be exact and the system would be periodic with a frequency proportional to @xmath240 times the individual polarization vector s oscillation frequency . using the multipoles as the primary variables , the same recurrence effects occur due to the truncation @xmath93 of the multipole series . the `` multipole wave '' of fig . [ fig : multipoleevolution2 ] can only propagate up to @xmath241 , is then reflected , and eventually returns back to @xmath86 . this is the moment of maximum recurrence of the overall polarization vector . for a more complicated example such as a realistic simulation of the supernova case , recurrence effects may not be so obvious as here . therefore , in a numerical treatment one must make sure that the largest relative angle developed between any two neighboring polarization vectors never grows to order unity , i.e. , one should monitor the largest relative angle between neighboring polarization vectors as a measure of accuracy of the calculation . in other words , kinematical decoherence is a property of the entire ensemble , but can also be a a property of individual bins if they are too coarse . one must make sure that within all individual bins kinematical decoherence remains negligible . we next turn to a symmetric system that is perfectly isotropic except for an arbitrarily small but nonzero initial anisotropy . we work in the multipole picture and assume that initially @xmath242 for all @xmath104 , @xmath243 for @xmath244 , @xmath245 and @xmath246 . note that @xmath247 corresponds to a flux term equivalent to the half - isotropic case . as a first example we show in fig . [ fig : nearisotropic ] the evolution of the overall polarization vector @xmath183 ( top ) for the normal hierarchy ( left ) and the inverted hierarchy ( right ) . the initial anisotropy is @xmath248 and @xmath249 . for both hierarchies , kinematical decoherence eventually obtains . in the bottom panels we show the evolution of @xmath250 ( solid line ) as a proxy for @xmath195 which is the primary quantity that causes decoherence . for the symmetric neutrino gas only the component @xmath251 is nonzero , but it can be negative and can change sign so that it is difficult to display on a logarithmic scale . we also show the evolution of the `` potential energy '' @xmath173 ( dotted line ) . for the normal hierarchy , @xmath251 grows exponentially after a short initial transient . while @xmath251 is small , the potential energy @xmath173 performs the usual oscillations of a harmonic oscillator , complemented by the opposite oscillations of the kinetic energy @xmath252 ( not shown ) that is entirely dominated by @xmath194 . as @xmath251 grows , it eventually dominates the kinetic energy and the end state is approximately @xmath253 and @xmath254 , with @xmath193 dominating @xmath174 . in other words , the end state is exactly as in the previous case with a large initial anisotropy . the crucial novel feature is that the flux term grows exponentially from a small value . this is a purely nonlinear effect caused by neutrino - neutrino interactions . for smaller mixing angles and/or other values of @xmath255 , the behavior is analogous . the rising part of the @xmath193 curve is the same , except that it shifts vertically in direct proportion to @xmath255 and @xmath256 . for the inverted hierarchy , the evolution is far more complicated . while @xmath251 grows in an average sense , this evolution is overlaid with oscillations and in fact @xmath251 changes sign at each spike in the @xmath193 curve . moreover , the evolution of the envelope of the @xmath193 curve is not monotonic . it can perform complicated motions , with growing , declining , and nearly flat phases , until finally decoherence obtains . from these numerical observations we tentatively conclude that an isotropic neutrino ensemble of equal densities of neutrinos and antineutrinos is not stable . a small flux term triggers a run - away evolution towards kinematical decoherence . as a starting point for an analytic understanding of this instability we use the equations of motion eq . ( [ eq : eom6 ] ) for the vectors @xmath196 and @xmath220 . initially only @xmath183 is of order one , @xmath257 , and all others vanish . as long as @xmath195 is sufficiently small , none of the higher multipoles can grow large so that we can limit our attention to the equations for @xmath86 and @xmath258 , i.e. , to eqs . ( [ eq : eom6a ] ) and ( [ eq : eom6b ] ) . we further observe that the @xmath86 equations are coupled to the higher multipoles only by the term @xmath259 , consisting of a product of two small quantities , i.e. , it is at least of order @xmath260 . therefore , the evolution of the near - isotropic ensemble is identical with that of the isotropic case until @xmath195 has grown sufficiently large . in other words , up to second order in @xmath255 the @xmath86 equation is that of the isotropic case . therefore , what remains to be solved is the @xmath258 equation . after neglecting terms involving @xmath244 all we need to study is @xmath261 where @xmath262 and @xmath263 are the solutions of the unperturbed pendulum equations . the equations simplify further in our case of equal @xmath11 and @xmath12 densities where symmetry dictates that all polarization vectors @xmath9 and @xmath146 evolve as each other s mirror images relative to the @xmath21-@xmath17-plane , the latter being singled out by our choice that @xmath19 lies in that plane . in this case all @xmath220 vectors are parallel to the @xmath155-axis , whereas all @xmath196 vectors are confined to the @xmath21-@xmath17-plane . as a consequence , the @xmath264 term drops out . the second equation then implies @xmath265 , and with the first equation yields @xmath266 + \omega\mu{\bf b}\times\left({\bf d}_0\times{\bf s}_1\right)\,.\ ] ] expanding the triple product , observing that in our case @xmath267 , and using @xmath268 this is @xmath269 we repeat that all @xmath220 are here parallel to each other and to the @xmath155-axis . in the absence of neutrino - neutrino interactions ( @xmath270 ) , @xmath195 obeys a harmonic - oscillator equation . in a dense neutrino gas , @xmath271 , we may ignore the term proportional to @xmath272 . since during the growth phase , @xmath273 and @xmath274 , the last term in eq . ( [ eq : ddot ] ) can also be neglected . a possible exponential growth is then accounted for by the term @xmath275 where we have used the bipolar oscillation frequency @xmath276 and we note that initially @xmath172 and that its length remains nearly constant if @xmath277 unless there is kinematical decoherence . if the mixing angle is small , the `` normal hierarchy '' is defined by neutrinos being essentially in the lower mass eigenstate or @xmath278 up to corrections of order @xmath279 . therefore , the critical part of the equation of motion for the flux term is approximately @xmath280 the solution includes growing modes of the form @xmath281 where @xmath282 is the dimensionless time variable used in all of our plots , i.e. , time in units of the bipolar oscillation period . this behavior is numerically verified in the lower - left panel of fig . [ fig : nearisotropic ] if we recall that the plotted quantity is @xmath283 $ ] . once more we note that the crucial absolute sign in eq . ( [ eq : expgrowth2 ] ) is traced back to the negative sign in the term @xmath284 in eq . ( [ eq : eom1 ] ) . therefore , the instability of the near - isotropic neutrino gas , for the normal hierarchy , is not just caused by different angular modes experiencing different neutrino - neutrino effects . this would also be the case for a hypothetical term @xmath285 that naively looks very similar , but causes a completely different overall behavior . in more complicated numerical situations , e.g. , for a full supernova simulation , it is easy to change this sign in order to diagnose the relevance of the term eq . ( [ eq : expgrowth1 ] ) for the overall behavior of the system . initially @xmath286 so that the exponential growth term alone is not enough . the initial evolution is dominated by the second line of eq . ( [ eq : eom6c ] ) that is initially @xmath287 where @xmath288 and we have used @xmath289 and @xmath290 . here the absolute sign is not critical because it does not matter if @xmath195 grows in the positive or negative @xmath155-direction . the first factor @xmath291 sets the overall scale for @xmath195 . from the expression for the energy we note that @xmath292 and @xmath293 appear in the combinations @xmath294 . the quantities @xmath196 are of order unity , whereas the natural scale for @xmath220 is @xmath291 . the initial evolution is @xmath295 where , again , @xmath296 is our usual measure of time . this linear growth is shown in fig . [ fig : initialevolution ] as a dotted line and agrees with the numerical examples . besides a transient caused by the other terms in the equation , the exponential growth takes over after a time of order @xmath177 . we conclude that the initial value of @xmath297 for its exponential growth is of order @xmath298 so that it grows to order unity within a time scale of order @xmath299 . for the inverted hierarchy , the initial growth of @xmath300 is the same ( dashed line in fig . [ fig : initialevolution ] ) . after that , however , the situation is entirely different because initially @xmath301 . this corresponds to the pendulum starting in a nearly upright position from where it starts almost full - circle oscillations . for our usual assumption @xmath156 the term @xmath302 is almost identical with @xmath179 shown in the upper - right panel of fig . [ fig : nearisotropic ] . this term changes sign during every swing of the pendulum . during the phases when @xmath303 , @xmath195 oscillates whereas during the phases @xmath304 exponential growth obtains . these are the relatively short phases when the pendulum is oriented downward , whereas it stays upright for long periods if the mixing angle is small . we illustrate this behavior in fig . [ fig : initialevolution2 ] where we show the evolution of @xmath300 overlaid with that of @xmath179 . we have chosen an extremely small mixing angle , @xmath305 , to obtain long plateaus for @xmath179 where the pendulum stands almost still in an almost upright position . during these phases we have @xmath306 so that eq . ( [ eq : expgrowth1 ] ) implies harmonic oscillations of @xmath300 with the frequency @xmath307 as borne out by the example in fig . [ fig : initialevolution2 ] . the frequency @xmath307 is not matched to the pendulum s oscillation period because the duration of the `` plateaus '' depend logarithmically on @xmath256 . therefore , the short exponential growth phases when @xmath308 occur at erratic instances relative to the harmonic @xmath195 oscillation . this interplay explains the erratic behavior of the @xmath195 evolution that is apparent in fig . [ fig : nearisotropic ] and that obtains in all numerical examples . this interplay also explains why small changes of parameters such as @xmath256 can completely change the overall @xmath195 evolution . therefore , we can not prove if there is some specific combination of parameters where kinematical decoherence will not occur , although this would have to be isolated parameter points that presumably have measure zero in parameter space if they exist at all . except for this caveat we conclude that both for the normal and inverted hierarchy kinematical decoherence is an unavoidable consequence of the nonlinear neutrino - neutrino interaction terms . an infinitesimally small , but nonzero , deviation from isotropy is enough to trigger an exponential evolution towards flavor equilibrium . therefore , the pendulum solution that describes the behavior of a perfectly isotropic gas is merely an unstable limit cycle of this nonlinear system . we have investigated multi - angle kinematical decoherence effects in a dense neutrino gas consisting of equal @xmath11 and @xmath12 densities . the current - current nature of the weak - interaction hamiltonian implies that a current of the background medium causes kinematical decoherence between neutrinos propagating in different directions . this simple effect becomes entirely nontrivial in the most interesting case when the `` background current '' is caused by the neutrinos themselves . we have shown that a neutrino gas of this sort is not stable in flavor space . if one prepares the ensemble in a given flavor state , even a small deviation from isotropy is enough to trigger an exponential evolution towards flavor equipartition , even when the mixing angle is small . up to logarithmic corrections , the time scale is determined by the bipolar oscillation period . we thus have to do with a self - induced macroscopic pair conversion process that proceeds much faster than ordinary pair processes that occur at order @xmath309 . with hindsight this is probably the same `` speed - up effect '' discussed in ref . however , the fast speed @xmath310 is not introduced by the multi - angle effect as is perhaps suggested in ref . rather , @xmath26 is the bipolar oscillation frequency that is inherent even in the isotropic system . for the normal hierarchy `` nothing '' seems to happen in the isotropic case , although this would be a misperception because the system moves with the speed @xmath26 , but the amplitude is small if @xmath311 . in any event , we agree with the importance of multi - angle effects and with the end result conjectured in ref . @xcite that complete flavor equipartition obtains on a time scale @xmath312 . however , we have found no indication that an even faster time scale @xmath313 plays any role , in contrast to what was conjectured in ref . @xcite . of course , flavor equipartition obtains only in a macroscopic sense with some degree of coarse graining in phase space . it is the nature of kinematical decoherence that information is not lost and the entropy does not increase . on the other hand , kinematical and dynamical decoherence are not always operationally distinguishable . it is not necessarily possible to distinguish between a neutrino gas in true chemical equilibrium and one where neighboring modes are simply `` de - phased '' relative to each other . either way , the `` self - maintained coherence '' of a dense neutrino gas represents only an unstable limiting form of behavior of the perfectly isotropic case . since no gas can be perfectly isotropic , self - induced decoherence is the generic behavior of a symmetric @xmath13 gas . contrary to a naive expectation , the fact that neutrinos propagating in different directions experience different weak potentials is necessary , but not sufficient , to cause multi - angle decoherence . the negative sign in the lorentz metric plays a crucial role because it determines that the flux term grows exponentially rather than oscillating . the occurrence of kinematical decoherence is a subtle feature of the nonlinear equations . we have studied the simplest possible case where decoherence effects obtain . a more realistic case is provided by neutrinos streaming off a supernova core . in this case the fluxes of neutrinos and antineutrinos are different and geometry implies that the neutrino - neutrino interaction strength declines with the fourth power of radius . the evolution of an asymmetric system is far more complicated , even in the isotropic case , where the simple pendular motion turns into one of a spinning top that can precess , nutate , or simply swing , depending on the @xmath13 asymmetry . in the supernova context , transitions between these modes occur as @xmath314 declines as a function of radius . multi - angle decoherence can not occur , for example , in regions where synchronized oscillations prevail . numerical simulations suggest that for neutrinos streaming off a supernova core , the collective behavior characteristic of an isotropic gas prevails in many situations , i.e. , kinematical decoherence occurs , but is not the dominating feature of the system , contrary to what we have found here for the symmetric @xmath13 gas . evidently it is not straightforward to apply the insights gained from our present study to the supernova case . we imagine , however , that the methods developed here may take us one step further to understand the supernova problem . we have formulated the equations of motion in a novel form adapted to the problem of multi - angle propagation , i.e. , we have used a multipole expansion rather than the usual momentum modes and we have used the particle number and lepton number polarization vectors ( or density matrices ) as the fundamental variables . in this way , the crucial exponentially growing quantity , our @xmath195 , could be isolated as the primary cause of kinematical decoherence . this approach may prove useful for understanding the supernova problem as well . for numerical three - flavor studies we have provided an expression for the conserved energy of the system in terms of the density matrices . the conserved energy and its three components @xmath173 , @xmath194 and @xmath193 provide a useful diagnostic tool for the behavior of the system . even simple nonlinear systems can exhibit a surprisingly rich phenomenology . it is fascinating that the neutrinos streaming off a supernova core provide an intriguing case in point where many aspects of its behavior remain to be understood . j. pantaleone , `` neutrino oscillations at high densities , '' phys . b * 287 * , 128 ( 1992 ) . s. samuel , `` neutrino oscillations in dense neutrino gases , '' phys . d * 48 * , 1462 ( 1993 ) . v. a. kosteleck , j. pantaleone and s. samuel , `` neutrino oscillation in the early universe , '' phys . b * 315 * , 46 ( 1993 ) . v. a. kosteleck and s. samuel , `` neutrino oscillations in the early universe with an inverted neutrino mass hierarchy , '' phys . b * 318 * , 127 ( 1993 ) . v. a. kosteleck and s. samuel , `` nonlinear neutrino oscillations in the expanding universe , '' phys . rev . d * 49 * , 1740 ( 1994 ) . v. a. kosteleck and s. samuel , `` self - maintained coherent oscillations in dense neutrino gases , '' phys . d * 52 * , 621 ( 1995 ) [ hep - ph/9506262 ] . v. a. kosteleck and s. samuel , `` neutrino oscillations in the early universe with nonequilibrium neutrino distributions , '' phys . d * 52 * , 3184 ( 1995 ) [ hep - ph/9507427 ] . s. samuel , `` bimodal coherence in dense selfinteracting neutrino gases , '' phys . d * 53 * , 5382 ( 1996 ) [ hep - ph/9604341 ] . v. a. kosteleck and s. samuel , `` nonequilibrium neutrino oscillations in the early universe with an inverted neutrino mass hierarchy , '' phys . b * 385 * , 159 ( 1996 ) [ hep - ph/9610399 ] . j. pantaleone , `` stability of incoherence in an isotropic gas of oscillating neutrinos , '' phys . d * 58 * , 073002 ( 1998 ) . r. f. sawyer , `` speed - up of neutrino transformations in a supernova environment , '' phys . d * 72 * , 045003 ( 2005 ) [ hep - ph/0503013 ] . s. pastor , g. g. raffelt and d. v. semikoz , `` physics of synchronized neutrino oscillations caused by self - interactions , '' phys . d * 65 * , 053011 ( 2002 ) [ hep - ph/0109035 ] . c. lunardini and a. y. smirnov , `` high - energy neutrino conversion and the lepton asymmetry in the universe , '' phys . d * 64 * , 073006 ( 2001 ) [ hep - ph/0012056 ] . a. d. dolgov , s. h. hansen , s. pastor , s. t. petcov , g. g. raffelt and d. v. semikoz , `` cosmological bounds on neutrino degeneracy improved by flavor oscillations , '' nucl . b * 632 * , 363 ( 2002 ) [ hep - ph/0201287 ] . y. y. y. wong , `` analytical treatment of neutrino asymmetry equilibration from flavour oscillations in the early universe , '' phys . d * 66 * , 025015 ( 2002 ) [ hep - ph/ 0203180 ] . k. n. abazajian , j. f. beacom and n. f. bell , `` stringent constraints on cosmological neutrino antineutrino asymmetries from synchronized flavor transformation , '' phys . d * 66 * , 013008 ( 2002 ) [ astro - ph/0203442 ] . j. t. pantaleone , `` neutrino flavor evolution near a supernova s core , '' phys . b * 342 * , 250 ( 1995 ) [ astro - ph/9405008 ] . y. z. qian and g. m. fuller , `` neutrino - neutrino scattering and matter enhanced neutrino flavor transformation in supernovae , '' phys . d * 51 * , 1479 ( 1995 ) [ astro - ph/9406073 ] . g. sigl , `` neutrino mixing constraints and supernova nucleosynthesis , '' phys . rev . d * 51 * , 4035 ( 1995 ) [ astro - ph/9410094 ] . s. pastor and g. raffelt , `` flavor oscillations in the supernova hot bubble region : nonlinear effects of neutrino background , '' phys . * 89 * , 191101 ( 2002 ) [ astro - ph/0207281 ] . a. b. balantekin and h. yksel , `` neutrino mixing and nucleosynthesis in core - collapse supernovae , '' new j. phys . * 7 * , 51 ( 2005 ) [ astro - ph/0411159 ] . g. m. fuller and y. z. qian , `` simultaneous flavor transformation of neutrinos and antineutrinos with dominant potentials from neutrino neutrino forward scattering , '' phys . d * 73 * , 023004 ( 2006 ) [ astro - ph/0505240 ] . h. duan , g. m. fuller and y. z. qian , `` collective neutrino flavor transformation in supernovae , '' astro - ph/0511275 . h. duan , g. m. fuller , j. carlson and y. z. qian , `` simulation of coherent non - linear neutrino flavor transformation in the supernova environment . i : correlated neutrino trajectories , '' phys . d * 74 * , 105014 ( 2006 ) [ astro - ph/0606616 ] . h. duan , g. m. fuller , j. carlson and y. z. qian , `` coherent development of neutrino flavor in the supernova environment , '' astro - ph/0608050 . s. hannestad , g. g. raffelt , g. sigl and y. y. y. wong , `` self - induced conversion in dense neutrino gases : pendulum in flavour space , '' phys . rev . d * 74 * , 105010 ( 2006 ) [ astro - ph/0608695 ] . a. friedland and c. lunardini , `` neutrino flavor conversion in a neutrino background : single- versus multi - particle description , '' phys . d * 68 * , 013007 ( 2003 ) [ hep - ph/0304055 ] . a. friedland and c. lunardini , `` do many - particle neutrino interactions cause a novel coherent effect ? , '' jhep * 0310 * , 043 ( 2003 ) [ hep - ph/0307140 ] . a. friedland , b. h. j. mckellar and i. okuniewicz , `` construction and analysis of a simplified many - body neutrino model , '' phys . d * 73 * , 093002 ( 2006 ) [ hep - ph/0602016 ] . a. d. dolgov , `` neutrinos in the early universe , '' yad . * 33 * , 1309 ( 1981 ) [ sov . j. nucl . * 33 * , 700 ( 1981 ) ] . m. a. rudzsky , `` kinetic equations for neutrino spin- and type - oscillations in a medium , '' astrophys . space science * 165 * , 65 ( 1990 ) . g. sigl and g. raffelt , `` general kinetic description of relativistic mixed neutrinos , '' nucl . b * 406 * , 423 ( 1993 ) . b. h. j. mckellar and m. j. thomson , `` oscillating doublet neutrinos in the early universe , '' phys . d * 49 * , 2710 ( 1994 ) .

the analysis of multiple populations ( mps ) in globular clusters ( gcs ) has become a forefront area of research in astronomy . multiple red giant branches ( rgbs ) , subgiant branches ( sgbs ) , and even main sequences ( mss ) have now been observed photometrically in many gcs , while broad abundance distributions of certain elements have been detected spectroscopically in most , if not all , gcs . uv photometry has been crucial in discovering and analyzing these mps , but the johnson u and the stromgren and sloan u filters that have generally been used are relatively inefficient and very sensitive to reddening and atmospheric extinction . in contrast , the washington c filter is much broader and redder than these competing uv filters , making it far more efficient at detecting mps and much less sensitive to reddening and extinction . here we investigate the use of the washington system to uncover mps using only a 1-meter telescope . our analysis of the well - studied gc ngc 1851 finds that the c filter is both very efficient and effective at detecting its previously discovered mps in the rgb and sgb . remarkably , we have also detected an intrinsically broad ms best characterized by two distinct but heavily overlapping populations that can not be explained by binaries , field stars , or photometric errors . the ms distribution is in very good agreement with that seen on the rgb , with% of the stars belonging to the second population . there is also evidence for two sequences in the red horizontal branch , but this appears to be unrelated to the mps in this cluster . neither of these latter phenomena have been observed previously in this cluster . the redder ms stars are also more centrally concentrated than the blue ms . this is the first time mps in a ms have been discovered from the ground , and using only a 1-meter telescope . the washington system thus proves to be a very powerful tool for investigating mps , and holds particular promise for extragalactic objects where photons are limited .

1406.0550

globular clusters ( gcs ) have historically been considered quintessential simple stellar populations , but a rapidly growing body of both spectroscopic and photometric work has shown that most , if not all , gcs are composed of multiple populations ( mps ) , with distinct chemical compositions and ages . for decades , low - resolution spectroscopy of limited samples in gcs has shown evidence for cn and ch abundance variations within some gcs ( e.g. , hesser et al . 1976 ; hesser et al . 1982 ) , but establishing that these were mps and not simply effects of stellar evolution required much more detailed evidence . mps have now been observed photometrically in a large number of gcs including omega cen ( bedin et al . 2004 ) , ngc 2808 ( piotto et al . 2007 ) , ngc 1851 ( milone et al . 2008 , hereafter m08 ; han et al . 2009 , hereafter h09 ; and lee et al . 2009 , hereafter l09 ) , ngc 104 , ngc 362 , ngc 5286 , ngc 6388 , ngc 6656 , ngc 6715 and ngc 7089 ( piotto et al . 2012 ) , and m4 ( marino et al . 2008 ) . in these clusters multiple red giant branches ( rbgs ) , sub giant branches ( sgbs ) , or even main sequences ( mss ) have been detected . additionally , using 8-meter class telescopes with multi - object spectrographs , studies of large samples of gc stars at high resolution and signal to noise are now common and show that gcs have intrinsically broad abundance ranges in many elements ( e.g. , carretta et al . 2011a for ngc 1851 ) . the na : o anticorrelation is the best studied characteristic of gc abundance spreads , and it has been found in virtually all clusters that have been observed with high - quality spectra of large samples of stars ( carretta et al . however , recent observations of ruprecht 106 ( villanova et al . 2013 ) , traditionally regarded as a gc due to its mass , age , metallicity , etc . , found that it is the first such object that does not exhibit an anticorrelation or even an abundance spread in na or o , or indeed any other element , although the sample size is small and more stars are needed to consolidate this result . the interplay between good photometric and spectroscopic data has been crucial in uncovering mps . clearly , spectroscopic data are superior when there exists a large enough sample of stars observed at high resolution and signal to noise , enabling the determination of detailed abundances for a variety of elements . however , this technique is large - telescope - time intensive . photometry enjoys a hefty advantage in this respect , allowing much larger samples covering a wider range of evolutionary stages . furthermore , it provides the atmospheric parameters required for follow up determination of spectroscopic abundances . in order to properly identify mps photometrically , the choice of filters is paramount . to date , the key has been using a combination of filters that include a uv bandpass . several studies , especially sbordone et al . ( 2011 ) and carretta et al . ( 2011b ) , have now shown that realistic abundance differences in the cno elements , as expected between mps , greatly affect uv filter bandpasses because of the strong cn , ch , and nh molecular bands present ( figure 1 ) . in contrast , the redder optical filters are much less affected . because these elements and their variations are representative of mps , a uv filter is now considered crucial in order to detect mps . the johnson u , stromgren u , or sloan u filters have been used for all previous ground - based detections of mps with uv photometry , but all of these filters suffer serious limitations . they are notoriously inefficient and are confined to a wavelength range that is very sensitive to both interstellar and atmospheric extinction . therefore , they require a significant amount of large telescope time to observe most gcs well , giving up a large part of the efficiency advantage that photometry enjoys over spectroscopy . the washington photometric system ( canterna 1976 ) was designed ( wallerstein & helfer 1966 ) to derive an accurate photometric temperature and metallicity for late - type giant stars using broad band filters . the @xmath1 and @xmath2 ( temperature ) filters are very similar to @xmath3 , providing a temperature index given by the @xmath4 color which is very similar to @xmath5 . the m ( metallicity ) filter , centered near 5000 , combined with an appropriate comparison bandpass such as @xmath1 , yields a metallicity indicator . finally , at the time the system was being setup , the phenomenon of cn and ch variations within gcs was being discovered ( e.g. hesser et al . 1976 ) and it was realized that the addition of another filter designed to measure the cn / ch strength independently of the metallicity would be of great use in disentangling these two abundances . _ thus , the c ( carbon ) filter was introduced , specifically to search for mps . _ to our knowledge , this was the first and still only such filter so designed . in order to maintain the broadband characteristics of the system , the filter needed to cover a wide wavelength range that included significant cn and ch bands . a design goal was to include both the uv cn band near 3600 as well as the ch g - band near 4300 . as shown in figure 1 , the washington c filter covers the same bandpasses and molecular bands as the other uv filters used to uncover mps , and in fact more of these latter than any other competitor . it is also both much broader ( fwhm @xmath6 1000 ) and centered significantly redward ( @xmath03900 ) of the other filters ( see table 1 ) , making it much more efficient and less reddening and extinction sensitive . it even includes some sensitivity to the nh band at 3360 . note that c has a very similar blue response as johnson u but is centered 300 redder and retains significant sensitivity to beyond 4500 , far beyond the red cutoff of u. the peak transmission is also higher in c than the other filters . it is thus much faster ( typically @xmath63 times faster at obtaining the same s / n ) than u and is many times more efficient than stromgren or sloan u , especially in red giants . .the washington c filter is shown to be centered redward and significantly broader than the other uv filters . these are filter specifications taken from ( 1 ) the ctio mosaic filter set and from ( 2 ) the kpno filter set . minor variations of all these parameters occur between different copies of the same filter . [ cols="^,^,^,^,^",options="header " , ] figure 2 shows 4 cmds from our final photometry . we show representative error bars on the side placed at each magnitude . these errors have been determined by both types of errors output by daomaster : the combined photometric measurement error output by allframe and the @xmath7 based directly on the observational scatter across the multiple images . we typically find the photometric error dominates in the brightest stars but the observational scatter dominates in the fainter stars , and for each magnitude we take the largest of these two errors to be the better representation of its final error . for errors in color we add in quadrature the final errors from each input magnitude . lastly , the representative errors shown in figure 2 are found by taking the median error of all stars in a 1 magnitude range . the brightest giants are saturated in our observations , so we have expanded our sample by taking advantage of the short photometric observations of ngc 1851 in c and t1 by geisler & sarajedini ( 1999 ) and the i observation from momany y. ( private communication ) . detailed comparisons of our observations to those of geisler & sarajedini ( 1999 ) show that their magnitudes are systematically brighter than ours by 0.006 in c and 0.029 in t1 . therefore , we have offset their results to be consistent and have used x symbols in the lower - right and upper - right panels of figure 2 to distinguish their data . for expanding our t2 magnitudes , we have used the brightest @xmath01000 stars ( 14.4@[email protected] ) that have both our t2 and i from momony to create a transformation relation . because of the strong similarities between these two filters the relation is linear across this full magnitude range with little scatter , which allows us to extrapolate to stars brighter than t2=14.4 without likely introducing significant systematics . using this relation we transform the i magnitudes for the brightest stars ( t2@xmath815 ) not in our sample to t2 , and we mark this data in red in the upper - left and lower - left panels of figure 2 . in the t1-t2 cmd ( upper - left panel of figure 2 ) there are no clear , separate branches visible , and the widths of the rgb and ms are similar to the representative errors displayed . this is consistent with previous ground - based studies that have shown that colors without a uv filter do not show either multiple branches or even a significantly broadened rgb . in contrast , things are very different in all cmds including the c filter . we plot the c - t1 color both against t1 ( upper right ) and c ( lower right ) . in both cmds we see that in addition to the primary population there is a smaller but still significant population of sgb stars that are fainter than the primary sgb and a population of rgb stars that are redder than the primary rgb . the sgbs are shown more clearly in the insets . in figure 3 we focus on the rgb and sgb in c - t1 versus c and for clarity color the two rgb branches . both the fainter sgb and the redder rgb appear to be related and to create a continuous branch , a second population . these observations of a second population are similar to those of h09 using johnson u and i. in direct comparison to h09 , it should be noted that our second population stars do not appear to create as distinct or as well defined of a sequence as found in h09 . they also appear to be less well defined than our primary population stars , suggesting that this is not the result of our observations having larger errors . therefore , we can not argue that in c the second population creates a photometrically distinct sequence from the primary sgb and rgb . similar to c , however , the stromgren observations in l09 also do not show a distinct second population sequence in the rgb . in further comparison to the u - i versus u observations of h09 , the mean color separation of the red and blue rgbs are quite similar to ours . above the observed rgb bump , where the separation is the largest , both their observations and ours in c - t1 versus c show color separations of @xmath00.25 . in the lower rgb , while our color separation is more difficult to define here , we both find a color separation of @xmath00.15 . this significant change in color separation between the two populations between the upper and lower rgb is of interest . in our observations this change is more appropriately analyzed in c - t1 versus t1 ( upper - right panel of figure 2 ) because here the two populations are shifted in color but not magnitude . we see in c - t1 versus t1 , in comparison to c - t1 versus c , that the true change in color separation between the upper and lower rgb is smaller but still is of significance . we expect a color separation increase , however , because the brighter giants are also increasingly cooler , leading to stronger molecular bands , giving that identical cno variations will produce greater c magnitude differences . comparing the numbers of our red and blue rgb stars marked in figure 3 gives that the number of the clearly distinct red rgb stars is only @xmath013.6% of the total rgb population . this is significantly smaller than the @xmath030% typically found for the second population when performing observations using johnson filters ( m08 ; h09 ) , where the second populations in the sgb and rgb appear more distinctly separated . therefore , this may be related to the more diffuse nature of this second population in our observations . this discrepancy and the second population s color distribution will be discussed in more detail in section 4.2 . for simplicity we will refer to these distributions from now on as the red and the blue rgb branches , and the bright and the faint sgb branches . in the c - t2 cmd we can still clearly see the separate branches of both the sgb and the rgb , but it is of interest that in this color we can no longer clearly detect the separate rhb branches observed with c - t1 . additionally , the two sgb and rgb branches are less defined . these differences may more be related to the increased errors for t2 , rather than differing characteristics between t1 and t2 , but these factors and the significantly shorter exposure times required for the t1 filter gives a clear advantage for the t1 over the t2 filter . therefore , c - t1 will be our primary color for our analysis in the rest of this paper . but it remains ideal to still observe both t1 and t2 to verify the observed populations with both filters versus c and to compare these colors to the narrow rgb that is shown only in t1-t2 . we have further analyzed the significance of these photometric features in the rgb by comparing the color errors of the stars to their color differences from the rgb fiducial . we do this in 1 magnitude bins based on the median values of these parameters . table 3 shows our results . in t1-t2 , the ratio of width to errors ranges from 0.83 to 1.18 , so the color errors are comparable to the observed width of the rgb and further strengthens our finding that there is no meaningful broadening in t1-t2 . we perform a similar comparison for c - t1 using c - t1 versus t1 . across the full range of magnitudes the rgb appears to be intrinsically broad . in all 5 magnitude bins the width to error ratio is meaningfully larger than that seen in t1-t2 , ranging from 1.5 to almost 3 @xmath7 . the comparison for c - t2 versus t2 also shows that on the rgb the width is significantly larger than the error , at 1.3 to 2.6 @xmath7 . the width of the broadened rgbs is the least significant in the faintest ( hottest ) rgb stars where the molecular bands will be the weakest . however , even at their weakest the rgb in c - t1 or c - t2 is still broader than that observed for t1-t2 . c c c c magnitude range & median width & median error & width to error ratio + + 15 - 16 & 0.018 & 0.022 & 0.83 + 16 - 17 & 0.023 & 0.020 & 1.18 + 17 - 18 & 0.019 & 0.019 & 1.04 + + 13 - 14 & 0.035 & 0.023 & 1.50 + 14 - 15 & 0.042 & 0.016 & 2.69 + 15 - 16 & 0.039 & 0.014 & 2.77 + 16 - 17 & 0.032 & 0.017 & 1.95 + 17 - 18 & 0.028 & 0.018 & 1.55 + + 14.4 - 15 & 0.056 & 0.022 & 2.59 + 15 - 16 & 0.035 & 0.021 & 1.64 + 16 - 17 & 0.034 & 0.020 & 1.70 + 17 - 17.9 & 0.027 & 0.021 & 1.34 + figure 4 shows 4 additional cmds based on combinations of our photometry and the b and v photometry . in the upper - left panel we look at v - t1 versus t1 , which as expected is very similar to the t1-t2 versus t1 cmd from figure 2 . the more limited errors given for the b and v magnitudes do not allow us to create a detailed and consistent presentation of representative errors , but similar to what is seen in t1-t2 the rgb does not present multiple branches or appear meaningfully broadened in this color . in the lower - right panel we look at c - v versus c , and this cmd shares many similarities with both the c - t1 and c - t2 versus c cmds from figure 2 . in particular , the rgb is heavily broadened with a more populous blue rgb and a sparser red rgb , and a faint sgb branch is also seen . the upper - right panel shows the very interesting b - t1 versus t1 figure . while the color does not involve the key c filter , the johnson b filter overlaps with the c filter and still contains some of the important cn and ch bands ( see figure 1 ) that are believed to create the c magnitude variations . consistent with this , the b - t1 versus t1 cmd exhibits a less significant but detectable population of red rgb stars that are similar to that seen in c - t1 versus t1 . this leads us to the lower - left panel , which analyzes the c and b magnitudes using c - b versus c. consistent with the previous observations , the rgb still is broadened , but less so than that observed in c - v . this is because both the c and b magnitudes are affected by the cno variations , leading to a weaker but still detectable color difference between the two populations . we can also search for mps on the ms of ngc 1851 , which have never been detected before , despite even being searched for in hst data ( see m08 ) . to limit our errors , we will first concentrate on stars well outside of the dense core but not apply too strict of an error cut or we will lose the faint ms stars . in figure 5 we only show stars that have a c - t1 color error of @xmath80.05 and that are in an annulus around the cluster center from 4 to 6.5 arcminutes . the half - mass radius from harris ( 1996 ) is 0.52 arcminutes , giving that this is 7.7 to 12.5 half - mass radii from the center . additionally , the representative error bars shown have been updated based on this subsample . we have plotted all of the colors versus t1 to increase the color difference between two potential populations . if we plotted versus c , a second redder population would also be shifted fainter , partly parallel with the ms , and decrease the apparent color difference at constant magnitude . remarkably , in the left panel of figure 5 we see clear signs for a broad ms with a substantial wing extending to the red that is not seen to the blue . this moderately redder ms population follows the same general shape as the primary population . additionally , at the primary population s turnoff this red population maintains its color separation and does not simply merge into the turnoff . indeed , it appears to transition into the faint sgb that we have previously identified . to help illustrate this second population we have colored its likely members red in figure 5 , where we have based this solely on c - t1 and for stars with t1@xmath619 . we must further analyze this red branch to test if it is a true second population or is simply a binary sequence , but we must also consider if it is a combination of both . in the left panel of figure 5 we first characterize the primary ms with a solid - black line . we also illustrate a corresponding equal - mass binary sequence with a black - dashed line , which characterizes where the upper envelope of the binary sequence would be . this shows that the binaries would merge into the primary turnoff and then have their own turnoff at a brighter magnitude . comparing to our data we find at the primary turnoff , even when considering errors , our observed second population still remains distinct and moderately red , and at fainter magnitudes the observed stars are increasingly too blue to be consistent with a binary sequence . furthermore , there are a negligible number of stars that are consistent with a second brighter turnoff of binaries . in figure 6 we further illustrate the characteristics of a binary sequence using population synthesis of a @xmath9 population with a binary fraction of 3% using the trilegal code ( girardi et al . we have input errors in this population synthesis based on our observed relation between photometric error and magnitude . to help distinguish the binaries in figure 6 we have colored red all binaries with a mass fraction @xmath60.6 , while binaries of lower mass fraction will not be photometrically distinguishable from single stars . this population synthesis in c - t1 further shows that the binaries will appear relatively very red at fainter magnitudes , quickly begin to converge in color at brighter magnitudes and merge into the primary turnoff , and then have their own weakly populated turnoff at brighter magnitudes . in contrast to this , a true second population would be shifted redder in c - t1 but have no meaningful change in t1 magnitude . the median c - t1 color difference between these two branches is 0.096 , and in figure 5 the solid - red line demonstrates that this color shift is more comparable to the red ms stars . that the brighter stars are moderately bluer and the fainter stars moderately redder than this uniform color shift is further consistency with the two populations in the rgb stars ( see figure 2 ) , which also exhibited changing color separations . in both cases the cooler stars , which will have stronger molecular bands , will create increased color separations at equivalent cno abundance variations . in the center and right panels of figure 5 we will now analyze this observed second ms branch in other colors by taking the selected sample from c - t1 ( colored red ) and also coloring them red in these cmds . in c - t2 ( center panel of figure 5 ) we see there is still a prominent redder population that is primarily composed of the same red stars observed in c - t1 . the scatter of the two populations appears larger in this color , but this is consistent with t2 s moderately greater error . the median c - t2 color difference between the marked black and red populations is 0.109 ( similar to the 0.096 in c - t1 ) , and similarly shows the color difference between the two populations moderately increases in the fainter ( cooler ) stars . comparing to the population synthesis for c - t2 ( center panel of figure 6 ) , we see the same binary characteristics found in c - t1 and the same differences when compared to our observed second branch in c - t2 . however , there is one key difference between the ms in c - t1 in comparison c - t2 that becomes readily apparent in their synthetic binary sequences : across the same t1 magnitude range the ms changes color in c - t2 more rapidly with decreasing magnitude , which greatly increases the apparent color separation between the single and binary sequences . therefore , the greater observed color dispersion in the red branch of c - t2 ( center panel of figure 5 ) may be indicative of at least minor binary contamination , but overall because the color difference between the two populations in c - t1 and c - t2 are in agreement , this suggests that this is predominantly a true second population created by a shift in c magnitude . we now mark these observed red - population stars in t1-t2 ( right panel of figure 5 ) and see that in this color there is no clear difference . the red - population stars observed in both c - t1 and c - t2 are distributed remarkably uniformly within this single sequence , and the median t1-t2 color difference is insignificant at 0.014 . in contrast to this , the right panel of figure 6 shows that in t1-t2 the binaries will still remain meaningfully redder than the single - star sequence . however , in our t1-t2 observations we do note a small number of color outliers are still seen in both populations . the total t1-t2 color distribution has a @xmath7 of 0.026 and we define the outliers as stars more than 2.5@xmath7 from the central sequence . 2.6% of the primary population ( black stars ) are t1-t2 color outliers , with 1.1% being blue outliers and 1.5% being red outliers . in comparison a more significant 10.4% of the red population are outliers , but with only 2.3% being blue outliers and a much larger fraction of 8.1% being red outliers . therefore , based on the small number statistics , in the primary population there is a comparable number of red and blue outliers , which are also comparable with the number of blue outliers in the red population . this suggests that @xmath03 - 4% of both populations are true photometric outliers , non - members , or stars with abnormally large errors , which will approximately be distributed evenly between red and blue outliers . correcting for this there remains an additional @xmath06 - 7% of the red population that are red outliers in t1-t2 . this small fraction of the red population ( @xmath00.65% of the total population ) may be true binaries that have a large mass ratio and will appear to be redder than the primary single - star population in all colors , while 93 - 94% of the red population remains strongly consistent with a second population of differing c magnitude . if this second population of stars is not caused by a binary sequence , why do we not see a meaningful population of binaries in this outer annulus of ngc 1851 ? the recent definitive analysis of binary populations in gcs by milone et al . ( 2012a ) may provide an answer . they observed the ms in the central region ( @xmath82.5 arcminutes from the center ) with acs / wfc in f606w and f814w , very similar to t1 and t2 , giving a color that should make our two purported populations virtually indistinguishable in the ms . they observed that ngc 1851 only has a binary fraction of [email protected]% for binaries with a mass ratio of greater than 0.5 . binaries with mass fractions of less than 0.5 were not able to be differentiated from single stars with their observations . based on their models they estimate that ngc 1851 has a total binary fraction of only [email protected]% in the observed central region . this is one of the lowest binary fractions of the 59 gcs they analyzed . furthermore , their observations find that a majority of the observed gcs , including ngc 1851 , have decreasing binary fractions at increasing distance from their center . this is consistent with expectations that binary systems , which dynamically act like a single star with mass equal to the sum of the binary components , will over time become more centrally concentrated in comparison to the less massive single stars . all of these factors suggest that in our observations of this outer annulus from 4 to 6.5 arcminutes , the binaries that would meaningfully contribute to a binary sequence ( mass [email protected] ) is less than 1% of the total population . this is remarkably similar to our estimate from figure 5 that in this outer annulus only @xmath00.65% of our ms stars are consistent with large - mass - ratio binaries . the number ratio for the red population shown in figure 5 is 9.5% , which is comparable to the 13.6% found for the red rgb in comparison to the blue rgb in section 3 . additionally , a comparable analysis of the sgb subsample shown in figure 5 gives that the faint branch of the sgb is 12.7% of the total population . the agreement of the number ratios in all 3 regions of the cmd further strengthens the conclusion that this extended red population in the ms is a second population consistent with the one observed in both the sgb and rgb . however , it should be noted again that these number ratios are based on the photometrically distinct red stars , and more reliable color distributions and population ratios will be discussed in section 4.2 . to further broaden our analysis , we can use the available b and v magnitudes . in figure 7 we show the same sample of stars from figure 5 ( when b and v are available ) using c - v , b - t1 , and b - v . as in figure 5 we again have marked the red population from c - t1 in all 3 cmds . in the left panel of figure 7 we see that in b - t1 the red ms population is not as distinct as in c - t1 or c - t2 , but it is still significantly redder than the primary ms by 0.076 . this is similar to what was seen with the rgb in b - t1 ( figure 4 ) , which illustrated that in this color the two populations still exhibit a significant but smaller color difference to that seen in colors with c. as with the rgb , this is expected since cno variations will also cause moderate differences in b magnitudes . the c - v color in the central panel is more surprising because even though the red ms still is consistently redder than the primary sequence , the difference is comparably weak at only 0.052 . lastly , in b - v we see that while there is not a large difference in color between the two populations , the red ms is still moderately redder than the primary ms by 0.030 . this is relatively a minor difference but it is not insignificant . both the central and right panels of figure 7 suggest that for these two ms populations there may also be a weak difference in v magnitude , which results in a weaker than expected color difference when comparing to both c and b. could this apparent difference in v magnitude for these two populations be indicative of additional effects beyond cno variations for the two populations , e.g. , possible differences in helium abundance , metallicity , or a large total c+n+o difference between these two populations ? this red ms population is very unlikely to be due to field stars , as it is centrally concentrated ( see section 6 ) , and the field contamination is shown to be relatively minor in the proper motion analysis by m08 . stars photometrically consistent with binaries appear to only represent a minor portion ( @xmath06 - 7% ) of this red ms , consistent with this cluster s small binary population ( milone et al 2012a ) . furthermore , the effects of photometric error and crowding do not seem to be artificially creating this second branch ( see the appendix for detailed discussion of this ) . therefore , figures 5 to 7 provide strong evidence that these two ms branches are predominantly two populations of differing c magnitude with no meaningful difference in either t1 or t2 , qualitatively consistent with the second population observed in both the sgb and rgb . with strong evidence for two ms populations found in an outer annulus subsample of @xmath01000 main - sequence stars , it is of great interest to look at a larger sample covering nearly the full field . beginning with all stars from our original sample shown in figure 2 , which only placed a stringent error cut on the stars within 2 arcminutes of the center , we will focus on the 3162 ms stars with t1@xmath619 and c - t1 color errors @xmath110.05 . with this large of a sample it is more informative to analyze the color distribution by taking a ms fiducial and creating a histogram of the color residuals in c - t1 versus t1 . the left panel of figure 8 shows this color distribution with bin sizes of 0.02 , and we see that there is a significant sample of broadly distributed stars in the red wing that are not seen in the blue wing . a kmm mixture modeling test ( ashman et al . 1994 ) indicates that the probability this distribution is only a single gaussian and not a bimodal distribution is essentially null . initial attempts to fit these two possible populations using two offset gaussians of equal sigma were not satisfactory . while the central peak is fit well by a gaussian with a @xmath7 of 0.033 , attempting to fit the redder population with a similarly narrow @xmath7 does a poor job of matching both the extended red wing and the blue wing . based on the analysis of synthetic cno variations and abundances by carretta et al . ( 2011b ) and ca11 , they find that in colors involving stromgren u the two rgb populations in ngc 1851 are not distinct sequences , similar to our c observations . in these papers they argue that the redder population is significantly broader in color than the blue population and the two populations are heavily overlapped , where the full redder population extends nearly as blue as the primary population . this suggests that our observed redder population may also only be the reddest wing of a broadly distributed population . our synthetic analysis of the effects of cno variations on the c magnitude , which will be discussed in detail in cummings et al . ( in prep . ) , shows the c filter also creates two heavily overlapping populations with a broader population extending well into the red . the left panel of figure 8 shows our fit of the color distribution when assuming these population characteristics , with a large but narrow population with @xmath7=0.031 and a smaller but significantly broader population with @xmath7=0.075 . this second population extends as blue as the first but also extends significantly redder creating the second population in the red wing . a clear advantage of this fit is that both the extremely blue and red stars are in agreement with the two gaussians , which was not the case when using two slightly overlapping gaussians of equal @xmath7 . for clarification this broad second population has partly been broadened further in this color distribution because its relative color difference between the primary population increases in the fainter ( cooler ) stars , so at any given magnitude its breadth will be smaller than 0.075 but still more significant than the narrow population with @xmath7 of only 0.031 . the peak heights and @xmath7 values of the blue and red distribution gaussians are 556 and 0.031 and 99 and 0.075 , respectively , giving that the broad and red ms is 30.1% of the total population . this is comparable to the ratio of 27.9% found for the percentage of blue hb to all hb stars , which are two far more reliably separated populations . this ratio is significantly larger than the population ratios found from figure 3 and 5 of 13.6% for the red rgb , 9.5% for the red ms , and 12.7% for the faint sgb , but as suggested these smaller red populations analyzed are only the extreme red wing of the full broadly distributed second population . for independent comparison , m08 found that their fainter sgb was @xmath030% of the total sgb population . similarly , h09 found in their u - i observations that the secondary red rgb and faint sgb branches were also @xmath030% of the total population . to further test these significantly overlapping gaussian fits , we have reanalyzed in c - t1 versus t1 the rgb from figure 2 using an identical color distribution method . the right panel of figure 8 shows the rgb color distribution for the 319 stars from 17.5@xmath8t1@xmath814 , and we can fit it quite well with two gaussians very similar to those we used for the ms . again , the kmm test very strongly favors a bimodal over a unimodal gaussian distribution . as discussed in section 3 , the color separation of the two rgb populations increases at brighter magnitudes , even more than that seen in the ms . therefore , even with significantly lower color errors , the redder population requires an even broader gaussian ( @xmath7 of 0.09 instead of 0.075 ) . the two rgb populations are also more separated in color , with the gaussian centers separated by 0.044 in the ms but by 0.064 in the rgb . consistent with this difference , we will discuss in detail in cummings et al . ( in prep . ) that cno variations will cause differences in c magnitude in both the ms and rgb , but the differences is greater in the typically cooler rgb stars . lastly , the peak heights and @xmath7 s of the blue and red distributions are 67.5 and 0.036 , and 12 and 0.09 , respectively , giving that the red rgb is 30.8% of the total rgb population . this is again in strong agreement with that found for the two ms populations and the two hb populations . this agreement adds further strength to characterizing the two populations with this method . in m08 , where the split sgb was first discovered in ngc 1851 , they also looked for evidence of mps in the ms . in contrast to what we have found , m08 did not find a second or even broadened ms , but they primarily searched for a clear split in the ms like that observed in ngc 2808 ( piotto et al . 2007 ) rather than two heavily overlapping populations such as we have found evidence for in our observations . furthermore , there are several key differences between their analysis and ours that may explain why they would not have found evidence for a second ms population . first , they analyzed the ms with f336w and f814w ( johnson u and i ) filters , but they did their analysis using only u - i versus u. plotting versus a uv magnitude , as we discussed in section 4.1 , diminishes the color separation of the two populations because the second population will be shifted both redward and fainter , partly parallel with the ms . while there remains signatures of a moderate redder population in their cmd ( see figure 4 of m08 ) , they do not appear as significant as those we detect . m08 do not mention these redder stars and they possibly assumed that they were a standard binary sequence . second , in their color distributions they removed all stars that were more than 4@xmath7 from the center of the distribution , likely removing many of these extreme red stars . third , their u - i color distribution is also quite broad , giving a large @xmath7 of 0.052 in their upper - ms stars and even greater @xmath7 in their fainter ms . in contrast , our primary ms exhibits a narrower @xmath7 of 0.031 , suggesting that their color errors are more significant . m08 also acknowledge that their u magnitude errors are large , but they do not test in detail if their errors can fully explain the large @xmath7 . all of these factors will have limited the significance of this second population . intrinsic differences between the u and c filters may also play an important role in causing this second ms to be undetected . in cummings ( in prep . ) we show that , assuming a constant c+n+o as found in v10 , variations in individual cno abundances that are consistent with the spectroscopic analyses of ngc 1851 ( gratton et al . 2012a and ca11 ) predict magnitude differences in c for upper - ms stars comparable to what we are observing here , and with no meaningful magnitude differences in r ( t1 ) or i ( t2 ) . when applying identical cno variations to f336w ( u ) we find that the predicted u magnitude variations are comparable to the variations in c magnitude . however , if we apply a moderate c and o anticorrelation , which due to the characteristics of u is important to consider , we find that the u magnitude variations will be relatively weaker ( @xmath080% ) than that predicted in the c filter . it should also be noted that if the difference in c abundances across the two populations is increased , a possibility due to the limited number of red ms stars observed in gratton et al . 2012a , the ability of the c filter to detect the two populations in the ms is greatly increased over the u filter . lastly , possible he and metallicity differences ( joo et al . 2013 and ca11 ) may also play an important role by how the two populations differently affect u versus c. mps in globular cluster mss have only been photometrically observed in a small number of clusters before , first with omega cen ( bedin et al . 2004 ) and ngc 2808 ( piotto et al . 2007 ) , and more recently with 47 tuc , ngc 6752 , ngc 6397 , and ngc 6441 ( anderson et al . 2009 , milone et al . 2010 , 2012b , c , bellini et al . however , all of these multiple mss have only been detected with hst photometry . therefore , it is quite remarkable to detect a possible second ms using relatively little telescope time on a ground - based 1-meter telescope . another important difference with our second ms is that , other than in the very complex case of omega cen , these previously observed secondary sequences are bluer than the primary ms . but it should be noted that the secondary rgb populations also are bluer in ngc 6752 , ngc 6441 , and 47 tuc , the opposite of what we observe in the rgb of ngc 1851 . carretta et al . ( 2010 ) also find from spectroscopic analysis of ngc 6397 that its o - poor / na - rich ( photometrically redder ) population also dominates . therefore , this suggests that in comparison to these other clusters ngc 1851 may have unique differences between its mps and that the mechanism of their formation may also be different , previous published photometry of the rhb for ngc 1851 has already shown it to be particularly broad in magnitude ( e.g. grundahl et al . 1999 , m08 , and h09 ) . in our observations shown in figure 2 in c - t1 versus c and versus t1 we find evidence for a split in the rhb . further analysis of this split structure can be performed by analyzing where the stars of the sequences fall in a variety of cmds . figure 9 shows all cmd combinations using our 3 filters with additional cmds based on v. the cmd of c - t1 versus t1 has the most significant split . based on the c - t1 versus t1 diagram , we have separated the rhb into 2 groups : a brighter and redder group and a fainter and bluer group . these identical color markers are applied to these same stars in all other cmds shown , and we have included representative error bars determined by the median magnitude and color errors of only the stars we have colored . quite strikingly , while the plots involving c - t1 show the clearest separated sequences , nearly all of the rhbs still show the 2 color groups with little overlap . these results indicate that while we do not observe distinct and separated groups in t2 magnitude , the two rhb groups have a meaningful difference in this magnitude similar to their difference in t1 . the lack of more distinct t2 magnitude differences may solely be the result of the t2 errors being nearly double those of t1 for these stars . the only combinations that do show significant overlap of the two groups are t1-t2 versus c and to a lesser extent versus v. the overlap in color is expected due to the consistency of the t1 and t2 magnitudes in these rhb stars , hence in the t1-t2 color the two groups have no meaningful difference . the overlap in magnitude suggests that for both c and v there is not as significant of a magnitude difference between the two groups . it is unclear what could cause the rhb stars to have differing sets of t1 magnitudes and a significant spread in t2 magnitudes ( if not distinct sets ) at consistent c. however , these figures further suggest that at least the significant spread , if not separate sequences , in the rhb are a real feature in t1 and t2 . while the bhb and the rhb already represent the two established populations in ngc 1851 , does the structure in the rhb suggest further possible differences within its single population ? a split rhb has been seen before in terzan 5 , 47 tuc , ngc 6440 , ngc 6569 , and ngc 6388 ( ferraro et al . 2009 , milone et al . 2012b , mauro et al . 2012 , and bellini et al . the split in terzan 5 , ngc 6569 , and ngc 6388 were all observed using j and k filters . our t1-t2 versus t1 and t2 cmds have similar characteristics to those observed in j and k : two clumps that have a moderate spread in color and are offset in magnitude . similar to ngc 1851 the brighter clumps in both ngc 6440 and terzan 5 are slightly redder than the faint clump , but in ngc 6569 there is no significant color difference between the two groups . in ngc 1851 the rhb clumps have a t1-magnitude difference of @xmath00.1 , which is similar to the k - magnitude difference observed in all other clusters besides terzan 5 , which has a striking @xmath00.5 k - magnitude difference . the split rhbs in 47 tuc and ngc 6388 were observed differently by using combinations of various uv filters and f435w , f606w , and f814w ( the last 3 of which are comparable to b , v , and i ) . the rhb structure in these filters also show similar characteristics to ours , and again this suggests that these split sequences of stars are not meaningfully different in u but are in i. however , we should note that in 47 tuc , milone et al . ( 2012b ) show that its double rhb is also visible but not clear in f275w - f336w . two groups have recently modeled the hb in ngc 1851 , which provides us important comparisons . first , kunder et al . ( 2013 ) performed hb synthesis based on the abundances described in gratton et al . ( 2012b ) , and they were able to create a very broad rhb in v - i versus v but with no split . in our similar v - t1 versus v diagram we also observe a broad rhb with no clear split , but figure 9 demonstrates that in this color the brighter extension of the rhb is composed of the brighter of the two split sequences we observe in c - t1 . kunder et al . ( 2013 ) considered in their synthesis the ba - rich , na - rich , and o - poor rhb subpopulation ( @xmath010% ) from gratton et al . ( 2012b ) , but they find that it is not related to the brighter rhb extension , which suggests abundance differences are not the key here . second , joo & lee ( 2013 ) did full cluster models for two populations in ngc 1851 , and they were able to reproduce the general photometric characteristics of ngc 1851 . as already well established , their two hb populations recreate the distinct rhb ( primary population ) and bhb ( second population ) with no overlap , but within the single rhb population they created in v - i versus v the heavily broadened rhb with even a weak possible split . their rhb structure was created through variations in stellar masses and mass - loss rates . consistent with this rhb split occurring in ngc 1851 without meaningful abundance variations , the hb abundances from gratton et al . ( 2012b ) show that the rhb stars have no meaningful spread in fe or ca . in contrast to this , the two rhbs in terzan 5 show a significant @xmath00.5 dex difference in [ fe / h ] ( ferraro et al . 2009 ; origlia et al . 2011 ) and a @xmath00.3 dex difference in [ @xmath12/fe ] . these abundance differences , combined with a possible large age difference , are likely the reason for the far more significant magnitude difference observed between its two rhbs . at a less significant magnitude , the metallicity analysis of ngc 6569 by valenti et al . ( 2011 ) suggests that overall it has a bimodal metallicity with a difference of @xmath00.08 dex , but this is based only on a very limited sample of 6 rgb stars and is not a direct comparison of stars in the two rhb groups . all of these split rhbs photometrically show many similar characteristics , in particular by being most prominent in red and ir photometry . additionally , their recent discovery in a number of gcs , including in already heavily studied clusters , suggests that it may be a more common feature than previously thought . searches for more of these split rhbs and detailed analyses for any abundance differences between the two clumps , as well as further understanding of the effects of mass loss , will be necessary to provide a complete picture for the cause or possibly multiple causes of this phenomenon . a useful method to analyze the potential formation mechanisms for the mps in ngc 1851 is to see if they have differing radial distributions . zoccali et al . ( 2009 ) have argued that the two different rgb branches have a differing radial distribution , with the redder rgb population being more centrally concentrated , but both milone et al . ( 2009 ) and olszewski et al . ( 2009 ) have found no evidence for this . the spectroscopic analysis of carretta et al . ( 2010 ) , in contrast , presents evidence for variations in radial distribution based on their metal - poor and metal - rich populations . in ca11 they discuss their disagreement with the conclusions of milone et al . ( 2009 ) and how the milone et al . analysis methods may have prevented them from observing a statistically meaningful difference . if the two populations have a differing radial distribution , this is consistent with the cluster having two star formation epochs , with an initial more extended population and a subsequent polluted population forming more centrally concentrated ( dercole et al . however , this only represents the initial radial distributions and the difference would be slowly washed out as the cluster dynamically relaxes . before analyzing our data for potential differences in radial distribution , we must consider possible effects of our zero - point photometric corrections ( see section 2 ) . in colors involving c a more centrally concentrated red population would create a redder average color near the core in comparison to stars at larger radii in both the ms and rgb . this could possibly affect our spatial corrections in c. in contrast , this would have no effect on the comparisons of the clearly separated red and blue hb , and this would not affect our corrections of either t1 or t2 . our analysis of the spatial correction in c with respect to v shows that while in total it is not radially dependent , it does have a radial component centered near the core of the cluster . however , this component shows the core was too blue rather than too red in c - v , and consistent radial components are also found in both t1 and t2 . this suggests that it is not an effect introduced by the two populations having differing radial distributions , or it would not similarly be seen in t1 and t2 . these radial components observed are more suggestive of the effects of increasing crowding at decreasing radii , which as seen will affect all 3 filters and on average artificially brighten the crowded stars . therefore , these photometric corrections will not affect our analysis of the radial distributions for these populations . for the radial distributions based on the two rgb branches , we have selected the blue and the red rgbs from 14@[email protected] ( see right - panel of figure 7 ) . we have defined all stars with @xmath13 from 0.06 to 0.30 as the red rgb ( 53 stars ) and all stars from -0.075 to 0.02 as the blue rgb ( 206 stars ) . this constrained color range helps to avoid regions with heavy population overlap . a ks test between these two rgb samples shows that they do not have a meaningful difference in radial distribution ( p - value=0.55 ) . a similar comparison of the radial distributions of the blue hb ( 36 stars ) and the red hb ( 93 stars ) , which are more reliably separated populations , yields that there also is no meaningful difference between the radial distributions ( p - value=0.394 ) . while the small number of available stars in the rgb and hb limits this analysis , it should be noted that combining their samples still gives no meaningful difference in their radial distributions . use of the larger ms sample ( see left - panel of figure 7 ) will greatly increase the number of available stars . to create a smoothly - varying radial profile we will remove the small number of uncut ms stars that are within 2 arcminutes of the center . to reliably assign stars to the two populations we will first avoid the regions with significant photometric overlap and define all stars with @xmath13 from -0.035 to 0.0 as the blue population ( 919 stars ) and all stars from 0.085 to 0.30 as the red population ( 257 stars ) . the upper panel of figure 10 compares these samples and we find at a significant level ( p - value=0.0 ) that their radial distributions are distinct , with the red population being more centrally concentrated . this radial comparison for the two ms populations also solidifies our supposition that the red ms is composed of cluster stars and is not a contaminating sample of field stars , since the field - star spatial distribution would not significantly vary . as a further test , we compare the radial distribution of this same blue ms population to a sample of redder stars from 0.02 to 0.04 ( 424 stars ) . even though this second sample is redder , figure 7 shows that it is still heavily dominated by the primary blue population . in the lower panel of figure 10 we show that these two blue ms samples show no significant difference in their radial distribution ( p - value 0.371 ) . this indicates that our data is not affected by a systematic color shift to the red at decreasing radii , which would similarly create our result in the upper panel of figure 10 but similarly create a higher central concentration for this redder wing of the blue ms . this further strengthens the significance of our two population analysis in the ms and that the red ms has a higher central concentration . we can also analyze the radial distributions of the two groups we have observed in the rhb , and we find that the brighter and redder rhb ( 43 stars ) does not show a meaningful difference ( p - value=0.346 ) in comparison to the fainter and bluer rhb group ( 51 stars ) . similarly , the two rhb groups in both ngc 6440 and ngc 6569 were not found to have differing radial distributions ( mauro et al . in contrast to this , the brighter group of rhb stars in both terzan 5 and tuc 47 were found to be more centrally concentrated ( ferraro et al . 2009 , milone et al . the reasons for the radial differences between the two rhb groups being observed in only some clusters may be another indication of different mechanisms for the formation of the two rhb groups . for example , the large abundance differences and radial distribution differences observed in terzan 5 may further indicate that the two rhbs are two distinct populations being observed . in ngc 1851 its two populations are already represented by the bhb and the rhb . its two rhb sequences likely do not have further abundance differences and are only composed of one population . considering all of these radial distributions , we can argue there is a higher central concentration for the redder population only for the ms stars . for the other evolutionary stages , there is no strong evidence . this may be the result of the smaller numbers ( 1230 for the ms versus 259 for the rgb and only 129 for the hb ) , but it most likely is indicative of other issues . for example , the binaries that we have shown to be a minor component of this red ms also clearly have a higher concentration in the core ( see milone et al . are these binaries biasing our radial distribution analysis in the ms ? possibly , but because they are a relatively small number of binaries , this likely is not at a level significant enough to fully explain our observations . the lack of detection in the rgb and hb may also be indicative of the effects of mass segregation / dynamical relaxation in this cluster , which will wash out initial differences in radial distributions for the two populations . this is also illustrated in the higher concentration of binaries in the core . do the low - mass ms stars provide a more reliable tracer of the initial radial distributions of these two populations , while the more massive rgb and hb stars have all become heavily concentrated in the center resulting in a weakened , if not completely removed , initial difference in radial distribution ? we have shown that the washington c filter that is very broad and specifically designed for cn / ch is a powerful tool for detecting mps in gcs . this provides a great advantage for photometrically studying mps because the previous methods have either used extremely high precision hst photometry or inefficient ground - based uv photometry using the narrower and bluer johnson u or stromgren or sloan u filter . this has previously made studying mps in gcs telescope time intensive and limited in depth . mps have been observed in the rgb and sgb of ngc 1851 before , which we have observed as well , but our data was obtained in a relatively short time on a 1-meter telescope . in addition , our observations with the washington c filter have provided strong evidence for mps in the ms of ngc 1851 . this is the first case of mps on the ms of a gc being uncovered via ground - based data , and the small aperture we used makes this doubly impressive . with the high - precision t1 photometry we have evidence for a double sequence in t1 and possibly t2 in the rhb . previously observed split rhb s have shown many similar characteristics to those we have observed . in some clusters it is believed the split is caused by the two rhb clumps having different metallicity and possibly age ( e.g. terzan 5 ) , but in all other clusters there remains no strong evidence for a metallicity difference within the rhb . when the split is only moderate ( i.e. a magnitude difference of @xmath00.1 ) and there is no meaningful metallicity or radial distribution difference they may solely be the result of variations in the mass - loss rates . detailed analysis of our two observed mss in all available filter combinations strongly argues that the redder sequence is caused by a fainter c magnitude for this population . this argues that these stars are representative of a second population and not a binary sequence . we further test the reality of mps on the ms by comparing to a binary sequence created through population synthesis , and find further distinction between our observations and a binary sequence . however , in our analysis of color outliers in t1-t2 we found that a subsample ( @xmath06 - 7% ) of the red population show characteristics consistent with a binary population that have high - mass ratios . in comparison to the total population this is @xmath00.65% , which is very small but similar to the high - mass - ratio binary fraction of [email protected]% observed in the core of ngc 1851 ( milone et al . 2012a ) . based on models for the photometric effects of varying cno abundance ( carretta et al . 2011b ; cummings et al in prep . ) , we have been able to fit the color distribution of the two populations on the ms with a narrow blue population and a heavily overlapping and broader population that is centered moderately redder . the red wing of this broader population extends well beyond the blue population creating the broadened ms and rgb and the clear signatures of a mp . fitting the full width of the broad second population shows that it is @xmath030% of the full population in both the ms and rgb . this is in remarkable agreement with the 27.7% ratio found from comparison of the bhb to the rhb numbers , where the bhb is believed to be representative of the evolved redder population in the hb . it is also consistent with the population ratios found by m08 and h09 . lastly , whether there is a difference in radial distribution between these two populations remains a heavily debated topic . independent analysis of our two photometric populations in all parts of the cmd show that in the high - mass stars ( but using relatively limited numbers ) in both the rgb and hb there is no meaningful difference found in the radial distributions of the two populations . however , in the low - mass ms stars there is a statistically significant difference in the radial distributions of the two populations , where the redder population is more centrally concentrated . is the differing result when comparing the high - mass and the low - mass stars a matter of small numbers , or is it indicative of the contaminating effects of binaries in the red ms that are known to have a higher central concentration , or does it indicate that the radial differences created during the formation of the cluster only remains in the low - mass stars and have already been dynamically washed out in the highest - mass stars ? the binary explanation at least seems remote given the very low incidence of these stars . these mps in the ms , the sgb , and the rgb are all observed by their variations in c magnitude and consistent t1 or t2 magnitude . the c magnitude variations are believed to be caused by variations in the strong cn , ch , and nh bands present in the c filter , suggesting a distinct difference between these populations in c or n abundance , if not both . this is consistent with previous spectroscopic analysis of ngc 1851 ( e.g. v10 ; gratton et al . 2012a , carretta et al . 2014 ) . we will discuss this in more detail in our following paper on ngc 1851 where we match available abundances to our photometry and analyze the possible correlations . we will also synthetically analyze the effects of cno , helium , and metallicity variations in the washington and johnson filter systems ( cummings et al . in prep ) . many questions remain about ngc 1851 , and further models and abundance analyses of a wide variety of elements in all evolutionary stages of this cluster are needed to help create a complete and self - consistent picture for not only what the differences between these populations are but how these populations formed . additionally , as with many questions in astronomy , analysis of mps in as wide a variety of clusters as possible will be helpful in shedding light on these remaining questions . with the considerable advantages that the washington c filter provides for photometrically discovering and analyzing these mps in massive star clusters , further washington observations of a broad range of gcs will be very helpful in this endeavor . the efficiency of the system for investigating mps means that studies of the host of magellanic cloud clusters , with their range in age and metallicity parameter space not covered by galactic gcs , are within reach of 4 to 8-meter - class ground - based telescopes . the inclusion of the f390w c filter in the wfc3 on board hst opens up the exciting possibility of exploring mps in m31 gcs . the authors would like to acknowledge the sharing of data by y. momany . dg and sv gratefully acknowledge support from the chilean centro de astrofisica fondap basal pfb-06/2007 and the chilean centro de excelencia en astrofisica y tecnologias afines ( cata ) . jc is thankful for the financial support from fondo gemini - conicyt 32100008 and the support by the national science foundation ( nsf ) through grant ast-1211719 . sv gratefully acknowledges the support provided by fondecyt n. 1130721 fig2cfinrfin3cricor622.pdf ( 58.8,69.5 ) ( 58.8,38.5 ) ( 58.8,60.5 ) ( 58.8,29.5 ) ( 28.5,38.5 ) ( 28.5,29.5 ) n1851crrgb.pdf fig3newtrim.pdf [email protected] with c - t1 color error @xmath110.05 and their difference from the ms fit shown in the left panel of figure 5 . two populations are fit by two gaussian distributions , which have characteristics qualitatively based on the models of carretta et al . ( 2011b ) . the sum of the two gaussians are also shown . this reliably fits the full range of the color distribution , including the heavily extended wings . the right panel shows the c - t1 color distribution of all of the rgb stars shown in figure 2 . we have applied similar gaussian fits as those shown in the ms . the gaussian fits in both panels give two populations of @xmath030 and 70% ratios with a smaller population that is typically redder and covers a broader range in color . ] section 4.1 provides strong evidence that we have a distinct second redder ms population . while we initially limited our ms sample to stars with relatively low c - t1 color error of @xmath80.05 , we should still test if photometric errors could be playing a role by falsely creating this red sequence . for the two populations we have compared their color errors to their color distribution and the corresponding blue ms @xmath7 . the blue ms has a median color error of 0.034 . its color distribution gives a @xmath7 of 0.028 with a median color residual of 0.020 , which can be sufficiently explained the blue ms errors . the red ms has a comparable median color error of 0.037 , which demonstrates its stars have no significant increase in errors . direct comparisons of their color residual to color error for individual red - population stars gives ratios that range from 0.94 to 7.42 with a median ratio of 2.26 . similarly , the red star color residuals range from 1.34 to 8.81@xmath7 with a median value of 2.67@xmath7 . therefore , based on both color errors and distribution analysis , a majority of these red population stars have significant color differences from the ms fiducial . as with all ground - based observations of gcs , the effects of crowding should be considered , even in our relatively uncrowded outer - annulus observations of the ms of ngc 1851 . based on the individual psfs from our images we have randomly placed 50 artificial ms stars ( 19.75@[email protected] ; 19.0@[email protected] ) across the full field with added poisson noise . this is a small enough number of stars to not further increase the crowding , and to increase our sample we repeated this 20 times in both c and t1 . our photometric methods have independently detected in c all 269 of the artificial stars that fell within our outer annulus from 4 to 6.5 arcminutes , and in t1 it detected all 241 of the artificial stars that fell within our outer annulus . for comparison , across the entire field our photometric methods have independently detected in c 99% of the input stars and in t1 99.4% of the input stars . one of the stars that was not detected fell directly on a bright field star , and all other artificial stars that were not detected fell in the core within 2.0 arcminutes of the center where it is more challenging to resolve faint ms stars . direct comparisons of the input and output magnitudes for our artificial stars provides a valuable tool to analyze our error characteristics . it should be noted that this is based only on single measurements of individual images ; therefore , the magnitude of these errors will not be representative of our final errors based on multiple combined measurements . in our outer annulus we find in c the residual distribution has a @xmath7 of 0.025 , and 5.2% of our sample have an observed difference greater than 0.1 ( @xmath14 ) from the input . similarly , in t1 we find the residual distribution has a @xmath7 of 0.018 , and 3.7% of our sample have an observed difference greater than 0.072 ( @xmath14 ) from the input . the larger effects seen in c are likely due to both the ms stars being fainter in c and the larger seeing in these observations , resulting in an increased probability of being affected by crowding . it should be noted that nearly all ( 87% ) of these magnitude outliers are found to be brighter than their input magnitudes , consistent with them being very near another star that adds light to its measurement . we take these low percentages to be our probability of being affected by crowding for the full sample of ms stars in this outer annulus . however , these magnitude outliers also exhibit moderately larger photometric errors and will have larger magnitude dispersions across multiple images , leading them to have been likely cut based on errors in our ms analysis here . furthermore , if these crowded stars are not cut by our error requirements , the increased effect of crowding on c over t1 will lead crowded stars to be preferentially bluer , in stark contrast to our clearly redder c - t1 population . consistent with this , the cut applied to our initial sample ( see section 2 ) of stars within 2 arcminutes of the center , where crowding is far more serious , shows that the cut stars with large errors were predominantly bluer than the primary rgb and ms . a further test for the potential effects of crowding is performed by directly marking the two populations in an image of the cluster . we see that in this outer annulus the red ms population are not more likely to have near neighbors that could cause problems and they are primarily isolated stars . reassuringly , there also is no clear difference in their spatial distributions azimuthally .

the subject of optomechanics involves interactions between optical and mechanical degrees of freedom , and is currently of great interest as an enabler of fundamental investigations in quantum mechanics , as well as a platform for ultrasensitive measurement devices . the majority of optomechanical configurations rely on the exchange of linear momentum between light and matter . we will begin this tutorial with a brief description of such systems . subsequently , we will introduce optomechanical systems based on _ angular _ momentum exchange . in this context , optical fields carrying polarization and orbital angular momentum will be considered , while for the mechanics , torsional and free rotational motion will be of relevance . our overall aims will be to supply basic analyses of some of the existing theoretical proposals , to provide functional descriptions of some of the experiments conducted thus far , and to consider some directions for future research . we hope this tutorial will be useful to both theorists and experimentalists interested in the subject .

1512.08989

optomechanical systems , which are based on the interaction between optical and mechanical degrees of freedom , are currently under intense research focus @xcite . much of the interest in these systems is due to their ability to access quantum mechanics at more macroscopic scales than ever before . for example , a major accomplishment of the field has been the cooling of a microfabricated harmonic oscillator , consisting of roughly @xmath0 atoms , to its quantum mechanical ground state @xcite . optomechanical systems therefore provide an avenue for investigating fundamental quantum mechanical issues such as wave function collapse @xcite , superposition @xcite , decoherence @xcite , entanglement @xcite , backaction @xcite and planck - scale physics @xcite . furthermore , these optomechanical systems hold great potential as platforms for ultrasensitive measurement technologies . indeed , optomechanical principles underlie cutting - edge displacement @xcite and force @xcite sensors , phonon detectors @xcite , gravitational wave interferometers @xcite , thermometers @xcite , accelerometers @xcite , and magnetometers @xcite , to cite a few examples . while a variety of physical configurations are used in the investigations mentioned above , in almost all of them , the optical and mechanical degrees of freedom interact via the exchange of _ linear _ momentum . however , light can also exchange _ angular _ momentum with matter @xcite . since the pioneering experiment of beth , it has been known that light can carry spin angular momentum ( i.e. polarization ) , which in turn can exert torque on bulk matter @xcite . subsequently , a seminal paper by allen et al . in 1992 pointed out that photons could also carry orbital angular momentum ( oam ) , encoded as spatial structure in the transverse profile of the corresponding optical beam @xcite . unlike spin , which can only take values @xmath1 , photonic oam can take any value @xmath2 of importance to this tutorial is the fact that oam - carrying beams can exert torque on matter as well , with experiments demonstrating optical rotation effects on nanoparticles @xcite , liquid crystals @xcite , and bose - einstein condensates @xcite , for example . also of relevance is the fact that , due to their intrinsic spatial inhomogeneities , beams with oam can sense the circulation of material particles , via the rotational doppler shift @xcite . this has been verified in velocimetry experiments @xcite ( which have also been implemented with circularly polarized beams @xcite ) . further , it is relevant to mention that the practically infinite hilbert space corresponding to the virtually unlimited oam available per photon has proved to be an important resource for quantum information processing @xcite . lastly , oam carried by an optical beam is associated with the presence of singularities , i.e. vortices , in the beam profile . these structures have their own interesting dynamics and place oam - carrying beams into the larger arena of ` singular optics ' @xcite . there are many other applications of oam - carrying beams ; we have mainly mentioned those which are relevant to our discussion . this tutorial aims to introduce optomechanical systems which are based on angular momentum exchange between light and matter . for the optical fields , this will imply the consideration of photons carrying either polarization or oam . for the mechanics , torsional and free rotational motion will be of importance . to our knowledge , existing introductions to , and reviews of , the micromanipulation of matter with angular momentum - carrying beams largely deal with liquid media @xcite . in contrast , the present tutorial will consider instead only systems under vacuum . with few exceptions , we will consider beams carrying well - defined optical angular momentum per photon , and will neglect the effects of linear photonic momentum on mechanical torsional or free rotational motion @xcite . it is worth stating explicitly the interest behind considering the two types of mechanical motion mentioned above . torsional oscillators are often the instruments of choice for precision measurement experiments . for example , they were used in historic measurements of the static electric force by coulomb @xcite , of the gravitational force by cavendish @xcite , and of optical polarization by beth @xcite . furthermore , a key difference between linear and torsional oscillations is that the linear vibrational motion of an object is affected by its total mass , while the torsional oscillations are affected not only by the total mass but also by how that mass is distributed , i.e. by the moment of inertia . for example , the mass of a sphere of radius @xmath3 varies as @xmath4 , but the moment of inertia goes as @xmath5 thus , smaller objects are easier to move linearly , but even easier to move torsionally . these distinctions make it worthwhile to study torsional oscillations on their own . on the other hand , free mechanical rotation presents some entirely novel features , compared to linear or torsional oscillatory motion , especially in the quantum regime . for example , it is well known that since the harmonic oscillator is a linear system ( i.e. its equations of motion are linear in the dynamical variables , or equivalently , the energy eigenvalues are evenly spaced ) , quantum effects are quite difficult to observe . in other words , the quantum oscillator behaves , for the most part , like a classical oscillator . to observe quantum effects , one usually needs to introduce some sort of nonlinear interaction , for example , by coupling the oscillator to the mode of an electromagnetic field , or to a two - level system @xcite . the free rotor , in contrast , is a nonlinear system ( i.e. its equations of motion are nonlinear in the dynamical variables , or equivalently , its energy spectrum is anharmonic , see below ) , which suggests that it might be easier to observe quantum effects in its behavior without the use of any auxiliary systems . one such quantum effect , which would be interesting to verify at much larger than atomic mass scales , is the quantization of angular momentum , one of the central predictions of quantum mechanics @xcite . a second intriguing feature of the rotor is that , unlike the harmonic oscillator ( and other elementary quantum systems such as the square well , the delta function well , or the morse potential ) , it has no ground state energy @xcite . this can be seen from the hamiltonian of a particle of mass @xmath6 rotating in a circle of radius @xmath3 , @xmath7 where @xmath8 is the angular momentum of the particle , and @xmath9 is its moment of inertia . the eigenvalues of @xmath10 are @xmath11 thus , the ground state @xmath12 is perfectly rotationless . finally , the prospect of observing counter - rotating superpositions of mechanically rotating states @xcite is quite tantalizing , and is the neutral mechanical analog of counter - propagating persistent electronic currents in superconducting circuits @xcite . from a more practical point of view , freely rotating mechanical systems could possibly be used for gyroscopy @xcite and for storing photonic oam @xcite . conversely , optical beams carrying oam could be used to probe mechanical rotation , in a quantum - limited manner @xcite . the remainder of the tutorial is organized in the following way . we first introduce optomechanics based on electromagnetic modes confined to optical resonators . section [ sec : vco ] describes cavity - based vibrational optomechanics , and lays the ground for the formalism , basic physics and applications to follow later . section [ sec : oam ] then introduces oam - carrying beams . the material presented thus far is then utilized to explain cavity - based torsional and rotational optomechanics in sections [ sec : torsionalopto ] and [ sec : rco ] respectively . accounting for the more recent interest in cavityless optomechanical systems , we have included some material on this topic in section [ sec : cavless ] . the tutorial ends with a conclusion in section [ sec : con ] and acknowledgments in section [ sec : ack ] . the format for each section is to first introduce a basic physical configuration . this is done typically , but not in all cases , using work published by our own group , simply because we understand our own work best . subsequently , reference is made to other works in the field . although we have tried to be inclusive in making such references , we apologize to all authors whose works are not adequately cited . in this section , we will provide a brief review of standard linear vibrational cavity optomechanics @xcite ; those who are familiar with the material may wish to proceed directly to section [ sec : oam ] . a number of comprehensive reviews of the topic already exist @xcite , so we will not be aiming to achieve detail or rigor in our presentation . rather , our discussion will be largely heuristic , focusing on the physics rather than the mathematical details , and aimed mainly at making the ensuing discussion of rotational optomechanics accessible . although we will be making reference to both classical and quantum regimes for the relevant optomechanical systems , we will largely employ the formalism of second quantization , which makes the notation compact and also reveals the microscopic physics of the system in terms of its fundamental quanta . we begin below by reviewing this formalism for massive particles as well as electromagnetic fields . consider a simple harmonic oscillator of mass @xmath6 oscillating at frequency @xmath13 , in one dimension . the corresponding classical mechanical hamiltonian is given by @xcite @xmath14 where @xmath15 and @xmath16 are the oscillator position and momentum , respectively . if we now consider the oscillator to be a quantum object , the two dynamical variables must obey the commutation relation @xmath17=i\hbar,\ ] ] where @xmath18 is planck s constant divided by @xmath19 . the dynamical evolution of any quantum mechanical operator @xmath20 can be found using the heisenberg equation @xcite @xmath21+\frac{\partial o}{\partial t},\ ] ] where @xmath22 is time , the dot denotes a total time derivative , and the last term is nonzero only if the operator carries explicit time - dependence . using this equation the heisenberg equations for @xmath15 and @xmath16 follow from eq . ( [ eq : hamsho ] ) @xmath23 @xmath24 in real systems , and certainly in optomechanical setups , it is important to account for damping and fluctuations experienced by the oscillator as a result of its coupling to the environment . the procedure for incorporating such effects in a way consistent with quantum mechanics is somewhat complicated and we will not describe it in detail . fortunately this procedure results in a straightforward prescription called the quantum langevin approach @xcite , which preserves eq . ( [ eq : shoqp1 ] ) but modifies eq . ( [ eq : shoqp2 ] ) , as @xmath25 where @xmath26 is the rate of mechanical damping and @xmath27 is a brownian stochastic force with zero mean which obeys the correlation function @xcite @xmath28,\ ] ] where @xmath29 is boltzmann s constant and @xmath30 is the ambient temperature . typically we will deal with high temperatures such that @xmath31 . in this limit , the correlation simplifies to @xmath32 working with the quantum mechanical variables @xmath15 and @xmath16 is convenient since they allow us to take the classical limit of the theory when required . however , sometimes it is more convenient to use instead a set of variables which describe the system in terms of its basic quanta . introducing the transformations @xcite @xmath33 where @xmath34 the hamiltonian of eq . ( [ eq : hamsho ] ) can be rewritten as @xmath35 where we have dropped a constant offset @xmath36 which corresponds to the ground state energy of the oscillator , and has no dynamical effect . in this formalism , the operator @xmath37 physically corresponds to the destruction ( creation ) of one quantum of the oscillator energy , i.e. one phonon . the combination @xmath38 is called the number operator , as its value denotes the number of quanta carried by the oscillator , with the energy per quantum being @xmath39 as per eq . ( [ eq : hamsqsho ] ) . from eqs . ( [ eq : commcan1 ] ) and ( [ eq : qptoaad ] ) , it follows that @xmath40=1.\ ] ] we now consider the quantization of a mode of the electromagnetic field , following the approach of ref . @xcite . by a mode we mean an electromagnetic wave of well - defined frequency @xmath41 . in order to isolate a single optical mode , it is convenient to use two mirrors facing each other , thus realizing a fabry - perot cavity , as shown in fig.[fig : p1 ] . the classical electromagnetic hamiltonian , which combines the electric and magnetic contributions , is given by @xcite . a laser injects light of wavelength @xmath42 into a fabry - perot cavity made from two highly reflecting mirrors . the cavity length is @xmath43.,scaledwidth=40.0% ] @xmath44,\ ] ] where @xmath45 and @xmath46 are the electric and magnetic fields , respectively , and the integration is over the cavity volume . as a useful idealization , the two mirrors , which are a length @xmath43 apart , are taken to be perfectly reflecting . this assumption will be relaxed later . the quantized electric and magnetic fields which satisfy the appropriate boundary conditions are @xmath47 @xmath48 where @xmath49 @xmath50 are sometimes referred to as the electric and magnetic fields per photon " . in eqs . ( [ eq : el ] ) and ( [ eq : ma ] ) , the wavenumber @xmath51 where @xmath52 is the speed of light , and the optical field frequency is given by @xmath53 which follows from the physical requirement that the length of the cavity has to be an integer number of half wavelengths to support the electromagnetic mode . using the fields of eqs . ( [ eq : el ] ) and eqs . ( [ eq : ma ] ) in eq . ( [ eq : hamop ] ) yields the harmonic oscillator hamiltonian @xmath54 the variables @xmath55 and @xmath56 correspond to destruction and creation operators for the photons confined to the cavity , and @xmath57=1.\ ] ] realistically , of course , the fabry - perot mirrors are not perfectly reflecting . in fact , some degree of transmission is required so photons can be injected into the cavity , typically by a laser . to account for this driving , the hamiltonian of eq . ( [ eq : hopt ] ) is modified to @xcite @xmath58 where @xmath59 is the frequency of the driving laser , @xmath60 is related to the input laser power @xmath61 by @xmath62 and @xmath63 is the rate at which photons leak from the cavity . the dynamical evolution of @xmath55 can be obtained from eq . ( [ eq : he ] ) ( we presently ignore the effects of @xmath63 on this equation , but see eq . ( [ eq : aeqn ] ) below ) @xmath64 which can be transformed using the substitution ( which corresponds to entering the frame rotating at the frequency of the input laser ) @xmath65 to the equation @xmath66 where @xmath67 is the detuning of the laser from the cavity . the presence of the cavity mirror transmissivity also introduces dissipation and noise into the cavity . the effect of these disturbances on the optical mode can be accommodated using the quantum langevin approach and changes the time evolution of eq . ( [ eq : har ] ) to @xcite @xmath68 where @xmath69 is the electromagnetic noise that comes in through the transmissive mirror , and has zero mean and the correlator @xcite @xmath70 at optical frequencies . the presence of the cavity decay @xmath63 introduces a nonzero width to the frequency @xmath41 of the electromagnetic field . with this inclusion of photon loss from the cavity , the optical field is referred to as a _ note that according to eq . ( [ eq : aeqn ] ) the field variables @xmath55 and @xmath56 each decay at the rate @xmath71 , while the oscillator energy decays at @xmath63 . in the classical limit , the operator @xmath55 can be replaced by the complex - valued dynamical variable @xmath72 , which corresponds to the amplitude of the classical electromagnetic field @xcite . in this limit , the number of photons in the cavity is @xmath73 we now consider a fabry - perot cavity where one mirror is fixed but the other is allowed to oscillate harmonically along the cavity axis . this can be accomplished by suspending it like a pendulum , as in fig . [ fig : p2](a ) , , but with one of the mirrors mounted as a harmonically oscillating pendulum with amplitude @xmath74 measured from its equilibrium position . ( b ) reversal of the photon momentum upon reflection from a perfectly reflecting mirror.,scaledwidth=60.0%,scaledwidth=20.0% ] or by mounting it on a spring , for instance . we will derive the hamiltonian for this system using heuristic semiclassical arguments @xcite ; a rigorous quantum mechanical description can be found in @xcite . if the mechanical oscillation frequency @xmath13 is much smaller than the spacing between two neighboring modes of the cavity , then we can consider a single optical mode , since photon scattering into neighboring modes is then negligible . let us assume first that the two mirrors are both perfectly reflecting , and consider the reflection of a single photon , as shown in fig . [ fig : p2](b ) . the photon momentum before reflection is @xmath75 and after reflection is @xmath76 . the net change of momentum of the photon is therefore @xmath77 the reflected photon returns after being reflected from the stationary mirror , after a time @xmath78 . the change of momentum per unit time , i.e. the force exerted by each photon is @xmath79 from which the optomechanical coupling constant may be written as @xmath80 or using eq . ( [ eq : resfreq ] ) as @xmath81 the total force due to all the photons present in the cavity is therefore @xmath82 . the work extracted by the optical field in moving the mechanical oscillator from @xmath43 to @xmath83 yields the energy of optomechanical interaction @xmath84 where all dynamical variables are quantum mechanical . the full optomechanical hamiltonian then combines eqs . ( [ eq : hopt ] ) , ( [ eq : hdrive ] ) , and ( [ eq : hint ] ) @xmath85 in order to remove the explicit time - dependence , a transformation @xmath86 into the frame rotating with the driving laser may be made , yielding , @xmath87 or explicitly , @xmath88 sometimes it is convenient to use the variables of eq . ( [ eq : qptoaad ] ) , @xmath89 where @xmath90 accounting for noise and dissipation , the quantum langevin equations for the optomechanical cavity are @xmath91 @xmath92 @xmath93 while the optomechanical cavity described above is associated with several physical phenomena , we will briefly discuss only two below : displacement sensing and cooling of mechanical motion . this abbreviated discussion will be sufficient for motivating the introduction of rotational optomechanics below . for readers interested in other effects , such as bistability , squeezing , etc . , we recommend several available reviews @xcite . displacement sensing using optomechanical platforms is an application that relies on the fact that a cavity photon picks up a phase that depends on the distance it travels , and hence on the displacement of the mechanical oscillator in fig . [ fig : p2 ] . mathematically , we can see from eq . ( [ eq : qleopto3 ] ) that the time evolution of the optical mode @xmath55 depends on the oscillator displacement @xmath15 . to investigate further , let us assume the so - called ` bad cavity ' limit , where @xmath63 is large and the photons leave the cavity so quickly that the optical mode can rapidly adjust to instantaneous position of the oscillator . in this regime , we can adiabatically eliminate the cavity mode , by setting @xmath94 in eq . ( [ eq : qleopto3 ] ) to zero . this procedure yields @xmath95 which shows the dependence of @xmath55 on @xmath15 . rewriting in polar form , and assuming @xmath96 for simplicity , @xmath97^{1/2 } } e^{i\tan^{-1}\left(-\frac{g'q}{\gamma_{0}/2}\right)},\ ] ] which shows that the oscillator position information is contained in both the amplitude as well as the phase of the cavity mode . in particular , the phase becomes more sensitive to @xmath15 if @xmath98 is large , i.e. for stronger optomechanical coupling and low - loss cavities ( although we are in the high loss limit ) . a more detailed analysis shows that in the ` good cavity ' limit @xmath99 small ) , the displacement sensitivity of the device improves with the cavity finesse @xcite . this principle is the basis of interferometers such as ligo which seek to transduce weak gravitational waves into detectable mechanical displacements @xcite . optical cooling of the mechanical motion is another application of the optomechanical cavity @xcite . neglecting dissipation , noise and driving for simplicity , eq . ( [ eq : homp ] ) reads @xmath100 the interaction term proportional to @xmath101 corresponds to cooling of the mechanical motion , since it describes a process in which a photon @xmath102 and a phonon @xmath103 are annihilated while a second photon @xmath104 is created . since energy has to be conserved during this process , the second photon must have higher energy than the first , by exactly @xmath39 , the phonon energy . this is allowed because the photons belong to a quasi - mode with nonzero linewidth , and need not both have exactly the same frequency . the key point is that the annihilation of a phonon cools the mechanical motion of the oscillator . using similar arguments , it can be shown that the interaction term proportional to @xmath105 corresponds to oscillator heating . a detailed treatment indicates that by tuning the driving laser to below the cavity resonance @xmath106 allows the cooling to dominate over the heating ; such a scheme has been used for laser cooling a mesoscopic mechanical oscillator to its quantum ground state @xcite . in this section , we will provide a brief introduction to optical beams carrying angular momentum @xcite . we will assume the reader is familiar with the fact that a photon can have spin angular momentum @xmath107 , corresponding classically to circularly polarized radiation . this is one way in which optical beams can carry angular momentum . another way is by encoding the appropriate structure in the beam phase @xcite . each photon in the beam can then be thought of as carrying an oam which can take unlimited values in principle , and quite high values in practice . although a clear separation of spin and orbital angular momentum degrees of freedom is generally a subtle matter @xcite , in this article we will operate under conditions where the requirements for effecting this separation are fulfilled . specifically , we will work in the paraxial regime of beam propagation , where the separation is valid . below we will introduce oam - carrying beams by discussing a specific method for their generation . based on this discussion we will consider some general properties of such beams , which come in several families @xcite . subsequently , we will consider a particular family , namely the laguerre - gaussian modes , for a more detailed exposition . there are several ways of generating beams carrying oam , such as using lasers , lenses , holograms , spatial light modulators and spiral phase plates @xcite . we will consider the relevant case of a spiral phase plate , shown in fig . [ fig : p3](b ) . ( b ) transmission of the wavefronts of a plane wave through a spiral phase plate of step height @xmath22 . the transmitted wavefronts in this case are helicoidal . adapted with permission from @xcite.,scaledwidth=80.0%,scaledwidth=40.0% ] to understand what such a phase plate does to an optical beam , we first consider the simpler case of the same beam passing through a uniform slab of thickness @xmath43 , as shown in fig . [ fig : p3](a ) . if the beam incident on the slab is modeled as a plane wave of amplitude @xmath108 , it will pick up a phase on transmission , leading to an output field @xmath109 where @xmath110 is the amplitude of the transmitted radiation , @xmath111 is the wavevector of the optical beam , and @xmath75 is the linear momentum per photon . the optical phase acquired by the beam therefore depends on the thickness @xmath43 of the slab . when the same beam passes through the spiral phase plate , which has a uniform azimuthal gradient in its thickness , it picks up a phase which depends on the angle @xmath112 about the light propagation direction , @xmath113 where @xmath114 @xmath22 is the step height of the phase plate , and @xmath42 the optical wavelength . analogous to the linear phase factor discussed above , where @xmath75 is the linear momentum , @xmath115 can be identified as the orbital angular momentum carried by each photon in the transmitted beam . this identification can be verified by rigorously calculating the angular momentum carried by the electromagnetic fields of the beam @xcite . by arranging for the plate thickness to be an integer multiple of the optical wavelength , we will restrict ourselves to integer values of @xmath116 in eq . ( [ eq : lint ] ) . if @xmath116 is not an integer , the transmitted beam consists of an infinite superposition of integer-@xmath117 waves , whose average oam is then the non - integer value given by eq . ( [ eq : lint ] ) . although we have considered transmission through the spiral phase plate in our example above , reflected photons pick up orbital angular momentum as well . in both cases the angular momentum is supplied by the spiral phase plate . usually the plate is clamped to an optical table and serves as a virtually infinite source or sink of angular momentum . lastly , we point out that the azimuthal phase factor @xmath118 in the field transmitted by the spiral phase plate is well - defined everywhere except on the @xmath119 axis , as @xmath112 is multiply defined on that line . in a consistent physical picture of the field , the amplitude vanishes on the @xmath119-axis , and the beam resultantly carries a singularity , namely a vortex . the intensity profile of the beam therefore presents a node at the core of the vortex , i.e. on the @xmath119 axis , see fig . [ fig : p4 ] below . the oam @xmath116 , which quantifies the beam vorticity , is sometimes referred to as the ` topological charge ' of the optical beam . we have presented an idealized description thus far , using a plane wave , which has infinite transverse extent . in practice , the beam has a transverse profile which decreases in intensity away from the propagation axis . we now move on to the discussion of such modes . an appropriate expression for an oam - carrying field is @xmath120 which accounts for the transverse coordinate dependence of the field . fields of this type form solutions to the paraxial wave equation in free space and are therefore bonafide modes of the electromagnetic field in that approximation @xcite . there are several families of beams that carry oam , corresponding to different forms for @xmath121 , including bessel , mathieu , and hypergeometric - gaussian modes @xcite . we consider the laguerre - gaussian fields for which @xmath122 and the normalized mode function is given by @xcite @xmath123^{|l| } e^{-\frac{\rho^{2}}{w^{2}(z ) } } l_{p}^{|l|}\left[\frac{2\rho^{2}}{w^{2}(z)}\right ] e^{\frac{-ik\rho^{2}z}{2\left(z_{r}^{2}+z^{2}\right)}}e^{il\phi } e^{i\left(2p+|l|+1 \right)\tan^{-1}\left(\frac{z}{z_{r}}\right ) } , \label{eq1}\ ] ] where the normalization @xmath124 the index @xmath116 corresponds to the number of azimuthal nodes , the index @xmath16 determines the number of radial nodes @xcite , @xmath125 is the beam waist , @xmath126 is the asssociated laguerre polynomial , and @xmath127 is the rayleigh range , @xmath42 being the optical wavelength . in eq . ( [ eq : lgmode ] ) , @xmath116 can be either a positive or negative integer , while @xmath16 can only assume positive integer values . the intensity profile of a laguerre - gaussian mode with @xmath128 has a node at the center , and appears shaped as an annulus , see fig . [ fig : p4](a ) for @xmath129 . the interference pattern between two lg@xmath130 beams consists of @xmath131 lobes arranged in a circle , see fig . [ fig : p4](c ) for the case where @xmath132 laguerre - gaussian beam ( b ) intensity pattern for an @xmath133 laguerre - gaussian beam ( c ) intensity pattern for a superposition of @xmath134 and @xmath135 laguerre - gaussian beams.,scaledwidth=50.0% ] with this brief introduction to oam - carrying beams , we are now ready to combine mechanical motion with photonic angular momentum . in this section , we will consider some optomechanical platforms that couple photonic angular momentum to mechanical torsional motion . we will first introduce a straightforward analog of the linear optomechanical cavity presented in section [ subsec : omcav ] , and then consider a variation on the same theme . we begin with a torsional analog to the standard linear vibrational optomechanical cavity of section [ subsec : omcav ] @xcite , as shown in fig . [ fig : p5 ] . in this analogy , the spherical mirrors used in a fabry - perot are replaced by spiral phase plates described in section [ sec : pwwat ] . one of the plates is fixed in place , while the other oscillates harmonically about the cavity axis , thus executing torsional motion . the angular displacement of the moving plate is measured by the angle @xmath112 . a beam with oam @xmath115 is incident on the cavity . the spiral phase plates reverse the linear momentum as well as the oam of the cavity photons upon reflection . this reversal of oam can be arranged by selecting the step heights and handedness on the spiral plates so that one of them adds , and the other subtracts , the same oam from the cavity photons , as shown in fig . [ fig : p5 ] . . adapted with permission from @xcite.,scaledwidth=60.0%,scaledwidth=40.0% ] the optomechanical coupling can be derived analogously to section [ subsec : omcav ] if we assume that the two spiral phase plates are perfectly reflecting . in this case , the moving plate reverses the photonic oam from @xmath115 to @xmath136 upon reflection , resulting in a momentum change of @xmath137 per photon . this happens once every cavity round trip time @xmath78 , if the cavity length is @xmath43 . the change of angular momentum per unit time , i.e. the torque exerted by each photon is given by @xmath138 which may be compared to eq . ( [ eq : omcoupling ] ) and from which the _ optotorsional _ coupling constant , may be obtained as @xmath139 which is analogous to eq . ( [ eq : gomvib ] ) . the total torque due to all the photons present in the cavity is @xmath140 . the work extracted from the torsional oscillator in deflecting it by an angle @xmath112 yields the energy of optomechanical interaction @xmath141 the entire optomechanical hamiltonian may now be assembled in analogy to eq . ( [ eq : homdriveframe ] ) @xmath142 where @xmath8 is the angular momentum of the oscillating spiral phase plate about the cavity axis , @xmath143 is its moment of inertia , and @xmath144 is the frequency of torsional oscillation . the first two terms describe the torsional energy of oscillation , the third term the optical mode energy , the fourth term the optomechanical interaction , and the fifth term the effect of external driving on the cavity . if the moving spiral phase plate has mass @xmath145 and radius @xmath146 , we may treat it as a disk and evaluate the moment of inertia @xmath147 about an axis passing through its center . in eq . ( [ eq : homt ] ) , all dynamical variables are quantum mechanical . specifically , the torsional variables obey the commutation relation @xmath148=-i\hbar.\ ] ] since eq . ( [ eq : homt ] ) has the same form as eq . ( [ eq : homdriveframe ] ) , we may expect to implement the same optomechanical physics on this new platform , i.e. detection of torsional displacements as well as cooling of torsional motion should be possible . interestingly , the optotorsional coupling in ( [ eq : gomtor ] ) depends on @xmath116 , which may be adjusted by the experimentalist . in recent experiments , upto @xmath149 has been used @xcite . unlike the standard fabry - perot , which has been investigated for more than a century , the spiral phase plate resonator appeared in the theoretical literature only once ( without the mechanical degree of freedom ) @xcite before the optomechanical version was proposed . therefore , much fundamental information about the spiral resonator remains to be discovered . as a first step in this direction , our group has recently analyzed the ray optics of this resonator @xcite ; related work has also recently been carried out for low finesse spiral etalons @xcite . no experimental realization of the spiral phase resonator seems to exist , possibly due to the low reflectivity @xmath150 and high diffractive losses presented by currently available spiral phase plates . these limitations preclude the implementation of efficient displacement sensing or torsional cooling , both of which require high - finesse cavities . interestingly , a design equivalent to a spiral phase resonator ( but without any mechanical degrees of freedom ) has been implemented in a situation where the mirror losses are offset by the presence of a gain medium - i.e. in a laser @xcite . in this section we will consider the proposal of romero - isart et al . @xcite for coupling a rod to oam - carrying cavity modes , and its subsequent generalization to nano - windmills @xcite . these proposals rely on the ability of light to optically trap dielectric particles at the intensity maxima of an optical beam . the trapping may be understood intuitively if the dielectric particle is smaller than an optical wavelength and thus may be assumed to behave as a point dipole . if the dipole moment is @xmath151 , the energy of interaction with an optical beam of electric field @xmath152 is @xmath153 if the particle is a neutral but linearly polarizable dielectric , with polarizability @xmath154 , then the dipole moment is an induced one , and is given by @xmath155 combining eqs . ( [ eq : hinttrap ] ) and ( [ eq : dipole ] ) , we obtain @xmath156 now since @xmath157 where @xmath158 is the light intensity , clearly @xmath159 is a minimum when the particle is located at an intensity maximum of the beam . at such maxima , in fact , the force from the optical beam can levitate the particle in space against the effect of gravity . in practice , it is quite common to use the focus of a gaussian beam as an optical trap for the particle @xcite . to lowest order in the particle displacement , such a trap provides harmonic confinement in all three directions in space . in their proposal , romero - isart suggest trapping a nanorod with multiple laguerre - gaussian cavity modes , in such a manner that the rod executes torsional oscillations about the cavity axis , see fig . [ fig : p6 ] . . ( a ) shows the trapping configuration along the cavity axis ( b ) shows the angular optical lattice in which the rod is trapped and executes torsional motion about the cavity axis . adapted with permission from @xcite.,scaledwidth=50.0%,scaledwidth=15.0% ] the trapping field is considered to be an experimentally changeable parameter and does not possess dynamics of its own . an additional laguerre - gaussian mode with @xmath160 is introduced to detect and cool the torsional oscillations . the corresponding hamiltonian is essentially of the same form as in eq . ( [ eq : homt ] ) , and therefore we do not repeat its full derivation here . we only note that the optomechanical coupling in the case of the rod was calculated using the bethe - schwinger perturbation theory formula @xcite @xmath161 which accounts for the variation of the cavity resonance @xmath162 due to the change in the angular position of the rod @xmath112 away from its equilibrium value @xmath163 and where @xmath164 and @xmath165 are the dielectric constant and volume , respectively , of the windmill , and @xmath166 is the volume of the optical cavity . the optomechanical coupling can then be found as @xmath167 from eq . ( [ eq : freqmode ] ) . the bethe - schwinger formula is applicable when the dimensions of the rod are small enough that its presence in the cavity shifts the optical resonance only by a small amount . interestingly , the authors point out that a suitable nanorod could be the tobacco mosaic virus , thus opening up a way to observe quantum effects in living organisms . the work of romero - isart et al . , has been generalized further by considering , instead of a single rod , a windmill - shaped dielectric , which may be thought of as a collection of rods , see fig . [ fig : p7](a ) @xcite . the advantage of a windmill shape is that it provides larger overlap with the cavity oam modes than a single rod , and hence higher optomechanical coupling . we note that micromachined windmills are available in the laboratory , and have been held and rotated in optical tweezers using radiation pressure from optical beams @xcite , see fig . [ fig : p7](b ) . while romero - isart considered only @xmath160 beams [ see eq . ( [ eq : lgmode ] ) ] , it was later found that using modes with nonzero @xmath16 yields better optomechanical coupling , since such modes match the shape of the windmill better @xcite . a recent experimental development along these lines was the demonstration of the transit of rotating nanorods through an optical cavity excited by a gaussian optical beam @xcite . more sophisticated theoretical proposals also exist , incorporating multiple nanodumbbells into an optical cavity and predicting optically induced ordering and nematic transitions analogous to those observed in liquid crystals @xcite . torsional optomechanics realized by coupling wave guide deformations with circularly polarized photons has also been proposed @xcite . thus far we have considered the coupling of optical angular momentum to mechanical torsional motion . we now proceed to examine the coupling of optical angular momentum with free mechanical rotation . in this section we introduce the coupling of optical angular momentum with unhindered mechanical rotation . first we consider a particle rotating in the evanescent field of an optical resonator , and then a configuration in which the particle experiences the intracavity field . in this section , we consider a subwavelength dielectric particle captured in orbital motion around a spherical whispering gallery mode resonator , as analyzed by rubin et al . @xcite , and shown in fig . [ fig : p8 ] ( a ) . . the large circle denotes a microsphere , the bright band along the equator is a whispering gallery mode which confines the optical field , and the small circle is a nanoparticle trapped in orbit around the sphere ( b ) surface field intensity at the microsphere equator as a function of time . reproduced with permission from @xcite.,scaledwidth=70.0% ] to our knowledge , this was the first configuration in which fully rotational cavity optomechanics was studied @xcite . the resonator itself is a dielectric microsphere . the optical mode is confined by total internal reflection near the surface of the resonator ; the authors specifically consider whispering gallery modes that are localized near the resonator equator , as in fig . [ fig : p8 ] . some of the light escapes from the spherical resonator in the form of an evanescent field . it is this light that is responsible for trapping the subwavelength dielectric particle and driving its motion . the optomechanical coupling in this case is somewhat complicated and will not be derived here . the complications arise because the particle stays quite close to the dielectric sphere and therefore modifies the charge distribution on the resonator , effects which are negligible in the optomechanical systems described thus far . the resulting particle dynamics are also quite involved . the radial , polar and azimuthal motions of the particle are driven by the radiation field , and are coupled to each other . in particular , the orbital rotation of the particle about the spherical resonator is driven by the optical torque , which turns out to be nonconservative . however , as in standard optomechanics , the particle motion acts back on the field and therefore can be deciphered by monitoring , for example , the field intensity at the resonator surface . figure [ fig : p8](b ) shows the surface field intensity along the resonator equator . the intensity variations along the horizontal axis represent radial particle oscillations , and from their non - uniform spacing the angular azimuthal acceleration of the particle can be found . similarly , the intensity variations along the vertical axis correspond to polar oscillations . while this configuration displays many interesting features , the role played by electromagnetic angular momentum is somewhat limited . assuming dipolar scattering of photons , the authors show that the nanoparticle couples only to three optical angular momenta , namely @xmath168 and @xmath169 . in this section we will analyze a rotating nanoparticle located _ inside _ an optical cavity , as shown in fig . [ fig : p9 ] @xcite . . a nanoparticle of mass @xmath6 rotates around an optical ring potential and is probed by an angular optical lattice inside a cavity . reproduced with permission from @xcite.,scaledwidth=30.0% ] unlike in the configuration of section [ subsec : nanorotout ] , the usual light - matter coupling applies . this is because the rotating particle is far away from the resonator boundaries . this configuration is the free rotational analog of the linear optomechanical system of section [ subsec : omcav ] , and we will examine it in some detail . our exposition below will be aimed at establishing the possibility of accurate rotation sensing , i.e. implementing velocimetry of the particle , using a cavity . we will restrict ourselves to a classical analysis . a quantum analysis of rotation is expected to be more subtle @xcite , the optical resonator in this case is a fabry - perot , with both mirrors fixed in place . the cavity is excited by a trap beam , which is a laguerre - gaussian mode with oam @xmath170 . this mode presents an intensity maximum on a ring [ see fig . [ fig : p9 ] and also fig . [ fig : p4](b ) ] of radius @xmath171 at the center of the cavity . a dielectric particle of diameter smaller than the optical wavelength is trapped on this ring via the optical gradient force , see eqs ( [ eq : hamdip ] ) . we consider this particle to be rotating on this ring at a uniform angular velocity @xmath172 set by the balance of an external torque and the damping produced by the background gas particles as realized experimentally @xcite . we assume conditions where the coupling to radial and polar motion are negligible , i.e. tight confinement on the ring , and do not consider these degrees of freedom any further . the trap mode is treated as a parametric quantity , without any associated dynamics . an additional optical field , a superposition of two degenerate counter - rotating laguerre - gaussian modes with oam @xmath173 and @xmath174 respectively , is introduced to probe the rotational motion of the nanoparticle . near the trapping ring , the modes set up an optical lattice with @xmath131 maxima , described by the function @xmath175 where @xmath176 is the probe wave - vector . figure [ fig : p4](c ) and fig . [ fig : p9 ] show the lattice for @xmath129 . since the particle dimensions are smaller than the cavity length @xmath43 and the optical wavelength @xmath42 , eq . ( [ eq : freqmode ] ) can be used to find the optorotational coupling @xmath177 , using eq . ( [ eq : probemode ] ) as the input . this yields @xmath178 with @xmath179 where @xmath180 is the relative dielectric constant of the nanoparticle and @xmath181 is a gamma function . the dynamics of the optical probe field can be described using the variable @xmath182 such that @xmath183 is the number of probe photons inside the cavity at any time @xmath22 , see the very end of section [ subsec : cavq ] . choosing the cavity axis as the @xmath119 direction , the angular momentum of the particle is @xmath8 . to describe the rotational coordinate of the particle we use the periodic variable @xmath184 where @xmath112 is the instantaneous angular displacement of the particle , see fig . [ fig : p9 ] . the coupled equations of motion in the frame of the laser driving the cavity with the @xmath185 fields can be written in analogy to eqs . ( [ eq : qleopto1])-([eq : qleopto3 ] ) @xmath186 -\frac{\gamma_{0}}{2}\right\}a+\sqrt{\gamma}a_{in},\\ \nonumber\end{aligned}\ ] ] where @xmath187 @xmath188 being the detuning between the @xmath189 driving laser at frequency @xmath190 and @xmath191 , @xmath63 is the cavity loss rate , @xmath192 is the dielectric particle s moment of inertia about the axis of rotation and @xmath193 , where @xmath194 is the input probe power . classical laser noise and the thermal contribution to the radiation mode have been assumed to be negligible . the rotor mechanical damping is @xmath26 , and @xmath195 is a langevin torque with zero mean and the two - time fluctuation correlation ( analogous to eq . [ eq : harmcorr2 ] ) @xmath196 signifying white noise and responsible for brownian motion . generally , we will assume that the externally applied torque is larger than the torque due to the optical lattice , i.e. @xmath197 comparing eq . ( [ eq : opticalfieldevolution ] ) to eq . ( [ eq : qleopto3 ] ) shows that the phase of the optical field is sensitive to the @xmath198 quadrature of the mechanical rotation , and that this sensitivity improves linearly with the cavity finesse , and may therefore be used to accurately detect the particle rotation . we elaborate on this procedure below . the steady - state of the system can be found readily by equating the time derivatives in eqs . ( [ eq : opticalfieldevolution])-([eq : angularmomentumevolution ] ) to zero , @xmath199 @xmath200 @xmath201^{1/2}}.\ ] ] we note from equation ( [ eq : steadystate ] ) that the torque due to the probe lattice makes no contribution , on average , to the rotation rate @xmath172 . we can examine the linear response of the system ( i.e. the response of the system to small changes away from the steady state ) by considering each variable as the sum of a mean value and a small fluctuation , e.g. @xmath202 for low probe powers , we find @xmath203 solving this set of equations given the initial conditions @xmath204 and @xmath205 , we find on resonance @xmath206 , and ignoring cavity losses @xmath207 , that the fourier transform of @xmath208 is @xmath209 where @xmath210 @xmath211 and @xmath212 the dirac delta at @xmath213 on the right hand side of eq . ( [ eq : sidebandspectrum1 ] ) corresponds to the carrier optical frequency , and the dirac deltas at @xmath214 to sidebands which are created as the particle rotating at the frequency @xmath172 encounters @xmath131 optical lattice sites . these sidebands fundamentally arise from the rotational doppler shift @xcite imprinted on the cavity photons by the mechanical motion . the rotation frequency peak at @xmath215 can thus be recovered by homodyning ( i.e. interfering with a piece which was split off from the probe beam before it enters the cavity ) the probe beam @xmath189 once it has exited the cavity @xcite . an investigation into the nonlinear regime of high probe power shows that the measurement accuracy eventually degrades as shown in fig . [ fig : p10 ] . as a function of probe lattice power @xmath216 in the setup of fig . ( [ fig : p9 ] ) . the detected frequency accurately mirrors the rate @xmath215 [ eq . ( [ eq : sidebandfrequency ] ) ] in the regime of linear response , i.e. for probe power less than @xmath217 , but is affected for higher powers , degrading the accuracy of the measurement . adapted with permission from @xcite.,scaledwidth=40.0% ] at very high powers , the nanoparticle is simply trapped in one of probe lattice wells . while the description above provides some initial exploration of the coupling between optical and particle oam , much remains to be accomplished in this area . particularly , the quantum regime needs to be explored . a situation where the quantum regime could be relevant is when the dielectric nanoparticle is replaced by a single atom . no experiment to date has been able to detect or manipulate the center - of - mass angular momentum of a single atom , see section [ subsec : rclo ] below . this is in contrast to other degrees of freedom , such as electronic and translational center - of - mass for which fine control has been established at the single atom level @xcite . such control is desirable for rotational degrees of freedom from a fundamental perspective , but also for applications such as quantum information processing @xcite . in this section we will consider the coupling of laguerre - gaussian optical cavity modes with acoustic waves on the surface of a highly reflective mirror , see fig . [ fig : p11 ] . . surface acoustic waves on the lower mirror couple to the optical mode supported by the cavity . reproduced with permission from @xcite.,scaledwidth=40.0% ] this mirror is one of two that make up a fabry - perot , of the kind described in section [ subsec : omcav ] . the acoustic excitations manifest as vibrations of the mirror surface along the cavity axis . importantly , the vibrational modes also carry azimuthal structure . to our knowledge , briant et al . were the first to experimentally demonstrate that standard optomechanical physics of eq . ( [ eq : homp ] ) could be realized by such vibrational modes , which do not involve any global motion of the mirror @xcite . they were able to detect the azimuthal profile of each mode by scanning a sharply focused gaussian beam , whose waist was much smaller than that of the surface acoustic mode , across the mirror surface and following the resulting phase shifts in the reflected cavity beam . the experimentally acquired mechanical mode vibrational frequencies and spatial mode profiles could be fit well by an analytical theory in the limit where the radius of curvature of the mirror was much larger than its thickness . in this limit , the elasticity equation , which determines the mechanical modes , took a form identical to the _ optical _ paraxial equation . this implied that each mechanical mode could be characterized as a laguerre - gaussian mode , as confirmed by the experimental data . in this case the mechanical phonons of the mirror surface acoustic field may be thought of as harmonic oscillator quanta which carry oam transverse to their direction of oscillation . they are therefore analogous to laguerre - gauss photons of the electromagnetic field which are also quanta of a harmonic oscillator , and also carry oam in the same manner . the work of briant et al . was extended in @xcite which subsequently explored the coupling of the mentioned mechanical laguerre - gaussian modes ( with indices @xmath117 and @xmath218 ) to optical cavity ( with indices @xmath116 and @xmath16 ) laguerre - gaussian modes . the interaction hamiltonian in this case is found by expanding the acoustic as well as optical fields in terms of laguerre - gaussian modes [ analogous to eqs . ( [ eq : el ] ) and ( [ eq : ma ] ) which expand the field in terms of sines and cosines ] , and yields @xmath219 which is exactly of the same form as eq . ( [ eq : hint ] ) , except for the coupling constant @xmath220 where @xmath221 is given in eq . ( [ eq : gp ] ) and the full expression for @xmath222 has not been reproduced here as it is quite lengthy . from the kronecker delta in eq . ( [ eq : gsaw ] ) we can deduce the oam selection rule @xmath223 for @xmath224 the factor of two in eq . ( [ eq : oamsel ] ) may be traced back to the fact that the optomechanical interaction is bilinear in the optical field , but only linear in the acoustic field . the selection rule of eq . ( [ eq : oamsel ] ) provides a path for targeting specific mechanical modes . in the original proposal this selectivity was exploited for storing photonic oam in mechanical vibrational modes @xcite . however , there is no analogous rigorous selection rule for @xmath16 and @xmath218 , only simply varying degrees of overlap between modes with various radial numbers , described by the factor @xmath222 . we note that for this optomechanical platform , angular momentum is carried and exchanged in a parametric way , and is not a dynamical variable with its own time evolution . in this section we briefly consider the interaction of mechanical degrees of freedom with electromagnetic modes which are not confined in a resonator , but propagate freely in space . such configurations constitute a rapidly growing class of optomechanical systems @xcite . below we will describe linear vibrational motion and then move on to torsional and then finally rotational motion . a typical configuration consists of a harmonically oscillating subwavelength dielectric particle trapped optically at the focus of a freely propagating gaussian optical beam , see fig . [ fig : p12 ] . . an optically levitated nanoparticle scattering light from the 1064 nm trapping laser . reproduced with permission from prof . a. n. vamivakas.,scaledwidth=40.0% ] when this configuration is placed inside a liquid medium , it is usually referred to as an ` optical tweezer ' @xcite . in the present case we will consider optical trapping in vacuum only . photons scattered by the trapped particle are used as a probe to detect its motion . at low background pressures , the particle motion is ballistic and simple harmonic to a very good approximation . the frequency of oscillation along each spatial direction is set by the intensity of the trapping beam . an example of an optomechanical effect in this system is cooling of the oscillator motion , using a technique different from that described in section [ subsec : omcav ] . cooling can be achieved by making the trapping beam intensity ( and therefore the mechanical trap frequency ) higher when the particle is moving away from the center of the trap , and lower when it is moving towards . using such a technique , cooling has been implemented from @xmath225k ( room temperature ) to @xmath226mk @xcite . the system can also be used as a thermometer @xcite , as a force sensor @xcite , and can carry other spin - like degrees of freedom @xcite . theoretical proposals exist for creating macroscopic superpositions in such systems @xcite . exchange of angular momentum between an optical field and a torsional mechanical oscillator in free space was first seen in a classic experiment by beth @xcite . by detecting the rotation of a quarter - wave plate suspended as a torsional oscillator due to the change in handedness of polarized light passing through it , beth was in fact able to experimentally show for the first time that light possesses angular momentum . the experiment has been repeated in modern times in the optical @xcite as well as microwave @xcite domains . in this section we describe several examples of the coupling of light carrying angular momentum to free mechanical rotation . charged flakes of graphene have been confined in an electrostatic trap under vacuum @xcite . the linear motion is to a good approximation harmonic along all three spatial directions . in addition , it was found that when the flake was irradiated with circularly polarized light , it spun at about @xmath227 mhz . the mechanism for angular momentum transfer was light absorption . optically trapped particles can exhibit interesting interplay between vibrational and rotational degrees of freedom , as demonstrated using an optomechanical gyroscope @xcite . the trapped nanoparticle in this case was slightly anisotropic in shape , and the linear degrees of freedom were harmonic , as in the examples above . as in the case of graphene , a circularly polarized light beam set the particle in rotation . an interesting effect was then demonstrated that involved both vibration and rotation , as shown in fig . [ fig : p14 ] . . ( a ) shows a trapped asymmetric particle executing simple harmonic motion with amplitude @xmath228 ( b ) shows the same particle spinning at a constant angular momentum @xmath8 about an axis transverse to the vibrational motion . the conservation of angular momentum results in a reduced amplitude of motion @xmath229 , which may be thought of as gyroscopic cooling of vibration @xcite.,scaledwidth=50.0% ] in the absence of rotation , the particle oscillated with a certain amplitude along each axis , as shown in fig . [ fig : p14 ] ( a ) . in the presence of rotation , however , the particle became resistant to changing the direction of its angular momentum , as for any spinning top . this led to a decrease in the oscillation amplitude of the vibrational particle motion , as shown in fig . [ fig : p14](b ) . this effect corresponds to ` gyroscopic cooling ' of vibrational motion , distinct from resonator - based ( section [ subsec : omcav ] ) and feedback ( section [ subsec : vclo ] ) cooling . in the experiment cooling of other rotational modes was also observed . for completeness we mention that there is a sizable literature associated with the transfer of optical angular momentum to the center - of - mass motion of hot @xcite , cold @xcite and degenerate @xcite atoms , molecules @xcite , and electrons @xcite . there are many phenomena of interest in this category including vortices , persistent currents , spin textures , and optical centrifuges . we will not describe these works in detail here , since they go back several years , if not decades . in this tutorial , we have presented some introductory examples of the rich optomechanics arising from the exchange of angular momentum between light and matter . we have attempted to relate the relevant theoretical tools to the earlier work on vibrational optomechanics , while trying to present a useful variety of experimental platforms on which the theory could be realized . we therefore hope our exposition will be of relevance to theorists as well as experimentalists . we are grateful to b. zwickl , m. vengalattore , s. agarwal and b. rodenburg for useful discussions . we would like to thank the research corporation for science advancement ( award no . 20966 ) and the national science foundation ( award no . 1454931 ) for support . chan j , mayer alegre t p , safavi - naeini a h , hill j t , krause a , grblacher s , aspelmeyer m and painter o 2011 laser cooling of a nanomechanical oscillator into its quantum ground state _ nature _ * 478 * 89 romero - isart o , pflanzer a c , blaser f , kaltenbaek r , kiesel n , aspelmeyer m , and cirac j i 2011 large quantum superpositions and interference of massive nanometer - sized objects _ phys . rev . lett . _ * 107 * 020405 lee k c , sprague m r , sussman b j , nunn j , langford n k , jin x m , champion t , michelberger p , reim k f , england d , jaksch d and walmsley i a 2011 entangling macroscopic diamonds at room temperature _ science _ * 334 * 1253 cohen j d , meenehan s m , maccabe g s , groeblacher s , safavi - naeini a h , marsili f , shaw m d , and painter o 2015 phonon counting and intensity interferometry of a nanomechanical resonator _ nature _ * 520 * 522 bliokh k y , ostrovskaya e a , alonso m a , rodrguez - herrera o g , lara d and dainty c , 2011 spin - to - orbital angular momentum conversion in focusing , scattering , and imaging systems _ opt . exp . _ * 19 * 26132 he h , friese m , heckenberg n r and rubinsztein - dunlop h 2001 direct observation of transfer of angular momentum to absorptive particles from a laser beam with a phase singularity _ phys . * 75 * 826 phillips d b , lee m p , speirits f c , barnett s m , simpson s h , lavery m p j , padgett m j and gibson g m , 2014 rotational doppler velocimetry to probe the angular velocity of spinning microparticles _ phys . a _ * 90 * , 011801(r ) lavery m p j , barnett s m , speirits f c and padgett m j , 2014 observation of the rotational doppler shift of a white - light , orbital - angular - momentumcarrying beam backscattered from a rotating body _ optica _ * 1 * , 2334 lanyon b p , barbieri m , almeida m p , jennewein t , ralph t c , resch k j , pryde g j , obrien j l , gilchrist a and white a g , 2009 simplifying quantum logic using higher - dimensional hilbert spaces _ nat . phys . _ * 7 * 134 mirhosseini m , magaa - loaiza o s , osullivan m n , rodenburg b , malik m , lavery m p j , padgett m j , gauthier d j and boyd r w , 2015 high - dimensional quantum cryptography with twisted light _ n. j. phys . _ * 17 * , 033033 tittonen i , breitenbach g , kalkbrenner t , muller t , conradt r , schiller s , steinsland e , blanc n and de rooij n f , 1999 interferometric measurements of the position of a macroscopic body : towards the observation of quantum limits _ phys . rev . a _ * 59 * , 1038 coulomb c a , 1788 premier mmoire sur llectricit et le magntisme . construction et usage dune balance lectrique , fonde sur la proprit quont les fils de mtal davoir une force de raction de torsion proportionnelle langle de torsion _ mmoires de lacadmie royale des sciences _ * 1788 * , 569 kuhn s , asenbaum p , kosloff a , sclafani m , stickler b a , nimmrichter s , hornberger k , cheshnovsky o , patolsky f and arndt m 2015 cavity - assisted manipulation of freely rotating silicon nanorods in high vacuum _ nano lett . _ * 15 * 5604 arnold s , keng d , shopova s i , holler s , zurawsky w and vollmer f 2009 whispering gallery mode carousel a photonic mechanism for enhanced nanoparticle detection in biosensing _ opt . express _ * 17 * 6230 veissier l , nicolas a , giner l , maxein d , sheremet a s , giacobino e and laurat j 2013 _ opt . _ * 38 * 712 marzlin k p , zhang w and wright e m 1997 vortex coupler for atomic bose - einstein condensates _ phys * 79 * 4728 kapale k t and dowling j p 2005 vortex phase qubit : generating arbitrary , counterrotating , coherent superpositions in bose - einstein condensates via optical angular momentum beams__phys . _ _ * 95 * 173601

in this work we develop and analyze a mathematical model of biological control to prevent or attenuate the explosive increase of an invasive species population in a three - species food chain . we allow for finite time blow - up in the model as a mathematical construct to mimic the explosive increase in population , enabling the species to reach disastrous " levels , in a finite time . we next propose various controls to drive down the invasive population growth and , in certain cases , eliminate blow - up . the controls avoid chemical treatments and/or natural enemy introduction , thus eliminating various non - target effects associated with such classical methods . we refer to these new controls as ecological damping " , as their inclusion dampens the invasive species population growth . further , we improve prior results on the regularity and turing instability of the three - species model that were derived in . lastly , we confirm the existence of spatio - temporal chaos . rana d. parshad , kelly black , and emmanuel quansah matthew beauregard

1502.02010

population dynamics are a fundamental aspect of many biological processes . in this paper , we introduce and investigate a mathematical model for the population dynamics of an invasive species in a three - species food chain model . exotic species are defined as any species , capable of propagating themselves into a nonnative environment . if the propagating species is able to establish a self sustained population in this new environment , it is formally defined as invasive . the survival and competitiveness of a species depends intrinsically on an individual s fitness and ability to assimilate limited resources . often invasive species possess the ability to dominate a particular resource . this allows them to expand their range via out - competing other native species . in the united states damages caused by invasive species to agriculture , forests , fisheries and businesses , have been estimated to be @xmath0120 billion a year @xcite . in the words of daniel simberloff : `` _ _ invasive species are a greater threat to native biodiversity than pollution , harvest , and disease combined . _ _ '' @xcite therefore understanding and subsequently attenuating the spread of invasive species is an important and practical problem in spatial ecology and much work has been devoted to this issue @xcite . more recently , the spread of natural and invasive species by nonrandom dispersal , say due to competitive pressures , is of great interest @xcite . however , there has been less focus , on the actual eradication of an invasive species , once it has already invaded . this is perhaps a harder problem . in the words of mark lewis : _ `` once the invasive species are well established , there is not a lot you can do to control them''_. @xcite it is needless to say however , that in many ecosystems around the world , invasion has already taken place ! some prominent examples in the us , are the invasion by the burmese python in southern regions of the united states , with climatic factors supporting their spread to a third of the united states @xcite . the sea lamprey and round goby have invaded the great lakes region in the northern united states and canada @xcite . these species have caused a severe decline in lake trout and other indigenous fish populations . lastly the zebra mussel has invaded many us and canadian waterways causing large scale losses to the hydropower industry @xcite . another factor attributed to the increase of an invasive population , is that the environment may turn favorable for the invasive species in question , while becoming unfavorable for its competitors or natural enemies . in such situations , the population of the invasive species may rapidly increase . this is defined as an _ outbreak _ in population dynamics @xcite . these rapid changes tend to destabilize an ecosystem and pose a threat to the natural environment . as an illustration , in the european alps certain environmental conditions have enabled the population of the larch budmoth to become large enough that entire forests have become defoliated @xcite . in most ecological landscapes , due to exogenous factors , one always encounters an invasive species and an invaded species . if the density of the former , be it invasive , disease causing , an agricultural pest , defoliator or other , undergoes a rapid transition to a high level in population , the results can be catastrophic both for local and nonlocal populations . biological and chemical controls are an adopted strategy to limit invasive populations @xcite . chemical controls are most often based on direct methods , via the use of pesticides @xcite . biological control comprises of essentially releasing natural enemies of the invasive species / pest against it . these can be in the form of predators , parasitoids , pathogens or combinations thereof @xcite . there are many problems with these approaches . for example , a local eradication effort was made by usgs through a mass scale poisoning of fish in order to prevent the asian carp ( an invasive fish species ) from entering the chicago sanitary and ship canal @xcite . the hope was to protect the fishing interests of the region . however , among the tens of thousands of dead fish , _ biologists found only one asian carp_. thus chemical control is not an exhaustive strategy . however , biological controls are also not without its share of problems . in fact , sometimes the introduced species might attack a variety of species , other than those it was released to control . this phenomena is referred to as a _ non - target effect _ @xcite , and is common in natural enemies with a broad host range . for example , the cane toad was introduced in australia in 1935 to control the cane beetle . however , the toad seemingly attacked everything else but its primary target @xcite ! in addition , the toad is highly poisonous and therefore predators shy away from eating it . this has enabled the toad population to grow virtually unchecked and is today considered one of australia s worst invasive species @xcite . in studies of biological control in the united states estimate that when parasitoids are released as biological controls that @xmath1 of the introduced species will attack non - targets , in canada these numbers are estimated as high as @xmath2 @xcite . in practise it is quite difficult to accurately predict these numbers . the current drawbacks make it clear that alternative controls are necessary . furthermore , that modeling of alternative controls is important to validate the effectiveness of a management strategy that hopes to avoid non - target effects . such modeling is essential to access and predict the best controls to employ , so that the harmful population will decrease to low and manageable levels . this then gives us confidence to devise actual field trials . we should note that in practice actual eradication is rarely achieved . thus there are clear questions that motivate this research : 1 . how does one define a `` high '' level for a population , and further , how well does an introduced control actually work , at various high levels ? 2 . is it possible to design controls that avoid chemicals / pesticides / natural enemy introduction , and are still successful ? this paper addresses these questions through the investigation of a mathematical model that : 1 . blows - up in finite time . given a mathematical model for a nonlinear process , say through a partial differential equation ( pde ) , one says finite time blow up occurs if @xmath3 where @xmath4 is a certain function space with a norm @xmath5 , @xmath6 is the solution to the pde in question , and @xmath7 is the blow up time . therefore `` highest '' level is equated with blow up , and the population passes through every conceivable high level of population as it approaches infinity . 2 . incorporates certain controls that avoid chemicals / pesticides / natural enemy introduction . the controls we examine are : 1 . the primary food source of the invasive species is protected through spatial refuges . the regions that offer protection are called _ prey refuges _ and may be the result of human intervention or natural byproducts , such as improved camouflage . 2 . an overcrowding term is introduced to model the movement or dispersion from high concentrations of the invasive species . densely populated regions have increased intraspecific competition and interference which cause an increase in the dispersal of the invasive species . this is an improvement to current mathematical models and will be seen to be beneficial if used a control . 3 . we introduce role reversing mechanisms , where the role of the primary food source of the invasive species , and the prey of this food source , is switched in the open area ( the area without refuge ) . this models situations where the topography provides competitive advantages to certain species . it will be seen that this also is beneficial if used a control . in effect , this uses the current ecosystem and by modifying the landscape a natural predator in the environment has an advantage in key areas . hence , the invasive species population will adversely be effected . it can also be thought of as introducing a competitor of the invasive species , to compete with it for its prey . note , none of the above rely on enemy release to predate on the invasive species , or a parasite or pathogen release to infect the invasive species . thus potential non - target effects due to such release can be avoided . in the literature finite time blow up is also referred to as an explosive instability @xcite . there is a rich history of blow up problems in pde theory and its interpretations in physical phenomenon . for example , this feature is seen in models of thermal runaway , fracture and shock formation , and combustion processes . thus blow up may be interpreted as the failure of certain constitutive materials leading to gradient catastrophe or fracture , it may be interpreted as an uncontrolled feedback loop such as in thermal runaway , leading to explosion . it might also be interpreted as a sudden change in physical quantities such as pressure or temperature such as during a shock or in the ignition process . the interested reader is referred to @xcite . blow up in population dynamics is usually interpreted as excessively high concentrations in small regions of space , such as seen in chemotaxis problems @xcite . our goal in the current manuscript is to bring yet another interpretation of blow up to population dynamics , that is one where we equate such an excessive concentration or blow up " of an invasive population with disaster for the ecosystem . furthermore , it is to devise controls that avoid non - target effects and yet reduce the invasive population , before the critical blow up time . in the following , the norms in the spaces @xmath8 , @xmath9 and @xmath10 are respectively denoted by @xmath11 in addition , the constants @xmath12 , @xmath13 and @xmath14 may change between subsequent lines , as the analysis permits , and even in the same line if so required . the current manuscript is organised as follows . in section [ 2 ] we formulate the spatially explicit model that we consider . in section [ 3 ] we describe in detail the modeling of the control mechanisms that we propose , and term ecological damping " . here we make three key conjectures [ c1 ] , [ c2 ] and [ c3 ] , concerning our control mechanisms . section [ 4 ] is devoted to some analytical results given via lemma [ lem : wsol ] , theorem [ thm : wsolimproved ] and [ thm : gattr ] . section [ 5 ] is where we explain our numerical approximations and test conjectures [ c1 ] , [ c2 ] and [ c3 ] numerically . in section [ 6 ] we investigate spatio - temporal dynamics in the model . we investigate the effect of overcrowding on turing patterns , and we also confirm spatio - temporal chaos in the model . lastly we offer some concluding remarks and discuss future directions in section [ 7 ] . a three species food chain model is developed , where the top predator , denoted as @xmath6 , is invasive . in our model , @xmath6 may blow up in finite time . although populations can not reach infinite values in finite time , they can grow rapidly @xcite . the blow up event indicates that the invasive population has reached `` an extremely high '' and uncontrollable level . naturally , that level occurs prior to the blow up time . therefore , as @xmath15 approaches infinity in finite time , it passes through every conceivable `` high '' level . the blow up time , @xmath7 , is viewed as the `` disaster '' time . we investigate mechanisms that attempt to lower and control the targeted population _ before time @xmath7_. this modeling approach has distinct advantages : 1 . there is no ambiguity as to what is a disastrous high level of population . 2 . there is a clear demarcation between when or if the disaster occurs . 3 . the controls that are proposed do not rely on a direct attack on the invasive species , as is the traditional approach , rather we attempt to control the food source of @xmath6 . this will avoid possible nontarget effects . the model provides a useful predictive tool that can be tuned and established through data , in various ecological settings . mathematical models are advantageous in many situations due to their cost - effectiveness and versatility . of course obtaining an analytical solution for a nonlinear model is virtually impossible , outside of special cases . however , a numerical approximation can be developed to accurately investigate the role and effect our controls have on the blow up behavior . suppose an invasive species @xmath6 has invaded a certain habitat and it has become the top predator in a three species food chain . hence , @xmath6 predates on a middle predator @xmath16 , which in turn predates on a prey @xmath17 . a temporal model is given for the species interaction , namely @xmath18 the spatial dependence is included via diffusion , @xmath19 defined on @xmath20 . here @xmath21 , @xmath22 and @xmath23 is the one or two dimensional laplacian operator . we define @xmath24 to be the spatial coordinate vector in one or two dimensions . the parameters @xmath25 , @xmath26 and @xmath27 are positive diffusion coefficients . neumann boundary conditions are specified on the boundary . the initial populations are given as @xmath28 are assumed to be nonnegative and uniformly bounded on @xmath29 . there are various parameters in the model : @xmath30 , @xmath31 @xmath32 @xmath33 and @xmath34 are all positive constants . their definitions are as follows : @xmath35 is the growth rate of prey @xmath17 ; @xmath36 measures the rate at which @xmath16 dies out when there is no @xmath17 to prey on and no @xmath6 ; @xmath37 is the maximum value that the per - capita rate can attain ; @xmath38 and @xmath39 measure the level of protection provided by the environment to the prey ; @xmath40 is a measure of the competition among prey , @xmath17 ; @xmath41 is the value of @xmath16 at which its per capita removal rate becomes @xmath42 ; @xmath34 represents the loss in @xmath6 due to the lack of its favorite food , @xmath16 ; @xmath43 describes the growth rate of @xmath6 via sexual reproduction . these models offer rich dynamics and were originally proposed in @xcite in order to explain why chaos has rarely been observed in natural populations of three interacting species . the model stems from the leslie - gower formulation and considers interactions between a generalist top predator , specialist middle predator and prey . the study of these models have generated much research @xcite . an interesting fact is if @xmath44 the spatially independent and spatially dependent models are easily seen to blow up in finite time @xcite . the spatially dependent model offers further rich dynamics , in particular the possibility of turing instabilities and non turing patterns @xcite . nevertheless , to avoid blow up it appears that one must restrict @xmath45 , where @xmath46 . this was established for the spatially independent model in @xcite and is offered here for convenience : [ thm : aziz ] consider the model - . under the assumption that @xmath47 all non - negative solutions ( i.e. solutions initiating in @xmath48 ) of - are uniformly bounded forward in time and they eventually enter an attracting set @xmath49 . we have recently shown the above result to be _ false _ in the ode and pde cases . that is , - may blow up in finite time , even under provided the initial data is large enough @xcite . it is clear that even if is maintained @xmath6 can blow up . this becomes more evident if we consider the coefficient @xmath50 on @xmath51 and the fact that if the fecundity of @xmath6 is large compared to @xmath52 then blow up will occur . this may happen in situations where : 1 . there is an abundance of @xmath16 , 2 . @xmath6 possesses certain abilities to out compete other species , and harvest enough @xmath16 , 3 . the environment has turned favorable for @xmath6 and unfavorable for its natural enemies or competitors . thus it can harvest @xmath16 unchecked . if no intervention is made we can envision @xmath6 growing to disastrous levels with adequate initial resources . the blow up time @xmath53 is viewed as the point of disaster in an ecosystem . thus one is interested in controlling @xmath6 via the use of biological controls , before the critical time @xmath53 . these observations motivate an interesting question . assume that both - and its spatially explicit form - blow up in finite time for @xmath54 for a given initial condition . can we modify - , via introducing certain controls , so that now _ there is no blow up _ , for the same initial condition ? it is clear that controlling the population of an invasive specie is advantageous , and most often necessary . however the avenues for which this is possible , whilst avoiding non - target effects , is not clear . here , we propose new controls that may delay or even remove blow - up in the invasive population . we refer to these controls as ecological damping " , akin to damping forces such as friction in physical systems , that add stability to a system . the crux of our idea is to use prey refuges , in conjunction with role reversal and overcrowding effects . there is a vast literature on prey refuge , spatial refuges as well as role reversal in the literature . the interested reader is referred to @xcite . however , to the best of our knowledge these have not been proposed as control mechanisms for invasive species . in the one dimensional case . this function decreases monotonically throughout its domain . the range is @xmath55 . if @xmath56 then the prey @xmath16 is protected while if @xmath57 then @xmath16 is unprotected . , title="fig : " ] we consider modeling a prey refuge . consider the continuous function @xmath58 . the region where @xmath59 or sufficiently close to one is defined as a _ prey refuge domain _ or _ patch_. we call the region where @xmath60 or sufficiently close to zero an open area , this is the region where @xmath6 can predate on @xmath16 . in figure [ fig : refugeplot ] we see a sharp gradient between the prey refuge domain and the open area . the inclusion of a prey refuge influences the equations for @xmath6 and @xmath16 , namely , @xmath61 posed on a bounded domain in one or two dimensions . neumann boundary conditions are specified . the equation for @xmath17 may be found in . the introduction of the prey refuge creates regions where it is impossible for @xmath6 to predate on @xmath16 . notice , that if the entire spatial domain is considered a prey refuge , that is , @xmath59 @xmath62 , then the coefficient of @xmath51 will depend on the sign of @xmath54 . if this is negative , then the invasive species dies off . likewise , in the absence of a prey refuge , that is , @xmath60 @xmath62 the equations for @xmath6 and @xmath16 collapse to our previous ones , that is , and , respectively . in such a case , it is know that blow up may still occur for sufficiently large enough data even if @xmath54 . this is because the coefficient of @xmath51 may still be positive , as one always has @xmath63 . the introduction of the refuge _ forces _ the coefficient of @xmath51 to change sign between the refuge and the open area . in the literature such problems are referred to as indefinite parabolic problems . the word _ indefinite _ refers to the sign of the coefficient being indefinite . although there is a vast literature on such problems for single species models @xcite , there is far less work for systems . there is also a large amount of literature for such switching mechanisms incorporated to understand competitive systems @xcite , particularly in the vein of human economic progress . this indefinite parabolic problem motivates a collection of questions : if blow up occurs for particular parameters in the case @xmath60 , will a prey refuge prevent blow up ? how does this depend on the geometry of the refuge ? what is the critical size or shape of a refuge that prevents blow up from occurring ? does this depend on the size of the initial condition ? in the situation that blow up still persists , how is the blow up time affected ? are these results influenced by multiple refuges ? in either case , we make the conjecture : [ c1 ] consider the three species food chain model - , a set of parameters with @xmath64 , and an initial condition @xmath65 such that @xmath6 which is the solution to , blows up in finite time , that is @xmath66 there exists a patch @xmath67 , s.t for any single patch of measure greater than or equal to @xmath68 , the modified model - , with the same parameter set and initial condition , has globally existing solutions . in particular the solution @xmath6 to , does not blow up in finite time . a proof of this conjecture can be established for certain special cases in one dimesion . assume - blows up at time @xmath69 , for a certain parameter set and an initial condition @xmath70 , such that @xmath71 . we consider now introducing a refuge at the right end , starting at some positive @xmath72 . now for @xmath73 $ ] the density of @xmath16 decreases , and the coefficient of @xmath51 is @xmath74 , hence @xmath6 is restricted on the right hand side of the domain . we assume that this is equivalent to introducing a dirichlet boundary condition somewhere in the closed interval @xmath75 $ ] . thus the modified model is equivalent to @xmath76 , \ \mbox{where } \ a < b < \pi,\ ] ] and @xmath77 and @xmath78 . here we assume the initial condition satisfies @xmath79 . let us now compare to @xmath80 , \ \mbox{where } \ a < b < \pi.\ ] ] the solution of the above is a supersolution to . however we know that there exists small data solutions to . this is easily seen via following the methods in @xcite . basically without loss of generality we may assume @xmath81 . we then multiply by @xmath82 and integrate by parts in @xmath83 to obtain @xmath84 thus @xmath85 we now define the functional @xmath86 since @xmath87 we obtain @xmath88 now we multiply by @xmath6 and integrate integrate by parts in @xmath83 to obtain @xmath89 this yields @xmath90 so if @xmath91 then @xmath92 and so @xmath93 this essentially yields @xmath94 setting @xmath95 . we derive the following differential inequality @xmath96 . and so @xmath97 which means the solutions exist for @xmath98 , where @xmath99 and then the solution @xmath6 blows up at time @xmath53 . this means that if @xmath100 , then the initial data is actually small enough to ensure globally existing solutions @xcite . what is required is @xmath101 this criteria can always be obtained for large enough refuge , that is , for @xmath102 small enough . since the boundary terms cancel , the norm here is equivalent to the @xmath103 norm , hence by sobolev embedding we have @xmath104 now since @xmath105 , for @xmath102 chosen small enough we obtain @xmath106 thus combining and we obtain @xmath107 this proves a particular case of conjecture [ c1 ] . in high population density areas a species should have greater dispersal in order to better assimilate available resources and avoid crowding effects such as increased intraspecific competition . these effects can be modeled via an overcrowding term , that has not been included in mathematical models of biological control . consider the improved mathematical model that includes an overcrowding effect of the invasive species @xmath6 , namely , @xmath108 with initial and boundary conditions as before . again , the equation for @xmath17 remains unchanged . the addition of @xmath109 represents a severe penalty on local overcrowding . this is interpreted as movement from high towards low concentrations of @xmath6 , directly proportional to @xmath6 . hence , @xmath6 attempts to avoid overcrowding and disperses toward lower concentrations . such models have been under intense investigation recently and are referred to as cross - diffusion and self - diffusion systems @xcite . the mathematical analysis of such models is notoriously difficult @xcite . we limit ourselves to the one - dimensional case in the forthcoming analysis and its numerical approximations . therefore the one dimensional laplacian is considered in . the presence of blow up is not affected if we maintain neumann boundary conditions at the boundaries . this can be seen in a straightforward manner . let us assume the classical model , that is - , blows up . therefore , without loss of generality , there exists a @xmath110 such that @xmath111 . this implies that if we set @xmath112 , then , @xmath113 leading to the blow up of @xmath114 . now , consider the integration over @xmath29 of . since the overcrowding term integrates to zero then blow up still persists . however , a combination of a prey refuge with the overcrowding effect may prevent blow up . further , we expect it takes a smaller refuge to accomplish the removal of blow up . this is precisely the following conjecture : [ c2 ] consider the three species food chain model - , a set of parameters , such that @xmath64 , and an initial condition @xmath65 such that @xmath6 which is the solution to , blows up in finite time , that is @xmath115 there exists a patch @xmath116 , and an overcrowding coefficient @xmath117 , s.t for any single patch of measure greater than or equal to @xmath118 , the modified model - , with the same parameter set and initial condition , has globally existing solutions . in particular the solution @xmath6 to , does not blow up in finite time . furthermore , @xmath119 , where @xmath120 is the patch found in conjecture . this is not difficult to see , as multiplying through by @xmath6 and integrating by parts yields @xmath121 so if @xmath91 then @xmath92 and so @xmath122 thus even if one has negative energy , @xmath123 so that blows - up , due to the presence of the positive @xmath124 term in , will not blow up for small data , thus yielding a global solution . thus there are initial data for which blow ups but does not , and so a smaller refuge would work in this case . the method follows by mimicing the steps in - , with the additional @xmath124 term . the prey refuge and overcrowding are included in the one - dimensional mathematical model . we also include a role - reversal of @xmath17 within the protection zone of the refuge . this models the scenario in which two species may prey on each other in various regions where it is advantageous . hence , the role - reversal of @xmath17 will compete with the invasive species @xmath6 . figure [ fig : refugeplotrolereversal ] depicts the scenario of a one dimensional prey refuge for which outside the protection zone both @xmath17 and @xmath6 may predate on @xmath16 . hence , the model we propose of this scenario is given below , @xmath125 in light of this model we propose conjecture [ c3 ] : [ c3 ] consider the three species food chain model - , a set of parameters such that @xmath64 , and an initial condition @xmath65 such that @xmath6 which is the solution to , blows up in finite time , that is @xmath126 there exists a patch @xmath127 such that the modified model - , with the same parameter set and initial condition , has globally existing solutions . in particular , the solution @xmath6 to does not blow up in finite time . furthermore , @xmath128 , where @xmath129 is the patch found in conjecture . all three conjectures are tested numerically in section [ sec : numresults ] . the nonnegativity of the solutions of - is preserved by application of standard results on invariant regions @xcite . this is due to to the reaction terms being quasi - positive , that is , @xmath130 since the reaction terms are continuously differentiable on @xmath131 , then for any initial data in @xmath10 or @xmath132 , it is easy to directly check their lipschitz continuity on bounded subsets of the domain of a fractional power of the operator @xmath133 , where @xmath134 the three dimensional identity matrix , @xmath135 is the laplacian operator and @xmath136 denotes the transposition . under these assumptions , the following local existence result is well known ( see @xcite ) . [ prop : ls ] the system - admits a unique , classical solution @xmath137 on @xmath138 . if @xmath139 then @xmath140 [ lem : gronwall ] let @xmath141 , and @xmath142 be nonnegative functions in @xmath143 . assume that @xmath144 is absolutely continuous on @xmath145 and the following differential inequality is satisfied : @xmath146 if there exists a finite time @xmath147 and some @xmath148 such that @xmath149 for any @xmath150 , where @xmath151 , and @xmath12 are some positive constants , then @xmath152 in this section we improve the global existence conditions that has been derived recently in @xcite . here , we provide the previous result , theorem [ thm : wsol ] . [ thm : wsol ] consider the three - species food - chain model described by - . for any initial data @xmath153 , such that @xmath154 , and parameters @xmath155 such that @xmath156 , there exists a global classical solution @xmath137 to the system . we now give an improved result for the ode case , which is easily modified for pde case . in essence , given any initial data ( however large ) , there is global solution , for @xmath36 appropriately large . this is summarised in following lemma , [ lem : wsol ] consider the three - species food - chain model described by - . as long as @xmath64 , given any initial data @xmath157 , however large , there exists a global solution @xmath137 to the system as long as the parameter @xmath36 is s.t @xmath158 consider the following subsystem @xmath159 the @xmath16 here grows faster than the @xmath16 determined by . similarly , the @xmath6 here blows up faster than @xmath6 in . in order to drive @xmath16 down below @xmath160 , we use the exact solution to , that is , @xmath161 assuming that @xmath162 , then for @xmath16 to be below @xmath160 implies that @xmath163 thus when @xmath164 , we shall have @xmath165 now blows up at time @xmath166 . so if we choose @xmath167 appropriately , we can make @xmath168 . this is done by choosing @xmath169 thus if @xmath167 is s.t @xmath170 , then @xmath171 , before @xmath6 blows up . therefore , if we consider the @xmath6 determined by then this will certainly not blown up by time @xmath69 . at this point however , @xmath172 is negative , so @xmath6 in , can not blow up after @xmath53 , and will decay . in this latter case the global attractor is a very simple , @xmath173 . hence , we have an extinction state for @xmath6 . in short , we see that for any initial condition that is specified , there exists an @xmath36 ( depending on the initial condition ) , s.t - has a global solution . we recap the following estimate from @xcite @xmath176 therefore , there exists time @xmath177 given explicitly by @xmath178 such that , for all @xmath179 , the following estimate holds uniformly : @xmath180 here @xmath181 is the compactification time of @xmath182 . we also recall the following local in time integral estimate for @xmath183 , from @xcite @xmath184 now we move to the @xmath185 estimate for the @xmath16 component . this can not be obtained directly . thus we use the grouping method again by multiplying by @xmath186 and by @xmath187 , adding the two together , and setting @xmath188 , we obtain : @xmath189 we add a convenient zero , @xmath190 , to to obtain @xmath191 by multiplying by @xmath192 , integrating by parts over @xmath29 , and keeping in mind the positivity of the solutions , we find that @xmath193 this yields @xmath194 such that @xmath195 now via estimates in @xcite we obtain @xmath196 for @xmath197 where @xmath198 is explicitly derived in @xcite , @xmath199 which implies that @xmath200 notice by multiplying by @xmath16 and integrating by parts one obtains , @xmath201 integrating the above on @xmath202 $ ] yields @xmath203 thus via the mean value theorem for integrals we obtain a time @xmath204 $ ] s.t @xmath205 in addition , we recap the uniform @xmath206 estimates found in @xcite , latexmath:[\[\label{eq : x1-h11 } @xmath206 compactification time of the solution . we next need to derive @xmath208 estimates for the @xmath16 component . this is tricky , since when we multiply by @xmath209 , integrate by parts , and then apply young s inequality with epsilon we obtain @xmath210 however , there is no global estimate on @xmath6 , as it may possibly blow up , but we know that there is always a local solution on @xmath211 $ ] , where @xmath53 is the blow up time of @xmath212 . furthermore , on this time interval the solution is classical . thus we make local in time estimates on this interval , using standard local in time estimates on @xmath213 from @xcite . using gronwalls inequality via integration on the time interval @xmath214 $ ] yields , @xmath215 for @xmath216 . note , lemma [ lem : wsol ] requires us to manipulate @xmath167 , in order to prove large data global existence . from a practical point of view this is not always easy as manipulating the death rate directly may not be condusive to realistic control strategies . we next provide the following improved result that is valid even if @xmath217 , without manipulating @xmath167 . [ thm : wsolimproved ] consider the three - species food - chain model described by - . for certain initial data @xmath153 , there exists a global classical solution @xmath137 to the system , even if @xmath218 , as long as the parameters are such that @xmath219 , by the @xmath220 absorption time of @xmath16 ( denoted @xmath221 ) , if @xmath221 satisfies @xmath222 , where @xmath53 is the blow up time of the following pde @xmath223 with the same initial and boundary conditions as in order to show that blow up can be avoided as stated above , even in @xmath224 our @xmath206 estimates via must be improved . this is because the embedding of @xmath225 is lost in @xmath224 . thus , if we are to control the @xmath226 norm in @xmath224 we need @xmath227 control . this is achieved in the next subsection . we will estimate the @xmath227 norms via the following procedure , we rewrite as @xmath229 we square both sides of the equation and integrate by parts over @xmath29 to obtain @xmath230 this result follows by the embedding of @xmath231 . therefore , we obtain @xmath232 due to classical local in time regularity results , see [ prop : ls ] , the solutions are @xmath233 ) , thus @xmath234 can not escape to @xmath235 , and is bounded below , upto the existence time . thus we obtain @xmath236 these constants may depend on @xmath237 since even though @xmath234 can not escape to @xmath235 in finite time the case of it decaying like @xmath238 is not precluded . we now make uniform in time estimates of the higher order terms . integrating the estimates of in the time interval @xmath239 $ ] , for @xmath240 , where @xmath241 is the maximum of the @xmath206 compactification times of @xmath242 yields @xmath243 similarly we adopt the same procedure for @xmath16 to obtain @xmath244 however , the above estimates for @xmath16 are made on the time interval @xmath245 $ ] . next , consider the gradient of . following the same technique as in deriving we obtain for the left hand side @xmath246 this follows via the boundary condition . thus we have @xmath247 this follows via young s inequality with epsilon , as well as the embedding of @xmath248 . this implies that @xmath249 we now use the uniform gronwall lemma with @xmath250 to obtain @xmath251 the estimate for @xmath16 is derived similarly , but again we assume we are on @xmath245 $ ] . thus , via elliptic regularity , @xmath252 in @xmath224 and @xmath253 , the following estimate is valid in @xmath224 and @xmath253 @xmath254 , if the parameters and the data @xmath255 are such that @xmath256 , ( for the @xmath12 in , where @xmath257 is the embedding constant for @xmath258 ) and @xmath259 then we obtain @xmath260 since @xmath261 the solution to @xmath262 has not blown up yet , and since this is a supersolution to @xmath6 solving , this @xmath6 has definitely not blown up . however , from this point on the coefficient of @xmath51 in is negative , as @xmath263 , hence @xmath6 solving can never blow up , from this point on . this proves theorem [ thm : wsolimproved ] . now we assume there are globally existing solutions , and we aim to investigate the regularity of the omega limit set . hence we assume we are working with initial data and parameters for which @xmath6 is globally bounded . to begin , we obtain useful integral in time estimates . in particular , we integrate in time from @xmath239 $ ] where @xmath264 , the @xmath220 absorbing time , to obtain @xmath265 similarly , an estimate for @xmath266 can be established , namely @xmath267 we now develop higher order estimates for the time derivatives . first , consider the partial derivative w.r.t @xmath237 of equations , multipling by @xmath268 , and integrate over @xmath29 we obtain @xmath269 using holder s inequality , young s inequality with epsilon , and our earlier estimates yield @xmath270 thus we obtain @xmath271 the application of the uniform gronwall lemma with @xmath272 gives us the following uniform bound latexmath:[\[\label{eq : uh1-t } methods applied to the equation for @xmath16 and @xmath6 yield @xmath274 we now show that the asymptotic state @xmath275 under certain parameter restrictions is an attractor with more regularity than was derived previously . essentially , we consider @xmath276 for which we have globally existing solutions , that is the maximal existence time for the solutions @xmath277 . then we consider the omega limit set for such solutions . @xmath278 we state the following result . [ thm : gattr ] consider the three - species food - chain model described by - . for initial conditions @xmath255 , such that there is a globally existing solution , there exists a parameter @xmath167 , ( depending on the initial condition ) , for which the omega limit set @xmath279 . furthermore this is an attractor with @xmath228 regularity . thus for suitably chosen @xmath167 , the attractor for such solutions is the extinction state of @xmath6 . we have shown that the system is well posed , under certain parameter , and data restrictions , via theorem [ thm : wsolimproved ] . thus , there exists a well - defined semigroup @xmath280 . also , via lemma [ lem : wsol ] we can find an @xmath167 such that @xmath6 decays to zero . the estimates via establish the existence of bounded absorbing sets in @xmath220 . now we rewrite the equation for a sequence @xmath281 as follows @xmath282 we aim to show the convergence of both the right hand side strongly in @xmath185 . due to the uniform bounds via , , and the embedding of @xmath283 and @xmath284 we obtain a subsequence @xmath285 stil labeled @xmath281 s.t @xmath286 as @xmath287 , as @xmath288 . similarly , @xmath289 as @xmath290 . thus , given a sequence @xmath291 that is bounded in @xmath292 , we know that , for @xmath293 , @xmath294 this yields the precompactness in @xmath228 , hence the closure in is in @xmath228 . a similar analysis can be done for the @xmath16 and @xmath6 components . hence the theorem is established . in this section a numerical approximation for the one and two dimensional models are developed in order to numerically demonstrate our earlier conjectures , , and in addition to exploring the rich dynamics the mathematical models exhibit . a one - dimensional spectral - collocation method is developed to approximate the one dimensional equations - which include overcrowding , refuges , and role - reversal . we then offer a second order finite difference approximation of the two - dimensional equations , , and that investigate the effect of a refuge . overcrowding and role - reversal effects are not investigated in the latter numerical approximation . the two dimensional approximation employs a peaceman - rachford operator splitting and techniques developed by one of the authors in @xcite . this approach takes advantage of the sparsity and structure of the underlying matrices and the known computational efficiency of operator splitting methods . without loss of generality the domain is scaled and translated to @xmath295 and @xmath296 in one and two - dimensions , respectively . we develop a spectral - collocation approximation of - . these equations can be written compactly as @xmath297 where @xmath298 and @xmath299 is the differential operator that includes the laplacian operator and reactive terms . the spatial approximation is constructed from a chebychev collocation scheme @xcite . the spatial approximation is constructed as a linear combination of the interpolating splines on the gauss - lobatto quadrature . the resulting system is then integrated in time using an implicit scheme , in particular a second order adams - moulton method . the chebychev collocation approximation allows for a high order spatial approximation @xcite . this offers the ability to capture fine resolution details with a relatively smaller number of degrees of freedom . however , there is a downside . the resulting matrices are full and ill conditioned @xcite . this is problematic for the inversion in the resulting linear solve . nevertheless , the spectrum associated with the second derivative operator is real , negative , and grows in magnitude like @xmath300 @xcite . the collocation scheme is constructed on the gauss - lobatto abscissa with respect to the chebychev weight , @xmath301 where @xmath302 . an approximation is then constructed as a linear combination of the lagrange interpolates on the abscissa , @xmath303 where @xmath304 is an approximation to the unknown populations and @xmath305 is a @xmath306 vector of fourier coefficients . the basis functions are defined by @xmath307 and each @xmath308 is a polynomial of degree @xmath309 @xcite . the basis functions also satisfy @xmath310 . this is used to develop an approximation to our differential operator and a system of equations is constructed via a discrete inner product , @xmath311 where @xmath312 is the projection operator onto the space of polynomials of degree @xmath309 . the discrete inner product is given below , and it is based on an inner product defined by @xmath313 the integral can be approximated as a sum over the gauss - lobatto quadrature for the chebychev weight and is exact for polynomials up to degree @xmath314 , @xmath315 the resulting norm results in an equivalent norm for polynomials up to degree @xmath316 @xcite . the abscissa , @xmath317 , are the same as those defined in equation , and the @xmath318 are the quadrature weights . the resulting numerical approximation is then constructed using this finite inner product @xmath319 the resulting approximation is an extension of that given in and is constructed using a galerkin approach , @xmath320 the basis functions , @xmath321 , are lagrange interpolants , and after substitution of the definition in equation ( [ eqn : numerics : approximationsum ] ) the result can be greatly simplified , @xmath322 where @xmath323 . the adams - moulton method used to advance the initial value problem in time necessitates solving a nonlinear system of equations at each time step . this is achieved through a newton - raphson method . the ensuing linearized problem in the newton - raphson method is solved through a restarted gmres @xcite . a second order approximation of , , and is developed . while the spectral - galerkin appoximation of the previous section may be extended to two dimensions , the computational cost of the nonlinear solve is troublesome due to the sensitivity of the matrices and the lack of their sparsity . here , the approximation is still of high order , while the underlying matrices are block tridiagonal or tridiagonal with diagonal blocks . this enables direct inversion techniques , in particular the thomas algorithm . the governing equations are written compactly as @xmath324 where @xmath298 , @xmath325 are the reactive terms that depend on space , time , and the species @xmath326 and @xmath6 , @xmath327 , @xmath328 is taken component - wise , and @xmath329 is a diagonal matrix with nonnegative entries @xmath330 and @xmath331 from semigroup theory the formal solution is @xmath332 where @xmath333 is the evolution operator associated with @xmath329 . a suitable quadrature is used to approximate the integral . here , a second order trapezoidal rule is used , that is , @xmath334 this motivates the implicit method , @xmath335 where @xmath336 . the exponential is approximated through a peaceman - rachford operator . this creates the second order implicit method , @xmath337 where @xmath338 is the variable time step , @xmath339 , and @xmath340 . the last reactive term does require the solution at the next time step . to avoid a required nonlinear solve , we use a first order approximation to avoid this . this simplification maintains the second order accuracy of the approximation . we write the solution in an alternative form , @xmath341 this can be conveniently solved through adi procedures @xcite , that is , by splitting the problem into , @xmath342 we see that the first equation keeps @xmath343 implicit while @xmath344 is explicit . we then take our intermediate solution , @xmath345 , and solve the second equation keeping @xmath344 implicit and @xmath344 explicit . now , the spatial operators may be approximated through the two - dimensional chebyshev approximation similar to that of the previous section , however we choose second order central differences . let @xmath346 and @xmath347 , where @xmath348 , @xmath349 , for @xmath350 , and @xmath351 let @xmath352 and @xmath353 be the second order approximations to the operators @xmath343 and @xmath344 . the approximation is utilized throughout the entire two - dimensional domain . at the boundary , we require an equation for @xmath354 at _ ghost _ points , that is locations of @xmath355 and @xmath356 . these are established using central difference approximations to the neumann boundary conditions . we then substitute these back into the system of equations . this maintains the second order accuracy and the tridiagonal structure of the equations . at each step , the tridiagonal equations in the adi method are directly solved using the thomas algorithm which comes at a expense of @xmath357 . we numerically demonstrate that there is good evidence for conjectures , , and . in the one dimensional setting , we translate and scale the domain to @xmath358 and use a refuge function of @xmath359 the refuge size is delineated by the value @xmath360 since that is the location where the gradient is steepest . as expected the blow up time is affected by the presence of a refuge . for instance , for a fixed parameter set we look at the effect on the blow up time with a refuge , refuge with role reversal , and role reversal and overcrowding as compared to the classical model . the results are shown in figure [ numblowupvsrefuge ] . interestingly , in the modified model with role reversal , it is found that blow up does not occur when @xmath57 or @xmath361 in the spatial domain . we also see that the blow up times are influenced by the size of the refuge . in fact , there is critical refuge size for which given any refuge size greater then the solutions will not blow up . this is evidenced in figure [ numblowupvsrefuge ] by the steep gradient of the curves around @xmath362 . therefore , the population of the invasive species can be controlled . in situations where there is only a spatial refuge the blow up time curve is always increasing . this is consistent with our two dimensional results . it is clear that there exists a critical refuge size for which refuges larger than this will prevent blow up . in the one - dimensional model , without role reversal or overcrowding effects , we investigate the effect of the critical refuge size versus the size of the initial condition of @xmath16 while maintaining the other initial conditions and parameters values . we use the same parameters as given in figure [ numblowupvsrefuge ] and vary the uniform initial condition on @xmath16 . interestingly , figure [ num : refugevsic ] shows a logarithmic dependency of the critical refuge size to prevent blow - up of @xmath6 , on the initial condition size of @xmath16 . for which blow up in the invasive specie population @xmath6 will occur . identical parameters and resolution are used as in figure [ numblowupvsrefuge ] . the domain was translated and scaled to @xmath358 . ] the size of the spatial refuge has an influence on the blow up time . however , the location and number of spatial refuges also influences the blow up time . in fact , blow up is sometimes eliminated depending on the initial condition , refuges , and their subsequent locations . for instance , if we consider a uniform initial condition in @xmath6 and compare the blow up time in a situation of a single refuge of width located near the boundary versus the case of evenly splitting this refuge then we find the blow up time is increased . this delay is exasperated if the two refuges are farther apart . of course , if we consider a different initial condition then blow may not just be delayed , rather removed ! for instance if we consider a concentrated initial population of @xmath6 for which the highest concentration of @xmath6 contains the spatial refuge then blow up can be eliminated ! hence , the concentration of @xmath6 inside the refuge may protect the species enough so that the population of @xmath6 decays sufficiently to avoid blow up in its population . that is , @xmath363 , is shown versus time . with a circular refuge , we see that blow up is avoided . this provides experimental evidence that the location and size of the refuge is important to avoid blow up in the invasive population . the dashed line represents the population in the case where there is no spatial refuge . the simulations were done on a @xmath364 uniform grid with a temporal step of @xmath365 . ] to illustrate this consider the two - dimensional model , , and with an initial condition of @xmath366 with parameters @xmath367 , @xmath368 , @xmath369 , @xmath370 , @xmath371 , @xmath372 , @xmath373 , @xmath374 , @xmath375 , @xmath376 , @xmath377 , @xmath378 , @xmath379 . clearly , the coefficient @xmath380 . if @xmath381 is zero throughout the entire spatial domain there exists blow up in the @xmath6 population . however , if we have a circular refuge such that , @xmath382 for @xmath383 then blow up is avoided . figure [ rpopulationavoided ] shows the total population versus time . we can see the population starts to increase rapidly , but the increase is attenuated as a result of the spatial refuge . hence , the population growth is not sustained and begins to decrease . hence , the size * and * location of the refuge and the initial condition play a delicate balance in preventing blow up . in two dimensions if we choose a particular shape of the spatial refuge it was conjectured that there exists a critical size for which blow up is avoided . here , we look at two situations : a square and circular refuges with increasing size centered in the middle of the spatial domain . we let @xmath59 inside the refuge while @xmath384 outside . this clearly delineates the protection zones . for a fixed parameter regime and initial conditions of @xmath385 we determine the critical refuge size . each calculation is carried out with a @xmath364 grid . the temporal step was fixed at @xmath386 . the parameters used : @xmath367 , @xmath387 , @xmath388 , @xmath371 , @xmath389 , @xmath373 , @xmath374 , @xmath375 , @xmath376 , @xmath377 , @xmath378 , @xmath379 . it is found that for a square the critical refuge area is @xmath390 roughly @xmath391 of the spatial domain . the critical refuge area for the circular refuge is approximately @xmath392 , roughly @xmath393 of the spatial domain . in this section we shall investigate the effects of overcrowding in the absence of refuges and role reversal , in the classical model . therefore , we focus on whether an appropriate choice of @xmath117 can induce turing instabilities . it is shown in @xcite that diffusion processes can destabilize the homogenous steady state solution when @xmath394 . for convenience , we restate the one - dimensional model given in equations , , and , @xmath395 consider the linearization of - about the positive interior equilibrium point @xmath396 where @xmath397 , @xmath398 is the the spatially homogenous steady state solution , and @xmath399 ( see @xcite ) . for instance , for the parameters given at the beginning of this section @xmath400 and @xmath401 . consider a small space time perturbation , that is , @xmath402 with @xmath403 as the positive interior equilibrium point given as @xmath404 and @xmath405 . substituting and collecting linear terms of order @xmath406 , we obtain @xmath407 where @xmath408 is the diffusion matrix and @xmath409 is the jacobian matrix associated with the ordinary differential equation part of model - . let @xmath410 where @xmath411 is the spatial coordinate in @xmath29 , @xmath412 is the amplitude , @xmath413 is the eigenvalues associated with the interior equilibrium point , @xmath404 and @xmath414 is the wave number of the solution . upon substituting , we obtain the characteristic equation @xmath415 is a @xmath416 identity matrix . the sign of @xmath417 indicates the stability , or lack thereof , of the equilibrium point @xmath404 . the dispersion relation is @xmath418 the coefficients of @xmath419 are determined by expanding , namely , @xmath420 to check for stability of the equilibrium solution @xmath404 , we use the routh hurwitz criterion . this state that for @xmath404 to be stable we need @xmath421 contradicting either of these statements ensures instability for @xmath404 . finally , for diffusion to cause a turing instability it is sufficient to require that around the equilibrium point we have @xmath422 we refer the reader to a well detailed analysis of this in @xcite . hence , for a turing instability to occur , we require that is satisfied when @xmath423 ( without diffusion ) and at least one of the equations in changes sign when @xmath424 ( with diffusion ) . by this we consider letting @xmath425 become negative when @xmath424 and satisfy when @xmath423 . this suggests that spatial patterns should be observed . our parameter search has not yielded a parameter set for which holds while @xmath426 when @xmath427 , at least one of these inequalities changes sign . thus we can not conclusively say that @xmath428 will induce or inhibit turing pattern . however , it is conclusive that @xmath428 certainly has an effect on the type of patterns that do form . in particular , the patterns fall into two types : spatial patterns and spatio - temporal patterns . the conditions for which type forms is succinctly described via table [ table : pa ] . .conditions on the signs of coefficients for different types of patterns . [ cols="^,^,^,^,^",options="header " , ] now when @xmath427 this clearly changes the spatio - temporal pattern , as evidenced by the sign changes from case 1 to 2 in table [ table : pa ] . in this situation the dispersion relation is shown in figure [ dispersionplot2 ] . the resulting patterns are shown figure [ spatiotemporalpatternsuvr](a)-(f ) . resembling the two cases in table [ table : pa ] . notice that when @xmath429 a band of wavenumbers are unstable . ] the long - time simulation yields ( a)-(c ) turing patterns ( @xmath394 ) , that are spatio - temporal , and ( d)-(f ) stripe turing patterns ( @xmath430 ) , which are purely spatial . the other parameters are : @xmath431 , @xmath432 @xmath433 @xmath434 @xmath435 grid points are used with a temporal step size of @xmath436 . ] if @xmath437 then - reduces to a model similar to the classical predator - prey model with a holling type ii functional response , for which we know there can not occur turing instability . there is one caveat , the death rate in @xmath16 now becomes nonlinear . so essentially the equations are @xmath438 such systems have been investigated with and without cross and self diffusion @xcite . in the regular diffusion case , it is known that a turing instability can exist @xcite , where the nonlinearity in death rate is due to cannibalism . however , to our knowledge , for the specific form of the death rate as above this is not yet known . thus we see that the classical model - can exhibit spatio - temporal patterns , apart from just the spatial patterns that were uncovered in @xcite . furthermore addition of the overcrowding term in - , can cause the spatio - temporal patterns to change into a purely spatial patterns . it is noted that the inclusion of overcrowding with a nonlinear death rate , such as in , can lead to turing instability in the classical predator - prey model with holling type ii response . the goal of this section is to investigate spatio - temporal chaos in the classical model - . spatio - temporal chaos is usually defined as deterministic dynamics in spatially extended systems that are characterized by an apparent randomness in space and time @xcite . there is a large literature on spatio - temporal chaos in pde , in particular there has been a recent interest on spatially extended systems in ecology exhibiting spatio - temporal chaos @xcite . however , most of these works are on two species models , and there is not much literature in the three - species case . in @xcite we showed diffusion induced temporal chaos in - , as well as spatial chaos when the domain is enlarged . in @xcite , various patterns non - turing were uncovered in - , in the case of equal diffusion , but spatio - temporal chaos was not confirmed . note , that the appearance of a jagged structure in the species density , as seen in @xcite , which seems to change in time in an irregular way , does not necessarily mean that the dynamics is chaotic . one rigoros definition of chaos means sensitivity to initial conditions . thus two initial distribution , close together , should yield an exponentially growing difference in the species distribution at later time . in order to confirm this in - , we perform a number of tests as in @xcite . we run - form a number of different initial conditions , that are the same modulo a small perturbation . the parameter set is chosen as in @xcite . we then look at the difference of the two densities , at each time step in both the @xmath226 and @xmath439 norms . thus we solve - with the following parameter set : @xmath440 , @xmath441 @xmath442 @xmath443 thus the steady state solution for the ode system is @xmath444 . the simulations use two different ( but close together in @xmath445 norms ) initial conditions . the first simulation ( which we call @xmath446 ) is a perturbation of @xmath447 by @xmath448 . the second simulation ( which we call @xmath449 ) is a perturbation of @xmath447 by @xmath450 . the densities of the species are calculated up to the time @xmath451 at each time step in the simulation we compute @xmath452 where @xmath453 are used . then , @xmath454 is plotted on a log scale . in doing son , we observe the exponential growth of the error . this grows at an approximate rate of @xmath455 . since this is positive then this is an indicator of spatio - temporal chaos . these numerical tests provide experimental evidence to the presence of spatio - temporal chaos in the classical model - . figure [ contourchaos ] shows the densities of the populations in the @xmath456-plane while figure [ contourchaoserror ] gives the error and its logarithm till @xmath457 . -plane for @xmath17 , @xmath16 , and @xmath6 from left to right . the long - time simulation yields spatio - temporal chaotic patterns . @xmath435 grid points are used with a temporal step size of @xmath436 . ] and its logarithm in the @xmath185 norm for the @xmath6 species , under slightly different initial conditions are shown . the difference is seen to grow at an exponential rate of approximately @xmath458 . comparable results were found for the @xmath226 and @xmath439 norms . these tests provide experimental evidence that there presence of spatio - temporal chaos in the classical model -.,title="fig : " ] and its logarithm in the @xmath185 norm for the @xmath6 species , under slightly different initial conditions are shown . the difference is seen to grow at an exponential rate of approximately @xmath458 . comparable results were found for the @xmath226 and @xmath439 norms . these tests provide experimental evidence that there presence of spatio - temporal chaos in the classical model -.,title="fig : " ] in this work we have proposed and investigated a new model for the control of an invasive population . an invasive species population is said to reach catastrophic population levels when its population reaches a particular threshold . the mathematical model uses the mathematical construct of finite time blow up which enables the model to examine the effect of controls for any particular threshold , especially since this level depends on the application . in effect , this construct demonstrates that if the invasive population has large enough numbers initially , it can grow to explosive levels in finite time : thus wreaking havoc on the ecosystem . hence , we are interested in the influence certain controls will have on the invasive population and if its population may be reduced below disastrous levels . this formulation yields a clear mathematical problem : assume that - blows - up in finite time for @xmath54 for a given initial condition . are there controls and features of the model that we can include to modify - so that now _ there is no blow up _ in the invasive population given the same initial condition ? this paper addresses this question , suggesting clear controls and improvements to the mathematical model . we then investigate these improvements numerically and theoretically . in traditional practice , biological control works on the enemy release hypothesis . that is releasing an enemy of the invasive species into the ecosystem will lead to a decrease in the invasive population . however , non - target effects are prevalent , hence this approach is problematic and may create even more devastating impacts on the ecosystem @xcite . here , we propose controls that do not use the release of biological agents . in particular , we introduce spatial refuges or safe zones for the primary food source of the invasive species . mathematically , this transforms - into an indefinite problem @xcite , that is , where the sign of the coefficient of @xmath51 switches between inside and outside of the refuge . we demonstrated numerically and in analytically , with some assumptions , that this control can prevent blow up and drive the invasive population down . our numerical experiments suggests that there is a delicate balance between the size and location of the refuge and the initial condition . in particular , we revealed a logarithmic dependence on the size of the initial condition for @xmath16 versus the critical refuge size to prevent blow - up of @xmath6 , see figure [ num : refugevsic ] . clearly , the balance is even more pronounced for multiple refuges and in higher dimensions . we also improved the mathematical model by incorporating overcrowding effects , which also may be used a control . we also examined the situation where a species may switch its primary food source based in regions of the domain , that is , a prey may switch to a predator , hence their roles are reversed . this models scenarios where influences on the landscape provide a competitive advantages to certain species . both , role - reversal and overcrowding act as damping mechanisms and also may prevent blow - up in the invasive population . in particular , smaller refuge sizes in conjunction with role - reversal and overcrowding are required to prevent blow up . of course , how does one enforce overcrowding in an ecosystem such that we obtain this desired effect ? can one devise mechanisms to facilitate this dispersal of population ? we suggest one approach . suppose we create a lure , such as a pheromone trap , that is placed in the refuge . this would lure the invasive species into the patch , where its growth would be controlled . of course , the species would eventually exit the refuge in search for a higher concentration of food . hence , in the future we plan to include spatially dependent diffusion constants to model this situation . in our mathematical models we confirmed spatio - temporal chaos . we also see that overcrowding can effect the sorts of turing patterns that might form . since , environmental effects are inherently stochastic , part of our future investigations introduces stochasticity into the model . it is not known what effect this will have on the spatio - temporal chaos or turing patterns that may emerge . it is clear that our new mathematical modeling constructs and results are useful analytical and numerical tool for scientists interested in control of invasive species . moreover , the results motivate and encourage numerous avenues of future exploration , many of which are currently under study and will be presented in future papers . we would like to acknowledge very helpful conversations with professor pavol quittner , professor philippe souplet and professor joseph shomberg , as pertains to the analysis of indefinite parabolic problems , as well as finite time blow - up in the superlinear parabolic problem , under various boundary conditions , and initial data restrictions . bampfylde , c.j . and lewis , m.a . , _ biological control through intraguild predation : case studies in pest control , invasive species and range expansion _ , bulletin of mathematical biology , 69 , 1031 - 1066 , 2007 . bryan , m.b . ; zalinski , d. ; filcek , k.b . ; libants , s. ; li , w. ; scribner , k.t . , _ patterns of invasion and colonization of the sea lamprey in north america as revealed by microsatellite genotypes _ , molecular ecology , 14(12 ) , 3757 - 3773 , 2005 . white , k. a. j. and gilligan , c.a . , _ spatial heterogeneity in three species , plant parasite hyperparasite , systems _ , philosophical transactions of the royal society of london . series b : biological sciences 353.1368 ( 1998 ) : 543 - 557 . loda , s.m . , pemberton , r.w . , johnson , m.t . and follet , p.a . _ nontarget effects - the achilles heel of biological control ? retrospective analyses to reduce risk associated with biocontrol introductions . _ , annual reveiw of entomology , 48 , 365 - 396 , 2003 . parshad , r. d. , kumari , n. and kouachi , s. , _ a remark on study of a leslie - gower - type tritrophic population model [ chaos , solitons and fractals 14 ( 2002 ) 1275 - 1293 ] _ , chaos , solitons @xmath459 fractals , 71(2 ) , 22 - 28 , 2015 . philips , b. ; shine , r. , _ adapting to an invasive species : toxic cane toads induce morphological change in australian snakes _ , proceedings of national academy of sciences usa , 101(49 ) , 17150 - 17155 , 2004 . ackermann , n. , bartsch , t. , kaplicky , . p and quittner , p. _ a priori bounds , nodal equilibria and connecting orbits in indefinite superlinear parabolic problems _ , transactions of the american mathematical society , 360(7 ) , 3493 - 3539 , 2008 . rodda , g.h . ; jarnevich , c.s . ; reed , r.n . , _ what parts of the us mainland are climatically suitable for invasive alien pythons spreading from everglades national park ? _ , molecular ecology , 14(12 ) , 3757 - 3773 , 2005 . upadhyay , r. k. , naji , r. k. , kumari , n. , _ dynamical complexity in some ecological models : effects of toxin production by phytoplanktons _ , nonlinear analysis : modeling and control , 12(1 ) , 123 - 138 , 2007 .

we examine the palindromic automorphism group of a free group , a group first defined by collins in which is related to hyperelliptic involutions of mapping class groups , congruence subgroups of , and symmetric automorphism groups of free groups . cohomological properties of the group are explored by looking at a contractible space on which acts properly with finite quotient . our results answer some conjectures of collins and provide a few striking results about the cohomology of , such as that its rational cohomology is zero at the vcd . _ dedicated to peter hilton _ : + doc , note , i dissent . a fast never prevents a fatness . i diet on cod .

math0112190

let @xmath3 be the automorphism group of a free group @xmath1 on @xmath4 generators @xmath5 . a reduced word @xmath6 is called a _ palindrome _ if it is equal to its reverse @xmath7 . in @xcite collins defines the _ palidromic automorphism group _ @xmath0 as the subgroup of @xmath3 consisting of all automorphisms @xmath8 for which @xmath9 is a palindrome for all @xmath10 . he showed that the group was generated by three types of automorphisms : * maps @xmath11 , @xmath12 , which send @xmath13 and fix all other generators @xmath14 . * maps @xmath15 which send @xmath16 and fix all other generators @xmath14 . * maps corresponding to elements of the symmetric group @xmath17 which permute the @xmath18 among themselves . the portion of @xmath0 generated by just the @xmath11 is called the _ elementary palindromic automorphism group _ of @xmath1 and denoted @xmath19 . note that @xmath20 collins showed that a set of defining relators for @xmath19 is given by relations of the form 1 . @xmath21 2 . @xmath22 3 . @xmath23 he remarked how similar this was to the relations for the pure symmetric automorphism group @xmath24 ( see gilbert s work in @xcite ) : 1 . @xmath25 2 . @xmath26 3 . @xmath27 where @xmath28 , @xmath12 , sends @xmath29 and fixes all other generators @xmath14 . on the basis of this , collins conjectured that one could find the virtual cohomological dimension of @xmath0 by employing the methods of @xcite , as he did for @xmath30 in @xcite . he also speculated that @xmath19 is torsion free , just as @xmath24 is . we are able to answer both of these questions in this paper , as well as obtaining several interesting facts about the cohomology of @xmath0 . [ c34 ] let @xmath0 be the palindromic automorphism group of the free group @xmath1 on @xmath4 letters and let @xmath19 be the subgroup of elementary palindromic automorphisms . then * a ) * the virtual cohomological dimension of @xmath0 is @xmath31 . * b ) * ( i ) for the prime 2 , the krull dimension of @xmath32 is @xmath4 . for odd primes @xmath33 , the krull dimension of @xmath34 is @xmath35 $ ] . ( ii ) in the range where the krull dimension of @xmath34 is 1 , the period is @xmath36 . * c ) * the group @xmath19 is torsion free . * d ) * the cohomology group @xmath37 . * e ) * if @xmath33 is an odd prime and @xmath38 , then the farrell cohomology of the palindromic automorphism group is the same as that of the symmetric group on @xmath33 elements : @xmath39 for analogous results concerning @xmath3 , see @xcite and @xcite . see @xcite for the definition of the farrell cohomology @xmath40 of a group @xmath41 of finite vcd with coefficients in a @xmath41-module @xmath42 and also see @xcite for several useful properties of these cohomology groups . the remainder of this paper is structured as follows . in section 2 , we discuss @xmath0 and note how it relates to some other groups , while in section 3 we introduce the space @xmath43 which @xmath0 acts on and prove parts a ) and d ) of theorem [ c34 ] . section 4 is concerned with a realization proposition which allows us to establish parts b ) ( i ) and c ) of the main theorem . finally , section 5 looks in more detail at the cohomology of @xmath0 at odd primes @xmath33 and establishes parts b ) ( ii ) and e ) of the main theorem . the authors would like to thank john meier for an enlightening conversation about symmetric automorphisms of free groups , aiding the presentation of this paper . let @xmath44 be the spine of auter space ( see @xcite,@xcite,@xcite ) and @xmath45 . let @xmath46 be the automorphism which sends @xmath16 for each @xmath10 . define the @xmath47-graph @xmath48 to be a graph with 2 vertices and @xmath49 edges , where each edge goes from one vertex to the other one . choose one of the two vertices of @xmath50 to be the basepoint @xmath51 , and define the rose @xmath52 to be the result of wedging together @xmath4 copies of @xmath50 at the basepoint . the petals of the rose @xmath52 can be identified with the generators @xmath53 of @xmath1 , so that @xmath54 . there is an action of @xmath55 on @xmath52 given by inverting each petal of the rose . this action realizes the subgroup @xmath56 in the sense of @xmath57}$ ] ( also cf . @xcite . ) an action of a group @xmath41 on a graph @xmath58 is _ without inversions _ if @xmath41 does not send any edge @xmath59 to its inverse @xmath60 , and an action is _ reduced _ if there are no @xmath41-invariant subforests in @xmath58 . the action of @xmath61 on @xmath52 is both without inversions and reduced . from now on , when we refer to a group action on a graph , it is assumed that the edges of the graph are subdivided as necessary to insure that the group acts without inversions . note that the palindromic automorphism group @xmath0 is just @xmath62 . this follows because an easy argument shows that every element of @xmath62 is palindromic , and because the generators of @xmath0 are all in @xmath62 . for example , the generators @xmath11 are just products of @xmath61-nielsen transformations ( see @xcite where the @xmath41-nielsen transformation @xmath63 of a @xmath41-graph @xmath58 has the same vertex and edge set as @xmath58 but where the terminal point of an edge @xmath64 , @xmath65 , in the new graph is the initial point of the edge @xmath66 in the original graph ; this induces a map @xmath67 from the fundamental groupoid of the first graph to that of the second where @xmath64 is sent to @xmath68 and all other edges are sent to themselves . ) that is , if the petal @xmath53 of the rose consists of the edges @xmath69 , then @xmath11 is the composition @xmath70 . as a note for the curious , it follows that @xmath0 and @xmath30 are distinct groups for @xmath71 , since a direct argument shows that @xmath30 has no element of order 2 in its center . in addition , @xmath19 and @xmath24 are also obviously distinct ( for @xmath72 ; for @xmath73 they are the same group with the same presentation , ) since the former abelianizes to an elementary abelian @xmath74-group of rank @xmath75 while the latter abelianizes to a free abelian group of rank @xmath75 . in addition to its formal palindromic properties , the group @xmath0 arises naturally from looking at hyperelliptic subgroups of mapping class groups ( cf . gries @xcite for corresponding homological properties . ) we have a commutative diagram @xmath76 where @xmath77 is the mapping class group of an orientable surface of genus @xmath78 , @xmath79 is the mapping class group of an orientable surface of genus @xmath78 with 1 puncture , and @xmath80 is the mapping class group of an orientable surface of genus @xmath78 with two punctures , where each puncture is fixed pointwise . the map from @xmath80 to @xmath81 is obtained first by taking an intersection basis @xmath82 for the fundamental group of the surface @xmath83 with two punctures . one of the punctures should serve as the basepoint for fundamental group considerations . the other is treated as an actual puncture , so that the fundamental group of the surface minus this point is a free group @xmath84 on @xmath85 generators . the map from @xmath80 to @xmath81 is now obtained by sending an element of @xmath80 to the automorphism of @xmath84 that it induces . the map from @xmath79 to @xmath86 is obtained similarly . let @xmath87 be a hyperelliptic involution ( see , for example , @xcite . ) then @xmath88 has @xmath89 fixed points , two of which are of course the punctures on the surface @xmath83 . choose loops @xmath82 , based at one of the punctures , which form an intersection basis for the surface @xmath83 ( say the ones described in @xcite in the section on hyperelliptic riemann surfaces . ) by going along the top row of diagram [ c24 ] and then projecting downward , we see that the image of @xmath88 in @xmath90 is @xmath91 . let @xmath92 be the image of @xmath88 in @xmath81 . our goal is to show that @xmath92 is conjugate to @xmath93 in @xmath81 . [ c25 ] if @xmath94 and the image of @xmath95 in @xmath96 is @xmath91 , then the image of any conjugate @xmath97 of @xmath95 , @xmath98 , is also @xmath91 . this follows directly since @xmath91 is in the center of @xmath96 . [ c26 ] if @xmath94 is an involution whose image in @xmath96 is @xmath91 , then @xmath95 can be realized on a marked graph whose underlying graph is the rose . realize @xmath95 on a reduced marked graph whose underlying graph is @xmath99 . first , we show that if @xmath59 is an edge of @xmath99 which is not fixed by @xmath95 , then we can assume that one endpoint of @xmath59 is the basepoint @xmath51 . let @xmath100 . choose a shortest path @xmath101 from @xmath59 to @xmath51 . since @xmath102 , @xmath103 for every @xmath104 in the path @xmath101 . so we can apply a sequence of nielsen transformations ( see @xcite ) and slide @xmath59 along @xmath101 to @xmath51 . note that since @xmath99 is reduced , @xmath105 now forms either a rose @xmath106 based at @xmath51 , or a @xmath50 . proceeding in this manner , we can slide all of the edges of @xmath99 not fixed by @xmath95 to the basepoint . by way of contradiction , suppose that two of these @xmath50-graphs ( that have been moved so that one vertex of each @xmath50 is the basepoint ) , say @xmath107 and @xmath108 , share another common vertex in addition to the basepoint @xmath51 . that is , suppose that there is a vertex @xmath109 and each of @xmath110 go from @xmath111 to @xmath51 . say @xmath112 is the marked graph , and recall that @xmath113 . by replacing @xmath95 by a conjugate if necessary ( see lemma [ c25 ] ) we can assume that @xmath78 sends the petal @xmath114 of @xmath52 to @xmath115 , the petal @xmath116 to @xmath117 , and the petal @xmath118 of @xmath52 to @xmath119 . so in @xmath120 we have @xmath121 @xmath122 and @xmath123 hence the first column of @xmath124 in @xmath96 is @xmath125 , the second column is @xmath126 , and the third column is @xmath127 . this contradicts the fact that @xmath128 . if the result of sliding to the basepoint @xmath51 all edges of @xmath99 not fixed by @xmath95 yields a rose @xmath52 , then we are done . otherwise , suppose by way of contradiction that there exist edges @xmath59,@xmath129 , and @xmath104 of @xmath99 such that * both @xmath59 and @xmath129 go from some vertex @xmath109 to @xmath51 . * @xmath130 . * @xmath104 goes from @xmath111 to @xmath111 . * @xmath131 . as before , say @xmath112 is the marked graph . by replacing @xmath95 by a conjugate if necessary ( see lemma [ c25 ] ) we can assume that @xmath78 sends the petal @xmath114 of @xmath52 to @xmath132 and the petal @xmath116 of @xmath52 to @xmath133 . so in @xmath120 we have @xmath134 and @xmath135 hence the first column of @xmath124 in @xmath96 is @xmath125 and the second column is @xmath136 . this contradicts the fact that @xmath128 . [ c27 ] if @xmath94 is an involution whose image in @xmath96 is @xmath91 , then @xmath95 is conjugate in @xmath3 to @xmath61 . realize @xmath95 on a marked graph @xmath137 . by replacing @xmath95 by a conjugate if necessary , we can assume @xmath138 for all @xmath10 . the involution @xmath95 of the graph @xmath52 must send the petal @xmath114 to some @xmath139 , since it is a graph automorphism . but since @xmath140 , we see that @xmath139 must be @xmath141 . similarly , we can see that the graph automorphism @xmath95 sends , for each @xmath10 , the petal @xmath53 to the petal @xmath142 . this means our current @xmath95 is equal to @xmath61 ( and thus our original @xmath95 , before we replaced it by a conjugate , was conjugate to @xmath61 . ) the following corollary is immediate : [ c28 ] the image @xmath92 of @xmath88 in @xmath81 is conjugate to @xmath93 . hence the hyperelliptic subgroup @xmath143 is conjugate to a subgroup of @xmath144 . as remarked in @xcite , the image of @xmath0 in @xmath96 is the subgroup of @xmath96 consisting of invertible matrices where each column has exactly one odd entry ( and the rest are even . ) the subgroup is the semidirect product @xmath145 where @xmath146 is the 2-congruence subgroup defined by the short exact sequence @xmath147 and @xmath17 is standard inclusion of the symmetric group @xmath148 we define a certain contractible space @xmath43 , related to auter space , which @xmath0 acts on with finite stabilizers and finite quotient . this allows us to obtain some cohomological results . a graph @xmath58 is a _ @xmath50-tree of rank @xmath4 _ if there exists a pointed tree @xmath149 such that @xmath58 is obtained by `` doubling '' every edge of @xmath149 into a @xmath50-graph . that is , the vertex set of @xmath58 is the same as the vertex set of @xmath149 and for every edge @xmath59 of @xmath149 going from @xmath111 to @xmath150 , @xmath58 has two edges @xmath151 and @xmath152 , both of which go from @xmath111 to @xmath150 . there is a natural @xmath153-action on such a graph @xmath58 , which is given by switching the two edges in each @xmath50-graph . note that the orbit space of @xmath58 under this action is just the tree @xmath149 . [ c1 ] the reduced graphs @xmath58 which realize the subgroup @xmath56 of @xmath3 are exactly the @xmath50-trees of rank @xmath4 , where @xmath61 acts on the trees via their natural @xmath153-action . we have already mentioned that the rose @xmath52 realizes @xmath61 . from theorem 2 of @xcite , the other reduced graphs @xmath58 which also realize @xmath154 are those that are nielsen equivalent to @xmath52 ( up to an equivariant isomorphism . ) if @xmath59 is an edge in one of the copies of @xmath50 in @xmath52 and @xmath129 is an edge in a different @xmath50 in @xmath52 , and both @xmath59 and @xmath129 point toward the basepoint , then nielsen transformation @xmath155 has the result of pulling the @xmath50-graph @xmath156 through the @xmath50-graph @xmath157 , so that now @xmath59 terminates at the initial vertex of @xmath129 , rather than at the basepoint @xmath51 . in other words , the result of applying one nielsen transformation to @xmath52 is that of sliding one of the petals of @xmath52 up though another petal . a basic induction argument now yields that the result of applying a series of nielsen transformations fo @xmath52 will be some @xmath50-tree @xmath58 . recall from @xcite that an edge @xmath59 of a @xmath41-graph @xmath58 is _ inessential _ if it is in every maximal @xmath41-invariant subforest of @xmath58 . a @xmath41-graph @xmath58 is _ inessential _ if it has at least one inessential edge and is _ essential _ if it is not inessential . let @xmath158 be the fixed point subspace of @xmath44 corresponding to some finite subgroup @xmath41 of @xmath3 . from @xcite ( cf . part iii of @xcite and @xcite ) , both the centralizer @xmath159 and the normalizer @xmath160 act on the contractible space @xmath158 with finite stabilizers and finite quotient . moreover , the space @xmath158 @xmath41-equivariantly deformation retracts to the space @xmath161 , where @xmath161 is constructed from @xmath158 by considering only essential marked graphs . hence @xmath161 is a good space to study if one wishes to calculate the cohomology of @xmath159 or @xmath160 . further recall that a @xmath41-graph @xmath162 is a @xmath41-equivariant _ blowup _ of a @xmath41-graph @xmath58 if some @xmath41-invariant subforest @xmath163 of @xmath162 can be collapsed away to yield @xmath58 . let @xmath164 be some reduced marked graph realizing @xmath56 . from claim [ c1 ] , @xmath58 is a @xmath50-tree . blow up @xmath58 @xmath61-equivariantly to some maximal essential blowup @xmath162 . [ c2 ] the fixed points / cells of the action of @xmath61 on @xmath162 are exactly the valence 2 vertices of @xmath162 . note that vertices in @xmath162 have valence 2 or 3 and that @xmath162 is obtained from @xmath58 by blowing up an oriented ideal forest ( see @xcite , @xcite , @xcite . ) briefly , ideal edges ( oriented ideal forests ) correspond to subsets of edges ( chains of subsets of edges ) which are pulled away from existing vertices in order to create new graphs which collapse down to the original graph . ) no edge is in @xmath165 because no edge is in @xmath166 and so blowing up ideal edges will not create any new edges that are fixed under the action of @xmath61 ( see @xcite page 229 . ) if a valence 3 vertex is in @xmath165 , then at least one edge of @xmath162 must be fixed by @xmath61 , which is a contradiction . all of the valence 2 vertices of @xmath58 are in @xmath166 . new valence 2 vertices created as ideal edges are blown up correspond to either : * old vertices of @xmath58 that used to be valence higher than 2 but have since had edges stripped ( pulled away ) from them . these are in @xmath165 because all vertices of @xmath58 are in @xmath166 . * new valence 2 vertices inserted to insure that @xmath61 acts on @xmath162 without inversions . these are also clearly in @xmath165 . note that @xmath167 . cut @xmath162 along each of its valence 2 vertices , yielding a graph @xmath168 with the same number of valence 3 vertices and edges as @xmath162 had , no valence 2 vertices , and twice as many valence 1 vertices as @xmath162 had valence 2 vertices . [ c3 ] @xmath169 the disjoint union of two trees @xmath170 and @xmath171 , where @xmath172 . there is a covering map @xmath173 obtained by mapping to the orbit space under the @xmath61-action . the forest collapse that sends @xmath162 to @xmath58 is @xmath61-equivariant , so it descends to a forest collapse of @xmath174 to @xmath175 . ( that the quotient of the forest upstairs in @xmath162 is also a forest in @xmath174 can be seen by an easy euler characteristic argument . ) but @xmath58 is a @xmath50-tree with a known @xmath61-action on it , and @xmath175 is a tree ( in fact , the underlying tree of the @xmath50-tree . ) hence @xmath174 is a tree . since @xmath173 is a covering map with fiber two points , @xmath168 is as described . let @xmath149 be a pointed tree with @xmath176 edges , all vertices valence either 1 or 3 , where @xmath51 is one of the valence 1 vertices . ( then @xmath149 has @xmath177 valence 1 vertices and @xmath31 valence 3 vertices . ) let @xmath178 and @xmath179 be two isomorphic copies of @xmath149 , and let @xmath180 be an isomorphism . define @xmath181 define a @xmath61-action on @xmath182 by @xmath183 [ c4 ] there is a bijective correspondence between trees @xmath149 as above ( with @xmath176 edges , etc ) and maximal , essential blowups @xmath162 of reduced @xmath61-graphs . the bijection is given by @xmath184 . from claims [ c2 ] and [ c3 ] , all blowups @xmath162 have the required form . finally , any @xmath182 can easily be reduced to a @xmath50-tree by collapsing edges , meaning that it is the blowup of such a graph . all maximal simplices in @xmath43 have the same dimension , from @xcite . maximal simplices in @xmath185 are constructed by taking chains of forest collapses from maximal blowups @xmath182 . alternatively , we can define a _ subforest _ of @xmath149 to be a collection @xmath83 of edges of @xmath149 such that there is no path in @xmath83 from one valence 1 vertex to another . ( if there were such a path , then @xmath186 would be a cycle in @xmath182 . ) in this way , we can think of maximal simplices as coming from chains of subforests of various trees @xmath149 . since @xmath0 acts on the contractible space @xmath43 with finite stabilizers and finite quotient , the vcd of @xmath0 is at most the dimension of a maximal cell from @xmath185 . such a cell comes from a chain of forest collapses of a tree @xmath149 with @xmath176 edges , @xmath177 valence 1 vertices , and @xmath31 valence 3 vertices . hence we can collapse at most @xmath31 of the valence 3 vertices into other vertices while doing forest collapses , resulting in maximal simplices of dimension @xmath31 . to show that the vcd of @xmath0 is at least @xmath31 , we note that the subgroup generated by @xmath187 for @xmath188 is isomorphic to @xmath189 . [ c6 ] let @xmath190 be the underlying graph of a particular marked graph in @xmath43 . hence @xmath190 comes equipped with a @xmath61-action . let @xmath191 be a simple closed curve in @xmath190 . then @xmath192 contains exactly two points , and @xmath193 is the curve @xmath194 , or @xmath191 with the opposite of its original orientation . @xmath190 can be blown up ( not necessarily uniquely ) to some maximal essential graph @xmath195 . @xmath58 is the union of two isomorphic copies @xmath178 and @xmath179 of @xmath149 , where @xmath178 and @xmath179 are attached along their corresponding valence 1 vertices . as we collapse from @xmath58 to @xmath190 , the trees @xmath178 and @xmath179 collapse to trees @xmath196 and @xmath197 . however , the attaching points for @xmath196 and @xmath197 are no longer necessarily just the valence 1 vertices , and could be other vertices as well . let @xmath198 be a taut path in @xmath196 from one attaching point @xmath199 to some other attaching point @xmath200 , where furthermore there are no attaching points in the interior of @xmath198 . let @xmath201 be the corresponding path in @xmath197 . then @xmath202 is a simple closed curve and @xmath203 ) = @xmath204 , or the original curve oriented in the other direction . hence the curve @xmath202 satisfies the conclusions of the lemma . our goal is to show that any simple closed curve @xmath191 takes this form . let @xmath205 be the portion of @xmath191 that is in @xmath196 and let @xmath206 be the portion of @xmath191 that is in @xmath197 . since @xmath191 is a cycle and yet both @xmath196 and @xmath197 are trees , both @xmath205 and @xmath206 are nonempty . in fact , there must be a path @xmath198 in @xmath205 from one attaching point @xmath199 to some other attaching point @xmath200 , where there are no other attaching points in the interior of the path . since @xmath191 is a simple closed curve , @xmath198 must be a taut path . let @xmath207 , some other taut path in @xmath190 from @xmath199 to @xmath200 . let @xmath208 be the map given by taking the quotient space under the action of @xmath61 . note that @xmath209 is a path from @xmath199 to @xmath200 . since @xmath196 is a tree and @xmath198 is the unique taut path in @xmath196 from @xmath199 to @xmath200 , this gives us that @xmath210 . hence if @xmath59 is an edge in @xmath198 , then @xmath211 is two edges , @xmath212 and @xmath213 . it follows that all of the oriented edges of the simple closed curve @xmath214 are in the simple closed curve @xmath191 . hence @xmath215 [ c7 ] let @xmath190 be the underlying graph of a particular marked graph in @xmath43 . hence @xmath190 comes equipped with a @xmath61-action . choose an edge @xmath59 in @xmath190 . among all simple closed curves @xmath216 which pass through @xmath59 , choose one curve @xmath191 for which the distance from the curve to the basepoint @xmath51 is minimal . then @xmath192 is two points @xmath199 and @xmath200 . one of these two points in the closest point in @xmath191 to the basepoint @xmath51 and one of them is the farthest point in @xmath191 to the basepoint @xmath51 . using the notation of the proof of lemma [ c6 ] , @xmath191 is the result of following some path @xmath198 in @xmath196 and then @xmath217 in @xmath197 , where @xmath198 goes from the attaching point @xmath199 to the attaching point @xmath200 . since @xmath198 is a path in a pointed tree , there is a unique vertex @xmath150 in @xmath198 which is closest to @xmath51 . by way of contradiction , suppose that @xmath218 . then @xmath150 is not an attaching point . let @xmath219 be the unique taut path in @xmath196 from @xmath51 to @xmath150 . let @xmath220 be the unique subpath of @xmath219 which contains @xmath150 and exactly one attaching point @xmath221 . now let @xmath222 be the path in @xmath196 which starts at @xmath221 , follows @xmath220 along to @xmath150 , and then either follows @xmath191 from @xmath150 to @xmath200 or @xmath194 from @xmath150 to @xmath199 ( where we choose whichever possibility insures that @xmath223 . ) then the simple closed curve @xmath224 is closer to @xmath51 than @xmath191 is , which is a contradiction . hence @xmath150 is @xmath199 or @xmath200 , and the lemma follows . [ c8 ] let @xmath190 be a graph which occurs as an underlying graph of marked graphs in @xmath43 . then there is only one possible @xmath61-action on @xmath190 . we see that our task is to show that a unique @xmath61-action is determined by the properties about simple closed curves listed in lemmas [ c6 ] and [ c7 ] . define an action @xmath225 on @xmath190 as follows . let @xmath59 be an oriented edge of @xmath190 . among all simple closed curves @xmath216 which pass through @xmath59 , choose one path @xmath191 for which the distance from the curve to the basepoint @xmath51 is minimal . let @xmath199 be a point on @xmath191 which is closest to the basepoint . let @xmath4 be the edge - path distance in @xmath191 from @xmath199 to @xmath59 . then there is an orientation @xmath226 such that if you traverse @xmath227 starting at @xmath199 and go @xmath4 edges , you get to @xmath59 . define @xmath228 to be the result of traversing @xmath229 starting at @xmath199 and then going @xmath4 more edges . by lemma [ c6 ] and [ c7 ] , the action @xmath225 is well defined and if any @xmath61 acts on @xmath190 , then the @xmath61-action and the @xmath225-action coincide . denote by @xmath230 the quotient space @xmath185 . [ c9 ] if two marked graphs in @xmath43 have the same underlying graph , then they correspond to the same vertex in @xmath230 . that is , the moduli space @xmath230 can be formed by looking only at the poset structure of the underlying graphs of marked graphs in @xmath43 . from proposition [ c8 ] , any underlying graph of a marked graph in @xmath43 has only one possible @xmath61-action . but from corollary 10.4 of @xcite , @xmath0 acts transitively on the set of marked @xmath61-graphs based on the same @xmath61-graph . the result follows . the simplices of @xmath43 group themselves into cubes , as described in 3 of @xcite . in @xcite , hatcher and vogtmann show that the quotients in @xmath231 of cubes in @xmath44 have the rational homology of balls . they use this to create a cubical chain complex which has the same rational homology as @xmath231 . our goal here is to establish a similar result for the cubes of maximal dimension in @xmath230 , where this time we want the quotients of cubes to have the @xmath232-cohomology of balls , where @xmath33 is any odd prime . following @xcite , we consider a maximal cube in @xmath43 . it is given by considering a maximal essential marked graph @xmath233 and considering some maximal subforest @xmath83 of @xmath149 . recall that by a subforest of @xmath149 we mean a subset @xmath83 of the edges of @xmath149 where there is no path in @xmath83 from one valence 1 vertex of @xmath149 ( or equivalently , one terminal edge of @xmath149 ) to another . from part a ) of theorem [ c34 ] , @xmath83 has @xmath31 edges in it . the cube corresponding to the pair @xmath234 can thus be thought of as imbedded in @xmath235 , where each coordinate vector is an edge of the cube , the graph obtained by collapsing each edge of @xmath236 is at the origin , and @xmath182 is at @xmath237 . let @xmath238 be the group of all ( pointed ) automorphisms of the tree @xmath149 which take @xmath83 to @xmath83 . the group @xmath239 acts linearly on the cube by permuting the coordinates of @xmath235 , and fixes the diagonal from @xmath240 to @xmath237 . ( the involution @xmath61 acts trivially on the cube , of course , since all of these cubes are coming from @xmath241 . ) hence , just as in @xcite , the quotient of the cube in @xmath230 is a cone with base @xmath242 , where @xmath243 is the boundary of the cube . [ c10 ] the finite group @xmath244 ( and hence its subgroup @xmath238 ) is all @xmath74-torsion . let @xmath245 . now @xmath246 must take the basepoint to the basepoint , and so it must take the unique edge attached to the basepoint to itself . for each @xmath4 , let @xmath247 be the edges in @xmath149 which are at most distance @xmath4 from @xmath51 . so @xmath248 is just one edge , and @xmath246 fixes it as already mentioned . since all nonterminal vertices of @xmath149 have valence 3 , an inductive argument yields that @xmath249 fixes @xmath247 pointwise . the following is an analog of proposition 3.1 of @xcite : [ c11 ] @xmath242 has the @xmath232-cohomology of an @xmath250-sphere or a ball . the latter possibility happens when there is an element of @xmath238 which induces an odd permutation of the edges of @xmath83 . the finite group @xmath238 , which is all 2-torsion , acts cellularly on @xmath243 , where the stabilizer of a cell fixes it pointwise . we use the spectral sequence for equivariant cohomology ( cf . @xcite vii 7 ) : @xmath251 \in \delta_n^r } h^s(stab(\delta ) ; { \mathbb{z}}_{(p ) } ) \rightarrow h_{aut(t , s)}^{r+s}(s^{n-2 } ; { \mathbb{z}}_{(p)})\ ] ] where @xmath252 $ ] ranges over the set @xmath253 of orbits of @xmath254-simplices @xmath255 in @xmath243 . since @xmath238 is all @xmath74-torsion and finite , so are all of the @xmath256 . hence if @xmath257 , @xmath258 . so the above spectral sequence converges to @xmath259 . but another filtration yields a spectral sequence with @xmath260 it follows that @xmath261 unless @xmath262 is @xmath263 or @xmath264 . hence @xmath265 and @xmath266 . the latter group of invariants is @xmath232 if the action of @xmath238 on @xmath243 preserves orientation and @xmath267 otherwise . the last assertion in the proposition follows from corollary 3.2 of @xcite . [ c14 ] the top dimensional cohomology group of @xmath230 vanishes . that is , @xmath268 . we show that the quotient of every maximal cube @xmath234 has a free face , so that the interior of the quotient of the cube can be collapsed away . if we can do this , then @xmath230 will have the same @xmath232-cohomology as an @xmath250-dimensional complex , and we will be done . in the degenerate case where there is an element of @xmath238 which induces an odd permutation of the edges of @xmath83 , then the quotient of the cube @xmath234 is not itself a cube . in this case , the diagonal from @xmath269 to @xmath270 is exposed in the quotient , and any @xmath250-dimensional simplex in the quotient which lies next to the diagonal is a free face . in the nondegenerate case , the quotient of the cube @xmath234 is itself a cube , although its boundary might be self indentified in various ways . since the subforest @xmath83 of @xmath149 is maximal , @xmath83 must contain at least one terminal edge @xmath59 . that is , one of the two vertices of @xmath59 is a valence 1 vertex or attaching point . let @xmath190 be the graph obtained from @xmath195 by collapsing the subforest @xmath271 . the graph @xmath190 has a maximal subforest corresponding to collapsing the edges @xmath59 and @xmath272 from the forest @xmath236 of @xmath58 . hence we see that collapsing @xmath59 gives us a face , which we will denote by @xmath273 , of the cube @xmath234 . it can be shown that this face corresponds to a ( nondegenerate , cubical ) face of the quotient of the cube @xmath234 because [ c15 ] there is an natural injection of @xmath274 into @xmath275 . define the lift @xmath276 of an automorphism @xmath277 by sending an edge @xmath129 to @xmath278 if @xmath279 and letting @xmath280 . denote by @xmath111 the valence 1 vertex of @xmath281 ( the attaching point ) and let @xmath150 be the other vertex of @xmath59 . in @xmath282 , @xmath283 . we must show that @xmath95 sends @xmath150 to @xmath150 . this follows automatically , however , as @xmath284 is the only valence 4 vertex of @xmath190 and so any automorphism of the graph must fix it . let @xmath129 and @xmath78 be the two other edges in @xmath149 which share the vertex @xmath150 . now if @xmath285 then @xmath95 could possibly exchange @xmath129 and @xmath78 , but this is fine as the lift @xmath276 also can . if @xmath109 , then one of @xmath129 or @xmath78 must be closer to the basepoint , and so @xmath95 must fix both @xmath129 and @xmath78 . regardless , @xmath276 can be defined as in the statement of the claim . warning : note that if @xmath59 is not a terminal edge , the above claim is false . collapsing an interior edge somtimes allows you to construct automorphisms with 3-torsion , which obviously can not be lifted to @xmath244 . no automorphism @xmath95 of @xmath273 can induce an odd permutation of the edges in @xmath286 , else the lift @xmath276 of @xmath95 to @xmath149 would induce an odd permutation of the edge of @xmath83 . since @xmath244 is all 2-torsion , it follows from the above claim that @xmath287 is also all 2-torsion . hence the same spectral sequence argument used in proposition [ c11 ] yields that the quotient of the cube corresponding to @xmath273 actually is a @xmath232-cohomology cube . it remains to be shown that the cubical face corresponding to @xmath273 is free . first , if another subforest @xmath288 with an edge @xmath289 of @xmath149 gives a cube with a face isomorphic to @xmath290 , then @xmath289 must also be a terminal edge of @xmath288 . hence the isomorphism @xmath291 maps the vertex that @xmath59 collapsed into to the the vertex that @xmath289 collapsed into , and so we can lift the isomorphism to one from @xmath292 . second , we must show that blowing up the vertex @xmath150 in @xmath190 only yields graphs isomorphic to @xmath58 . this follows by considering the ways that the vertex @xmath150 in @xmath190 can be blown up . say that the edges @xmath129 , @xmath78 , @xmath293 , and @xmath294 are the ones incident to @xmath150 . if the ideal edge orbit @xmath295 is blown up , we get back @xmath58 exactly , and if @xmath296 is blown up , we get a graph isomorphic to @xmath58 . as these are the only ways to blow up the graph @xmath61-equivariantly into another essential graph , we are done . [ c16 ] @xmath297 . that @xmath298 follows immediately from theorem [ c14 ] . recall that @xmath0 acts with finite stabilizers and finite quotient @xmath230 on the contractible space @xmath43 . since the stabilizers are finite , their rational cohomology vanishes , and the standard equivariant spectral sequence yields that @xmath299 . note that part d ) of theorem [ c34 ] follows from the above corollary . as a final remark for this section , we show that @xmath43 is an @xmath300 ( cf . @xcite ) ; that is , for finite subgroups @xmath41 of @xmath0 , the fixed point subcomplex @xmath301 is contractible . this follows directly from the corresponding property of @xmath3 . the following proposition is unneccesary in the specific case of @xmath43 , since ( proof omitted ) @xmath43 actually equals @xmath302 . this does not normally happen ( for example , the spaces @xmath303 mentioned later in fact [ c21 ] are not equal to the corresponding fixed point space of @xmath44 ) , however , and thus it seems worth noting the more general fact . [ ptwo1 ] let @xmath83 be a finite subgroup of @xmath3 and let @xmath304 be either @xmath305 or @xmath306 . let @xmath307 be the retract , defined by krstic and vogtmann and consisting of essential marked graphs , of the fixed point subcomplex @xmath308 of the spine of auter space @xmath44 . then @xmath307 is an @xmath309 space . let @xmath310 be a finite subgroup of @xmath304 and let @xmath41 be the ( finite , because @xmath311 ) subgroup generated by @xmath310 and @xmath83 . then @xmath312 , and @xmath158 is contractible from @xcite . it remains to be shown that @xmath313 deformation retracts to @xmath314 . given a marked graph @xmath58 representing a vertex of @xmath315 , we must show ( see proposition 3.3 of @xcite ) that for every edge @xmath59 in @xmath58 and every @xmath316 , @xmath59 is @xmath83-inessential if and only if @xmath317 is @xmath83-inessential . this follows automatically from corollary 4.5 of @xcite , which characterizes essential edges by looking at the stabilizers ( in @xmath83 ) of paths in @xmath58 . since @xmath318 , the stabilizers in @xmath104-translates of such paths are still in @xmath83 and are isomorphic ( conjugate by @xmath104 ) to those of the original path . let @xmath319 be a finite subgroup of @xmath0 and let @xmath320 be the ( finite ) subgroup generated by @xmath319 and @xmath61 . by zimmerman s @xcite realization theorem , we can realize @xmath320 by an action on an @xmath320-reduced graph @xmath58 . from the proposition below , @xmath58 is also @xmath56-reduced ; that is , @xmath58 is a @xmath50-tree . note that the corresponding statement is not true in @xmath321 ( have @xmath322 act on a @xmath47-graph @xmath323 and certainly would not be true in @xmath3 if the @xmath61-action were replaced by some other @xmath153-action . [ c18 ] let @xmath324 be a finite subgroup of the palindromic automorphism group with @xmath325 . realize @xmath320 by an action on an @xmath320-reduced marked graph @xmath164 . then @xmath164 is also a @xmath56-reduced marked graph . as before , let @xmath326 and identify the petals of the rose @xmath52 with the generators @xmath53 . note that @xmath58 has no separating edges , else it would not be @xmath320-reduced . in this proof , when we refer to concepts such as the number of times an edge @xmath59 of @xmath58 occurs in some @xmath327 , we mean that we should take the unique taut path in @xmath58 , starting and ending at @xmath51 , which is homotopic to the path @xmath327 in @xmath58 , and then count the number of times @xmath59 occurs in this taut path . by way of contradiction , suppose @xmath58 is not @xmath56-reduced . let @xmath328 be an edge of minimal distance to the basepoint @xmath51 such that @xmath329 is a forest . * case 1 : * @xmath330 . since @xmath151 is not a separating edge of @xmath58 , we can choose a nontrivial cycle @xmath331 , starting and ending at @xmath51 , which has just one occurence of @xmath151 and none of @xmath332 . if for all @xmath333 , the cycles @xmath327 have an even number of occurences of @xmath334 , then we could not write @xmath331 as a product of them and their inverses . so some @xmath335 has an odd number of occurences of @xmath336 in it . say that the exponent sum of @xmath151 in @xmath335 is @xmath337 , @xmath337 odd . then the exponent sum of @xmath151 in @xmath338 is still @xmath337 , but the exponent sum of @xmath339 is @xmath340 . this contradicts the fact that @xmath341 * case 2 : * @xmath342 . let @xmath8 be a shortest length path from @xmath51 to @xmath151 . say without loss of generality that @xmath151 is the oriented edge from @xmath111 to @xmath150 and that @xmath8 goes from @xmath51 to @xmath111 . let @xmath343 . then @xmath344 is a shortest length path from @xmath51 to @xmath345 . now @xmath346 , else we could write @xmath347 and get @xmath348 as a @xmath61-invariant forest closer to @xmath51 than @xmath107 is . ( if @xmath349 , then @xmath285 and so @xmath350 necessarily . ) so we have both @xmath8 and @xmath344 are paths from @xmath51 to @xmath111 moreover , @xmath351 ( else @xmath329 is not a forest . ) now @xmath352 is not a forest , since @xmath58 is @xmath320-reduced . hence we can choose some simple closed curve @xmath331 in @xmath353 that contains @xmath151 . there must exist some @xmath354 , @xmath355 , such that @xmath356 . otherwise we could deformation retract @xmath331 to the set of vertices @xmath357 , which contradicts the fact that @xmath331 is a simple closed curve . now @xmath358 , as @xmath359 . hence @xmath360 is a nontrivial cycle starting and ending at @xmath51 which contains exactly one occurence of @xmath151 and none of @xmath345 . so there must be a @xmath361 , @xmath362 , which contains an odd number of occurences of @xmath334 and an even number of occurences of @xmath363 . ( if we had some even / odd @xmath335 , then we could act by @xmath61 to get odd / even , and this would be a @xmath339 . otherwise , all @xmath327 are all even / even or odd / odd , and so combine together just to get more even / even or odd / odd loops . ) this is a contradiction , however , because @xmath364 still has an odd number of occurences of @xmath334 while @xmath365 has an even number of occurences of @xmath334 . from the action of @xmath366 on the rose @xmath52 such that the @xmath10th generator inverts the @xmath10th petal and leaves all others fixed , we know that the krull dimension at the prime 2 is at least @xmath4 . similarly , there is an action of @xmath367}$ ] on @xmath52 where the first @xmath368 rotates the first @xmath33 petals , the second @xmath368 rotates the next @xmath33 petals , etc . hence the krull dimension at the prime @xmath33 is at least @xmath35.$ ] let @xmath320 be a maximal rank elementary abelian subgroup of @xmath0 . from proposition [ c18 ] , we can realize @xmath320 by an action of @xmath320 on a @xmath61-graph @xmath58 which is both @xmath320-reduced and @xmath61-reduced . that is , we have an action of @xmath320 on a pointed @xmath369-tree @xmath58 . since elements of @xmath320 must preserve basepoints , the action of @xmath320 on the tree @xmath175 does not invert edges . hence we have inclusions @xmath370 the result ( for 2 or odd primes @xmath33 ) now follows from standard facts about @xmath17 ( cf . theorem 1.3 in chapter vi of @xcite . ) we sketch the proof , which uses standard methods . suppose that some @xmath371 lies in @xmath19 . from proposition [ c18 ] , we can realize @xmath320 by an action of @xmath320 on a @xmath61-graph @xmath58 which is both @xmath320-reduced and @xmath61-reduced . let @xmath164 be the corresponding marked graph . let @xmath372 , a pointed tree with an @xmath320-action on it . first , suppose that the @xmath320-action on @xmath149 is nontrivial . ( this will always be the case if @xmath33 is odd . ) then there are two edges @xmath151 and @xmath152 , both oriented so that their terminal vertices are closer to the basepoint than their initial vertices , of @xmath58 such that a generator of @xmath320 rotates the edge @xmath373 $ ] into the edge @xmath374 $ ] in @xmath149 . some generator @xmath375 of @xmath1 must be such that @xmath376 contains an odd number of occurences of @xmath151 in it . choose @xmath377 similarly . then @xmath378 is a palindromic word in the edges of @xmath58 with either @xmath379 or @xmath380 in the middle of the palindrome . the generator of @xmath320 ( thought of as an element of @xmath19 ) must send @xmath375 to a palindrome with either @xmath377 or @xmath381 in the center of it . this contradicts the fact that all elements of @xmath19 send generators @xmath53 to palidromes with @xmath53 in the center of them . the only remaining case is where @xmath382 and @xmath320 acts trivally on @xmath149 . so @xmath320 is a subgroup of the group @xmath366 of graph automorphisms of @xmath58 which act by inverting the @xmath369 s in the @xmath369-tree @xmath58 . hence the generator of @xmath320 corresponds to an element of @xmath0 which , for at least one @xmath10 , sends @xmath53 to a palindrome with @xmath142 in its center . as none of these automorphisms are in @xmath19 , we again have a contradiction . let @xmath33 be an odd prime ( as will always be the case from now on in this paper . ) we wish to calculate the farrell cohomology of @xmath0 using ken brown s @xcite normalizer spectral sequence , which states that @xmath383 where @xmath41 is a group with finite virtual cohomological dimension , @xmath384 is the poset of nontrivial elementary abelian @xmath33-subgroups of @xmath41 , @xmath385 is the poset of conjugacy classes of nontrivial elementary abelian @xmath33-subgroups of @xmath41 , and @xmath386 is the set of @xmath254-simplices in the realization @xmath387 . a first step toward performing such a calculation is calculating @xmath387 . in other words , we wish to calculate conjugacy classes of elementary abelian subgroups @xmath388 . by proposition [ c18 ] , we can realize such finite groups @xmath389 by reduced actions on @xmath50-trees . if @xmath390 , define a particular subgroup @xmath391 of @xmath0 by letting @xmath392 act on the rose @xmath52 by rotating its first @xmath33 leaves and leaving the last @xmath393 leaves fixed . that is , @xmath392 corresponds to automorphisms which rotate the first @xmath33 generators @xmath394 and leave the remaining generators fixed . we show that @xmath392 is the only conjugacy class of nontrivial elementary abelian @xmath33-subgroups that is in @xmath0 . by proposition [ c18 ] , we see that an arbitrary nontrivial elementary abelian @xmath33-subgroup @xmath320 comes from some action on a @xmath50-tree with @xmath33-symmetry . since @xmath395 , the only possibility is that @xmath320 acts on a @xmath50-tree @xmath58 by rotating @xmath33 of the @xmath50-leaves and leaving the other @xmath393 @xmath50-edges in the tree fixed . but it is clear that a product of @xmath397-nielsen transformations takes the rose @xmath52 to the graph @xmath58 , and hence we see that @xmath320 and @xmath392 are conjugate to each other in @xmath0 . the @xmath406 in the above decomposition comes from automorphisms of @xmath1 which permute the first @xmath33 generators and leave the remaining @xmath407 fixed . the @xmath403 being acted upon in the semidirect product structure above has @xmath10th generator @xmath408 . the @xmath404 is the involution which inverts the first @xmath33 generators of @xmath1 and leaves the remaining @xmath407 fixed . finally , the @xmath402 comes from automorphisms which fix the first @xmath33 generators of @xmath1 and act on the last @xmath407 generators by identifying the subgroup @xmath409 with @xmath403 . consider the action of @xmath410 on the rose @xmath52 . @xmath392 rotates the first @xmath33 petals . label the first @xmath33 petals of the rose as @xmath394 as before , but label the last @xmath407 petals as @xmath411 . since @xmath412 is an odd prime , @xmath413 and in lemma 5.1 of @xcite , we calculated @xmath414 where the first @xmath403 in @xmath415 is the free group on the @xmath392-nielsen transformations @xmath416 for @xmath417 and the latter @xmath403 is the free group on the @xmath392-nielsen transformations @xmath418 , @xmath417 . note that @xmath419 acts on @xmath415 via @xmath420 and @xmath421 . in other words , if @xmath422 then @xmath423 . let @xmath41 be the subgroup @xmath424 of @xmath425 , where @xmath403 is the free group on the generators @xmath426 for @xmath427 , and @xmath428 is included in @xmath429 in the obvious way . it follows directly that @xmath430 to prove the proposition , we must show that they are equal . take an arbitrary @xmath431 say @xmath432 , where @xmath433 , @xmath422 , @xmath434 , and @xmath435 . since @xmath436 , @xmath437 also . so @xmath438 and @xmath439 this means that the map @xmath440 sends @xmath53 to @xmath142 for @xmath441 and @xmath442 to @xmath443 for @xmath444 . now both @xmath61 and @xmath59 restrict to maps in @xmath445 and moreover @xmath446 and @xmath447 both restrict to the identity map in @xmath445 . hence for @xmath444 , we have @xmath448 and we see that @xmath449 restricts to @xmath405 in @xmath429 . as @xmath435 , this means @xmath450 . hence @xmath451 also . since @xmath452 , this gives @xmath453 . in other words , we have @xmath454 and @xmath455 it follows that @xmath456 so @xmath457 and @xmath458 . in summary , we have shown that an arbitrary element @xmath459 has @xmath460 and @xmath450 . thus @xmath461 , as desired . define a _ @xmath33-admissible tree _ @xmath149 to be a triple @xmath465 where @xmath149 is a pointed tree , @xmath466 is a vertex of @xmath149 ( which may be the basepoint @xmath51 ) , @xmath320 is a subset of the vertices of @xmath149 called the _ set of attaching points _ , @xmath467 , and all valence 1 vertices of @xmath149 are in @xmath320 . for a @xmath33-admissible tree @xmath149 , define the _ corresponding graph _ @xmath182 as follows : take two isomorphic copies @xmath178 and @xmath179 of the tree @xmath149 , and let @xmath180 be an isomorphism . then let @xmath468 be the graph @xmath469 let @xmath470 be a @xmath47-graph with @xmath33 edges and two vertices @xmath199 and @xmath200 . let @xmath471 be the @xmath466-vertex in @xmath178 and let @xmath472 be the @xmath466-vertex in @xmath179 . finally , let @xmath473 if @xmath474 , then say @xmath149 is a @xmath33-admissible tree of _ rank _ @xmath4 . if @xmath149 is a @xmath33-admissible tree of rank @xmath4 , define a @xmath56-action on the edges of @xmath182 by @xmath475 since this action inverts the edges of the @xmath47-graph in @xmath182 , we then need to subdivide these edges so that the group acts without inversions . next , define a @xmath392-action on @xmath182 by having @xmath392 fix @xmath468 and rotate the edges of @xmath470 cyclically . in this way , @xmath182 is a @xmath397-graph . a @xmath33-admissible tree is @xmath149 _ reduced _ if the corresponding @xmath397-graph @xmath182 is reduced ; that is , if all vertices of @xmath149 are attaching points . similarly , a @xmath33-admissible tree @xmath149 is a _ maximal _ if the attaching points of @xmath149 are exactly its valence 1 vertices , the valence @xmath74 vertices of @xmath149 consist of just the point @xmath466 , and @xmath149 has no vertices with valence 4 or more . as before , a _ subforest _ of @xmath149 is a collection of edges @xmath83 of @xmath149 such that there is no path in @xmath83 from one attaching point to another . lastly , isomorphisms of @xmath33-admissible trees must be graph isomorphisms which take @xmath51 to @xmath51 , @xmath466 to @xmath466 , and @xmath320 to @xmath320 . 1 . there is a bijective correspondence between reduced @xmath33- admissible trees of rank @xmath4 and the underlying graphs of @xmath397- reduced marked graphs , given by @xmath476 . 2 . there is a bijective correspondence between maximal @xmath33-admissible trees of rank @xmath4 and the underlying graphs of maximal essential marked @xmath397-graphs , given by @xmath476 . the virtual cohomological dimension of @xmath462 is @xmath477 . 4 . let @xmath58 be a graph which occurs as the underlying graph of a marked graph in @xmath463 . then there is only one possible @xmath61-action on @xmath58 . if two marked graphs in @xmath463 have underlying graphs which correspond to the same @xmath33-admissible tree , then they correspond to the same vertex in @xmath478 . that is , we can form the moduli space @xmath478 by looking only at the poset structure of the @xmath33-admissible trees corresponding to marked graphs in @xmath463 . the top dimensional cohomology class of @xmath478 , with coefficients in @xmath232 , vanishes . that is , @xmath479 . @xmath480 . note that ( 4 ) and ( 5 ) above are a little bit different from their analogs proposition [ c8 ] and corollary [ c9 ] . basically , the underlying graphs @xmath58 always have just one possible @xmath61-action , as before , but it is conceivable ( for example , if the graph contains two or more copies of @xmath470 inside it and we must decide which one @xmath392 rotates ) that it might have several possible @xmath392-actions . that is why we talk about @xmath33-admissible trees instead in ( 5 ) , since the vertex @xmath466 in the tree determines where the @xmath33 edges that @xmath392 rotates are located . we use the equivariant cohomology spectral sequence for @xmath462 acting on the contractible space @xmath463 with finite stabilizers and finite quotient @xmath478 . the equivariant cohomology spectral sequence for this action is @xmath482 \in \delta_{n}^r } \hat h^s(stab_{n_{aut(f_n)}(p_n \times \langle\sigma_n\rangle)}(\delta ) ; { \mathbb{z}}_{(p ) } ) \hfill \cr \hfill \rightarrow \hat h^{r+s}(n_{aut(f_n)}(p_n \times \langle\sigma_n\rangle ) ) ; { \mathbb{z}}_{(p ) } ) \cr}\ ] ] where @xmath252 $ ] ranges over the set @xmath483 of orbits of @xmath254-simplices @xmath255 in @xmath463 . from the decomposition @xmath484 we see that @xmath485 has @xmath33-rank 0 . since @xmath406 acts trivially on marked graphs in @xmath463 by permuting the edges of the @xmath47-graph attached at @xmath466 , it follows that for every simplex @xmath255 we have @xmath486 the @xmath487-page of the spectral sequence is @xmath267 in the rows where @xmath488 and a copy of the cellular cochain complex with @xmath368-coefficients of the @xmath489-dimensional complex @xmath478 in rows @xmath490 . it follows that the @xmath491-page has the form : @xmath492 hence we see that the spectral sequence converges at the @xmath491-page . * case 2 : * m=1 . then @xmath478 is a contractible 1-dimensional complex with 3 vertices and two edges . define the maximal @xmath33-admissible tree @xmath149 of rank @xmath4 to be the tree with three vertices @xmath51 , @xmath466 , @xmath111 and two edges @xmath151 , @xmath152 where @xmath151 goes from @xmath51 to @xmath466 and @xmath152 goes from @xmath466 to @xmath111 . the middle vertex of the 1-dimensional complex @xmath478 corresponds to the graph @xmath182 . the other two vertices and two edges @xmath478 correspond to the two possible ways that @xmath182 can be collapsed equivariantly . * case 3 : * m=2 . then @xmath478 is a 2-dimensional complex with 13 vertices , 28 edges , and 16 two - simplices . there are two maximal graphs in @xmath478 . simplices coming from the first graph are listed in figure [ hyper1 ] and simplices from the second graph are listed in figure [ hyper2 ] . in figures [ hyper1 ] and [ hyper2 ] , the maximal graphs are listed in the center . these maximal graphs can be collapsed in various ways , and these are listed around the periphery of the figures . in the graphs , a solid dot indicates the basepoint @xmath51 and the hollow dots represent attaching points @xmath466 for the @xmath47-graph @xmath470 . if there is only one hollow dot in a graph , both ends of the @xmath47-graph should be attached to that one vertex . upon identifying the boundaries of the simplices listed in figures [ hyper1 ] and [ hyper2 ] , we obtain the complex @xmath478 pictured in figure [ hyper3 ] . the complex is homeomorphic to the fletching of a dart , three half disks , all identified along a common line in their boundary . this complex is clearly contractible . d. j. collins , _ palindromic automorphisms of free groups _ , _ combinatorial and geometric group theory ( edinburgh , 1993 ) _ , london math . lecture note ser . , 204 , cambridge university press cambridge ( 1995 ) 63 - 72 .

we present a model for the radio outbursts of microquasars based on the assumption of quasi - continuous jet ejection . the jets are ` lit up ' by shock fronts traveling along the jets during outbursts . the shocks accelerate relativistic particles which emit the observed synchrotron radiation . the observed comparatively flat decay light curves combined with gradually steepening spectral slopes are explained by a superposition of the radiation of the aging relativistic particle population left behind by the shocks . this scenario is the low energy , time - resolved equivalent to the internal shock model for grbs . we show that this model predicts energy contents of the radiating plasma similar to the plasmon model . at the same time , the jet model relaxes the severe requirements on the central source in terms of the rate at which this energy must be supplied to the jet . observations of ` mini - bursts ' with flat spectral slopes and of infrared emission far from the source centre suggest two different states of jet ejections : ( i ) a ` mini - burst ' mode with relatively stable jet production and weak radio emission with flat spectra and ( ii ) an outburst mode with strong variations in the jet bulk velocities coupled with strong radio emission with steeper spectra . we also show that the continuous jets in microquasars should terminate in strong shocks and possibly inflate radio lobes similar to extragalactic jet sources . we investigate the possibility of testing the predictions of this model with resolved radio observations . finally , we suggest that doppler - shifted x - ray iron lines , and possibly h - alpha lines , may be emitted by the jet flow of microquasars if thermal instabilities analogous to those in ss433 exist in their jets .

astro-ph0001501

the number of stellar mass black hole candidates known to produce jets has increased considerably over recent years . while ss433 was for a long time thought to be a rather exotic object , advances in x - ray astronomy have dramatically increased the number of known galactic x - ray binaries ( van paradijs 1995 ) . at least nine of these sources subsequently showed evidence for the production of relativistically moving jets . mirabel & rodr ' iguez ( 1999 ) give an excellent review on the observational and theoretical status of these objects , the microquasars . the presence of relativistically moving material in these sources was discovered during radio observations . during times of strongly enhanced radio emission , which in the following we will refer to as radio outbursts , the emission region can often be resolved into at least two components . these components are observed to separate in opposite directions over a fews tens of days ( i.e. mirabel & rodrguez 1994 ) . the projected velocity of the component traveling on a trajectory towards the observer can exceed the speed of light , if its intrinsic velocity is large ( e.g. rybicki & lightman 1979 ) . in addition to the apparently superluminal nature of the jet components , their propagation has been observed to slow down only in one case ( xte j1748 - 288 , hjellming et al . this constant expansion speed led to the interpretation of practically ballistic trajectories of discrete plasmon ejections as explanation for the observed radio components ( e.g. mirabel & rodrguez 1999 ) . although suggested for the jets of ss433 ( hjellming & johnston 1988 ) , models with quasi - permanent jet production have received little attention in the case of microquasars . this is somewhat surprising given the large number of similarities they show with jet producing extragalactic sources like quasars ( hence the name microquasars ) and radio galaxies ( e.g. mirabel & rodrguez 1998 ) . in these cases there is little doubt that apart from some possible minor intermittency of the jet production mechanism for some sources ( i.e. reynolds & begelman 1997 ) the jet flow is practically continuous . in this paper we endeavour to close this gap by the development of a continuous jet model for microquasars to explain the radio outbursts observed in these sources . the model is based on the idea of internal shocks in jets ( rees 1978 ) which was successfully applied to gamma ray bursts ( grb ) ( rees & meszaros 1994 ) . in sect . [ sec : pj ? ] we briefly review the plasmon model and discuss possible improvements on this within the jet picture . we develop the treatment of the relativistic jet flow in sect . [ sec : dyn ] and discuss the evolution of the synchrotron emission resulting from the internal shock in sect . [ sec : syn ] . the model is then applied to the probably best studied radio outburst of any microquasar , the march 1994 event in grs 1915 + 105 ( mirabel & rodrguez 1994 ) in sect . [ sec : app ] . the properties of the jet of grs 1915 + 105 like its energy content are derived from the model in sect . [ sec : prop ] . in sect . [ sec : end ] we consider the implications of a continuous jet for the interaction of microquasars with their environment . finally , in sect . [ sec : test ] we summarise some observational consequences of our model which may be used to test the model with future observations . atoyan & aharonian ( 1999 ) developed a model which is intended to reconcile the idea of discrete magnetised plasmon ejections during radio outbursts with the observed lightcurve and spectral behaviour in the radio waveband . they find that a single population of relativistic particles accelerated at the time of the ejection of the plasmons can not explain the observations . the energy losses of the relativistic particles due to synchrotron radiation would lead to a sharp cut - off in the radio spectrum moving to lower frequencies as the plasmons expand and travel outwards . this is quite different from the observed rather gentle steepening of the spectrum . atoyan & aharonian ( 1999 ) also show that a continuous replenishment of relativistic particles to the plasmons alone can not solve this problem because in this case the spectral cut - off moves to higher frequencies with time . they therefore postulate that the relativistic particles in the plasmons during the march 1994 event in grs 1915 + 105 were continuously replenished , presumably by a shock at the side of the plasmons pointing towards the source centre , but also suffered energy dependent escape losses . to fit the observations the scenario proposed by atoyan & aharonian ( 1999 ) requires that the mean free path of the most energetic relativistic particles in the magnetised plasmons is comparable to or exceeds the physical dimensions of the plasmons . this implies that these particles travel through the plasmons producing synchrotron emission but then leave them without scattering once off irregularities in the magnetic field or other particles . this is difficult to reconcile with the requirement that in order to be accelerated to relativistic velocities in the shock regions the mean free path in these regions must be short to ensure many shock crossings . if the accelerating shocks are close to the plasmons , this then means that the properties of the plasma change dramatically over short distances . moreover , it is not clear in this scenario why the synchrotron emission is not completely dominated by the contribution of the shocks themselves . in this paper we propose a different scenario to explain the observed properties of the radio emission of microquasars during outbursts . this is based on the assumption that microquasars may produce continuous jets for a long time before the actual outburst occurs and may well do so permanently ( see also levinson & blandford 1996a , b ) . the outbursts in our model are then caused by two shocks traveling along these continuous jets which accelerate the required relativistic particles in situ . after the shock has passed a particular region in one of the jets , this region continues to contribute to the total emission until the cut - off in the specific spectrum of this region moves below the observing frequency . the jet components observed in microquasars are in general not resolved and so the measured flux is the integrated emission from all the jet regions passed by the shock which are still emitting at the relevant frequency . this implies that the observed spectrum is steeper than that of the jet region immediately behind the shock where radiation losses are still negligible . the variation of the strength of the magnetic field and of the number of relativistic particles accelerated by the shock along the jet then give rise to a slowly steepening radio spectrum . this effect was discussed in the case of the radio hot spots of powerful extragalactic radio sources by heavens & meisenheimer ( 1987 ) . our investigation is based on the jet model of blandford & rees ( 1974 ) and marscher & gear ( 1985 ) . a similar approach to explain the synchrotron self - compton emission of extragalactic jets was taken by ghisellini et al . they develop a numerical scheme to follow the evolution of the energy spectrum of the relativistic particles downstream of the jet shock taking into account radiative as well as adiabatic energy losses . they also include synchrotron self - absorption in this calculation . since we are mainly interested in the radio emission of the jets of microquasars on rather large scales ( @xmath0 m ) , we can neglect any compton scattering and absorption effects on the energy spectrum . in this case only adiabatic and synchrotron losses are important and we can use the analytic solution for the evolution of the energy spectrum of the relativistic electrons derived by kaiser et al . ( 1997 ) . the underlying physical processes of the model presented here are very similar to the internal shock model proposed as explanation for grb ( rees & meszaros 1994 ) . the same scenario has also been invoked to explain the x - ray and @xmath1-ray emission of extragalactic jets ( ghisellini 1999 ) . in the internal shock model the energy of the shock traveling along the jet is thought to be supplied by the collision of fast shells of jet material with slower ones ( rees 1978 ) . in the case of grb this energy is released practically instantaneously leading to the extremely short duration of the observed bursts of emission . in extragalactic jets the shell collision may take longer but the large distance to these objects makes it difficult to separate the contributions of multiple collision to the total emission . as outlined above , we argue in this paper that the jets of microquasars provide us with the possibility to observe the development of internal shocks in jets resolved both in space and time . we follow marscher & gear ( 1985 ) and ghisellini et al . ( 1985 ) in the assumption that all relevant physical quantities of the jet and the jet material are simple power law functions of the unprojected distance from the source centre , @xmath2 . the shock traveling along the jet will compress the jet material but we will assume that it does not change the behaviour of the physical quantities as a function of the distance from the source centre . the radius of the cross section of the jet is assumed to follow @xmath3 . for a freely expanding , conical jet @xmath4 . the energy density of the magnetic field as measured in the rest frame of the jet material is then given by @xmath5 , where energy flux conservation would require @xmath6 . here and in the following dashes denote quantities measured in the rest frame of the jet material moving at relativistic speeds while quantities measured in the frame of the observer are undashed . the two frames of reference , the rest frame of the observer and that of the shocked jet material , are defined such that the origins of both coincide when the radio outbursts starts , i.e. when the shocks are formed in the centre of the source and start traveling outwards . consider a section of the jet after the shock has passed through it . at time @xmath7 this section is located at @xmath8 in the rest frame of the observer . here , @xmath9 is the angle of the jet to the line of sight and @xmath2 is the unprojected distance of the jet section from the centre of the source , i.e. the origin of the observer s rest frame . @xmath10 is the deprojected velocity of the shock as measured in the rest frame of the observer and @xmath11 is the deprojected velocity of the shocked jet material in this frame in units of the speed of light . the expression in brackets in eq . ( [ simple ] ) takes account of the doppler shifted time measurements taken in the observer s frame caused by the receding ( @xmath12 ) or approaching ( @xmath13 ) component of the motion of the jet material ( e.g. rybicki & lightman 1979 ) . transforming @xmath7 and @xmath2 to the frame comoving with the jet material we find @xmath14 where @xmath15 is the lorentz factor corresponding to the velocity of the shocked jet material , @xmath16 . for the origin of the comoving rest frame , @xmath17 , we recover from eqs . ( [ tdash ] ) and ( [ rdash ] ) the well - known result @xmath18 with the usual relativistic doppler factor @xmath19 ^{-1}$ ] . since the shock moves along the jet , most of the observed emission is not produced at the origin of the comoving frame but at @xmath20 . suppose the section of the jet introduced above was passed by the shock at time @xmath21 as measured in the frame comoving with the shocked gas . since this section is at rest in this frame , it is subsequently located at @xmath22 , where @xmath23 is the velocity of the shock as measured in the shocked gas frame . using this and substituting @xmath2 from eq . ( [ rdash ] ) in eq . ( [ tdash ] ) yields @xmath24 this expression illustrates the fact that the emission we observe at a given time @xmath7 from different parts of the jet was not produced simultaneously in the frame of the shocked gas . we assume that only those parts of the jet contribute to the total emission which have been passed by the shock . the emission regions within the jet can therefore be labeled with their ` shock time ' , @xmath25 , and are observed at the intrinsic time @xmath26 given by eq . ( [ tcom ] ) . since @xmath27 , this implies @xmath28 . for convenience we introduce the ratio @xmath29 , where we have used eqs . ( [ tdash ] ) and ( [ rdash ] ) . we assume that the shock passing through the jet material accelerates a population of relativistic electrons and/or positrons . during the acceleration process and afterwards these relativistic particles are subject to energy losses due to the approximately adiabatic expansion of the jet material and synchrotron radiation . to determine the exact form of the energy spectrum of the relativistic particles in a given jet region the kinetic equation including acceleration and energy terms must be solved . heavens & meisenheimer ( 1987 ) present analytic and numerical solutions for some simplified cases . they find that the energy spectrum follows a power law with a high energy cut - off . the cut - off occurs at the energy for which energy gains due to shock acceleration balance the synchrotron energy losses ( e.g. drury 1983 ) . the cut - off becomes steeper further downstream from the shock . for simplicity we assume that the relativistic particles in a given jet region are initially accelerated at a time @xmath21 during a short time interval @xmath30 to a power law spectrum with a sharp high energy cut - off at @xmath31 . in terms of the number of particles this can be expressed by @xmath32 here @xmath33 is the rate at which relativistic particles are accelerated in the jet by the shock at time @xmath21 . the normalisation of this energy spectrum and the position of the high energy cut - off depend on the local conditions for diffusion in the jet ( e.g. drury 1983 ) . these are not straightforward to estimate and we therefore assume for simplicity that the initial high energy cut - off of the relativistic particles freshly accelerated at time @xmath21 is independent of @xmath34 . in the internal shock model for grb the shock is caused by the collision of shells of jet material moving at different velocities . for grb it is implicitly assumed that the collision energy is dissipated very close to instantaneously . in the case of microquasars the propagation of the shock is resolved in time . for the normalisation of eq . ( [ spec ] ) we therefore assume @xmath35 where @xmath36 is the position of the shock at time @xmath21 and @xmath37 a model parameter . this implies that the rate at which the collisional energy is dissipated and partly conferred to the relativistic particles is almost constant for @xmath38 and decreases exponentially at larger distances . the onset of the exponential behaviour then signifies the point at which almost all of the collisional energy has been dissipated and the shock starts to weaken significantly . this would coincide with the time at which the two colliding shells have practically merged into one . alternatively , in only intermittently active sources the exponential decrease may be caused by the shock , and therefore the fast shell causing the shock , reaching the end of the jet . after the passage of the shock the relativistic particles continue to loose energy . the rate of change of the lorentz factor of these particles due to the nearly adiabatic expansion of the jet is given by ( e.g. longair 1981 ) @xmath39 where @xmath40 is the volume of the jet region the particles are located in . we assume that the bulk velocity of the shocked jet material is constant and this implies that the jet is only expanding perpendicular to the jet axis , i.e. @xmath41 . changing variables from @xmath26 to @xmath42 then yields @xmath43 energy losses due to synchrotron radiation give @xmath44 where @xmath45 is the thompson cross section and @xmath46 the rest mass of an electron . by summing eqs . ( [ adlo ] ) and ( [ sylo ] ) and integrating we find the lorentz factor @xmath1 at time @xmath26 of those electrons which had a lorentz factor @xmath47 at time @xmath21 ( see also kaiser et al . 1997 ) @xmath48 with @xmath49\ ] ] and @xmath50 the number of relativistic particles with a lorentz factor in the range @xmath1 to @xmath51 in the jet region overtaken by the jet shock at time @xmath21 is therefore given by @xmath52 with @xmath53^{p-2 } \tau'(t'_{\rm s})^{2/3 a_1 ( p-1 ) } .\ ] ] note here that the high energy cut - off , @xmath54 , also evolves according to eq . ( [ evol ] ) . the synchrotron emission of the relativistic particles at time @xmath26 in the region of the jet overtaken by the shock at time @xmath21 is given by @xmath55 where @xmath56 is the synchrotron emission spectrum of a single electron with lorentz factor @xmath1 . here we assume that the magnetic field in the jet is tangled on scales smaller than the radius of the jet . this then implies that @xmath57 , where @xmath58 is one of the synchrotron integrals normalised to give @xmath59 ( e.g. shu 1991 ) . to get the total emission of the jet behind the jet shock we have to sum the contributions of all the regions labeled with their shock times @xmath21 within the jet . from eq . ( [ cspec ] ) we see that this implies integrating eq . ( [ syn ] ) over @xmath21 . this integration must be performed numerically . finally , to compare the model results with the observations we have to transform to the rest frame of the observer , @xmath60 . note here that the luminosity inferred from the observations at frequency @xmath61 was emitted in the gas rest frame at a frequency @xmath62 . as mentioned above , the steepening of the radio spectrum of the superluminal jet components observed during outbursts of microquasars is caused by the integrated emission from an extended region of the jet . the further away from the shock the emission is created in the jet , the lower the cut - off in the energy spectrum of the relativistic electrons will be . however , if the jet region contributing to the total emission is not too large the properties of the energy spectrum will not change dramatically within this region . in this case we may estimate the approximate location of the break in the radio spectrum beyond which the spectrum steepens significantly . most of the energy lost by relativistic particles is radiated at their critical frequency , @xmath63 , where @xmath64 is the larmor frequency . defining the break frequency , @xmath65 , as the critical frequency of the most energetic particles just behind the jet shock we get @xmath66 where @xmath67 is the elementary charge and @xmath68 is the magnetic permeability of the vacuum . the steepening of the radio spectrum of the observed outbursts of microquasars strongly suggest that the observing frequency , @xmath61 , is close to the break frequency , i.e. @xmath69 . therefore , if most of the observed emission comes from the region just behind the shock , we expect from eq . ( [ break ] ) that @xmath70 . this implies that @xmath71 and @xmath72 are not independent parameters of the model but that they are correlated . the number of parameters in the model outlined above is large . in order to reduce this number we assume that the jets in microquasars are freely expanding , i.e. @xmath73 , and that the flux of magnetic energy through the jet is conserved , i.e. @xmath74 . since we assume the bulk velocity of the jet material to be constant , this implies that the ratio of the kinetic energy and the energy of the magnetic field is constant as well . furthermore , we impose symmetry between the approaching and receding sides of the source in the sense that the model parameters describing the jet are the same on both sides . this may be a poor assumption as the jets of gro j1655 - 40 are observed to be asymmetric ( hjellming & rupen 1995 ) . the model then depends on five free parameters : the e - folding distance of the number of relativistic particles within the jet accelerated by the shock , @xmath37 , the bulk velocity of the shocked jet material , @xmath16 , the maximum lorentz factor up to which relativistic particles are initially accelerated , @xmath71 , the slope of the initial power law energy spectrum of these particles , @xmath75 , and the energy density of the magnetic field at @xmath76 , @xmath77 . the acceleration rate of relativistic particles at the normalisation radius , @xmath78 , is in principle also a free parameter . however , from eq . [ syn ] ) we note that it is only a multiplicative factor in the calculation of the total radio emission of the jet . we therefore use it to normalise the model in such a way that for a given set of model parameters @xmath79 is such that the difference between the model predictions and the observational data is smallest for this given set of parameters . many radio outbursts of a number of microquasars have been observed . but to constrain the model parameters in a meaningful way , we would ideally need radio observations at two or more frequencies which clearly resolve the approaching and the receding jet component . furthermore , the resolution should be sufficient to decide whether one of these jet components consists of multiple subcomponents , i.e. multiple shocks , the emission of which may be blended in observations of lower resolution . to date there are very few simultaneous multi - frequency observations which come even close to this ideal situation . the best studied radio outburst of any microquasar is still that of march 19th 1994 of grs 1915 + 105 ( mirabel & rodrguez 1994 ) . this is also the outburst studied by atoyan & aharonian ( 1999 ) . to test our model we will use the comparatively large data base accumulated during this event . the radio outburst of grs 1915 + 105 which occurred in march 1994 was one of the strongest recorded for this object . the flux density at 1.4 ghz exceeded 1 jy which is at least ten times higher than the radio flux in quiescence ( rodrguez et al . 1995 ) . the outburst was observed with the vla in a - array at 8.4 ghz during 7 epochs covering almost 42 days . except for the first of these , the two jet components were resolved at this frequency ( mirabel & rodrguez 1994 ) . for the first unresolved and one further epoch measurements with the vla are also available at 4.9 ghz and 15 ghz . in addition , grs 1915 + 105 was monitored during this time by the nancay telescope at 1.4 ghz and 3.3 ghz ( rodrguez et al . . there are 24 flux measurements at each frequency but the source is unresolved at these frequencies . after a reconfiguration of the vla four more observations of grs 1915 + 105 of lower resolution were obtained at 8.4 ghz in b - array ( rodrguez & mirabel 1999 ) . these measurements cover roughly another 28 days but there is a gap of about 40 days between the end of the a - array observations and the start of the b - array campaign . from the resolved vla observations at 8.4 ghz mirabel & rodrguez ( 1994 ) determined a velocity for the jet components of 0.92 c and an angle of the jets to the line of sight of 70@xmath80 . this assumes that the approaching and the receding component travel at the same velocity in opposite directions . furthermore , it is assumed that grs 1915 + 105 is located 12.5 kpc away from us . fender et al . ( 1999 ) observed another radio outburst of grs 1915 + 105 in october 1997 and found a higher intrinsic velocity which is inconsistent with a distance of 12.5 kpc . they argue that the most likely distance for this object is 11 kpc which then implies that the velocity of the jet components in march 1994 was 0.86 c and the angle of the jets to the line of sight is 68@xmath80 . in the following we will adopt these later values . extrapolating back the trajectories of the two jet components mirabel & rodrguez ( 1994 ) find that the outburst started at 20 hours on march 19 . figs . [ fig : res8.4 ] and [ fig : res1.4 ] show all available flux density measurements as a function of time . rodrguez & mirabel ( 1999 ) note that another outburst of grs 1915 + 105 occurred on april 21 . the jet components of this new outburst are clearly visible as an unresolved emission peak coincident with the source centre in the vla radio map of epoch 6 . the approaching component of this new outburst is also distinctly visible on the map of the following observing epoch while the receding component is probably blended with that of the previous outburst of march 19 . the signature of this later outburst as a sudden increase of the radio flux is not very distinct at 1.4 ghz and 3.3 ghz . even at 8.4 ghz the situation in terms of the total flux density is somewhat unclear . however , the outburst can be easily identified as a separate event from the one of march 19 because the vla maps reveal an emission peak distinct from those of the earlier outburst . the later observation epochs at the vla with lower resolution detect the jet components of the april 21 burst while the components of the march 19 event were not detected ( rodrguez & mirabel 1999 ) . this is rather puzzling as the extrapolation of their lightcurves from the earlier observations indicate that they should still have been visible during these later observations . a close examination of the total radio flux at 8.4 ghz reveals another sudden increase around observation epoch 4 ( april 9 , fig . [ fig : res8.4 ] ) . this increase is also detected at 1.4 ghz and 3.3 ghz by the nancay telescope on april 5 ( rodrguez et al . 1995 , fig . [ fig : res1.4 ] ) . atoyan & aharonian ( 1999 ) point out that this may be caused by a sudden additional injection of fresh relativistic electrons in the ` blobs ' of gas they consider in their model . an alternative interpretation , which we will adopt here is that another , smaller outburst occurred shortly after april 4 . in this case , the radio emission caused by the propagation of a new shock down the jet is most likely blended with that of the earlier event of march 19 because of the limited resolution of the observations . if the smaller outburst occurred at april 5 , 0:0 hours , and assuming that the apparent shock velocity of this outburst is equal to that of the march 19 event , i.e. 17.5 mas day@xmath81 ( rodrguez & mirabel 1999 ) , then the distance of the shock from the source centre on the approaching jet side at observing epoch 4 would be roughly 0.1 `` . the position of the emission peak on april 9 is given as 0.36 '' from the source centre ( mirabel & rodrguez 1994 ) . the resolution of the vla in a - array at 8.4 ghz is @xmath820.3 " and this means that a secondary peak caused by a later , somewhat weaker outburst as proposed here could not be detected as an individual structure . the situation on the receding jet side with its lower expansion velocities is even worse . however , the radio emission of the additional outburst would contribute to the total radio flux of the source and we believe that this has been detected here . in summary , we will assume that the observations outlined above cover three separate radio outbursts of grs 1915 + 105 . the first and strongest occurred on march 19 . the jet components of this burst are detected until the end of the first set of vla observations ( april 30 ) . by the time of the second vla campaign ( june 13 ) they had vanished although the extrapolation of their earlier lightcurves suggested that they should still be observable . the second much weaker outburst occurred shortly after april 4 , since the nancay data show a sudden increase in radio flux on april 5 but not on april 4 . the radio emission of this event is most likely blended with that of the first outburst and there is no sign of this second burst in the second set of vla observations . the third outburst finally started on april 21 and was intermediate in strength . the radio emission caused by this event is clearly detected in the source centre on april 23 and the approaching jet component can be seen in the vla map obtained on april 30 . both jet components are clearly detected in all four vla observing epochs in june and july . in the continuous jet model for microquasars outlined above , only one shock is thought to travel outwards in each jet . because of this , only observational data from observing epochs during which we can be sure that there was only one shock per jet contributing to the radio emission can be used in constraining the model parameters . from the above discussion it is clear that we can only use the three unresolved vla measurements at 4.9 ghz , 8.4 ghz and 14.9 ghz from march 24 and the two following resolved observations at 8.4 ghz . these later observation provide us with separate flux measurements for the approaching and receding jet components . we also use the measurement of the receding jet component of april 9 , since it seems likely that all the flux of the second outburst was attributed to the approaching component by mirabel & rodrguez ( 1994 ) . alternatively , this measurement can be taken as an upper limit . of the nancay observations we use all flux measurements starting march 24 through to april 4 . there are eight observations during this time at 1.4 ghz and 3.3 ghz . for all 24 measurements used to constrain the model we assume the conservative error of 46% suggested by rodriguez & mirabel ( 1999 ) as opposed to the original error of 5% quoted by rodrguez et al . ( 1995 ) . in practice we did not use the eight 3.3 ghz data points as their inclusion led to significantly worse fits of the model to the observational data . see the next section for a discussion of this point . we use the model outlined above to calculate the expected radio flux at the times grs 1915 + 105 was observed during the outburst starting march 19 . the ` goodness of fit ' of the model for a given set of free parameters to the observations was assessed by calculating the sum of the @xmath83-differences at the times the source was observed . the best - fitting model was then found by minimising this @xmath84-value using a 4-dimensional downhill simplex method ( press et al . the remaining fifth model parameter , the energy density of the magnetic field at the normalisation radius , @xmath72 , was set ` by hand ' to five different values . the normalisation radius , @xmath76 , was set to @xmath85 m which corresponds to the unprojected distance of the jet shock from the source centre at the time of the first vla observation , i.e. march 24 . the results of the model fits are summarised in table [ tab : res ] . the model fits the observational data equally well for all five adopted values for the strength of the magnetic field . as expected from eq . ( [ break ] ) we find that the maximum lorentz factor up to which relativistic particles are accelerated correlates strongly with the value of the energy density of the magnetic field . the values found for @xmath71 in the model fits follow almost exactly a @xmath86 law . this shows that the model presented here can not be used to constrain the strength of the magnetic field within the jet independently of the maximum energy of the relativistic particles . to proceed we adopt in the following @xmath87 as our fiducial model . this corresponds to the equipartition value of the magnetic field of @xmath88 t ( 0.1 g ) found by atoyan & aharonian ( 1999 ) for the chosen value of @xmath76 . note however , that the energy density of the magnetic field and that of the relativistic particles does not stay in equipartition for all times in our model . cccccccc & @xmath89 & @xmath37 & @xmath16 / c & @xmath90 & @xmath75 & @xmath84 & @xmath78 + + & @xmath91 & 2.8 & 0.61 & 3.2 & 1.4 & 0.26 & @xmath92 + & @xmath93 & 2.7 & 0.61 & 2.9 & 1.4 & 0.26 & @xmath94 + without 3.3 ghz data & @xmath95 & @xmath96 & @xmath97 & @xmath98 & @xmath99 & 0.26 & @xmath100 + & @xmath101 & 2.7 & 0.61 & 2.4 & 1.4 & 0.26 & @xmath102 + & @xmath103 & 2.6 & 0.61 & 2.2 & 1.4 & 0.26 & @xmath104 + all data & @xmath95 & 4.2 & 0.56 & 2.8 & 1.9 & 1.02 & @xmath105 + figs . [ fig : res8.4 ] and [ fig : res1.4 ] show the model predictions of the fiducial model compared to the observational data . also shown are the predictions of the model for the best - fitting parameters when using also the 3.3 ghz data . if we use all available data , the model predictions at the higher observing frequencies are rather low at early times . for the vla measurements at 4.9 ghz and 14.9 ghz ( not shown in figs . [ fig : res8.4 ] and [ fig : res1.4 ] ) on march 24 the model predicts flux densities of 759 mjy and 385 mjy respectively . this is much lower than the measured flux densities of 887 mjy and 514 mjy at these frequencies . a closer inspection of fig . [ fig : res1.4 ] also shows that the measurement at 4.9 ghz actually exceeds those taken at 3.3 ghz at comparable times , which is unlikely to be real . table [ tab : res ] also shows that the fit obtained including the 3.3 ghz data is much worse in terms of the reduced @xmath84-values than that excluding them . also the flux densities predicted by our fiducial model which excludes the 3.3 ghz data , 926 mjy at 4.9 ghz and 499 mjy at 14.9 ghz , are much closer to the observations . finally , the predictions of this model at 3.3 ghz fit the observations well at this frequency apart from the early observing epochs ( see fig . [ fig : res1.4 ] ) . we therefore believe that the model parameters found using our fiducial model and excluding the 3.3 ghz data are more reliable than those found when including these additional measurements . to estimate the expected error of the model parameters of our fiducial model we calculated the @xmath84-value for a large set of combinations of the 4 free model parameters . the uncertainties quoted in table [ tab : res ] are 1-@xmath106 errors corresponding to those parameter ranges for which @xmath107 . note that the uncertainties of the model parameters , particularly those of @xmath37 and @xmath75 , are large while the light curves predicted by the model pass the data points well within the error bars of the flux measurements ( see figs . [ fig : res8.4 ] and [ fig : res1.4 ] ) . this suggests that the quoted errors of the observed fluxes , at least for the vla data points , are too conservative which also results in an overestimation of the uncertainties of the model parameters . the exponential function which describes the change in the acceleration rate of relativistic particles as a function of the position of the shock in the jet , eq . ( [ expo ] ) , implies that the radio flux caused by the first outburst on march 19 decreases quickly once @xmath108 exceeds @xmath37 . this effect is clearly visible in fig . [ fig : res8.4 ] at 8.4 ghz . the model predicts that without the additional blended radio emission caused by the second and third outbursts around april 5 and april 21 the two jet components would have faded much more rapidly than is observed . this effect is less pronounced at lower frequencies ( see fig . [ fig : res1.4 ] ) . however , even for these the predicted and the observed light curves steepen somewhat roughly 15 days after the start of the first outburst . the continued steepening of the lightcurves of the two jet components at 8.4 ghz can also explain why they were not detected during the second observing campaign at the vla ( see fig . [ fig : res8.4 ] ) . we tried replacing the exponential in eq . ( [ expo ] ) with a simple power law . the data can be adequately fitted with this modified model as well . however , this change in the temporal behaviour of @xmath78 also leads to a much increased flux at low observing frequencies at later times . using this modified model we found a flux at 1.4 ghz 40 days after the start of the first outburst exceeding the observations by a factor of at least 1.3 even without considering the possible contribution from later outbursts . this supports the picture of two colliding shells of jet material of finite width causing the internal shock . the rate at which energy is dissipated is roughly constant during the collision and decreases rapidly once the two shells have merged . comparing the model lightcurves with the observational data the signature of the second outburst starting around april 5 , about 16 days after the start of the first outburst , can clearly be detected at 8.4 ghz . the increase in the radio emission caused by the second event is less dramatic at 3.3 ghz and 1.4 ghz but can still be seen in fig . [ fig : res1.4 ] . the third outburst of april 21 , 32 days after the first burst , is seen as excess emission at 8.4 ghz and 3.3 ghz but is less obvious at 1.4 ghz . we note that in general the smooth lightcurve predicted by our model fits the vla observations much better than the flux measurements at 1.4 ghz and 3.3 ghz taken with the nancay telescope . this may imply larger errors for the low frequency data which hide to some extent the signatures of the second and third outburst which are much weaker than the first . using the model parameters of our fiducial model we now derive some of the physical properties of the jets of grs 1915 + 105 during the outburst of march 19 . the rate at which energy is transfered by the jet shock to the relativistic particle population at time @xmath21 is @xmath109 where we have assumed that only electrons and/or positrons are accelerated and that the initial energy spectrum of the relativistic particles extends down to @xmath110 . the rate at which energy is transported in the form of magnetic fields can be estimated by @xmath111 where @xmath9 is the opening angle of the conical jet and @xmath23 is the speed of the shock in the frame of the shocked jet material . for our fiducial model @xmath112 c. fender et al . ( 1999 ) find that during another outburst of grs 1915 + 105 in 1997 @xmath9 was smaller than @xmath113 . this would then imply an upper limit to @xmath114 of @xmath115 w. note that the strength of the magnetic field in the jet after the passage of the shock is used here . this estimate does not imply that the unshocked jet material carries a magnetic field of this strength . some or all of the magnetic field may be generated in the shock itself . finally , we can derive an lower limit for the bulk kinetic energy transported by the jet material . we know the number of relativistic light particles in the jet and so @xmath116 this is only a strict lower limit , since we do not know whether the jets also contain thermal material and/or protons . in the case that there is one proton for each relativistic electron we find that the numerical constant in eq . ( [ kinetic ] ) increases to @xmath115 . note that this then is identical to the energy carried in the form of magnetic fields for @xmath117 . the estimates for the energy transported along the jet in various forms presented above are lower by about a factor 10 than the estimates of fender et al . ( 1999 ) for the weaker outburst in 1997 . however , it should be noted that their estimates are based on the assumption that the radio emission is caused by two ` blobs ' of relativistic plasma which were ejected by the central source within about 12 hours . the continuous jet model presented here requires that the estimated energy supply to the jet is sustained by the central source for at least 42 days ; the length of the first observing campaign . this means that the total amount of energy produced by grs 1915 + 105 is predicted by our model to be at least an order of magnitude greater during the march 1994 outburst than it was in the case of discrete ejections assumed for the september 1997 event . these estimates illustrate that a continuous jet model can not decrease the total amount of energy needed for a given radio outburst but the rate at which this energy is produced is much lower than in a model assuming discrete ejection events . this is the case because much , if not most , of the energy needed to produce the radio emission observed is ` stored ' in the material of the continuous jet . this material was ejected by the central source during comparatively long period well before the process which led to the formation of the jet shock took place . only the acceleration of relativistic particles at the jet shock then ` lights up ' the jet and we are able to detect it . since all jet properties are assumed to scale with distance from the source centre in our conical jet , it is clear that at some early time in the outburst the jet material was opaque for radio emission because of synchrotron self - absorption . the absorption coefficient in the rest frame of the emitting gas is given by ( e.g. longair 1981 ) @xmath118 where in our notation @xmath119 and @xmath120 is of order unity . we only consider the region just behind the jet shock where the energy distribution of the relativistic particles is completely described by a power law of exponent @xmath75 . for a photon emitted at the centre of the jet the optical depth in the radial direction is then @xmath121 . for our fiducial model we then find that the jet material becomes transparent at 8.4 ghz roughly 2 hours after the start of the outburst when the shock has reached a distance of @xmath122 m from the source centre . mirabel et al . ( 1998 ) find that for the much weaker ` mini - bursts ' of grs 1915 + 105 the jets become transparent about 30 minutes after the start of the burst . bearing in mind that the mini - bursts may be quite different in their properties compared to the major outburst considered here , our value is therefore in good agreement with their findings . several groups have reported the detection of infrared emission from grs 1915 + 105 ( i.e. sams et al . 1996 , mirabel et al . 1996 , 1998 ) . in the case of the mini - bursts simultaneous flux measurements at radio frequencies and in the k - band are available at times of about 10 to 20 minutes after the start of the bursts ( mirabel et al . 1996 ) . because of the uncertainties in the dust corrections in the k - band towards grs 1915 + 105 it is difficult to estimate the spectral behaviour from radio to infrared wavelengths . however , for the mini - bursts flat spectra , @xmath123 , regardless of the exact magnitude of extinction are observed very early during the bursts ( mirabel et al . the slope of the initial energy distribution of the relativistic particles in combination with the rather low high - energy cut - off of this distribution we found for our fiducial model is inconsistent with such flat emission spectra . however , this model does predict an unobscured , optically thin infrared flux of about 14 mjy in the k - band for a time about 15 minutes after the start of the burst . this may be enough to be detected in future observations of large outbursts . the timing requirements for such an observations are however difficult to meet , since the predicted infrared flux very quickly becomes undetectable at only slightly later times . more puzzling is the detection of a resolved jet component with k - band flux of at least 1.8 mjy about 0.3 " away from the centre of grs 1915 + 105 by sams et al . the shock on the approaching side of our jet model would need 24 days to reach such a large distance from the source centre . by this time our fiducial model predicts no synchrotron emission in the k - band at all . we have estimated whether this infrared emission may be caused by radio photons which are inverse compton scattered to such high frequencies within the jet plasma . however , we find that this can not explain the observations since the density of the relativistic particles in the jet in our model is orders of magnitude too low . the observation of k - band emission far away from the core of grs 1915 + 105 and the flat spectral indices of the mini - bursts suggest that two different types of outbursts may occur in the jets of this source . the strong radio bursts like the one of march 1994 are caused by jet shocks which produce large numbers of relativistic particles with a steep energy distribution . the weaker mini - bursts involve shocks which accelerate less particles but produce a flatter energy distribution which may also extend to higher energies than in the stronger bursts . the ` mini - burst mode ' may correspond to a phase of relative stable jet production with only small variations in the bulk velocity of the jet material . such flat spectra extending to millimeter wavelengths , possibly coupled with the continuous ejection of a jet , have been observed in cygnus x-1 ( fender et al . the strong radio outbursts then probably mark phases of more violent changes in the central jet production mechanisms . fender ( 1999 ) points out that this proposed behaviour may also be reflected in the x - ray signature of the accretion disk . in any case , other sources of infrared emission in the close vicinity of the jets like dust illuminated by the disk and/or the jet may further complicate the situation ( mirabel et al . 1996 ) . to test the validity of the proposed scenario resolved observations of outbursts of grs 1915 + 105 and other galactic jet sources from radio to infrared frequencies would be necessary . the energy transported by the jets of microquasars is enormous . this energy will be continuously deposited at the end of the jets and may lead to significant radiation from this region depending on how it is dissipated . in the following we investigate the fate of the energy transported by the jets of microquasars as predicted by our model . the discussion is based on the work by leahy ( 1991 ) . in order for the jets to expand they have to accelerate the surrounding ism and push it aside . the velocity of the contact surface between the front end of the jet and the ism , @xmath124 , is given by balancing the momentum or ` thrust ' of the jet material with the ram pressure of the receding ism @xmath125 ^{-1 } , \label{balance}\end{aligned}\ ] ] where @xmath126 and @xmath127 are the adiabatic indices of the jet material and the ism respectively , @xmath128 is the internal mach number of the jet flow , @xmath129 is the mach number of the contact surface with respect to the sound speed in the ism and @xmath130 . here @xmath131 is the mass density of the jet material while @xmath132 is the density of the ism . this expression is strictly valid only for non - relativistic jet velocities . however , since the bulk velocity of the shocked jet material , @xmath16 , is only mildly relativistic in our fiducial model , @xmath133 , we take eq . ( [ balance ] ) to be a good approximation . note that the velocity of the jet material in front of the shock is even lower than @xmath16 . for @xmath134 the jet material does not decelerate strongly at the end of the jet . this implies that little of the kinetic energy transported by the jet is dissipated . even for large internal mach numbers it is then unlikely that a strong shock will develop in the jet flow close to the contact surface this occurs when the jet is overdense , i.e. @xmath135 and so the jet flow is close to being ballistic . for underdense jets , @xmath136 , the ratio @xmath137 can become considerably smaller than 1 . in this case a strong deceleration of the jet ensues and much of its kinetic energy is dissipated . for @xmath138 a strong shock will form and can act as a site of efficient acceleration of relativistic particles . examples for this are the powerful extragalactic radio sources of type frii ( fanaroff & riley 1974 ) with their very bright radio hot spots at the end of their jets . the diffuse radio lobes enveloping their jets are the remains of the shocked jet material left behind by the advancing contact surface . in the transonic regime , @xmath139 , only weak shocks may form at the jet end and particle acceleration is less efficient . the less powerful jets of fri objects fall in this class . in the model developed in the previous sections we have assumed the jets of microquasars to be conical with a constant opening angle . this implies @xmath140 and , because of the adiabatic expansion of the jet material , @xmath141 . the bulk velocity of the jet material is high in our fiducial model and unless the jet material is very hot ( @xmath142 k in the case of a proton - electron jet ) the internal mach number of the jet flow will always greatly exceed 1 . since @xmath143 is a strongly decreasing function of @xmath2 , we expect from eq . ( [ balance ] ) that the ratio @xmath144 will always fall significantly below unity for large values of @xmath2 . this means that the jets of microquasars should end eventually in strong shocks which may be detectable in the radio . in the source xte j1748 - 288 a region of bright radio emission was observed to slow down and brighten at the same time some distance from the centre of the source ( hjellming et al . 1999 ) . in our model this is interpreted as an internal shock reaching the end of the jet where the termination shock further boosts the relativistic particle population which was pre - accelerated by the internal shock . after passing through the termination shock the jet material may inflate a radio lobe very similar to extragalactic frii objects if @xmath145 ( see also levinson & blandford 1996a , b ) . it has been suggested that the diffuse radio emission region w50 around ss433 is the radio lobe inflated by the jets of this source ( begelman et al . other radio lobes were detected around 1e 1740.7 - 2942 ( mirabel et al . 1992 ) , grs 1758 - 258 ( rodrguez et al . 1992 ) and possibly gro j1655 - 40 ( hunstead et al . 1997 ) , but not in the vicinity of grs 1915 + 105 ( rodrguez & mirabel 1998 ) . the absence of a radio lobe in grs 1915 + 105 may indicate that the jet in this source is relatively young and has not yet reached the point at which it becomes underdense with respect to the ism . in the following we estimate the distance out to which the jets in this source may travel without the formation of a strong termination shock . a lower limit for @xmath143 can be derived from our fiducial model assuming that the jets consist only of the relativistic particles responsible for the synchrotron emission plus the particles needed for charge neutrality . thus @xmath146 where @xmath147 is the mass of the average particle in the jet . an upper limit for @xmath143 can be derived from the assumption that the rate at which mass is ejected along the jet can not exceed the mass accretion rate within the disk powering the jet . fits to the x - ray spectrum of grs 1915 + 105 suggest an accretion rate of order @xmath148 kg s@xmath81 ( belloni et al . we then find @xmath149 note that this upper limit does not depend on the nature of the jet material . using eqs . ( [ lower ] ) and ( [ upper ] ) and assuming @xmath150 kg m@xmath151 , corresponding to a particle density of 1 cm@xmath151 , we find @xmath153 for a proton - electron jet and @xmath154 for a pair plasma jet . the lower limit for the pair plasma jet assumes that the pairs are cold . because of pair annihilation it is unlikely that the material of a pair plasma jet is cold ( e.g. gliozzi et al . 1999 ) and so this lower limit is used here for illustrative purposes only . relativistic thermal motion of the pairs would raise this lower limit . for the reasonable assumptions @xmath155 and @xmath138 fig . [ fig : balance ] shows the ratio @xmath156 as calculated from eq . ( [ balance ] ) . we see that even if the mass transport rate of the jet is equal to the mass accretion rate a termination shock should form about @xmath157 m away from the core of the source . this distance is reached by the jet material traveling at @xmath158 c in less than two years . grs 1915 + 105 was discovered as a bright x - ray source on the 15th of august 1992 ( castro - tirado et al . 1992 ) . given the availability of x - ray monitoring satellites before 1992 it is not likely that this source was very active before this date . the subsequent radio monitoring with the green bank interferometer ( e.g. foster et al . 1996 ) shows that after its discovery grs 1915 + 105 produced radio outbursts every few months . in the frame of the internal shock model described here this implies that jet production must have been reasonably steady since 1992 . assuming that the bulk velocity of the jet material did not vary strongly , the end of the jet must have reached a distance of roughly @xmath159 m from the core by the end of 1997 . this is the time of the radio observation of the large scale surroundings of grs 1915 + 105 by rodrguez & mirabel ( 1998 ) who did not find any evidence for a termination shock of the jet or radio lobes ( but see levinson & blandford 1996a ) . the estimation of the position of the termination shock depends crucially on the overdensity of the jet material with respect to the ism . it is possible that the gas density in the vicinity of grs 1915 + 105 is lower than assumed here . however , given its location in the galactic plane this is rather unlikely . a further possibility is that the jets of microquasars are not conical for their entire length . the jets of extragalactic frii objects are believed to pass through a very oblique reconfinement shock which brings them into pressure equilibrium with their environment ( e.g. falle 1991 ) . these shocks are not very efficient in accelerating relativistic particles and so are often undetectable . this scenario is also confirmed for frii sources by numerical simulations of their jets ( e.g. komissarov & falle 1998 ) . the same process may recollimate the jets of microquasars as well . in this case they may stay overdense with respect to the ism much longer and this would enable them to travel out to much larger distances before terminating in a strong shock . in this respect it is interesting to note that rodrguez & mirabel ( 1998 ) found a compact non - thermal emission region located 16.3 away from grs 1915 + 105 . the feature is elongated and its major axis is aligned with one of the jets . if this feature is caused by the jet pointing in its direction then it must have been ejected by the core roughly 280 years ago . this may be the time scale on which grs 1915 + 105 becomes active and produces jets . in our model the jet flow initially consists of non - relativistic hot plasma . due to adiabatic expansion losses of thermal energy during the propagation of this material along the conical jet with given opening angle , the temperature of the plasma decreases . internal shocks as envisioned above lead to local heating and acceleration of relativistic particles but do not change this general picture . in reality the situation is very similar to that in the well - known source ss433 . in ss433 we observe bright optical recombination balmer , paschen and brackett lines of hydrogen which are blue - shifted in the approaching jet and red - shifted in the receding jet . the velocity of the jet flow in ss433 is equal to 0.26 c. due to the precession of the jet the observed red and blue shifts are strong functions of time ( margon 1984 ) . in our case the inclination angle of the jets of grs 1915 + 105 is known from the observations by mirabel & rodrguez ( 1994 ) . this angle , @xmath160 , and the bulk velocity of 0.6 c of the jet material found in our fiducial model permits us to estimate the red and blue line shifts of the emitting jet material : @xmath161 note that any line emission coming from the jet approaching the observer is hardly shifted in wavelength at all . assuming that the bulk velocity of the jet material in the jets of gro j1655 - 40 is also close to 0.6 c , we find for this source that the emission lines are redshifted for the approaching jet ( @xmath162 ) as well as for the receding jet ( @xmath163 ) . this is caused by the large viewing angle , @xmath164 , of the jets in this object ( hjellming & rupen 1995 ) . in both cases the very large inclination angles result in a strong predicted asymmetry in the line shifts for the two jets . measuring these shifts will permit us to estimate both the velocity of the jet bulk flow and the viewing angle @xmath9 . furthermore , any jet precession as in the case of ss433 could be detected . measuring the predicted line shifts is complicated by the low density of the material in the jet flow of ss433 and grs 1915 + 105 which prevents the production of bright recombination lines . however , we know that in the case of ss433 there is a strong thermal instability in the flow which leads to the formation of small , dense cloudlets ( panferov & fabrika 1997 ) . this increases the recombination rate and effective emission measure of the plasma in the flow . if there is a similar instability in the jets of grs 1915 + 105 and gro j1655 - 40 we have a good chance to observe recombination lines from both of these sources . another problem is the strong obscuration of grs 1915 + 105 by interstellar dust . therefore , it is only possible to look for recombination lines of hydrogen in the k - band . in the case of gro j1655 - 40 obscuration is low and there is a chance to detect lyman and balmer lines . unfortunately , the mechanical power of the jet in gro j1655 - 40 is smaller than in grs 1915 + 105 or ss433 . this will lead to a smaller density and emission measure of the jet material and therefore also a smaller intensity of the lines . it is important to bear in mind that in ss433 the emission lines are extraordinarily bright and but modern observational techniques permit us to look for blue and red - shifted lines which are weaker by many orders of magnitude . in ss433 asca discovered red and blue - shifted x - ray k - lines of iron with a rest energy of roughly 6.7 kev and similar lines of hydrogen- and helium - like sulphur and argon ( kotani et al . 1997 ) . in our case the cooling jet flow with an initially very high temperature must lead to the emission in similar lines of recombining high - z ions . again , the mechanical energy of the flows in grs 1915 + 105 and gro j1655 - 40 is smaller than in ss433 and , therefore , the lines should be weaker in these objects . however , the new x - ray spacecraft , xmm , chandra , astro - e , constellation - x and xeus , may be able to detect such emission in red and blue - shifted x - ray lines . note in this respect the detection of shifted iron lines in gro j1655 - 40 reported by bauciska - church & church ( 2000 ) with rxte which the authors attribute to the accretion disk but may very well originate in the continuous jets of this source . all predictions for the production of line emission in the jets are based on the assumption that the jet flow consists of matter with a high but non - relativistic temperature moving as a whole with relativistic bulk velocities . there are two other obvious possibilities : ( i ) the jet matter consists of a pair plasma and ( ii ) the jets consist only of ultra - relativistic plasma with no cold electrons present . in case ( i ) we have to consider the possibility of a bubble around an x - ray source filled with a huge amount of positrons . if these positrons become non - relativistic due to adiabatic or other energy losses inside the jets and they cool down to sufficiently low temperatures , we may observe a blue and red - shifted recombination line of positronium in the optical and uv wavebands . this line has a wavelength twice that of the ly-@xmath165 line of hydrogen . much more important in this case , annihilation lines could be observed again red and blue - shifted relative to the rest energy of 511 kev . the strong red - shift but weak blue - shift of this line predicted by our model leaves a unique signature which will be observable with integral . a luminosity only a few times smaller than the mechanical power of the jets will be emitted in the electron - positron annihilation line in this case . this large luminosity should make the annihilation lines observable despite the unfavorable angle of the jets to our line of sight . in the case of only ultra - relativistic plasma in the jets , case ( ii ) , no recombination or annihilation lines should be observable . the internal shock models of grbs ( rees & meszaros 1994 ) attribute the formation of the shock traveling along the jet to the collision of shells of jet material with different bulk velocities . in the non - relativistic limit the velocity of the resulting shock is governed by the same momentum balance , eq . ( [ balance ] ) , as the velocity of the termination surface of the jet . * all quantities in that equation with subscript ` c ' now refer to the slower jet material in front of the jet shock while those with subscript ` j ' denote properties of the faster jet material driving the shock . the density ratio @xmath143 is now simply given by the densities of the faster jet material driving the shock , @xmath166 , and that of the slower gas in front of the shock , @xmath167 . we already pointed out in sect . [ sec : mom ] that the jet shock is likely to be strong and so both mach numbers in eq . [ balance ] are significantly greater than unity . therefore @xmath168 . since in our model all material is assumed to be part of the conical jet structure , we find @xmath169 and therefore @xmath170 . this implies that within the limitations of the model presented here the velocity of the shock is constant as well which is confirmed by the observations ( e.g. mirabel & rodrguez 1999 and references therein ) . * once the energy of the shell collision is spent , the shock emission fades rapidly . it is therefore possible that we can observe the shock reaching the end of the jet only in special cases ( xte j1748 - 288 , hjellming et al . 1999 ; see above ) . we would then expect that the superluminal component should brighten , as well as decelerate rather abruptly . in the plasmon model the observed constant superluminal motion is taken to indicate a large mass and consequently large kinetic energy of the plasmon . if a plasmon is observed to slow down because of the growing mass of ism it sweeps up , then this deceleration should be rather gradual unless the plasmon encounters a local overdensity in the ism . the observed deceleration of the superluminal component in xte j1748 - 288 occurred rather rapidly at a distance of about 1 `` from the core after a phase of expansion with practically constant velocity . furthermore , the emission region is still detected in recent observations ; 15 months after the start of the burst ( rupen , private communication ) . during this time it appears to have advanced only slowly at a velocity of about 0.01 '' per month or roughly 5000 km . this slow motion and persistent radio emission may be interpreted as arising from the shock at the end of a continuous jet ( see the previous section ) . some interesting predictions can be made from the model for future radio observations in the case that these can resolve the approaching and receding jet components along the jet axis . because of the way in which the rate of acceleration of relativistic particles in the jet by the shock varies with time , the peak of the radio emission is not coincident with the position of the shock . this off - set depends on the observing frequency in the sense that the lower this frequency the more the emission peak lags behind the leading shock . this is illustrated in fig . [ fig : movpeak ] where we plot the distance of the emission peak on the approaching jet side as a function of time for two different frequencies . note also that this effect predicts that we should measure slightly different advance velocities of the emission peaks at different frequencies . this will not be observed in the case of discrete plasmon ejections . the steepening of the radio spectrum of the jets in microquasars in this model is explained by the superposition of the contribution to the total emission from various regions within the jet . in resolved radio maps of the jet components this should be visible because the model predicts the radio spectral index to change along the jet axis . this behaviour is shown in fig . [ fig : spec ] for the approaching jet . * fig . [ fig : spec ] also shows the distribution of the flux along the jet axis . the relatively uniform distribution is caused by the decrease of the magnetic field strength further out along the jet counteracting the injection of newly accelerated particles by the shock . the jet region over which the spectrum steepens is small and the emission originating in this region also weakens considerably in the direction away from the jet shock . this may make a detection of the spectral steepening along the jet difficult . however , the apparent shortening of the emission region along the jet at higher observing frequencies may be detectable . * the decrease in the strength of the magnetic field combined with the high energy cut - off of the energy spectrum of the relativistic particles leads to an overall steepening with time of the radio spectrum along the jet axis . this is also shown in fig . [ fig : spec ] and should be observable if the jet components can be resolved at more than one frequency . note also that the length of the region along the jet axis which is emitting radiation at a given frequency increases with time . although the fraction of the distance of the shock from the source centre subtended by the emitting region shrinks for later times ( see fig . [ fig : spec ] ) , the absolute extent of this region will grow . this may also be detectable in future radio observations of sufficient surface brightness sensitivity . another prediction of the model is that the lightcurves of radio outbursts in microquasars at a given observing frequency should have a fairly constant slope for a few tens of days . after that they steepen rapidly once the critical frequency of the most energetic relativistic particles moves below the observing frequency . this steepening occurs earlier at higher frequencies . at the same time that the steepening of the lightcurves occurs , the flux ratio of the approaching and receding jet components should decrease . this effect is seen in fig . [ fig : res8.4 ] . the jet components of the second outburst of april 21 observed during the second observing campaign with the vla ( rodrguez & mirabel 1999 ) show a much steeper lightcurve than those of the first outburst and , at the same time , a smaller flux ratio . since these components were observed later in their evolution than those of the first outburst , this is in agreement with the predictions of the model . we developed a continuous jet model for the radio outbursts of galactic microquasars . the model naturally explains the observed rather flat decaying lightcurves of these bursts as the signature of synchrotron radiation of relativistic particles accelerated by internal shocks in the conical jets . the comparatively long duration of the bursts implies that this model is a ` time - resolved ' version of the internal shock model proposed for grb ( rees & meszaros 1994 ) , though the synchrotron emission is produced at much lower frequencies . the gradual steepening of the radio spectrum is explained by a superposition of the radiation of different populations of relativistic particles with different ages . this spectrum of ages results from the shock traveling along the jet with older populations of accelerated particles left behind . we find that only a roughly constant rate of acceleration of relativistic particles followed by an exponential decay can explain the observed light curves for the strong outburst of grs 1915 + 105 in 1994 . we interpret this behaviour as the signature of two colliding shells of jet material , as in the internal shock model for grb . a consequence of this is the continued steepening of the lightcurves of a given outburst coupled with a decreasing flux ratio of the emission observed from the approaching to that from the receding jet side . the energy requirements of the continuous jet model for producing radio outbursts are similar to those of the plasmon model . however , much of the energy underlying the outbursts may be stored in the continuous jet while the passage of the internal shock only ` lights up ' the jet . this implies that the rate at which the energy of the outburst is supplied by the central engine to the jet is much lower than in the plasmon model . the occurrence of mini - bursts in microquasars with flat spectra up to infrared frequencies ( mirabel et al . 1998 ) and the observation of k - band emission in the jet a considerable distance away from the core ( sams et al . 1996 ) suggest different modes of jet production : ( i ) a stable ` mini - burst ' mode with relative little variation in the bulk jet speed and therefore also only weak internal shocks . ( ii ) a more variable outburst mode with strong variations in the jet speed and strong internal shocks ( see also fender 1999 ) . the weaker flavour internal shocks seem to produce flatter relativistic particle spectra extending to higher energy compared to the strong shocks . however , the total number of accelerated particles must be much larger in the strongly variable phase . we show that the properties of the continuous jets of microquasars should lead to strong shocks at their ends where they are in contact with the surrounding ism . this is consistent with the recent observations of the decelerating radio emission region of xte j1748 - 288 ( hjellming et al . 1999 ) and its persistence for 15 months after the start of the original outburst ( rupen , private communication ) . the shocked jet material may subsequently inflate a low density cavity around the jets similar to the radio lobes in extragalactic jet sources of type frii . this is observed in ss433 ( dubner et al . 1998 ) and may be in a few other microquasars . the absence of such shocks and radio lobes in grs 1915 + 105 may indicate that the jets in this source are young and/or that they recollimate because of the pressure of their environment . if this is the case then the detection of a non - thermal emission region in the more extended environment of grs 1915 + 105 ( rodrguez & mirabel 1998 ) may imply a recurrence time of the jet activity scale of @xmath171 years . many of the predictions of this model for microquasars can be tested observationally . however , to clearly distinguish between this model of continuous jets and the plasmon model it would be necessary to spatially resolve the superluminal emission regions during outbursts , preferentially at more than one radio frequency . additional support for the scenario of continuous jets may come from further high resolution observations of the cores of microquasars during quiescence . these should show at least some spatial extension of the radio emission along the jet axis as observed in cygnus x-1 ( fender et al . these observations can potentially provide us with valuable information on the properties of the jets which otherwise we can only study during strong outbursts when strong shocks pass through them . the authors would like to thank g. ghisellini and the referee , l.f . rodrguez , for valuable discussions which improved the paper . this work was partly supported by ec grant erb - chrx - ct93 - 0329 within the research network ` accretion onto compact objects and proto - stars '

andreev reflection of quasiparticle excitations from quantized line vortices is reviewed in the isotropic b phase of superfluidhe in the temperature regime of ballistic quasiparticle transport at . the reflection from an array of rectilinear vortices in solid - body rotation is measured with a quasiparticle beam illuminating the array mainly in the orientation along the rotation axis . the result is in agreement with the calculated andreev reflection . the andreev signal is also used to analyze the spin down of the superfluid component after a sudden impulsive stop of rotation from an equilibrium vortex state . in a measuring setup where the rotating cylinder has a rough bottom surface , annihilation of the vortices proceeds via a leading rapid turbulent burst followed by a trailing slow laminar decay from which the mutual friction dissipation can be determined . in contrast to currently accepted theory , it is found to have a finite value in the zero temperature limit : . # 11exto 8.5cm0.3cm1ex = ~^ # 1*#1 * ) ] # 1_#1

1408.6145

andreev reflection @xcite , the celebrated phenomenon which alexander andreev introduced in 1964 to explain the increased resistance in heat flow through a normal metal superconductor interface , took three decades to be demonstrated in superfluid @xmath0he - b @xcite . nevertheless , during the more recent past it has become one of the prime tools to study the zero temperature limit , @xmath5 , of this charge - neutral p - wave fermion system . if the temperature is sufficiently low , so that collisions between quasiparticle excitations are practically absent in the bulk volume and ballistic propagation prevails , then measurements with vibrating sensors typically prominently display characteristic signatures from andreev reflection . rotation is another central research tool of superfluid @xmath0he . during the past decade it has been applied at ever lower temperatures so that andreev reflection measurements have become possible even in rotating flow . research on rotating @xmath0he superfluids has been centered in helsinki since the early 1980ies and has been vital for exploring the many different forms of quantized vorticity , both the structure as well as the dynamics . alexander andreev was one of the original founders of this research effort . a well - known example of andreev reflection , originally demonstrated with vibrating wire resonators by the lancaster group in 2001 @xcite , is shown in fig . [ andreevdemo ] . here two mechanical oscillators in close proximity to each other are vibrating in a bath of superfluid @xmath0he - b . one of them is driven at high displacement amplitude such that its strong vibrations generate a turbulent tangle of quantized vortices which shrouds both vibrators . the second vibrator is driven at low amplitude such that its output measures the damping of its oscillations by the thermal ( but ballistic ) quasiparticles . in spite of the heat input by the generator and the resulting increased overall number of excitations , the detector , shielded by the cloud of tangled vortices , displays reduced damping : it is hit by fewer quasiparticles from the surrounding cloud of thermal excitations and thus records a lower apparent temperature . this counterintuitive result is caused by the andreev retroreflection shadow cast by the vortex tangle . he - b at 0.20__t__@xmath6 . a drive with square wave envelope at high amplitude is fed to one of the forks which generates a turbulent vortex tangle . the tangle spreads around the two forks and casts an andreev shadow which is recorded as reduced damping by the second sensor fork driven at a one order of magnitude lower excitation amplitude . ] the example in fig . [ andreevdemo ] @xcite demonstrates the possibilities of using andreev reflection for the study of a clean fermi system in the very low temperature limit . this work and the development of the appropriate techniques has long been the domain of the lancaster group @xcite . more recently the ultra - low temperature regime of @xmath0he - b has moved in the forefront of general interest when it was realized that conventional sub - mk refrigeration techniques by means of adiabatic demagnetization cooling of copper are quite adequate if the total heat leak to the sample volume can be reduced to below @xmath7pw . this recognition has led to the development of new measuring techniques for the ballistic regime in superfluid @xmath0he which are based on the use of mechanical vibrating resonance devices @xcite or the existence of a novel coherently precessing nmr mode @xcite . such work has been driven by the hope to reveal explicit new information on andreev reflection , or on the existence of andreev bound states on surfaces , interfaces , and vortex cores , or the expectation to identify the majorana character in the spectrum of the bound state excitations @xcite . today the experimental tools for making use of andreev reflection consist of a quasiparticle radiator , a box with a heating element coupled to the @xmath0he - b bath via a small orifice which defines the beam , and a sensor in the bath , which traditionally has been a highly sensitive vibrating wire resonator . more recently the mass - produced quartz tuning fork oscillator @xcite has been found to have sufficient sensitivity as a quasiparticle detector and because of its easier use and insensitivity to magnetic fields it has gained in popularity . the theoretical basis of andreev reflection in superfluid @xmath0he , with emphasis on the phenomena arising from the spin - triplet and orbital p - wave pairing , was laid out in an early review of kurkijrvi and rainer . the present brief overview discusses measurements on andreev reflection in the rotating b phase , describing how andreev reflection can be used to study quantized vortex lines and their dynamics . this is an obvious area where andreev reflection measurements have great potential , especially concerning quantum turbulence , the peculiar characteristic of superfluid flow in the limit of weak mutual friction damping , which has been in the focus of recent interest @xcite . originally andreev reflection measurements in rotation became necessary as a means to calibrate reflection from quantized line vortices @xcite . this provides the starting point for the discussion below . the same experimental setup can also be used to record the dynamic response of a rotating vortex array . the most common type of such measurements is the determination of the response to a sudden stop of the rotation drive . based on our knowledge of superfluid flows in @xmath8he , it has been thought that laminar flow of quantized line vortices becomes unstable in the limit of vanishing mutual friction : when @xmath9 any minute perturbations in the flow can lead to instabilities and cause a tangle of vortices to be formed , so that new dissipation mechanisms become available , primarily fueled by reconnections between neighbouring vortices . andreev reflection provides one of the means to monitor the flow of vortices . recent measurements on the rate of vortex annihilation after a sudden reduction of the rotation velocity have shown that in @xmath0he - b turbulence is not necessarily the only form of response @xcite . depending on the geometry , surface properties , etc . , one finds that the rate of vortex decay in cylindrically symmetric flow may correspond to the slow mutual friction damping of laminar flow or to a faster process brought about by additional turbulent dissipation mechanisms . some of these studies will be discussed below . the theory of andreev reflection is particularly straightforward in the ballistic temperature regime , as has been demonstrated by barenghi et al . @xcite , who using hamiltonian mechanics calculate the trajectory of a quasiparticle scattered by a single rectilinear vortex ( fig . [ andreevscatter ] ) . . ] the quasiparticle moves with the kinetic energy @xmath10 with respect to the fermi energy @xmath11 , where @xmath12 is the linear quasiparticle momentum and @xmath13 its effective mass ( we refer here to the 29bar pressure of the @xmath0he liquid in the measurements of sec . [ results ] with @xmath14 , where @xmath15 is the bare mass of the @xmath0he atom ) . the superfluid circulation @xmath16 trapped around the vortex core corresponds to an azimuthally circulating superfluid flow with the velocity @xmath17 . in this flow field the energy of the quasiparticle is transformed to @xmath18 where @xmath19 is the energy gap in the zero temperature limit ( @xmath20 ) . the trajectory of the particle can be traced from the hamiltonian equations of motion @xmath21 ( @xmath22 , the group velocity of the excitations ) and @xmath23 . two length scales are involved : namely ( 1 ) describing the change in the order parameter amplitude at the vortex , the vortex core radius , measured in terms of the superfluid coherence length @xmath24 nm , while ( 2 ) the scattering process is characterized by an angular momentum @xmath25 ( where @xmath26 ) , which is a constant of the motion and defines the second length scale , the impact parameter @xmath27 . in the ballistic regime the quasiparticle energy spectrum resembles the roton minimum in superfluid @xmath8he , since @xmath28 an incoming quasiparticle with its energy above the minimum in the range @xmath29}\ ] ] is found to follow a usual straight trajectory past the vortex retaining its particle nature . for particle excitations this is the case on that side of the vortex where @xmath30 , while on the opposite side there are no allowed states for a particle in the range @xmath31}\ ] ] and it is found to retroreflect @xcite , _ ie . _ the particle - hole symmetry is broken and the excitation changes character ( and sign ) from a particle to a hole . it also almost retraces its trajectory of incidence , being deflected only by a very small angle @xmath32 the distance of closest approach to the vortex core , where the retroreflection occurs , is found to be @xmath33 and is thus inversely proportional to the kinetic energy @xmath34 of the incoming particle while the length scale is the geometric mean of @xmath35 and @xmath27 . the upper limit for the impact parameter , where retroreflection can still occur , is @xmath36 its magnitude is of order @xmath37 for thermal quasiparticles ( with @xmath38 ) and thus the andreev shadow of the vortex becomes experimentally significant , even though it is restricted to only one side of the vortex for one species of excitations . in practice rotating measurements proceed in the low - density limit where the andreev shadows of neighbouring vortices can be considered approximately additive . for denser 3-dimensional turbulent vortex structures the problem of tracing the excitation trajectories becomes complex @xcite , the total andreev shadow can not be estimated from the contribution of single vortices , and one has to resort to numerical monte - carlo - type simulation calculations . for quantitative experimental measurements a controlled setup with an oriented beam of quasiparticles and a well - known distribution of vortices is required . in the experiment of fig . [ andreevdemo ] , for instance , the usual assumption is that the vortex tangle is homogeneous and isotropic in all directions @xcite . this is a simplification , since the prongs of the tuning fork oscillator vibrate in antiphase in the plane in which they are contained . when the fork excitation is increased above a critical value , vortex rings are generated and shed off from the prongs . the rings propagate with their self - velocity @xmath39 , and because of the spread in their size distribution , they ultimately collide forming a tangle via reconnections @xcite . owing to the oriented motion of the prongs , the ejection pattern of the rings is not isotropic and thus the orientational homogeneity or the spatial extent of the tangle are not expected to be uniform @xcite . as andreev reflection depends on both the configuration and density of vortices with respect to the incident quasiparticle beam , a better controlled measurement is needed for calibration purposes . this can be achieved using the rotating equilibrium state in a cylindrical container ( fig . [ expscheme ] ) . here the configuration is fixed and the number of vortices @xmath40 can be externally adjusted by manipulating the angular rotation velocity @xmath41 , which controls the aerial density @xmath42 and thus the inter - vortex distance @xmath43 of the rectilinear line vortices in the rotating cylinder of radius @xmath44 . = 0 ) on the left , all excitations which are not scattered back through the orifice ( owing to diffuse scattering processes on the walls ) are thermalized in the heat exchanger at the bottom . in rotation ( @xmath45 ) on the right , part of the beam is andreev reflected from the equilibrium vortex state below the orifice and the density of excitations above the orifice is increased . ] in rotation metastability in the total number of vortices can arise from the presence of a non - negligible critical angular velocity increment @xmath46 . it determines the flow velocity at the cylindrical wall at which a vortex is formed , while rotation is increased : @xmath47 @xcite . this process controls the total number of vortices in increasing rotation , while the density is fixed to the equilibrium rotation value @xmath42 . in contrast , annihilation of vortices on the cylindrical container wall is not associated with any appreciable energy barrier @xcite . thus the reference state is preferably formed making use of the threshold to annihilation and not of vortex formation , which means that our rotating reference state is the equilibrium vortex state where the centrally located vortex cluster is surrounded and separated from the cylindrical wall by a vortex - free confining annulus of width @xmath48 which in practice is usually the minimum possible width @xcite and only slightly larger than the inter - vortex distance : @xmath49 . so far in the very low temperature measurements the inter - vortex distance is much larger , with @xmath50 mm , than the radius of the andreev shadow [ eq . ( [ r - c ] ) ] . the recipe to create the equilibrium vortex state with @xmath51 vortices is to increase rotation well above the target velocity and then to decrease @xmath41 so that vortices have annihilated when the desired value of @xmath41 is reached . these operations are preferably performed at high mutual friction above @xmath52 and subsequently the sample is cooled at constant rotation to the desired temperature . this precaution becomes necessary since after annihilation the remaining vortex array relaxes to its equilibrium configuration which at the lowest mutual friction values is a slow process requiring hours . in the setup of fig . [ setup ] , the rotating cylinder is compartmentalized with two division walls in three sections . the lower division wall with a small orifice of 0.3 mm diameter is the main thermal resistance along the cylindrical tower . it is used to measure the change in the quasiparticle density above the orifice as a function of @xmath51 . the upper large orifice of 0.75 mm diameter restricts the flow of vortices from the middle section to the uppermost compartment . in the uppermost section the critical velocity is defined by the smooth fused quartz walls of the cylinder so that it is characterized by a high value @xmath53 rad / s . it can be rotated in the metastable vortex - free state , the so - called landau state , up to @xmath54 . in contrast , the lowermost section has a rough bottom of sintered copper powder which reduces its critical velocity to @xmath55 rad / s . likewise , the middle section with the two quartz tuning fork oscillators and their leads has a reduced @xmath46 . thus after rotational deceleration , when vortices have been annihilated , one will find the equilibrium vortex state in the two lower sections , while in the uppermost section the vortex number can be adjusted to have any value between @xmath56 @xcite . in fig . [ setup ] , the heat flow carried by quasiparticle excitations moves along the tower towards the sintered heat exchanger which is always the coldest place , the heat sink . the dominant thermal resistance which it encounters is the orifice of 0.3 mm diameter in the lower division plate . it defines the quasiparticle beam for the andreev reflection measurement . when the heater fork is activated , a net heat current @xmath57 is carried through the orifice to the lowermost section . this current is produced by the residual heat leak @xmath58 and the heater power @xmath59 . in rotation the current is reduced by the andreev reflection from vortices @xmath51 in the lowermost section . the measurement of a carefully prepared reference state with @xmath51 vortices is started by demagnetizing the nuclear cooling stage until the thermometer above the lower orifice stops cooling at about @xmath60 , _ ie . _ when its temperature is fixed by the residual heat leak @xmath61 and the thermal resistance of the 0.3 mm aperture . thereafter the demagnetization cooling is continued at a much reduced rate , to maintain constant conditions . in this situation the temperature below the orifice is much lower , typically @xmath62 , as measured in the absence of the 0.3 mm orifice , owing to good thermal contact via the heat exchange sinter to the nuclear cooling stage . thus the reverse current of thermal excitations upwards from the lowest section can be neglected and the temperature rise above the orifice depends on the heating delivered by the heater fork and on the rotation velocity , which controls both the andreev reflection from the vortices below the orifice and the rotation - dependent heat leak to the two compartments above the orifice . by recording the temperature rise as a function of the power @xmath59 fed to the heater fork , it becomes possible to extract the thermal resistance at the given rotation velocity @xmath41 and by comparing resistances measured at different @xmath41 , ultimately the andreev reflection . the thermometer fork is selected to have small effective mass and a small intrinsic resonance width . in the ballistic temperature regime its damping at low excitation level measures the exponentially vanishing quasiparticle density . following the analysis of the vibrating wire thermometer , as introduced in ref . @xcite , the resonance width @xmath63 of the fork output is approximated as @xmath64 , where @xmath65 is the intrinsic zero temperature resonance width . in practice , in the regime of linear response with well - behaved lorentzian resonance line shape it is often sufficient to monitor the resonance amplitude , _ i.e. _ the in- and out - of - phase signals so that the resonance amplitude and frequency can be recovered , instead of performing complete frequency sweeps across the resonance to measure the width @xmath66 directly . the zero temperature width @xmath67 is typically measured in vacuum at the lowest possible temperature here at about 10mk . its value depends on the type of device , its preparation , mounting , and measuring circuitry , since at best @xmath68mhz which corresponds to a q value as high as @xmath69 or more . thus its measurement is a delicate matter , as well as its stability after changes in mounting or simply from one cool down to the next . in ref . @xcite @xmath67 was determined from in situ measurements of the relaxation rate of trapped magnon condensates as a function of temperature : the result @xmath70mhz was found to agree well with the low - temperature vacuum value of @xmath67 . in ref . @xcite a tuning fork of similar size and properties was deduced to have @xmath71hz by comparison against a vibrating wire resonator . this value was twice higher than 0.07hz measured in vacuum at 1.5k . for the temperature calibration one needs additionally one reliable temperature reading to fix the second calibration constant @xmath72 . accurate and at the same time convenient temperature calibration of the fork thermometer is not a straightforward task since a temperature reading from the ballistic regime requires a stable calibration point around @xmath73 or less , where the inevitable temperature difference between an outside thermometer and the @xmath0he sample can be large and difficult to estimate . here @xmath72 was estimated by extrapolating readings at around @xmath74 , by comparison to a @xmath0he melting pressure thermometer , which is thermally anchored to the copper nuclear cooling stage , or to b - phase nmr frequency shifts measured in the top @xmath0he compartment in fig . [ setup ] @xcite . in ref . @xcite the calibration was established by comparison to a simultaneously measured vibrating wire resonator . a 32khz fork with prongs of @xmath75mm@xmath76 cross section ( where @xmath77 mm is the width perpendicular to the direction of motion ) and 2.3 mm length was concluded to have a calibration constant @xmath78khz ( at zero pressure ) . one might think that other similar forks would faithfully display the same value , since it is expected to depend only on the density of quartz , on geometrical factors , and the fermi momentum @xmath79 . however , the matter appears to be more complicated : two different forks with similar dimensions and properties as that of ref . @xcite were used for thermometry in the present rotating measurements with @xmath80khz and 9.4khz at 0.5bar pressure , _ i.e. _ a sensitivity reduced by a factor of 4 . in ref . @xcite a larger fork with prongs of @xmath81mm@xmath76 cross section and 3.1 mm length was found to have sensitivity reduced by an order of magnitude from that expected according to the calibration recipe offered in ref . @xcite . in contrast , the pressure dependence appears to scale @xmath82 as expected , since the fork with @xmath80khz at 0.5bar was calibrated at 29bar pressure to have @xmath83khz . the ratio of these two values is close to the expected number 1.54 . in fig . [ tempresponse ] we see an example of the temperature response when the excitation of the heater fork is suddenly switched from zero to 6pw . the time constant is slow , about 25s , determined by the thermal rc time constant of the bolometer , and not by the one order of magnitude faster response which would correspond to the inverse of the resonance width of the thermometer fork the temperature rise settles at a value of roughly @xmath84k which is small compared to the total temperature difference across the orifice . and @xmath85 . ] assuming thermal equilibrium above the orifice and neglecting the orders of magnitude lower density of quasiparticles below the orifice , one may express the thermal balance across the orifice as @xmath86 here @xmath61 is the rotation dependent residual heat leak , introduced by mechanical vibrations and other uncontrolled sources to the two compartments above the orifice . @xmath59 is the heating power dissipated by the heater fork in the liquid , while @xmath87 is the heat flow carried by excitations through the orifice , which because of andreev retroreflection is assumed rotation dependent . a transparent derivation @xcite of @xmath57 starts from the kinetic theory of gases where the particle flux from the half space above a pin hole in a thin plate is given by @xmath88 here @xmath89 is the thermal average of the excitation density @xmath90 multiplied by the group velocity @xmath91 of the excitations , @xmath92 is their density of states , and @xmath93 the fermi distribution function , when @xmath94 . thus the particle flux represents an energy flow @xmath95 , where @xmath96 and @xmath97 is the effective area of the orifice . ( [ balance-1 ] ) can now be given in the form @xmath98 andreev reflection is thus measured in terms of the effective orifice area @xmath97 , which is reduced when excitations are retroreflected . the reduction is conveniently expressed in terms of a reflection coefficient @xmath99 , defined as @xmath100 using eq . ( [ balance-2 ] ) , results from measurements as a function of the heating power at different rotation velocities are plotted in fig . [ thermalcurrentdata ] . the data points in this plot are obtained by changing at fixed @xmath41 the applied power level from one value to the next and by averaging the corresponding equilibrium temperature readings for @xmath101 min at each power level . the intercept of the linear fit with the power axis gives the residual heat leak @xmath61 to the sample volume above the 0.3 mm orifice . it proves to vary from 12pw at @xmath85 to 18pw at @xmath1021.8rad / s . it should be pointed out that achieving pw - level heat leaks in a large superfluid @xmath0he sample housed within a massive rotating nuclear demagnetization cryostat is a major `` tour de force '' . a large part of the heat leak is caused by residual mechanical vibrations which are enhanced when the rotation velocity is increased , especially on approaching mechanical resonances at certain velocities or owing to the general loss of stability caused by rotational imbalance at high @xmath41 . the @xmath41 values used in the measurement had to be carefully selected , to be sufficiently spaced from any mechanical resonances . in addition at high @xmath41 values the rotation - induced heat leak fluctuated with variations of about 1pw . thus it is understandable that both the heat leak and the scatter of the data in fig . [ thermalcurrentdata ] increase with @xmath41 . ) as a function of the applied heating power at three different rotation velocities . ] the inverse of the slope of the lines in fig . [ thermalcurrentdata ] gives the effective area @xmath97 of the orifice . the measurement with no vortices gives @xmath103 mm@xmath76 which is less than half of the measured geometrical area of the orifice . a hole in an 0.7 mm thick division plate is not an ideal aperture and diffuse scattering of the excitations from the walls of this channel account for the difference . in any case , it is the relative change in eq . ( [ reflcoef ] ) which determines the andreev reflection which is plotted in terms of the reflection coefficient @xmath99 in fig . [ reflcoefvsomega ] as a function of the rotation velocity . the reflection coefficient increases with the total number of vortices @xmath104 approximately linearly , which would be the case if the andreev shadows are additive . unfortunately , the scatter is large and does not allow resolving whether in the measured arrays the andreev shadows start to overlap . the main difficulty is not believed to be mechanical vibrations but an unstable heater fork . in principle , if its power calibration would have been time independent , then the calibration would not have affected the measurement of @xmath99 . it is appropriate to remind that the middle compartment above the orifice is also in the equilibrium vortex state . these vortices cause changes in the quasiparticle trajectories above the orifice , but do not give rise to thermal gradients comparable to those created across the orifice , since the orifice is the dominant resistance . thus sufficient thermal equilibrium is preserved above the orifice also in rotation and the presence or absence of vortices in the upper sections of the tower has little effect on the andreev measurements . of quasiparticle excitations andreev reflected back upward through the orifice , as determined from the measured steady - state temperature increase above the orifice . the temperature is @xmath105 above the orifice and@xmath106 below . the statistical uncertainty in the value of @xmath107 is smaller than the size of the data points @xmath108 , but systematic error sources are larger , as discussed in the text . the measured data are in reasonable agreement with simulation calculations applying diffuse scattering from the cylindrical quartz wall . ] [ thermalcurrentdata ] and [ reflcoefvsomega ] are based on the measured steady state temperature increment when the heater fork is activated . consistency with the transient response in fig . [ tempresponse ] can be compared by noting that in the thermal time constant @xmath109 the resistance is dominated by the orifice , @xmath110^{-1 } \propto 1/a(\omega)$ ] , and the heat capacity by the specific heat of @xmath0he - b in the volume @xmath111 above the orifice @xcite , @xmath112 where @xmath113 is the density of states at the fermi level . using the @xmath114 value from fig . [ thermalcurrentdata ] gives a time constant of 32s , which is in reasonable agreement with the 25s value in fig . [ tempresponse ] . thus overall these measurements of andreev reflection and their analysis are believed to be consistent and to agree with expectations . the results in fig . [ reflcoefvsomega ] have also been checked in numerical calculations with monte carlo simulation of individual quasiparticle trajectories . using a more rigorous formulation , the flux through the orifice is obtained in the form @xmath115 where the transmission @xmath116 is set equal to one if an excitation above the orifice at position @xmath117 and moving in the direction @xmath118 reaches the sinter and zero if it is andreev reflected back . the integration is performed over the cross section of the orifice [ while @xmath119 , @xmath120 and @xmath121 . thus by sampling the trajectories individually by integrating eq . ( [ flux ] ) numerically and then solving equations ( [ balance-2 ] ) and ( [ reflcoef ] ) for @xmath107 , one collects results for the reflection coefficient . the simulations were calculated assuming either diffusive or specular quasiparticle scattering from the quartz - glass walls . as seen in fig . [ reflcoefvsomega ] , it is the former which are in better agreement with the measurements . in the presence of diffusive scattering a wall collision radically reduces the probability of such trajectories which would reflect the excitation back through the orifice . in contrast , with specular scattering from the quartz walls the reflection coefficient is substantially enhanced in the presence of vortices : an excitation scattering from the wall and subsequently andreev reflected by a vortex is sent right back through the orifice . the porous sintered heat exchanger surface is assumed to have zero reflectivity . overall we might conclude that the calculated data has little scatter , the deviation from linear dependence reflects partial screening of individual andreev shadows in the vortex array , as analyzed for 2-dimensional random vortex arrays in ref . @xcite , and the agreement of the measurements with the calculations using diffusive wall scattering and no adjustable fitting parameters is quite satisfactory . andreev reflection leads to variations in the local thermal quasiparticle density if changes occur in the surrounding vortex configuration . thus andreev reflection can be used to monitor vortex motions . a well - known example is the measurement of the free decay of turbulent vortex tangles , which have been created with a vibrating wire @xcite or a vibrating grid resonator @xcite in a quiescent bath of @xmath0he - b . in rotating flow it has been used to record the evolution of well characterized initial states of quantized vorticity in the presence of a time dependent rotation drive . so far , rotating measurements have concentrated around an elementary question , whether superfluid flow necessarily becomes turbulent in the zero temperature limit , when the bulk mutual friction dissipation @xmath9 ? this is the general conclusion which one derives from studies of flow in superfluid @xmath8he . however , the @xmath8he - ii vortex core diameter is of atomic size ( @xmath122 nm ) and on this length scale most surfaces tend to be rough so that strong surface pinning can generally be expected to influence dynamic measurements at low velocities and high surface friction to be present at higher velocities . in @xmath0he - b pinning is less conspicuous in fact , measurements so far in a cylindrically symmetric fused quartz container , which has been carefully screened with respect to imperfections , have not displayed clear implications from dissipative surface interactions . at high mutual friction dissipation @xmath123 at temperatures @xmath124 dynamic responses are laminar and turbulence is observed only at lower temperatures @xcite . still , the intriguing possibility remains whether laminar response might be stable in @xmath0he - b in the most ideal conditions at zero temperature . rad / s@xmath76 and @xmath125 is reached at @xmath126 . the rapidly developing large initial peak is generated when the rotating flow is destabilized by turbulence . the turbulent instability appears owing to the breakdown of the rotationally symmetric flow pattern , caused by dissipative interaction with the rough sintered bottom surface . the instability triggers a burst of reconnections , leading to an increase in total vortex length and the formation of a localized tangle near the bottom surface , which together remove about one third of the kinetic energy . the remaining later spin down is predominantly laminar . the two measurements characterize different situations for spin - down from @xmath127rad / s : a ) 0.5bar liquid @xmath0he pressure , measuring setup similar to that of fig . [ andreevdemo ] at @xmath128 ; b ) 29bar pressure , measuring setup of fig . [ setup ] where @xmath129 below the orifice and @xmath60 above . for more explanations about the comparison of these two fork outputs see sec . [ turbburst ] . ] this hypothesis has been tested in both nmr @xcite and andreev reflection measurements @xcite with similar conclusions . a straightforward well - defined measurement is obtained by recording the response to a sudden impulsive stop of rotation . initially the sample cylinder is rotated at constant rotation @xmath130 in the equilibrium vortex state . rotation is then suddenly stopped and the free decay of the rotating superflow is monitored while the cylinder itself is at rest , @xmath125 . in practice the bulky nuclear demagnetization cryostat of @xmath131 kg can not be stopped instantaneously without introducing additional heating . instead , the deceleration to zero rotation is done smoothly at the rate @xmath132rad / s@xmath76 . the point when the cryostat comes to rest at @xmath85 is here referred to as @xmath133 . in laminar decay the rotating vortex cluster consists of predominantly straight line vortices which move on a spiral trajectory outward until they annihilate on the cylindrical wall . the outward bound motion is damped only by mutual friction dissipation and therefore the slow laminar response can take hours at the lowest temperatures . if turbulent processes intervene , then additional dissipation mechanisms are coupled in and the decay speeds up . the motivation driving the research of superfluid dynamics has largely been the question what these mechanisms exactly are . it has been concluded from contactless nmr measurements on the top compartment of the sample cylinder in fig . [ setup ] that its spin - down response is fully laminar at least down to @xmath60 @xcite . it turns out that to obtain turbulence in such a cylinder , the presence of a highly dissipating surface is needed , such as the ab phase boundary separating one section of the cylinder filled with @xmath0he - b from another filled with @xmath0he - a @xcite , or a rough surface as in the inset of fig . [ andreevdemo ] @xcite , or deviations from cylindrically symmetric flow , such as the obstructions presented by the tuning forks themselves in the compartment in the middle of the cylinder in fig . [ setup ] . in such cases of weak turbulent perturbation , laminar flow is initially destabilized at higher vortex density and turbulent disturbances start to evolve : a burst of turbulent tangle formation and of accelerated decay follows , which efficiently reduces the kinetic energy in the early stages of the decay . the turbulent burst is the origin for the initial large peak in the tuning fork output in fig . [ turbulentsignal ] . this peak is short - lived , and as it subsides , the later response becomes more and more laminar . we describe the laminar signal next and postpone the discussion of the turbulent signal to later . of the superfluid component as a function of time after stopping the rotation of the cryostat from @xmath134rad / s . the measurement is fitted to the laminar response in eq . ( [ step ] ) : the dashed line represents @xmath135 and gives @xmath136 . the two insets show the oscillations of the fork output at the frequency @xmath137 at two different points in time in the laminar regime of the spin - down decay of fig . [ turbulentsignal ] . these oscillations of small amplitude arise from the periodic variation in the thermal excitation density around the fork owing to andreev reflection from the precessing vortex bundle with slight asymmetry . the measurement was performed at 0.5bar and @xmath138 in a setup similar to that of fig . [ andreevdemo ] . ] a clearly identifiable signature from laminar flow is present in fig . [ turbulentsignal ] in signal ( a ) . this signal records the resonance width of the thermometer fork as a function of time in a setup similar to that in the inset of fig . [ andreevdemo ] during the spin - down of the superfluid component . the signal displays a superimposed oscillating component of small amplitude which becomes visible on the falling slope of the rapidly evolving turbulent peak , but grows more prominent after the peak has died at @xmath139s . the oscillations are generated by andreev reflection from vortices which still are left over and rotate with respect to the stationary cylinder walls . the frequency of the oscillations @xmath137 monitors the rotation period of the azimuthally circulating flow which is created by this remaining rotating vortex cluster . an oscillating component arises if the precessing cluster deviates from perfect rotational symmetry , either with respect to its structure or its central confinement within the cylinder , so that andreev reflection from the periodically changing vortex configuration modulates the thermal excitation density in the neighbourhood of the tuning fork resonator . although the amplitude of the oscillating signal is small , its presence shows that the rotating bundle of vortices in free spin - down at @xmath85 does not possess perfect axial symmetry . for instance , a residual misalignment of the rotation and the sample cylinder axes of @xmath140 is expected . in simulation calculations with the vortex filament method @xcite such a small inclination has been concluded to be much below the large tilt angles needed to destabilize laminar flow , so that precession and laminar flow can coexist in the presence of weak breaking of axial symmetry . oscillating signals from the precession of a vortex bundle with rotational asymmetry have also been observed in nmr experiments @xcite . analysis of the angular rotation velocity @xmath137 from the oscillating signal component shows that it follows the laminar decay expected for the vortex density with solid - body density distribution @xmath141 in mutual - friction - damped spin - down of a vortex cluster composed of straight line vortices . its time dependence follows from the euler equation for inviscid rotationally symmetric flow @xcite @xmath142 . \label{de}\ ] ] assuming a step change of the rotation drive at @xmath126 from an angular velocity @xmath143 to @xmath125 , the solution of eq . ( [ de ] ) is given by @xmath144 where the time constant for the decay of laminar flow is @xmath145 . its value is typically a sizeable fraction of an hour below @xmath73 . thus the laminar decay of rotating superflow conserves the solid - body - like vortex line distribution so that both the density and the total vortex length decay @xmath146 . in contrast the kinetic energy of the rotating superfluid decays at a steeper rate @xmath147 , since the flow energy of the rotating superfluid with density @xmath148 is given by @xmath149 so that using eq . ( [ step ] ) the rate can be expressed as @xmath150 here @xmath44 and @xmath151 are the radius and the height of the rotating cylinder . returning to our example in fig . [ turbulentsignal ] , an excellent match between the measured @xmath137 and eq . ( [ step ] ) is obtained if one uses @xmath152 and @xmath153 as fitting parameters ( fig . [ laminarsignal ] ) . runs at different initial rotation velocities @xmath154 1.5rad / s and temperatures @xmath155 @xmath156 yield rather uniformly @xmath157 and as expected @xmath158 . this means that the relative change from @xmath159 to @xmath160 is not strongly dependent on @xmath159 or temperature : roughly one fifth of the vortices are removed in the initial turbulent burst , independently of the initial conditions , while the remaining vortices are predominantly straight , oriented along the rotation axis , and decay in a laminar fashion . their motion towards annihilation at the cylinder wall is resisted by mutual friction only , which here arises from the scattering of thermal excitations from the vortex - core - bound quasiparticles @xcite . in the limit @xmath5 , plotted as a function of the measured resonance width @xmath161 of the thermometer fork in the temperature range 0.14 @xmath60 . the value of @xmath162 at 0.5bar pressure has been worked out from the precession frequency in the fork width @xmath163 , as shown in fig . [ laminarsignal ] , by fitting the measured response to eq . ( [ step ] ) , and by averaging over 2 4 measurements at different rotation velocities in the range @xmath154 1.5rad / s at any given temperature . the 9.5bar data have been obtained from similar recordings of the nmr output . the uncertainty in the @xmath162 value is estimated to be of order @xmath164 . the vertical intercept @xmath165 appears not to vary with liquid @xmath0he pressure in the range 0 10bar . a finite value of the intercept is a common feature of many similar measurements of zero - temperature " extrapolations involving @xmath0he - b vortices , which now needs a solid explanation . ] in fig . [ andreevalpha ] the dissipative mutual friction coefficient @xmath166 , as derived from the fitted results , is shown as a function of temperature on an exponential scale , plotted in terms of the resonance width of the thermometer fork . a linear dependence is obtained , since both mutual friction and the fork width depend on the exponentially temperature dependent excitation density @xmath167 . the result is consistent with earlier measurements with a mechanical membrane resonator above @xmath168 @xcite if , in contrast to these authors , no renormalization is applied to the bulk - liquid gap @xmath169 , but instead it is allowed to obey the same temperature dependence as measured by the resonance width @xmath170 of the fork ( see ref . @xcite for details ) . the result is also consistent with nmr data down to @xmath60 at a high pressure of 29bar @xcite . the intriguing feature is the non - zero intercept @xmath165 in the @xmath171 limit . the source of this dissipation has not been unequivocally identified , but one explanation involves heating of the vortex - core - bound excitations when vortices are moving with nonzero acceleration , as suggested in ref.@xcite . another possibility could be surface interactions at the rough bottom wall . other measurements of different type ( see eg . ref.@xcite ) also point to the existence of new dissipation mechanisms involving vortices , when extrapolated to zero temperature on an exponential scale . the interesting conclusion from fig . [ andreevalpha ] is that mutual friction appears to have a lower bound on cooling towards the @xmath5 limit , at least if viewed on an exponentially developing temperature scale . this was not known earlier , but has now important implications on the dynamics of flows in @xmath0he - b . the rapidly evolving initial overshoot in the tuning fork damping in fig . [ turbulentsignal ] is the signature which we associate with turbulence . the overshoot is an immediate reaction to the deceleration , its maximum is generally reached at the moment when the cryostat rotation @xmath41 stops , while at high initial vortex density @xmath172 the signal may peak already during the deceleration . if the spin - down decay is laminar as in ref . @xcite , the large but short - lived overshoot is missing , while in the presence of a surface with high vortex friction , like the ab interface in ref . @xcite , a pronounced overshoot in total vortex length and overall dissipation is observed . thus we expect that the origin for the overshoot of the two signals in fig . [ turbulentsignal ] , which were recorded in somewhat different setups , is a turbulent instability caused by interactions of the vortices in the expanding vortex cluster with the rough sintered bottom surface . during rapid vortex decay , when the vortex ends move quickly over the sinter surface , the resulting motion resembles that of uniform surface friction on the ab interface . a second source for instabilities are deviations from cylindrical flow symmetry owing to the presence of tuning forks or their wires . it is the two turbulent responses in fig . [ turbulentsignal ] which we discuss next . as the tuning fork monitors the evolution of the local excitation density , other sources than the direct increased dissipation from turbulent vortex decay can also contribute to the signals . for instance , after deceleration the residual heat leak from mechanical vibrations is reduced by @xmath173 because of its @xmath41-dependence , as seen in fig . [ thermalcurrentdata ] . this reduction is little if at all visible in signal ( a ) , but explains about half of the drop from the initial to the final signal level in the case of ( b ) . the second half of the overall signal level change in ( b ) is explained by the andreev reflection from the initial equilibrium vortex state , which is not present in the final state at @xmath174 . to create the initial sharp peaks from sources other than the turbulence requires that the deceleration itself generates a pulse of heating . for instance , the rotation of the cryostat displays mechanical resonances with increased vibrational heating at certain fixed rotation velocities . however , these resonances require time to build up which is not available during the rapid deceleration . one can assure oneself about the correctness of such considerations by recording the spin - down signals at different deceleration rates . in short we conclude that the rapidly evolving peaks predominantly characterize the turbulence , but their quantitative analysis is both complicated and different for the two cases in fig . [ turbulentsignal ] . if the sinter surface is the main destabilizing effect , then the turbulence is spatially restricted to the vicinity of the bottom surface , as shown by simulation calculations of the free spin down at the ab interface @xcite . in a long cylinder the volume filled with turbulent tangle and displaying increased dissipation may be assumed to be of similar extension above the ab interface and above a rough sinter surface . however , as the two signals in fig . [ turbulentsignal ] have been measured in different setups , they display the turbulence differently , primarily since the thermal coupling to the heat sink differs by more than two orders in magnitude . the slower setup ( b ) was found to have a thermal time constant of 25s in fig . [ tempresponse ] . as seen in fig . [ turbulentsignal ] , the recovery from peak output is not obeying this time constant of the bolometer but takes place on a slower time scale which is determined by the turbulent decay . the bolometric design of setup ( b ) with the 0.3 mm diameter orifice emphasizes the increase in fork damping owing to andreev reflection from the turbulent tangle . thus this signal is more closely related to the time evolution of the density and polarization of the decaying vortices , in particular to the total line length deflected towards the plane transverse to the original rotation axis . since the overall polarization of the vortices changes continuously , while the tangle evolves , the andreev reflection signal has a complicated origin . in the more open geometry of setup ( a ) good thermal contact ensures rapid response @xcite and direct heat generation by the turbulence is the dominant source for this output . summarizing we note that the short burst of turbulence leads to an accelerated decay of vortices , when compared to the slow laminar decay which dominates the later spin down . the turbulent instability sets in when deceleration is started and causes vortex lines to reconnect and to form a turbulent tangle . this increases the heat release and the total vortex length . both features are monitored by the fork signals : ( 1 ) the peak in the reconnection rate and dissipation is reached early , already during the deceleration or immediately at its end , depending on the initial rotation @xmath172 . subsequently the turbulent dissipation decreases rapidly , as seen from signal ( a ) in fig . [ turbulentsignal ] . ( 2 ) the peak in turbulent vortex length develops slower as well as its subsequent decay , as monitored by signal ( b ) . the decay of the turbulence proceeds by means of a series of mechanisms . at the outer length scale of the flow disturbance and the turbulent instability , the kinetic energy resides in eddies formed from bundles of approximately aligned vortices . these decay to smaller eddies and bundles by reconnections so that the kinetic energy ends up cascading down the length scales with a kolmogorov spectrum @xmath175 where the kolmogorov constant @xmath176 . assuming that the kinetic energy is dissipated by some means at the length scale of the inter - vortex distance @xmath177 , where @xmath178 is the turbulent vortex density , the dissipation can be assumed on dimensional grounds to be of the form @xmath179 where @xmath180 is a phenomenological constant , an effective kinematic viscosity . equating the cascading energy flux @xmath181 from eq . ( [ kspec ] ) with eq . ( [ eps ] ) one obtains the time dependence of the turbulent vortex density @xmath182 where @xmath183 denotes the long - wave - length cutoff , the length scale of the flow disturbance at the wave vector @xmath184 . these arguments apply to the free decay of a homogeneous and isotropic vortex tangle . it would seem that spin - down in a cylinder with mostly laminar flow of the superfluid component has little to do with full - blown homogeneous turbulence , since reconnections and tangle formation are substantially reduced owing to the high degree of polarization of the vortices along the cylinder axis . surprisingly the recovery of both signals from peak damping in fig . [ turbulentsignal ] obeys the @xmath185 time dependence . to extract quantitative estimates proves problematic . depending on whether one assumes the vortices to be mainly parallel or perpendicular to the beam of excitations , one can work out upper or lower bounds on @xmath178 from signals of type ( b ) , based on numerical calculations of andreev reflection . this gives a factor of 3 difference in the estimate for the maximum vortex density created in the turbulent burst @xcite . more sophisticated evaluation of the turbulence in fig . [ turbulentsignal ] requires comparison to andreev reflection from numerically calculated simulations of the actual vortex line distribution in the given experimental settings , to which the measurements can be compared . moreover , the rate of the decay ( eq . ( [ dens ] ) ) in terms of one single fitting parameter @xmath186 has so far not yielded consistent results in the @xmath5 limit : andreev reflection measurements on the free decay of tangles created with a vibrating grid in @xmath0he - b suggest @xmath187 @xcite , while ion transmission measurements on the free spin down of superfluid @xmath8he in a cubic container with rough walls yield @xmath188 @xcite . the magnitude of this disagreement is uncomfortable , but reflects the fact that the measurements and their analyses are not compatible . to fit the decay rates in fig . [ turbulentsignal ] requires a further reduction of @xmath186 by two orders of magnitude or more , @xmath189 . it is to be expected that the homogeneous model of eq . ( [ dens ] ) can not be appropriate for the weak turbulence with nearly straight vortices in a rotating cylinder with a rough bottom . the analysis of measurements of the opposite case , of how vortices fill a long rotating cylinder at constant @xmath41 , show that the proper description of the precessing and propagating turbulent vortex front , which is then formed , requires two friction coefficients @xcite . the two coefficients are different by two orders of magnitude : a larger bulk friction describes energy dissipation while the smaller accounts for the slow removal of the angular momentum . thus in the presence of finite vortex polarization , which develops to varying degree in the above mentioned measurements , this decoupling effect , the large difference in decay rates , needs to be taken into account @xcite . clearly , for the signals in fig . [ turbulentsignal ] these considerations apply : a single energy - dissipation coefficient @xmath186 is not sufficient to explain the slow rate of angular momentum removal from flow which is only weakly perturbed from cylindrical symmetry . considering the above objections , the fact that the two signals in fig . [ turbulentsignal ] appear to follow the @xmath185 time dependence looks to be a coincidence and clearly not a measure of an effective kinematic viscosity . nevertheless , putting all evidence together , one might argue that a reasonably consistent qualitative interpretation emerges of the interplay of turbulent and laminar flows in @xmath0he - b . one particularly pertinent measurement would be a repetition of the experiment in fig . [ expscheme ] , but with the cylinder below the orifice replaced by a tube with a square or rectangular circumference and rough walls . in this environment the free spin - down after an impulsive stop of rotation might mimic more closely homogeneous isotropic turbulence and would perhaps provide a more reliable calibration of the andreev reflection signal from ideal superfluid turbulence , when started from a controlled vortex state . analogous experiments have been performed with superfluid @xmath8he @xcite , where turbulent responses dominate . a direct comparison with @xmath0he - b would be valuable , in order to compare to a situation when turbulence is less prevalent . the zero temperature limit of the fermi superfluid @xmath0he - b , with ballistic quasiparticle transport and no bulk - volume impurity scattering , is in the forefront of current fermion physics in condensed matter , since many new phenomena are here expected to come together . the most recent rush is to search for such alluring predictions as andreev surface states with excitations of majorana character . in this effort andreev reflection is one of the promising experimental tools . it took a long time before andreev reflection could be verified in @xmath0he - b , but now the practical devices for quasiparticle radiators and sensors have been developed . further improvement is expected , if arrays of quartz tuning forks @xcite or of resonators fabricated as micro - electro - mechanical systems ( mems ) become available as quasiparticle detectors . during the past decade it has been demonstrated that the zero temperature limit of @xmath0he - b can also be reached and explored in rotation . this advance promises to shed new light on the study of the excitations of the vortex core in this spin triplet orbital p - wave condensate . to search for new evidence one needs extreme low temperatures below @xmath190 and sensitive measuring techniques which function in this temperature range . so far andreev reflection in rotation has been employed to calibrate reflection from a well - controlled vortex state and to monitor the free spin - down of the superfluid component in cylindrical flow environments . this calibration measurement is the first and only one where the theory of andreev reflection from quantized vortices has been quantitatively tested . while the results agree with expectations , more measurements of other well - understood vortex configurations would be valuable . in nmr measurements the spin - down dynamics have been found to be laminar in ideal cylindrically symmetric flow conditions , which now is understood to result from the decoupling of energy and angular momentum dissipation . andreev reflection has been applied to measurements where the influence of perturbations is examined , such as the presence of a rough surface which interacts with the end of a moving vortex . the measured responses display an early phase of accelerated turbulent dissipation , with a modest overshoot caused by a burst of reconnecting and tangle formation , and a late phase of slow laminar decay at low vortex density . unlike in superfluid @xmath8he , where the vortex core diameter is of atomic size , in @xmath0he - b with two to three orders of magnitude larger core diameters pinning and surface friction are reduced , which helps to reduce dissipation and the duration of turbulence in the dynamic responses . this made it possible to record the angular rotation velocity of a vortex bundle in free spin down , when the damping of a tuning fork resonator is modulated by the variation in the local thermal excitation density , owing to a precessing non - axisymmetric andreev shadow cast by the bundle . this oscillating signal with slowly increasing periodicity gives directly the laminar vortex friction . it is found to have a small , but nonzero extrapolation at @xmath191 . this observation has implications on the vortex dynamics of @xmath0he - b . for instance , the current view about superfluid turbulence holds that the kinetic energy is transported from large length scales to ever smaller scales owing to a series of different mechanisms . on scales larger than the inter - vortex distance this cascade is structured in a similar manner as in classical turbulence of viscous fluids . on the smaller quantum length scales " it is believed that at very low mutual friction dissipation a kelvin wave cascade becomes possible when nonlinear interactions between kelvin waves allow the kinetic energy to propagate to smaller wave lengths along a single line vortex . such a hypothesis has not been rigorously confirmed experimentally . numerical simulations at very low mutual friction with sufficient spatial and temporal resolution are notoriously difficult and time consuming . however , recent numerical work @xcite concludes that a kelvin cascade might become possible in the regime @xmath192 . combining this result with fig . [ andreevalpha ] means that , unlike in superfluid @xmath8he , in the fermi superfluid @xmath0he - b with new sources of dissipation , the kelvin wave cascade may not be an important component in turbulent dissipation after all . note , however , that kelvin wave excitations as such , damped by mutual friction , are most important as a dissipation mechanism . whether the inference about the absence of the cascade in @xmath0he - b holds up in future research will be an interesting question . many central questions in vortex dynamics remain unanswered and require further researching . here rotation and quasiparticle beam techniques , perhaps combined with nmr measurement , will provide new possibilities for noninvasive vortex monitoring . _ * epilogue:*_rotating superfluid @xmath0he research is deeply indebted to sasha andreev who is one of its founding fathers . the first research programme ever to study @xmath0he superfluids in rotation was set into motion in 1978 by three academicians alexander andreev , elephter andronikashvili , and olli lounasmaa . this effort , known by the acronym rota , was supported by both the academy of finland and the soviet academy of sciences . it produced the first rotating nuclear demagnetization refrigerator which became operational in mid 1982 . since then the cryostat and its later modifications have been churning out research on quantized vorticity in @xmath0he superfluids applying many different measuring techniques . much of this work has been nmr based and here many generations of students and researchers from the kapitza institute for physical problems have been participating . sasha andreev had the role of an influential godfather , who advanced and promoted the research . in 2012 he was awarded the lounasmaa memorial prize for his achievements in research and its advancement . _ * acknowledgements*_:we thank our friend and colleague the late nikolai kopnin @xcite for promoting and advancing the research on fermion excitations in rotating @xmath0he superfluids . this work was financially supported by the aalto school of science , its research infrastructure low temperature laboratory , and the academy of finland ( centers of excellence programme project no . 250280 ) . andreev , zh . 99 * , 1823 ( 1964 ) [ sov . jetp * 99 * , 1228 ( 1964 ) ] . enrico , s.n . fisher , a.m. gunault , g.r . pickett , k. torizuka , phys . lett . * 70 * , 1846 ( 1993 ) . fisher , a.j . hale , a.m. gunault , g.r . pickett , phys . lett . * 86 * , 244 ( 2001 ) . m. blazkova , m. loveko , v.b . eltsov , r. de graaf , j.j . hosio , m. krusius , d. schmoranzer , w. schoepe , l. skrbek , p. skyba , r.e . solntsev , d.e . zmeev , j. low temp . phys . * 150 * , 525 ( 2008 ) . fisher , g.r . pickett , in _ prog . _ , vol xvi , ed . m. tsubota ( elsevier publ . , amsterdam 2009 ) . ahlstrom , d.i . bradley , s.n . fisher , a.m. gunault , e.a . guise , r.p . haley , s. holt , o. kolosov , p.v.e . mcclintock , g.r . pickett , m. poole , r. schanen , v. tsepelin , a.j . woods , j. low temp . phys . in print ( 2014 ) . heikkinen , s. autti , v.b . eltsov , r.p . haley , v.v . zavjalov , j. low temp . phys . * 175 * , 681 ( 2014 ) . y. okuda , r. nomura , j. phys : cond . mat . * 24 * , 343201 ( 2012 ) . r. blaauwgeers , m. blazkova , m. loveko , v.b . eltsov , e. gazo , r. de graaf , j.j . hosio , m. krusius , d. schmoranzer , w. schoepe , l. skrbek , p. skyba , r.e . solntsev , j. low temp . phys . * 146 * , 537 ( 2007 ) . j. kurkijrvi , d. rainer , in _ helium three _ , ed . by w.p . halperin and l.p . pitaevskii , elsevier science publishers b.v . ( amsterdam , 1990 ) . hosio , v.b . eltsov , m. krusius , j.t . mkinen , phys . b * 85 * , 224526 ( 2012 ) . j.j . hosio , v.b . eltsov , r. de graaf , m. krusius , j. mkinen , d. schmoranzer , phys . b * 84 * , 224501 ( 2011 ) . barenghi , y.a . sergeev , n. suramlishvili , phys . b * 77 * , 104512 ( 2008 ) . see illustrations in : g.r . pickett , j. theor . ( jetp ) , this issue . barenghi , y.a . sergeev , n. suramlishvili , p.j . van dijk , phys . b * 79 * , 024508 ( 2009 ) barenghi , y.a . sergeev , n. suramlishvili , p.j . van dijk , europhys . lett . * 90 * , 56003 ( 2010 ) . bradley , d.o . clubb , s.n . fisher , a.m. gunault , r.p . haley , c.j . matthews , g.r . pickett , v. tsepelin , k. zaki , phys . lett . * 96 * , 035301 ( 2006 ) . s. fujiyama , a. mitani , m. tsubota , d.i . bradley , s.n . fisher , a.m. gunault , r.p . haley , g.r . pickett , v. tsepelin , phys . b * 81 * , 180512 ( 2010 ) . a. nakatsuji , m. tsubota , and h. yano , phys . b * 89 * , 174520 ( 2014 ) . ruutu , . parts , j.h . koivuniemi , n.b . kopnin , m. krusius , j. low temp . phys . * 107 * , 93 ( 1997 ) . ruutu , j.j . ruohio , m. krusius , b. plaais , e.b . sonin , physica b * 255 * 27 ( 1998 ) . s.n . fisher , a.m. gunault , c.j . kennedy , g.r . pickett , phys . * 63 * , 2566 ( 1989 ) . heikkinen , s. autti , v.b . eltsov , j.j . hosio , m. krusius , v.v . zavjalov , j. low temp . phys . * 175 * , 3 ( 2014 ) . bradley , p. crookston , s.n . fisher , a. ganshin , a.m. gunault , r.p . haley , m.j . jackson , g.r . pickett , r. schanen , v. tsepelin , j. low temp . phys . * 157 * , 476 ( 2009 ) . r. de graaf , v.b . eltsov , p.j . heikkinen , j.j . hosio , m. krusius , j. low temp . phys . * 163 * , 238 ( 2011 ) . bradley , m. loveko , e. gao , p. skyba , j. low temp . phys . * 152 * , 147 ( 2008 ) . s.n . fisher , a.m. gunault , c.j . kennedy , g.r . pickett , phys . lett . * 69 * , 1073 ( 1992 ) . c. buerle , bunkov , s.n . fisher , h. godfrin , phys . b * 57 * , 14381 ( 1998 ) . finne , t. araki , r. blaauwgeers , v.b . eltsov , n.b . kopnin , m. krusius , l. skrbek , m. tsubota , g.e . volovik , nature * 424 * , 1022 ( 2003 ) . d.i . bradley , d.o . clubb , s.n . fisher , a.m. gunault , r.p . haley , c.j . matthews , g.r . pickett , v. tsepelin , k. zaki , phys . lett . * 95 * , 035302 ( 2005 ) . eltsov , r. de graaf , p.j . heikkinen , j.j . hosio , r. hnninen , m. krusius , v.s . lvov , phys . 105 * , 125301 ( 2010 ) . p.m. walmsley , v.b . eltsov , p.j . heikkinen , j.j . hosio , r. hnninen , m. krusius , phys . b * 84 * , 184532 ( 2011 ) . r. hnninen , j. low temp . phys . * 156 * , 145 ( 2009 ) . r. hnninen , v.b . eltsov , a.p . finne , r. de graaf , j. kopu , m. krusius , r.e . solntsev , j. low temp . phys . * 155*,98 ( 2009 ) . hosio , v.b . eltsov , r. de graaf , p.j . heikkinen , r. hnninen , m. krusius , v.s . lvov , g.e . volovik , phys . lett . * 107 * , 135302 ( 2011 ) . m. krusius , j.s . korhonen , y. kondo , e.b . sonin , phys . b * 47 * , 15113 ( 1993 ) . kopnin , _ theory of nonequilibrium superconductivity _ ( clarendon press , oxford 2001 ) . bevan , a.j . manninen , j.b . cook , a.j . armstrong , j.r . hook , h.e . hall , j. low temp . * 109 * , 423 ( 1997 ) . eltsov , r. de graaf , p.j . heikkinen , j.j . hosio , r. hnninen , m. krusius , j. low temp . phys . * 161 * , 474 ( 2010 ) . silaev , phys . lett . * 108 * , 045303 ( 2012 ) . hosio , v.b . eltsov , p.j . heikkinen , r. hnninen , m. krusius , v.s lvov , nat . comm . * 4:1614 * ( 2013 ) . l. kondaurova , v. lvov , a. pomyalov , i. procaccia , arxiv:1405.5177 ( 2014 ) . p.m. walmsley , a.i . golov , h.e . hall , a.a . levchenko , and w.f . vinen , phys . lett . * 99 * , 265302 ( 2007 ) . eltsov , r. hnninen , m. krusius , proc . sci . * 111 * , supplement 1 , 4711 ( 2014 ) . p.m. walmsley , d. zmeev , f. pakpour , a.i . golov , proc . * 111 * , supplement 1 , 4691 ( 2014 ) .

we show that the fourth order nonlinear ode which controls the pole dynamics in the general solution of equation compatible with the kdv equation exhibits two remarkable properties : 1 ) it governs the isomonodromy deformations of a matrix linear ode with polynomial coefficients , and 2 ) it does not possesses the painlev property . we also study the properties of the riemann hilbert problem associated to this ode and find its large asymptotic solution for the physically interesting initial data .

1301.7211

the study of relationships between the theory of isomonodromic deformations and the theory of differential equations satisfying the so - called painlev property is a well - established branch of the analytic theory of differential equations in complex domain ( see below a brief summary of the most important results obtained in this direction ) . one of the outputs of the present paper suggests that the above mentioned relationship is less straightforward than it was traditionally believed . we illustrate the point with an example of a fourth order ode for a function @xmath2 @xmath3 this equation appeared in @xcite in the study of pole loci of solutions to a degenerate garnier system . our observation is that eq . governs isomonodromic deformation of certain linear differential operator with polynomial coefficients ( see eq . however , this equation does not satisfy the painlev property as its general solution has third order branch points of the form @xmath4 ( cf . @xcite ) where the location of the branch points depends on the choice of the solution ( the so - called _ movable critical singularities _ ) . in spite of such a somewhat surprising phenomenon the method of isomonodromic deformations proves to be almost as powerful in the study of solutions to the equation [ pi21 ] as in the case of classical painlev equations . namely , it is possible to derive the large @xmath1 asymptotics of solutions and , moreover , to describe the branching locus of a given solution in terms of a kind of a spectral problem for a quintic anharmonic oscillator . the equation [ pi21 ] is of interest on its own . namely , it describes the behavior of poles of solutions to another fourth order ode @xmath5 usually denoted as . it is the second member of the so - called hierarchy . the coefficients of this equation depend on @xmath1 as on a parameter . it is well known ( see , e.g. , @xcite ) that the equation is compatible with the korteweg - de vries ( kdv ) dynamics @xmath6 the laurent expansion of solutions @xmath7 to the system , near a pole @xmath8 has the form @xmath9 where the function @xmath10 solves eq . [ pi21 ] . in this way one associates a multivalued solution @xmath10 to eq . [ pi21 ] with any solution to eq . . the branch points of @xmath10 correspond to the triple collisions of the poles of solutions to the . thus , the above mentioned isomonodromy realization of eq . [ pi21 ] provides one with a tool for studying the kdv dynamics of poles of solutions to the equation . particular solutions to the equation are of interest . we will concentrate our attention on one of them , namely , the one that has no poles on the real axis @xmath11 ( see @xcite , @xcite , @xcite , @xcite about importance of this special solution to the eq . ) . such a solution @xmath12 exists @xcite for any real @xmath1 and it is uniquely determined by its asymptotic behavior for large @xmath13 . using the developed techniques along with the isomonodromic description @xcite of the special solution to , we arrive at the asymptotic description for large @xmath1 of poles of this special solution . before proceeding to the formulation of main results let us first briefly recall the basic notions of the present paper , namely the painlev property and the isomonodromic deformations . in the description of the historical framework we will manly follow the paper @xcite . in 1866 , l. fuchs @xcite has shown that all the singular points of solutions to a linear ode are among the singularities of its coefficients and thus are independent from the initial conditions . in the nonlinear case , the reasonable problem is to look for odes defining the families of functions , called the _ general solutions _ , which can be meromorphically extended to the universal covering space of a punctured riemann surface with the punctures determined by the equation . in other words , the problem is to find odes whose general solutions are free from the branch points and essential singularities depending on the specific choice of the initial data . this property is called now the _ painlev property _ , or the _ analytic painlev property _ , and is obviously shared not only by the linear odes but also by the odes for elliptic functions . in @xcite , l. fuchs started the classification of the first order polynomial in @xmath14 and @xmath15 odes , @xmath16 , with single - valued in @xmath17 coefficients with respect to the painlev property . h. poincar @xcite and p. painlev @xcite accomplished the analysis not finding new transcendental functions . in @xcite , p. painlev revisited the l. fuchs idea extending the program of classification to the second order odes of the form @xmath18 with @xmath19 meromorphic in @xmath17 and rational in @xmath14 and @xmath15 . in the course of classification of the 2nd order 1st degree odes modulo the mbius transformation @xcite , it was found 50 equations which pass the painlev @xmath20-test and thus called now the _ gambier equations_. it occurs that all these equations can be either integrated in terms of the classical linear transcendents , or elliptic functions , or reduce to one of the six exceptional _ classical painlev equations _ . the monodromy group to a linear ode was first considered by b. riemann @xcite , h. schwarz @xcite and h. poincar @xcite . apparently , it was l. fuchs in @xcite who for the first time set the problem of the deformations of the coefficients in a linear equation that leave the monodromy group unchanged . namely , assuming that solutions of a linear ode depend on an additional variable , he obtained a system of the first order pdes the solutions must satisfy . more modern treatment of the isomonodromy deformations was developed by r. fuchs in @xcite . he has shown that the monodromy group of a scalar linear ode with four fuchsian singularities at @xmath21 and an apparent singular point at @xmath22 does not depend on the location @xmath17 of the fourth fuchsian singular point if the location of the apparent singularity @xmath14 depends on @xmath17 according to a nonlinear 2nd order ode . later , the fuchs isomonodromic deformation equation was included as the sixth painlev equation into the list of the classical painlev equations . in 1912 , l. schlesinger @xcite generalized the fuchs approach finding the equations of the isomonodromic deformations for arbitrary linear fuchsian odes , while r. garnier @xcite presented the scalar second order linear odes with irregular singular points whose isomonodromic deformations are governed by the lower classical painlev equations . in 1980 , m. jimbo , t. miwa and k. ueno @xcite extended the theory of garnier to the linear odes with generic irregular singularities . it is interesting that to the date of the mentioned above achievements , the fact that these equations do indeed possess the painlev property was not proved rigorously even for the case of the classical painlev equations . thus the term `` painlev equation '' loosely refers to various equations among which we mention the higher order odes in hierarchies associated with the classical painlev equations , the higher order odes in the classifications by f. bureau @xcite and c. cosgrove @xcite based on the use of some painlev tests , as well as the differential , difference , @xmath23-difference and elliptic - difference equations found in the course of the study of the symmetries and geometry of the classical painlev equations , see e.g. @xcite and @xcite . a general elegant approach to the painlev property of the equations of the isomonodromic deformations was presented by t. miwa @xcite and b. malgrange @xcite in the early 80s of the last century ( recall that the direct proof of the painlev property of the classical painlev equations appeared even later , see @xcite ) . the approach by t. miwa and b. malgrange is based on the use of the zero curvature representation and the riemann hilbert correspondence . in fact , they have proved the analytic painlev property of the isomonodromic deformation equations for arbitrary linear fuchsian odes and for equations with unbranched irregular singular points . in this respect we also mention the papers of m. inaba and m. saito @xcite who developed an algebro - geometric approach to the _ geometric painlev property _ of the isomonodromic deformations of the logarithmic and unramified irregular connections which also implies their analytic painlev property . one can expect that the isomonodromy deformation equations for arbitrary rational connections possess the painlev property . however , the nave induction does not work . namely , and this is our first result , we show that the polynomial in all variables 4th order 1st degree ode governs the isomonodromic deformations of the following linear differential equation with polynomial coefficients @xmath24 here @xmath2 , @xmath25 , @xmath26 are some smooth functions . the monodromy of the stokes multipliers of the system - see below . ] of the system does not depend on @xmath1 if and only if the coefficients have the form @xmath27 a ' + a-40 \alpha_3 ^ 0 t + \alpha_1 ^ 0 \nonumber\\ & & \alpha_0=-\frac9{20 } a'''-\frac{54}5 { a'}^4 + 432 \alpha_3 ^ 0 { a'}^3 + 18 \left [ t-330 ( \alpha_3 ^ 0)^2 - \alpha_2 ^ 0 \right]{a'}^2 \nonumber\\ & & + 3\left [ a -130 \alpha_3 ^ 0 t + 10800 ( \alpha_3 ^ 0)^3 + 90 \alpha_2 ^ 0 \alpha_3 ^ 0 + \alpha_1 ^ 0\right]\ , a ' -30 \alpha_3 ^ 0\ , a \nonumber\\ & & + 2100(\alpha_3 ^ 0)^2 t -54000 ( \alpha_3 ^ 0)^4 -900\alpha_2 ^ 0(\alpha_3 ^ 0)^2 -30 \alpha_1 ^ 0 \alpha_3 ^ 0 \nonumber\end{aligned}\ ] ] where @xmath28 , @xmath29 , @xmath30 are some constants , together with the following ode for the function @xmath2 @xmath31 a '' -\frac{200}3 { a'}^2 + \frac{200}9 t \nonumber\\ & & + \frac{4000}3 \alpha_3 ^ 0 a ' -6000 ( \alpha_3 ^ 0)^2 -\frac{200}9 \alpha_2 ^ 0=0.\end{aligned}\ ] ] after a change @xmath32 the equation reduces to the [ pi21 ] equation . so , our claim is that the equation [ pi21 ] describes the isomonodromic deformations of the system but , however , it does not possess the painlev property as being satisfied by a 4-parameter puiseux series in powers of @xmath33 . nevertheless , as we show , this equation can be effectively analyzed using the isomonodromy deformation techniques developed during last decades . in particular , the ramification points @xmath34 ( see eq . above ) can be determined by a kind of a spectral problem for the quintic anharmonic oscillator @xmath35 here @xmath36 , @xmath37 , @xmath38 are the coefficients of the puiseaux expansion of @xmath10 ( see eq . below ) . namely , the coefficients of the quintic polynomial must be chosen in such a way that the equation possesses solutions exponentially decaying on certain contours in the complex plane , look for details in remark [ scattering_problem ] . in section 2 we discuss the second member of the hierarchy , equation , as an equation that describes the isomonodromic solutions to the kdv equation , and find a nonlinear ode , equation , that controls its pole dynamics . in section 3 , we derive the above isomonodromic representation for the eq . presenting a regularization of the linear system for along its singularity locus . in section 4 , we consider the puiseux series solution for and the singularity reduction of the corresponding linear system at the branch points . in section 5 , we set the riemann hilbert problems for all the previously introduced wave functions and discuss existence and uniqueness of their solutions . in section 6 , we present the asymptotic analysis of the riemann hilbert problems corresponding to a physically interesting special solution of and implementing the steepest - descent method introduced by deift and zhou . our main result in this section is the description of the large @xmath1 asymptotics of the singularity locus as a theta - divisor on a modulated elliptic curve , @xmath39 where @xmath40 , @xmath41 and @xmath42 , the function @xmath43 is determined by equations , @xmath44 and @xmath45 are the period integrals on the elliptic curve . in section 7 , we find explicitly the pole dynamics of the special solution in a vicinity of the attracting point @xmath46 , @xmath47{3}}{2\sqrt[4]{5}\sqrt{7 } } \ln\bigl ( t^{-7/4}(m+\tfrac{1}{2})\tfrac{3^{11/4}}{2^{6}5^{3/4}7^{5/2}e } \bigr ) \\ + i\pi(n+\tfrac{1}{2})\tfrac{\sqrt[4]{3}}{\sqrt[4]{5}\sqrt{7 } } + { \mathcal o}(t^{-7/2}\ln^2t),\end{gathered}\ ] ] where @xmath41 and @xmath40 enumerate the points of the pole lattice . thus , asymptotically for large @xmath48 , the poles of the special solution to never collide . it would be interesting to study their collisions for finite time . in section 8 , we briefly discuss our results and various open problems . the 4th order 1st degree polynomial ode , @xmath49 usually denoted as , is the second member of the so - called hierarchy . this equation governs the isomonodromy deformations of a linear polynomial ode described by the system @xmath50 where the connection matrices are explicitly written using the generators of @xmath51 , @xmath52 , @xmath53 and @xmath54 , [ abc_p12 ] @xmath55\sigma_3 \\ & \quad+\bigl [ 8\lambda^2 + 8u\lambda+(12u^2 + 2u_{xx}-120 t ) \bigr]\sigma_+ \\ & \quad+\bigl [ 8\lambda^3 - 8u\lambda^2 -(4u^2 + 2u_{xx}+120t)\lambda \\ & \hskip3 cm + ( 16u^3 - 2u_x^2 + 4uu_{xx}+240x ) \bigr]\sigma_- \bigr\ } , \end{split } \\ \label{b_p12 } b=&\sigma_+ + ( \lambda-2u)\sigma_- , \\ \label{c_p12 } c=&\tfrac{1}{6}u_x\sigma_3 -\tfrac{1}{3}(\lambda+u)\sigma_+ -\tfrac{1}{3 } ( \lambda^2 -u\lambda -2u^2 -\tfrac{1}{2}u_{xx})\sigma_-,\end{aligned}\ ] ] besides ( [ p12 ] ) which is the compatibility condition @xmath56=0\ ] ] of ( [ a_p12 ] ) and ( [ b_p12 ] ) , the system ( [ abc_p12 ] ) also implies the kdv equation , @xmath57 and equations which follow from ( [ p12 ] ) and ( [ kdv ] ) . the solutions of ( [ p12 ] ) compatible with the kdv equation ( [ kdv ] ) are called the _ isomonodromic _ solutions to kdv . it is known that equation passes painlev tests as being presented in the list by c. cosgrove @xcite under the symbol f - v with @xmath58 and the parameters @xmath59 , @xmath60 and @xmath61 . see @xcite for a proof of the painlev property based on the riemann hilbert correspondence . below , we are especially interested in the 4-parameter series solution to with the following initial terms ( because is polynomial in all its variables , the construction of the complete formal series via a recurrence relation is straightforward ) , @xmath62 this 4-parameter series is compatible with the kdv equation ( [ kdv ] ) if the coefficients @xmath63 depend on @xmath1 in a particular way described by the function @xmath10 , @xmath64 in its turn , the pole position @xmath10 must satisfy the equation , @xmath65 besides the pole dynamics , the function @xmath8 parametrizes the singularity locus for the linear system ( [ lax_pair_p12 ] ) , ( [ abc_p12 ] ) . equation admits also the 3-parameter series solution , @xmath66 which , however , is not compatible with the kdv equation ( [ kdv ] ) . clearly it corresponds to the triple collisions of the poles . it is remarkable , that the linear system for can be regularized along the singularity locus @xmath8 : let @xmath67 be the 4-parameter laurent series solution ( [ laurent_gen ] ) for equation _ _ . then the gauge transformation @xmath68 with the gauge matrix @xmath69 regularizes the connections @xmath70 and @xmath71 along the singularity locus @xmath8 and , if @xmath72 it regularizes also the connection @xmath73 along the same locus . the proof is straightforward by inspection . the above gauge transformation is constructed as a sequence of the so - called shearing transformations . namely , substituting the initial terms of the relevant laurent series instead of @xmath14 into the matrix @xmath74 , one finds the leading order singular term proportional to the sum @xmath75 . conjugation by @xmath76 `` shares '' the singularity order among the off - diagonal entries . the matrix coefficient of the term of order @xmath77 becomes nilpotent , and its conjugation with @xmath78 turns it to the normal @xmath79-form . this brings into play the lower order singularities , and the procedure can be repeated . it is known that the singularity reductions exist also for the linear systems associated with the classical painlev equations . however the construction of the corresponding gauge transformation does not rely on a sequence of the shearing transformations . in contrast , the required gauge transformations simply turn the 1st order rank 2 differential systems into scalar 2nd order linear odes . [ singularity_reduction_existence ] the singularity reduction exists for an arbitrary isomonodromy system . the linear system for regularized along its singularity locus @xmath8 yields the linear system for equation . equation _ _ controls the isomonodromy deformations of a linear matrix ode with the polynomial coefficients described by the system @xmath80 with the coefficient matrices @xmath81 \sigma_+ -\tfrac{1}{3}\sigma_- . } \beta_2=2a',\quad \beta_1=10t-9(a')^2,\quad \beta_0=36(a')^3 -60a't -10a.\end{gathered}\ ] ] by inspection , the compatibility condition , @xmath82=0 $ ] , yields @xmath83 and @xmath84 i.e. equations for @xmath85 in ( [ cj_via_a_in_kdv ] ) and ( [ a_ode_from_kdv ] ) . it is easy to verify that equation admits the formal series solution in powers of @xmath33 with the following initial terms ( again , since the equation is polynomial in all variables , the recurrence relation for its coefficients is straightforward ) , @xmath86 the system ( [ a_lax_pair ] ) , ( [ ac_a_lax_pair ] ) is singular , however regularizable , at the branch point @xmath34 of @xmath10 : let @xmath10 be the 4-parameter puiseux series solution ( [ a_puiseux_series ] ) for equation . then the rational in @xmath87 and @xmath1 gauge transformation @xmath88 with the gauge matrix @xmath89 regularizes the connection @xmath90 at the branch point @xmath34 . here @xmath91 , @xmath92 , @xmath93 are pauli matrices , @xmath94 the new function @xmath95 satisfies the linear matrix ode equivalent to the anharmonic oscillator equation , which does not admit any continuous isomonodromic deformation , @xmath96 [ scattering_problem ] the anharmonic oscillator equation can be regarded as a stationary schrdinger equation . thus it is possible to introduce a scattering problem and to map the set of the parameters @xmath97 , @xmath36 , @xmath37 , @xmath38 to the set of the scattering data . indeed , let us introduce the jost solutions @xmath98 uniquely determined by the asymptotics @xmath99 here , we define @xmath100 on the @xmath87 complex plane cut along the positive part of the real line and choose its principal branch . then the incident wave @xmath101 produces several reflected and transmitted waves , @xmath102 where the coefficients @xmath103 and @xmath104 are called the transmission and reflection coefficients , respectively . only @xmath105 of the @xmath106 coefficients are independent . indeed , there exist six relations of the form @xmath107 , @xmath108 , @xmath109 , which come from the preservation of the amplitude of the dominant solution in the sectors @xmath110 $ ] , @xmath111 ; the equality @xmath112 normalizes the amplitude of the incident wave ; two conditions @xmath113 mean the solution is subdominant as @xmath114 ; the last condition @xmath115 means the continuity of this subdominant solution across the positive part of the real line , @xmath116 the four independent scattering coefficients , say , @xmath117 , can be used to reconstruct the parameters @xmath97 , @xmath36 , @xmath37 and @xmath38 determining the potential @xmath118 . in ( [ tk_rk_rel ] ) , the numbers in parentheses specify the values of the free transmission and reflection coefficients corresponding to the triple collisions of the poles of the special solution to of our interest , see below . along the oscillatory directions , the wave function corresponding to this special potential is given by the jost solutions @xmath119 ( @xmath120 ) , @xmath121 ( @xmath122 ) and by the sum @xmath123 as @xmath124 . this implies another simple characterization of the potential of our interest . defining @xmath100 on the plane cut along the negative part of the real axis , the schrdinger equation @xmath125 has the solution with the uniform in the sector @xmath126 asymptotics @xmath127 . existence of a solution with such asymptotics is tantamount to vanishing of four stokes multipliers @xmath128 ( see below ) . below , however , we characterize @xmath95 adopting the riemann - hilbert problem instead of the scattering problem . observe the successive reduction of the number of the deformation parameters the @xmath87-equations depend on : for the connection @xmath70 , the space of the deformation parameters @xmath129 is 2-dimensional , for @xmath90 , the deformation parameter space reduces to the 1-dimensional space of @xmath1 , and for @xmath130 , it is 0-dimensional . besides the continuous isomonodromic deformations considered above , it is possible to introduce discrete poincar - like isomonodromy mappings from one to another branch of @xmath10 and from one to another point of the lattice @xmath131 . it is natural to denote the above listed connections by the symbols , and , respectively , to reflect explicitly the number of the continuous deformation parameters involved . the equations of the isomonodromic deformations are thus denoted by the symbols ( same as ) , and ( the latter equation is a poincar - like mapping of the lattice @xmath131 ) . the boundary riemann hilbert ( rh ) problem consists in finding a piece - wise holomorphic function by its prescribed analytic properties : \i ) asymptotics at some marked points ; \ii ) discontinuity properties across a piece - wisely oriented graph . each of the above linear matrix odes for @xmath135 , @xmath136 and @xmath137 has one irregular singularity at @xmath138 and no other singular points . we introduce the formal series solutions that represent the asymptotic behavior of the genuine solutions to these linear odes in the interior of particular sectors near @xmath138 . since all the canonical asymptotics differ from each other in a rational left diagonal multiplier , it is convenient to unify the notations and distinguish all three cases introducing a parameter @xmath139 : @xmath140 here , the principal branch of the square root of @xmath87 is chosen . using ( [ nu_canonical_as ] ) for the asymptotic solutions @xmath141 and @xmath142 , one has to take into account the reductions @xmath8 for @xmath143 and @xmath34 , @xmath144 for @xmath145 . below however , using ( [ nu_canonical_as ] ) to set the corresponding rh problems , we assume that the exponential @xmath146 depends on arbitrary complex deformation parameters @xmath1 and @xmath17 because it is not allowed to use the unknown values @xmath10 and @xmath147 as the data in the rh problems formulated below . as the result , it is necessary to remember that the solvability domains for the rh problems shrink to the lines @xmath8 for @xmath143 and to a lattice @xmath148 for @xmath145 . the canonical solutions differ from each other in a triangular right matrix multipliers called the stokes matrices , @xmath149 moreover @xmath150 since the gauge transformations @xmath151 and @xmath152 are rational , they do not affect the stokes matrices . the stokes multipliers @xmath153 are the first integrals of equations ( [ p12 ] ) and ( [ a_ode_from_kdv ] ) . the solutions of the rh problems below are constructed using collections of the canonical solutions in the above presented sectors in a way to have the uniform asymptotics in a vicinity of infinity . the normalization of the rh problem however is rather subtle thing . to this aim , observe the prolongated canonical asymptotics @xmath154 @xmath155 and @xmath156 in the above right hand side exponential factors , the parameters @xmath157 can be expressed in terms of the coefficients of the relevant equation . for instance , @xmath158 in ( [ psi_canonical_as_long ] ) is one of two hamiltonians of . however , these particular relations are not important for us at this stage . now , we have the following [ nu_rhp ] given the complex values of the parameters @xmath17 , @xmath1 and @xmath153 , @xmath159 , satisfying ( [ sk_struct ] ) , and denoting @xmath160 find the piece - wise holomorphic @xmath0 matrix function @xmath161 , @xmath162 , with the following properties : 1 . @xmath163 + moreover * there exist a constant @xmath158 such that @xmath164 * if @xmath165 , then there exist constants @xmath158 , @xmath166 such that @xmath167 with some constant matrix @xmath168 ; * if @xmath145 , then there exist constants @xmath169 and @xmath170 such that @xmath171 with some constant matrix @xmath172 ; 2 . @xmath173 as @xmath174 ; 3 . on the union of eight rays @xmath175 , where @xmath176 , @xmath177 , and @xmath178 , all oriented towards infinity , the jump condition holds true , @xmath179 where @xmath180 and @xmath181 are limits of @xmath161 on @xmath182 from the left and from the right , respectively , and the piece - wise constant matrix @xmath183 is given by equations + [ jump_matrices ] @xmath184 the solution of the rh problem [ nu_rhp ] is unique . indeed , the scalar function @xmath185 is continuous across the set of rays @xmath182 and is bounded at the origin . therefore , @xmath185 is an entire function , and taking into account ( [ phi_as_def ] ) and applying the liouville theorem , @xmath186 . assume for the moment that there are two solutions of the rh problem [ nu_rhp ] , @xmath161 and @xmath187 . consider their ratio , @xmath188 . as it is easy to see , @xmath189 is continuous across all the rays of the set @xmath182 and is bounded at the origin , thus @xmath189 is an entire function . using ( [ phi_as_def ] ) , @xmath190 since all the entries @xmath191 are entire functions in @xmath87 , the liouville theorem yields the ambiguity @xmath192 where @xmath193 are arbitrary constants . these ambiguities however are eliminated using the asymptotic conditions in ( [ phi_c0_triviality_condition ] ) , ( [ z_c1_triviality_condition ] ) and ( [ x_c2_triviality_condition ] ) . given jump matrices , the set of points @xmath194 at which the rh problem is not solvable is called the malgrange divisor . it coincides with the zero locus of a holomorphic @xmath195-function by miwa and with the singularity locus of the isomonodromy deformation equation . for what follows , it is convenient to articulate our assumptions on the singularities and critical points of equations and . [ pp_conjecture ] 1 ) equation has no other movable singularities except for the movable poles ( [ laurent_gen ] ) satisfying ( [ cj_via_a_in_kdv ] ) , ( [ a_ode_from_kdv ] ) , and their triple collisions ( [ laurent_deg ] ) . 2 ) equation has no other movable singularities or critical points except for the branch points ( [ a_puiseux_series ] ) . these assumptions mean that the smooth branches of the malgrange divisor for the rh problem [ nu_rhp ] with @xmath196 ( corresponding to ) are parameterized by equations @xmath8 and thus coincide with the solvability set for the rh problem [ nu_rhp ] with @xmath143 ( corresponding to ) . similarly , the vertices of the malgrange divisor of the rh problem with @xmath196 correspond to the branch points @xmath34 of @xmath10 and therefore coincide with the solvability set of the rh problem with @xmath145 . in other words , conjecture [ pp_conjecture ] implies the following remarkable properties of the domains of solvability of all three rh problems : 1 . these domains do not pairwise intersect ; 2 . all these domains together cover the deformation parameter space @xmath197 . in this section , we describe another interesting feature of the family of three rh problems . roughly speaking , any solution of the _ non - homogeneous _ rh problem with the bigger value of @xmath198 allows one to construct infinitely many solutions of the _ homogeneous _ problems with the smaller value of @xmath198 . precise formulation of this property requires some more accuracy . first of all , let us introduce the rh problems equivalent to those above but unbranched and normalized at infinity . to this aim , write @xmath199 and define the piece - wise holomorphic functions , @xmath200 these functions solve the non - homogeneous rh problems on the union of rays @xmath201 , @xmath202 with a singular point at the origin and normalized to the unit at infinity . the corresponding homogeneous rh problem differs from the non - homogeneous counterpart in the asymptotics of @xmath203 that vanishes as @xmath204 . slightly abusing our notations , we formulate the non - homogeneous and homogeneous rh problems as follows : [ hat_phi_rhp ] find a piece - wise holomorphic function @xmath205 @xmath206 , @xmath162 , with the following properties : 1 . @xmath207@xmath208 ; 2 . across the rays @xmath209 , @xmath210 , all oriented towards infinity , the following jump conditions hold , @xmath211 3 . @xmath212,as @xmath213 . the following theorem holds true any solution @xmath214 to the non - homogeneous rh problem [ hat_phi_rhp ] with @xmath143 yields infinitely many solutions @xmath215 to the homogeneous rh problem [ hat_phi_rhp ] with @xmath196 . any solution @xmath216 to the non - homogeneous rh problem [ hat_phi_rhp ] with @xmath145 yields infinitely many solutions @xmath217 to the homogeneous rh problem [ hat_phi_rhp ] with @xmath218 . we prove the first part of the theorem . the proof of the second part is similar . consider a solution @xmath214 to the rh problem [ hat_phi_rhp ] for @xmath143 . it has the asymptotics @xmath219 where @xmath220 is a constant matrix , @xmath221 , and @xmath222 . then @xmath223 has the same jump properties as @xmath214 and the asymptotics of the form , @xmath224 thus @xmath225 is the solution of the homogeneous rh problem [ hat_phi_rhp ] with @xmath196 . in this section , we construct large @xmath1 asymptotic solution to the rh problem [ nu_rhp ] for @xmath196 and @xmath143 corresponding to a special solution of equation . this special solution @xmath12 has the asymptotics @xmath226{6|x|}$ ] as @xmath227 and is real and regular on the real line for any @xmath228 . the physical importance of this solution for @xmath229 from the point of view of string theory was justified in @xcite . in @xcite , this solution appeared in the study of the problem of universality of critical behavior of solutions to hamiltonian perturbations of hyperbolic pdes . the above mentioned properties uniquely distinguish this special solution . in @xcite , the characterization of this solution in terms of the stokes multipliers of the associated linear system was found , @xmath230 assume that @xmath231 is large and @xmath232 is such that @xmath233 our starting point is the large @xmath1 asymptotics of the spectral curve , @xmath234 , where @xmath235 , i.e. @xmath236 the asymptotic analysis of various degenerated solutions of performed in @xcite , shows that the topological properties of the asymptotic spectral curve are significantly different in the interior of the domains @xmath237 and @xmath238 , see figure [ fig3a ] . for @xmath239 , the spectral curve has genus 0 , @xmath240 where the double branch points @xmath241 and the simple branch point @xmath242 satisfy the conditions , @xmath243 and the ambiguity in choice of the root of the cubic equation for the simple branch point @xmath242 is fixed demanding that , for @xmath244 , @xmath245 the conditions above are consistent with the quasi - stationary asymptotic behavior of the special solution to ( [ p12 ] ) , @xmath246{6x_0},\quad x_0\to\pm\infty,\end{gathered}\ ] ] where the real on the real line branch of @xmath247{6x_0}$ ] is chosen . let us indicate some interesting points in the @xmath248 complex plane . at the point @xmath249 , two double branch points of the asymptotic spectral curve coalesce , so that @xmath250 and @xmath251 . generically , at the asymptotically quadruple branch point , the local solution of the rh problem can be approximated using the garnier miwa @xmath135 function for the second painlev transcendent , @xcite . for the rh problem we are studying , the relevant local approximate solution corresponds to a hastings mcleod solution to @xcite . at any @xmath252 including its boundary , the spectral curve has one simple and two double branch points . the precise asymptotic location of these branch points can be found for @xmath253 , where @xmath254 , @xmath255 and @xmath256 . at this point , as well as at any point of @xmath257 including its boundary , the leading order asymptotic solution of the rh problem can be expressed in elementary functions , cf . @xcite for the real line case . observe also the point @xmath258 where two double branch points coalesce . the asymptotic branch points corresponding to this point are @xmath259 and @xmath260 . generically , the quadruple degeneration corresponds to the appearance of . however , for the rh problem we consider here , the relevant painlev function is trivial , and the asymptotic solution to the rh problem remains elementary . in the interior part of the domain @xmath238 , the large @xmath1 asymptotic solution to the rh problem is constructed on the model elliptic curve , @xmath261 where the branch points @xmath262 , @xmath263 , are determined by the values of @xmath264 and @xmath248 as the roots of the cubic equation @xmath265 the double branch point @xmath264 is determined as a function of @xmath266 by the system of boutroux equations , see @xcite , @xmath267 supplemented by the described above boundary conditions on @xmath268 . as @xmath248 approaches the real segment @xmath269 , all asymptotic branch points become real and satisfy the inequalities @xmath270 . thus the second of the equations ( [ boutroux_tt ] ) trivializes while the first of these equations turns into the condition @xmath271 obtained in @xcite in the analysis of the whitham equations and used in @xcite to study the same special solution on the real line . the strategy of the steepest - descent asymptotic analysis by deift and zhou @xcite of the rh problem involves several standard steps : 1 ) transformation of the jump graph to the steepest - descent directions of a suitable @xmath272-function ; 2 ) construction of the local approximate solutions ( parametrices ) ; 3 ) matching all the parametrices into a global parametrix ; 4 ) proof that the global parametrix indeed approximates the genuine solution to the original rh problem . since all the above mentioned steps are well explained in the literature ( see e.g. @xcite ) , below , we omit unnecessary details . according the steepest - descent strategy , we first transform the jump contour for each of the rh problems to the steepest - descent graph for the exponential @xmath273 , see figure [ fig2 ] . observe that , in the special case ( [ tri - truncated_stokes ] ) , @xmath274 , @xmath275 , the jump graph depicted on figure [ fig1 ] can be transformed to the one shown in figure [ fig3 ] , where @xmath276 the large @xmath1 asymptotics of @xmath277 corresponding to the stokes multipliers ( [ tri - truncated_stokes ] ) is constructed on the riemann surface @xmath278 of the model elliptic curve @xmath279 glued of two copies of the complex @xmath280-plane cut along @xmath281\cup[\xi_3,-\infty)$ ] , see figure [ fig4 ] . define the complete elliptic integrals @xmath282 and abelian integrals @xmath283 we define the integral @xmath284 on the upper sheet of the riemann surface @xmath278 cut along the sum of intervals @xmath281\cup[\xi_4,\xi_3]\cup[\xi_3,-\infty)$ ] . the boutroux conditions ( [ boutroux_tt ] ) imply @xmath285 observe the following properties of @xmath284 and @xmath286 : 1 . as @xmath287 , @xmath288 2 . @xmath284 and @xmath286 are discontinuous across the broken line @xmath281\cup[\xi_4,\xi_3]\cup[\xi_3,-\infty)$ ] oriented from @xmath289 to infinity , moreover @xmath290 in this subsection , following @xcite , we solve the permutation rh problem on the segments @xmath281\cup[\xi_3,-\infty)$ ] whose solution gives a leading order contribution to the solution of the above rh problem . [ quasi_permutation_rhp ] given @xmath291 and the complex value of the parameter @xmath248 , find a piece - wise holomorphic @xmath0 matrix function @xmath292 with the following properties : @xmath293 where @xmath294 across the union of segments @xmath295 oriented as indicated , the jump condition holds true , @xmath296 where @xmath297 and @xmath298 are limits of @xmath292 on the segments from the left and from the right , respectively , see figure [ fig3 ] . we do not impose any conditions on the behavior of @xmath292 at the points @xmath299 , @xmath263 . as the result , the solution of the model rh problem [ quasi_permutation_rhp ] is determined up to a left rational matrix multiplier with possible poles at @xmath299 , @xmath263 , and certain asymptotics at infinity . below , we will use this rational multiplier to prove or disprove the asymptotic solvability of the original rh problem [ nu_rhp ] . consider the principal branches of the functions @xmath301 , @xmath302 both defined on the @xmath280 complex plane with cuts along the broken line @xmath281\cup[\xi_4\xi_3]\cup[\xi_3,-\infty)$ ] . these functions solve the scalar rh problems : 1 . @xmath303 , as @xmath287 ; 2 . the discontinuity of @xmath301 across the oriented contour @xmath281\cup[\xi_4\xi_3]\cup[\xi_3,-\infty)$ ] is described by the conditions @xmath304 define one more function , @xmath305 where the parameter @xmath306 is defined by @xmath307 this function @xmath308 has the following obvious properties , \(1 ) as @xmath309 , @xmath310 \(2 ) @xmath308 is discontinuous across the broken line @xmath281\cup[\xi_4,\xi_3]\cup[\xi_3,-\infty)$ ] oriented from @xmath289 to infinity , moreover @xmath311 using the riemann theta - function , @xmath312 , define the matrix function @xmath313 , @xmath314 here the parameters @xmath315 , @xmath316 and the factors @xmath317 , @xmath318 are defined by @xmath319 it can be shown that @xmath320 , the function @xmath313 ( [ phi_ba_def ] ) satisfies ( [ phiba_as_def ] ) and ( [ psiba_jump_rel ] ) and thus is one of the solutions of the rh problem [ quasi_permutation_rhp ] . any other solution to this rh problem has the form of the product @xmath321 where @xmath322 is rational with poles at @xmath299 , @xmath263 , and satisfies the asymptotic condition @xmath323 as @xmath287 . as it is well known , near a single branch points , the relevant boundary - value problem can be solved using the classical airy functions , see e.g. @xcite . define the wronsky matrix of the airy functions @xcite , @xmath325 where @xmath326 besides @xmath327 , introduce auxiliary functions @xmath328 by construction @xcite , @xmath329 assemble the piece - wise holomorphic functions @xmath330 , @xmath324 , @xmath331 where @xmath332 observing that the jump matrices of @xmath330 coincide with those across the lines emanating from the node points @xmath262 , @xmath324 , in figure [ fig3 ] , we are ready to construct the relevant local parametrices , @xmath333 here @xmath334 , @xmath324 , are holomorphic in some finite neighborhoods of @xmath299 matrices , and @xmath335 are changes of variables biholomorphic in some finite neighborhoods of @xmath299 . this biholomorphic change of variables has to be chosen in a way to ensure that the global parametrix , see below , has small enough jumps as @xmath48 , namely it must satisfy the condition @xmath336 where @xmath337 has the form ( [ spectral_curve_p12 ] ) , ( [ spectral_curve_genus1 ] ) with the branch points @xmath338 , @xmath339 the unique abelian differential holomorphic on the elliptic curve @xmath340 is @xmath341 thus the 1-parameter deformation of the model degenerated curve that respects the degeneration is generated by the vector field @xmath342 finally , we find the elliptic curve satisfying the asymptotic conditions , @xmath343 the global approximate solution to the rh problem for @xmath292 , @xmath344 , is a piece - wise analytic matrix function @xmath345 defined as follows , @xmath346 see figure [ fig13 ] . the exact solution is constructed using the correction function @xmath347 , @xmath348 that satisfies the following rh problem : \(1 ) the limit @xmath349 exists and is diagonal ; \(2 ) across the contour @xmath182 shown in figure [ fig14 ] , the jump condition holds , @xmath350 where @xmath351 to apply the @xmath352-theory to the latter rh problem @xcite , all jump matrices @xmath353 across the jump contour @xmath182 in figure [ fig14 ] must satisfy the estimate @xmath354 as @xmath355 . for the infinite tails emanating from the circles , this fact holds true because the relevant jumps are uniformly exponentially small . the jumps across the circles centered at the branch points @xmath262 , @xmath263 , can be made small as @xmath1 is large if one adjusts the rational matrix @xmath322 , @xmath344 , and the holomorphic matrices @xmath356 , @xmath344 , @xmath324 , in an appropriate way . omitting the straightforward but tedious computations , we present the result : @xmath357 the relevant correction function satisfies the estimates @xmath358 in the domain of the parameter @xmath359 described by @xmath360 in contrast , the computation of the left rational multiplier @xmath322 for @xmath143 yields @xmath361 together with the additional condition , @xmath362 or , equivalently , @xmath363 in this case , the `` external '' parametrix is elementary , @xmath364 and the correction function @xmath347 satisfies the estimates ( [ chi_estim ] ) . the above computation shows that , as @xmath48 , for each value of the deformation parameter @xmath17 , @xmath365 , the rh problem [ nu_rhp ] is solvable either with @xmath196 or @xmath143 . thus there is no room for the solvability of the rh problem with @xmath145 . therefore , at least in the large @xmath1 limit , the special solution to has no triple pole collisions and the corresponding solutions to have no branch points . the relevant malgrange divisor consists of smooth branches only . to compute the large @xmath1 asymptotic distribution of poles of the special solution to , we use the phase shift given in ( [ v_for_poles ] ) and the definition of @xmath315 in ( [ v_phi_nu_cj_cj*_def ] ) , @xmath366 using the canonical dissection of the riemann surface , the difference @xmath367 is expressed in terms of a single contour integral over the contour @xmath368 depicted in figure [ fig15 ] , @xmath369 inflating the contour @xmath368 , we transform it to a contour encircling the infinite point which is a branch point of the curve . then expanding the integrand at the infinity and using the residue theorem , we find @xmath370 and the pole distribution formula ( [ pole_distribution_as ] ) yields @xmath371 since @xmath264 is determined by @xmath248 via the boutroux equations ( [ boutroux_tt ] ) , equation ( [ pole_distribution_as_fin ] ) determines the pole position , @xmath372 as a transcendent function of two integers @xmath373 . as @xmath48 , the particular pole @xmath374 approaches the attracting point satisfying the equation @xmath375 although it is not known how to solve the transcendent equation ( [ attractor ] ) , the problem of finding the attractor to the pole distribution significantly simplifies observing that the pole attractors correspond to the quasi - stationary solutions of equation , @xmath376 substituting ( [ a_quasi_stat_as ] ) into ( [ a_ode_from_kdv ] ) , we find an algebraic equation for the parameter @xmath377 , @xmath378 with the roots @xmath379 . however only one of the roots , @xmath380 is consistent with the described above properties of the large @xmath1 asymptotic spectral curve for the special solution of . the linearization of equation at the 0-parameter power series solution with the leading order term coefficient ( [ a_infty_sol ] ) has four linearly independent solutions . two of them are exponential , @xmath381 $ ] , and we set them aside . two other solutions of the linearized equation , @xmath382 and @xmath383 , are relevant to our quasi - stationary behavior of the poles . using them , we form the 2-parameter series @xmath384 let us relate the free parameters @xmath385 in ( [ quasi_stat_series ] ) with the integers @xmath386 and @xmath387 in ( [ pole_distribution_as_fin ] ) . recall that , along the boundary @xmath388 , two branch points @xmath389 and @xmath289 of the asymptotic spectral curve coalesce . namely , if @xmath48 , the limiting values corresponding to the attracting point are @xmath390 the asymptotics of the branch points in the model elliptic spectral curve compatible with the expansion ( [ quasi_stat_series ] ) is given by @xmath391{15}\sqrt{a_{10}}\,t^{-7/8 } \\ -\tfrac{9}{28 } \bigl ( a_{01 } + 11a_{10 } + a_{10}\ln ( t^{-7/4 } ) \bigr)t^{-7/4 } + { \mathcal o}(t^{-21/8}\ln^2t).\end{gathered}\ ] ] the asymptotics of the periods @xmath392 as @xmath393 have the forms @xmath394 thus , for large @xmath1 , @xmath395{3}}{\sqrt{7}\sqrt[4]{5 } } ( 1+{\mathcal o}(t^{-7/8 } ) ) , \\ \shoveleft \omega_b= -\tfrac{\sqrt[4]{3}\sqrt{7}}{4\sqrt[4]{5 } } \ln t + \tfrac{2\sqrt[4]3}{\sqrt{7}\sqrt[4]{5 } } \ln\tfrac{3^{3/2}\sqrt{-a_{10 } } } { 2^{3/2}7^{3/2}\sqrt[4]{15 } } + { \mathcal o}(t^{-7/8}\ln t),\quad t\to+\infty.\end{gathered}\ ] ] finally , the coefficients @xmath396 and @xmath397 determining the asymptotic series for @xmath10 ( [ quasi_stat_series ] ) , @xmath398 follow using the asymptotic formula ( [ pole_distribution_as_fin ] ) , @xmath399{3}}{\sqrt{7}\sqrt[4]{5 } } + \tfrac{7}{12 } ( m+\tfrac{1}{2 } ) \tfrac{\sqrt[4]{3}}{\sqrt[4]{5}\sqrt{7 } } \bigl [ \ln ( t^{-7/4 } ) + \ln\tfrac{3^{3}a_{10 } } { 2^{5}7 ^ 2\sqrt{15 } } \bigr ] + \dots\end{gathered}\ ] ] equating coefficients of @xmath400 , we find @xmath401 , @xmath402{3}}{2\sqrt[4]{5}\sqrt{7}},\quad m\in{\mathbb z}_+,\ ] ] while the constant terms yield the family of values of @xmath403 , @xmath404{3}}{2\sqrt[4]{5}\sqrt{7 } } \ln\bigl [ ( m+\tfrac{1}{2 } ) \tfrac{3^{11/4}}{2^{6}5^{3/4}7^{5/2}e } \bigr ] + i\pi(n+\tfrac{1}{2})\tfrac{\sqrt[4]{3}}{\sqrt[4]{5}\sqrt{7 } } , \\ m\in{\mathbb z}_+,\quad n\in{\mathbb z}.\end{gathered}\ ] ] the formulas ( [ a10_sol ] ) and ( [ a01_sol ] ) with ( [ quasi_stat_series ] ) yield the asymptotic formula ( [ pole_dynamics_sol ] ) for @xmath405 which implies that the poles of the special solution to in a vicinity of the attracting point @xmath46 form a regular lattice with the slowly modulated intervals and a boundary formed by the line of poles corresponding to the values @xmath406 and @xmath40 . in particular , the interval between two the most right vertical lines of poles is given by @xmath407{3}}{2\sqrt[4]{5}\sqrt{7 } } \ln\bigl [ t^{-7/4}\tfrac{3^{17/4}}{2^{7}5^{3/4}7^{5/2}e } \bigr ] + { \mathcal o}(t^{-2}\ln^2t),\ ] ] note that the boundary @xmath388 formally corresponds to @xmath408 , @xmath409 , and the distance between the first vertical line of poles and @xmath388 is @xmath410{3}}{4\sqrt[4]{5}\sqrt{7 } } \ln\bigl [ t^{-7/4 } \tfrac{3^{11/4}}{2^{7}5^{3/4}7^{5/2}e } \bigr ] + { \mathcal o}(t^{-2}\ln^2t).\ ] ] above , we have presented and studied equation , to our best knowledge , the first of the differential equations that controls the isomonodromy deformations of a linear ode with rational coefficients and does not possess the painlev property . for the first glance , its existence contradicts the theorem by miwa and malgrange . however , it is not the case . the absence of the painlev property in is related to the fact that the domain of the solvability of the corresponding rh problem in the 2-dimensional complex space with the coordinates @xmath129 is restricted to the malgrange divisor of , i.e. to the set of the complex lines @xmath411 which are allowed to intersect . actually , this fact provides us with the important information on the nontrivial analytic structure of the malgrange divisor for which forms a riemann surface with the infinite number of sheets and third order branch points . besides this , the discovery of equation provides us with a new wide field of research . for instance , it is interesting to explore the possibility of existence of similar equations associated with other isomonodromic solutions of kdv or other integrable pdes like nonlinear schrdinger or pohlmeyer lund regge equations . other interesting problem not discussed above is the structure in induced by the singularity reduction of the hamiltonian structure in ( and , in the case of a successful extension of the singularity reduction methodology to the hierarchies associated with other painlev equations , the structures induced by the weyl symmetries ) . finally , we mention the problem of characterization of such initial data to ( and other isomonodromy deformation equations ) whose singularity reductions do not have branch points . in the case , we conjecture that its special solution considered above does not have merging poles for any @xmath1 and therefore the relevant solution of does not have branch points . if this conjecture is true , it can serve as one more characteristic property of this special solution . the authors are grateful to prof . y. ohyama for the reference to the paper @xcite . this work is partially supported by the european research council advanced grant from - pde , by the russian federation government grant no . 2010 - 220 - 01 - 077 and by prin 2010 - 11 grant `` geometric and analytic theory of hamiltonian systems in finite and infinite dimensions '' of italian ministry of universities and researches . thanks the staff of sissa for hospitality during his visit when this work was done . bureau , differential equations with fixed critical points . i , _ annali di matematica _ , * 64 * ( 1964 ) 229364 ; ii , ibid . * 66 * ( 1964 ) 1116 ; to3emquationes differentielles du second ordre en y et du second degr en y do nt lintegrale gnrale a ses points critique fixes , _ annali di matematica _ , * 91 * ( 1972 ) 163281 . cosgrove , higher - order painlev equations in the polynomial class i. bureau symbol p2 , _ stud . * 104 * ( 2000 ) 1 - 65 ; to3emhigher - order painlev equations in the polynomial class ii . bureau symbol p1 , _ stud . _ * 116 * ( 2006 ) 321 - 413 . b. dubrovin , on universality of critical behavior in hamiltonian pdes , _ geometry , topology , and mathematical physics _ , ( 2008 ) 59109 , _ amer . soc . transl . _ ser . 2 , * 224 * , amer . soc . , providence , ri , 2008 ; arxiv:0804.3790 . fuchs , zur theorie der linearen differentialgleichungen mit vernderlichen coefficienten , _ jahrsber . gewerberschule _ , berlin , ostern ( 1865 ) : _ werke _ i , 111 - 158 ; _ j. math . _ * 66 * ( 1866 ) 121 - 160 : _ werke _ , i , 159 - 204 . fuchs , ber lineare differentialgleichungen , welche von parametern unabhngige substitutiongruppen besitzen , _ sitzungsberichte knigl . preuss akad . _ , berlin , 1892 , pp . 157 - 176 : _ werke _ , iii , 117 - 140 ; to3ember lineare differentialgleichungen , welche von parametern unabhngige substitutiongruppen besitzen , _ sitzungsberichte knigl . preuss akad . _ , berlin , einleitung und no . 1 - 4 , 1893 , pp . 975 - 988 ; no . 5 - 8 , 1894 , s. 1117 - 1127 : _ werke _ , iii , 169 - 198 . r. garnier , sur les quations diffrentielles du troisme ordre do nt lintgrale gnrale est uniforme et sur une classe dquationnes nouvelles dordre suprieur do nt lintgrale gnrale a ses points critiques fixes , _ ann . sci . ecole norm . * 29 * ( 1912 ) 1 - 126 . m. inaba , moduli of parabolic connections on a curve and riemann hilbert correspondence , arxiv : math/0602004v2 [ math.ag ] ; m. inaba and m. saito , moduli of unramified irregular singular parabolic connections on a smooth projective curve , arxiv:1203.0084v2 [ math.ag ] a.a . kapaev , monodromy deformation approach to the scaling limit of the painleve first equation , _ crm proc . notes _ , * 32 * ( 2002 ) 157 - 179 ; http://arxiv.org/abs/nlin.si/0105002 to3em , monodromy approach to the scaling limits in isomonodromy systems , _ theor . math . _ , * 137 * , no . 3 ( 2003 ) 1691 - 1702 ; http://arxiv.org/abs/nlin.si/0211022 b. malgrange , sur les dformations isomonodromiques , i : singularits rgulires , in sminaire ens , mathematics and physics ( paris , 1979/1982 ) , 401426 , _ progr . in math . _ * 37 * , birkhuser - verlag , boston , 1983 ; to3emsur les dformations isomonodromiques , ii : singularits irrgulires , mathematics and physics ( paris , 1979/1982 ) , 427438 , _ progr . in math . _ * 37 * , birkhuser - verlag , boston , 1983 . m. noumi , affine weyl group approach to painlev equations , _ proceedings of the international congress of mathematics _ iii ( beijing , 2002 ) 497509 , higher ed . press , beijing , 2002 ; to3em_painlev equations through symmetry _ , translations of math . monographs , vol . 223 , ams , providence , rhode island , 2004 . b. riemann , beitrge zur theorie der durch die gausssche reihe @xmath413 darstellbaren functionen , _ aus dem siebenten band der abhandlungen der koniglichen gesellschaft der wissenschaften zu gttingen _ , 1857 . schwarz , ueber diejenigen falle , in welchen die gaussische hypergeometrische reihe eine algebraische function ihres vierten elementes darstellt , _ j. math . _ * 75 * ( 1872 ) 292 - 335 : _ abhandlungen _ , * 11 * , 211 - 259 .

the masses and radii of low - magnetic field neutron stars can be measured by combining different observable quantities obtained from their x - ray spectra during thermonuclear x - ray bursts . one of these quantities is the apparent radius of each neutron star as inferred from the x - ray flux and spectral temperature measured during the cooling tails of bursts , when the thermonuclear flash is believed to have engulfed the entire star . in this paper , we analyze 13,095 x - ray spectra of 446 x - ray bursts observed from 12 sources in order to assess possible systematic effects in the measurements of the apparent radii of neutron stars . we first show that the vast majority of the observed x - ray spectra are consistent with blackbody functions to within a few percent . we find that most x - ray bursts follow a very well determined correlation between x - ray flux and temperature , which is consistent with the whole neutron - star surface emitting uniformly during the cooling tails . we develop a bayesian gaussian mixture algorithm to measure the apparent radii of the neutron stars in these sources , while detecting and excluding a small number of x - ray bursts that show irregular cooling behavior . this algorithm also provides us with a quantitative measure of the systematic uncertainties in the measurements . we find that those errors in the spectroscopic determination of neutron - star radii that are introduced by systematic effects in the cooling tails of x - ray bursts are in the range% . such small errors are adequate to distinguish between different equations of state provided that uncertainties in the distance to each source and the absolute calibration of x - ray detectors do not dominate the error budget .

1103.5767

the thermal spectra of neutron stars during thermonuclear x - ray bursts have been used during the last three decades in numerous attempts to measure the neutron - star masses and radii ( e.g. , van paradijs 1978 , 1979 ; foster , fabian & ross 1986 ; sztajno et al . 1987 ; van paradijs & lewin 1987 ; damen et al . 1989 , 1990 ) . such efforts were often hampered by large systematic uncertainties in the estimated distances to the bursters and in the theoretical models for their x - ray spectra . moreover , the relatively small number of x - ray bursts observed by early satellites from each individual source made it impossible to assess systematic uncertainties related to the degree of anisotropy of the thermonuclear burning on the neutron - star surface , or of the obscuration by and the reflection off the accretion flow . the situation has changed dramatically in the last few years . the distances to several globular clusters that contain bursting neutron stars has been narrowed down with the use of observations by the hubble space telescope ( see , e.g. , kuulkers et al . 2003 ; zel , gver , & psaltis 2009 ; gver et al . the distances to x - ray bursters in the galactic disk and halo have also been determined using alternate methods ( e.g. , gver et al . theoretical models of the x - ray spectra of bursting neutron stars have been greatly improved and can account for the subtle effects of the presence of heavy metals in their atmospheres ( e.g. , london , taam , & howard 1986 ; foster , fabian , & ross 1986 ; ebisuzaki 1987 ; madej , joss , & raska 2004 ; majczyna et al . 2005 ; suleimanov , poutanen , & werner 2011 ) . finally , the high signal - to - noise observations of the x - ray spectra of several hundreds of bursts with the rossi x - ray timing explorer ( galloway et al . 2008a ) allow for the formal uncertainties of individual measurements to be substantially reduced . the combination of these developments led recently to the measurement of the masses and radii of several neutron stars ( zel et al . 2009 ; gver et al . 2010a , 2010b ; steiner , lattimer , & brown 2010 ) which are already sufficient to provide broad constraints on the equation of state of neutron - star matter ( zel , baym , & gver 2010 ) . in some cases , the formal uncertainties in the spectroscopic measurements are as low as a few percent ( gver et al . 2010b ) suggesting that their accuracy might be limited by systematic effects rather than by counting statistics . in this series of articles , we use the large database of x - ray bursts observed with the rxte in order to assess the systematic effects in various spectroscopic measurements of their properties . in the first article , we focus on the measurements of the apparent surface areas of 12 neutron stars as inferred from their x - ray spectra during the cooling tails of the bursts . our goal is to quantify the degree to which : _ ( i ) _ the x - ray burst spectra observed in the rxte energy range can be described by blackbody functions ( the so - called color correction arising from atmospheric effects are then applied a posteriori ) ; _ ( ii ) _ the entire surface area of each neutron star burns practically uniformly during the cooling of the bursts ; _ ( iii ) _ the accretion flows make minor contributions to the emission during the bursts . lccccc 4u 0513@xmath140 & 05 14 06.60 & @xmath140 02 37.0 & 6 & 0.014@xmath2 & gc + 4u 1608@xmath152 & 16 12 43.00 & @xmath152 25 23.0 & 26 & [email protected]@xmath4 & x - ray edges + 4u 1636@xmath153 & 16 40 55.50 & @xmath153 45 05.0 & 162 & 0.44@xmath5 & x - ray edges + 4u 1702@xmath1429 & 17 06 15.31 & @xmath143 02 08.7 & 46 & 1.95 & x - ray continuum + 4u 1705@xmath144 & 17 08 54.47 & @xmath144 06 07.4 & 44 & [email protected]@xmath6 & x - ray edges + 4u 1724@xmath1307 & 17 27 33.20 & @xmath130 48 07.0 & 3 & 1.08@xmath2 & gc + 4u 1728@xmath134 & 17 31 57.40 & @xmath133 50 05.0 & 90 & 2.49 @xmath30.14@xmath7 & x - ray edges + ks 1731@xmath1260 & 17 34 12.70 & @xmath126 05 48.5 & 24 & 2.98 & x - ray continuum + 4u 1735@xmath144 & 17 38 58.30 & @xmath144 27 00.0 & 6 & 0.28@xmath8 & x - ray edges + exo 1745@xmath1248 & 17 48 56.00 & @xmath124 53 42.0 & 22 & [email protected]@xmath9 & x - ray continuum + 4u 1746@xmath137 & 17 50 12.7 & @xmath137 03 08.0 & 7 & 0.36@xmath10 & gc + sax j1748.9@xmath12021 & 17 48 52.16 & @xmath120 21 32.4 & 4 & 0.79@xmath10 & gc + sax j1750.8@xmath12900 & 17 50 24.00 & @xmath129 02 18.0 & 4 & 4.97 & x - ray continuum + 4u 1820@xmath130 & 18 23 40.45 & @xmath130 21 40.1 & 5 & 0.25 @xmath30.03@xmath11 & x - ray edges + aql x@xmath11 & 19 11 16.05 & @xmath1200 35 05.8 & 51 & [email protected]@xmath13 & counterpart [ sourcestable ] we base our study on the x - ray burst catalog of galloway et al . ( 2008a ) . we chose 12 out of the 48 sources in the catalog based on the following criteria : _ ( i ) _ we considered sources that show at least two bursts with evidence for photospheric radius expansion , based on the definition of the latter used by galloway et al . ( 2008a ) . this requirement arises from our ultimate aim , which is to measure both the mass and the radius of the neutron star in each system using a combination of spectroscopic phenomena ( as in , e.g. , zel et al . 2009 ) . _ ( ii ) _ we excluded dippers , adc sources , or known high - inclination sources . this list includes exo 0748@xmath1676 , mxb 1659@xmath1298 , 4u 1916@xmath105 , grs 1747@xmath1312 , 4u 1254@xmath169 , and 4u 1710@xmath1281 , for which it was shown that geometric effects related to obscuration or reflection significantly affect the flux from the stellar surface that is measured by a distant observer ( galloway , zel , & psaltis 2008b ) . _ ( iii ) _ we did not consider the known millisecond pulsars sax j1808.4@xmath13658 and hete j1900.1@xmath12455 because the presence of pulsations in their persistent emission implies that their magnetic fields are dynamically important and , therefore , may affect the properties of x - ray bursts . _ ( iv ) _ we excluded the sources grs 1741.9@xmath12853 and 2e 1742.9@xmath12929 as well as a small number of bursts from aql x-1 , 4u 1728@xmath134 , and 4u 1746@xmath137 , for which there is substantial evidence that their emission is affected by source confusion ( galloway et al . 2008a ; keek et al . 2010 ) . _ ( v ) _ for each source , we considered only bursts for which the flux in the persistent emission prior to the burst is at most 10% of the inferred eddington flux for each source , i.e. , @xmath14 , as calculated by galloway et al . ( 2008a ) . because we subtract the pre - burst persistent emission from the decay spectrum of each x - ray burst , this requirement reduces substantially the systematic uncertainties introduced by potential changes in the emission from the accretion flow during the burst . table [ sourcestable ] lists all the x - ray bursters that fulfill the above requirements , together with the number of bursts observed by rxte for each source . this is the complete list of sources for which the masses and radii can be measured , in principle , using the spectroscopic method of zel et al . ( 2009 ) , with currently available data . our analyses of the bursts from exo 1745@xmath1248 , 4u 1608@xmath152 , and 4u 1820@xmath130 have been reported elsewhere ( zel et al . 2009 ; gver et al . 2010a , 2010b ) and will not be repeated here . we analyzed the burst data for the 12 sources shown in table [ sourcestable ] following the method outlined in galloway et al . ( 2008a ; see also zel et al . 2009 ; gver et al . 2010a , 2010b ) . for each source , we extracted time resolved [email protected] kev x - ray spectra using the _ seextrct _ ftool for the science event mode data and the _ saextrct _ ftool for the science array mode data from all of the rxte / pca layers . science mode observations provide high count - rate data with a nominal time resolution of 125@xmath15s in 64 spectral channels over the whole energy range ( 2@xmath160 kev ) of the pca detector . following galloway et al . 2008a , we extracted spectra integrated over 0.25 , 0.5 , 1 , and 2 s time intervals , depending on the source count rate during the burst , so that the total number of counts in each spectrum is roughly constant . ( in a few cases , data gaps during the observations result in x - ray spectra integrated over shorter exposure times ) . we took the spectrum over a 16 s time interval prior to the onset of each burst as the spectrum of the persistent emission , which we subtracted from the burst spectra as background . we generated separate response matrix files for each burst using the pcarsp version 11.7 , heasoft release 6.7 , and heasarc s remote calibration database and took into account the offset pointing of the pca during the creation of the response matrix files . this latest version of the pca response matrix makes the instrument calibration self - consistent over the pca lifetime and yields a normalization of the crab pulsar that is within 1@xmath16 of the calibration measurement of toor & seward ( 1974 ) for that source . in 4 , we discuss in some detail the effect of the uncertainties in the absolute flux calibration on the measurement of the apparent surface area of neutron stars . finally , we corrected all of the x - ray spectra for pca deadtime following the method suggested by the rxte team . to analyze the spectra , we used the interactive spectral interpretation system ( isis ) , version 1.4.9 - 55 ( houck & denicola 2000 ) . for each fit , we included a systematic error of 0.5% as suggested by the rxte calibration team . we fit each spectrum with a blackbody function using the _ bbodyrad _ model ( as defined in xspec , arnaud 1996 ) and multiplied it by the _ tbabs _ model ( wilms , allen , & mccray 2000 ) that takes into account the interstellar extinction , assuming ism abundances . the model of the x - ray spectrum for each burst , therefore , depends on only three parameters : the color temperature of the blackbody , @xmath17 , the normalization of the blackbody spectrum , @xmath18 , and the equivalent hydrogen column density , @xmath19 of the interstellar extinction . allowing the hydrogen column density @xmath19 to be a free parameter in the fitting procedure leads to correlated uncertainties between the amount of interstellar extinction and the temperature of the blackbody . moreover , for every 10@xmath20 @xmath21 overestimation ( underestimation ) of the column density , the inferred flux of the thermal emission is systematically larger ( smaller ) by @xmath22 10@xmath23 erg @xmath21 s@xmath24 ( galloway et al . reducing the uncertainties related to the interstellar extinction for each burster is , therefore , very important in controlling systematic effects . a recent analysis of high resolution grating spectra from a number of x - ray binaries ( miller , cackett , & reis 2009 ) showed that the individual photoelectric absorption edges observed in the x - ray spectra do not show significant variations with source luminosity or spectral state . this result strongly suggests that the neutral hydrogen column density is dominated by absorption in the interstellar medium and does not change on short timescales . in order to reduce this systematic uncertainty and given the fact that there is no evidence of variable neutral absorption for each system , we fixed the hydrogen column density for each source to a constant value that we obtained in one of the following ways . _ ( i ) _ for a number of sources , the equivalent hydrogen column density was inferred independently using high - resolution spectrographs . _ ( ii ) _ in cases where only an optical extinction or reddening measurement exists , we used the relation given by gver & zel ( 2009 ) to convert it to the equivalent hydrogen column density . _ ( iii ) _ finally , for three sources ( 4u 1702@xmath1429 , ks 1731@xmath1260 , sax j1750.8@xmath12900 ) there are no independent hydrogen column density measurements published in the literature . in these cases , we fit the x - ray spectra obtained during all the x - ray bursts of each source allowing the n@xmath25 value to vary in a wide range . we then found the resulting mean value of n@xmath25 and used this as a constant in our second set of fits to all the x - ray bursts . in table [ sourcestable ] , we show the adopted values for the hydrogen column density together with the measurement uncertainties , the method with which the values were estimated , and the appropriate references . future observations of these sources with x - ray grating spectrometers onboard chandra and xmm@xmath1_newton _ can help decrease the uncertainty arising from the lack of knowledge of the properties of the interstellar matter towards these sources . our goal in this article is to study potential systematic uncertainties in the measurement of the apparent area of each neutron star during the cooling tails of x - ray bursts . hereafter , we adopt the following working definition of the cooling tail . it is the time interval during which the inferred flux is lower than the peak flux of the burst or the touchdown flux for photospheric radius expansion bursts . for the purpose of this definition , we use as a touchdown point the first moment at which the blackbody temperature reaches its highest value and the inferred apparent radius is lowest . in order to control the countrate statistics , we also set a lower limit on the thermal flux during each cooling tail of 5@xmath2610@xmath23 erg s@xmath24 @xmath21 ( or 5@xmath2610@xmath27 for the exceptionally faint sources 4u 1746@xmath137 and 4u 0513@xmath1401 ) . in 3 and 4 , we discuss in detail our approach of quantifying the systematic uncertainties in the measurements of the apparent radii during the cooling tails of thermonuclear bursts , using the sources ks 1731@xmath1260 , 4u 1728@xmath134 , and 4u 1636@xmath1536 as case studies . for another analysis of the cooling tails of x - ray bursts from 4u 1636@xmath1536 , see zhang , mendez , & altamirano ( 2011 ) . in 5 , we repeat this procedure systematically for seven additional sources from table [ sourcestable ] . finally , the cooling tails of the bursts observed from 4u 0513@xmath140 and aql x-1 show irregular behavior , and we report our analysis of them in the appendix . our first working hypothesis is that the spectra of neutron stars during the cooling tails of thermonuclear bursts can be modeled by blackbody functions in the observed energy range . the fact that x - ray spectra can be described well with blackbody functions has been established since the first time resolved x - ray spectral studies of thermonuclear bursts ( see , e.g. , swank et al . 1977 ; lewin , van paradijs , & taam 1993 ; galloway et al . 2008a and references therein ) . under that assumption , the temperature measured by fitting blackbodies to the spectra are then corrected for the effects of the atmosphere by applying a numerical prefactor called the color correction factor @xmath28 , which is the ratio of the color to the effective temperature . this approach is expected to be only an approximation for a number of reasons . first , theoretical models of the atmospheres of neutron stars in radiative equilibrium invariably show that the emerging spectra are broader than blackbodies , especially at low temperatures and towards low photon energies ( e.g. , madej et al . 2004 ; majczyna et al . moreover , for neutron stars that are rapidly spinning , the spectra measured by an observer at infinity will be broader and more asymmetric with respect to those at the stellar surface ( zel & psaltis 2003 ) . the expected degree of broadening and asymmetry will be at most comparable to @xmath29 where @xmath30 is the spin frequency of the neutron star and @xmath31 is its radius . this is an uncertainty that can , in principle , be corrected for by fitting the x - ray data with theoretical models of the x - ray spectra emerging from the atmospheres of rapidly spinning neutron stars . second , the accretion flow in the vicinity of the neutron star may alter in a number of ways the observed spectrum . it may act as a mirror , reflecting a fraction of the emerging radiation towards the observer and , therefore , increasing the apparent surface area of emission . it might occult a fraction of the stellar surface , reducing the apparent surface area of emission ( see galloway et al . if the accretion flow is surrounded by a hot corona , comptonization of the surface emission will alter its spectrum as well as the relation between energy flux and apparent surface area ( e.g. , boutloukos , miller , & lamb 2010 ) . finally , the residual emission of the accretion flow , which we attempt to remove by subtracting the pre - burst spectrum of the source , might change during the duration of the burst , introducing systematic changes in the net spectrum . all of the above effects have the potential of changing the shape of the x - ray spectrum of a burster and , perhaps more importantly , lead to stochastic changes in the inferred apparent radii . their overall effect , however , is expected to be significantly reduced by the fact that the intense radiation field during each x - ray burst should either disrupt the inner accretion flow or cool any corona to the compton temperature of the radiation . such a phenomenon has been observed during a superburst from 4u 1820@xmath130 ( ballantyne & strohmayer 2004 ) . moreover , these effects are expected to be the strongest for high inclination systems ( as was inferred by , e.g. , galloway et al.2008b ) . this is why we not only limited our sample to bursts for which the ratio of the persistent flux prior to the burst to the eddington flux @xmath32 in order to minimize accretion - related uncertainties but also selected the known high inclination sources out of our sample ( see 2 ) . values obtained from fitting 1309 , 2519 , and 4596 x - ray burst spectra observed from the sources ks 1731@xmath1260 , 4u 1728@xmath134 , and 4u 1636@xmath1536 , respectively . the solid lines show the expected @xmath33/d.o.f . distributions for the number of degrees of freedom used during the fits . the vertical dashed lines correspond to the highest values of @xmath34/d.o.f . that we considered as statistically acceptable for each source . the vast majority of spectra are well described by blackbody functions.,title="fig : " ] /d.o.f . values obtained from fitting 1309 , 2519 , and 4596 x - ray burst spectra observed from the sources ks 1731@xmath1260 , 4u 1728@xmath134 , and 4u 1636@xmath1536 , respectively . the solid lines show the expected @xmath33/d.o.f . distributions for the number of degrees of freedom used during the fits . the vertical dashed lines correspond to the highest values of @xmath34/d.o.f . that we considered as statistically acceptable for each source . the vast majority of spectra are well described by blackbody functions.,title="fig : " ] /d.o.f . values obtained from fitting 1309 , 2519 , and 4596 x - ray burst spectra observed from the sources ks 1731@xmath1260 , 4u 1728@xmath134 , and 4u 1636@xmath1536 , respectively . the solid lines show the expected @xmath33/d.o.f . distributions for the number of degrees of freedom used during the fits . the vertical dashed lines correspond to the highest values of @xmath34/d.o.f . that we considered as statistically acceptable for each source . the vast majority of spectra are well described by blackbody functions.,title="fig : " ] a final source of systematic uncertainties in the determination of the x - ray spectral shape of a burster is related to errors in the response matrix of the detector . these are expected to be @xmath35% , according to the rxte calibration team ( see 2.2 ) . in order to quantify the degree to which any of these effects influence the spectra of x - ray bursters , we fit all cooling - tail spectra from each source with a blackbody model , allowing for a level of systematic uncertainty @xmath36 , determined in the following way . we reduced the degree of freedom in this procedure by setting @xmath36 in each spectral bin to be a constant fraction @xmath37 of the poisson error @xmath38 and fixed the value of @xmath37 to a constant for each source . we then added this systematic uncertainty to the formal poisson error of each measurement in quadrature , i.e. , @xmath39 , and defined a statistic @xmath34 such that @xmath40 here , @xmath33 is the standard statistic calculated for each fit . the posterior distribution of the new statistic @xmath34 may not be the same as that of the @xmath33 statistic and will depend on the nature of the systematic uncertainties . in order to explore the degree and nature of systematic uncertainties , we fit for each source the @xmath34 distribution with the @xmath33 distribution expected for the number of degrees of freedom used in the fits , with @xmath37 as a free parameter . note that the expected distribution that we use is formally correct if the countrate data in the individual spectral bins have uncorrelated errors . the expected @xmath33 and the observed @xmath34 distributions for the optimal @xmath37 value for ks 1731@xmath1260 , 4u 1728@xmath134 , and 4u 1636@xmath1536 are shown in figure [ fig : chi2 ] . the @xmath37 value required to make the observed @xmath34 distributions for ks 1731@xmath1260 consistent with the expected ones is 0.55 ( see table [ chi2table ] ) . this value corresponds to systematic uncertainties that are very small . as an example , we consider a typical 0.25 s integration for ks 1731@xmath1260 , when its spectrum is characterized by a color temperature of 2.5 kev . in this case , the average number of counts in each spectral bin in the 3 - 10 kev energy range is 110 ct and the corresponding poisson uncertainty is @xmath41% . multiplying this by the inferred value @xmath42 for this source leads to the conclusion that the systematic uncertainties required to render the observed spectrum consistent with a blackbody are @xmath43% . for the case of 4u 1728@xmath134 , which is brighter than ks 1731@xmath1260 , the formal uncertainties are @xmath44% and the systematic uncertainties amount to @xmath45% . figure [ fig : chi2 ] shows that the resulting @xmath34 distributions can be well approximated by the @xmath33 distribution but have weak tails extending to higher values . these tails most likely arise from the inconsistency of only a small number of spectra with blackbody functions even though we can not rule out the possibility that the @xmath34 statistic follows a different posterior distribution than @xmath33 . figure [ fig : chi2 ] allows us also to identify the x - ray spectra that are statistically inconsistent with blackbodies . for each source , there is a maximum value of @xmath34 per degree of freedom beyond which the distribution of @xmath34 values deviates from the theoretical expectation . for the case of ks 1731@xmath1260 , this limiting value of the reduced @xmath34 is 1.7 , for 4u 1728@xmath134 it is 1.7 , and for 4u 1636@xmath1536 is 1.9 ( see table [ chi2table ] ) . the spectra with high reduced @xmath34 values often occur at the late stages of the cooling tails , where the subtraction of the persistent emission is the most problematic . however , unacceptable @xmath34 values may also be found in other , seemingly random places of the cooling tails . nevertheless , the fraction of spectra that is inconsistent with blackbodies are @xmath46% , @xmath47% , and @xmath48% for ks 1731@xmath1260 , for 4u 1728@xmath134 , and for 4u 1636@xmath1536 , respectively . hereafter , we consider only the spectra that we regard to be statistically acceptable , given the values of the reduced @xmath34 . lcccc source & number of & @xmath37 & @xmath34/dof limit & fraction of + & spectra & & & acceptable spectra + 4u 0513@xmath140 & 94 & 0.44 & 1.7 & 93.6% + 4u 1636@xmath153 & 4579 & 0.59 & 1.9 & 99.3% + 4u 1702@xmath1429 & 285 & 0.21 & 2.0 & 99.3% + 4u 1705@xmath144 & 694 & & 1.5 & 87.5% + 4u 1724@xmath1307 & 58 & 0.44 & 1.5 & 91.4% + 4u 1728@xmath134 & 2532 & 0.57 & 1.7 & 93.4% + ks 1731@xmath1260 & 1312 & 0.52 & 1.7 & 98.2% + 4u 1735@xmath144 & 40 & & 2.0 & 97.5% + 4u 1746@xmath137 & 187 & & 1.6 & 85.0% + sax j1748.9@xmath12021 & 104 & 0.26 & 1.9 & 99.0% + sax j1750.8@xmath12900 & 82 & 0.54 & 1.7 & 100% + aql x@xmath11 & 2191 & 0.45 & 1.6 & 92.1% + [ chi2table ] our second working hypothesis is that , during the cooling tail of each burst , the entire neutron star is emitting uniformly with negligible lateral temperature variations . this assumption is again expected to be violated at some level for a number of reasons . the non - uniformity of accretion onto the neutron star ( e.g. , inogamov & sunyaev 1999 ) , the finite time of propagation of the burning front around the star ( see , e.g. , nozakura , ikeuchi , & fujimoto 1984 ; bildsten 1995 ; spitkovsky , levin , & ushomirsky 2002 ) , as well as the excitation of non - radial modes on the stellar surface ( heyl 2004 ; piro & bildsten 2005 ; narayan & cooper 2007 ) are all expected to lead to some variations in the effective temperature of emission at different latitudes and longitudes on the stellar surface . the characteristics of burst oscillations observed during the cooling tails of x - ray bursts , however , imply that any variations in the surface temperatures of neutron stars can only be marginal . indeed , any component of the variation in the surface temperature that is not symmetric with respect to the rotation axis leads to oscillations of the observed flux at the spin frequency of the neutron star . such oscillations have been observed in the tails of bursts from many sources ( galloway et al . the r.m.s . amplitudes of burst oscillations in the tails of bursts can be as large as 15% , although the typical amplitude is @xmath43% ( muno , zel , & chakrabarty 2002 ) . the stringent upper limits on the observed amplitudes at the harmonics of the spin frequencies can be accounted for only if the temperature anisotropies are dominated by the @xmath49 mode in which exactly half the neutron star is hotter than the other half ( muno et al . moreover , the rather weak dependence of the r.m.s.amplitudes on photon energy ( muno , zel , & chakrabarty 2003 ) requires that any temperature variation between the hotter and cooler regions of the neutron star is @xmath50 kev , even for the bursts that show the largest burst oscillation amplitudes . all of the above strongly suggest that the expected flux anisotropy during the cooling tail of an x - ray burst is @xmath51% and , therefore , the expected systematic uncertainties in the inferred apparent stellar radius will be half of that value , since the latter scales as the square root of the flux . on the color temperature of the x - ray spectrum for two sequences of models by madej et al . ( 2004 ) , majczyna et al . ( 2005 ) , and suleimanov et al . ( 2011 ) with different metal abundances and surface gravities . between 1 kev and 2.5 kev , the color correction factor depends weakly on the color temperature . + + ] a final inherent systematic uncertainty in the spectroscopic measurement of the apparent surface area of a neutron star arises from the dependence of the color correction factor on the effective temperature of the atmosphere . in figure [ fig : fcolor ] we show the predicted evolution of the color correction factor as a function of the color temperature of the spectrum as measured by an observer at infinity , for calculations by different groups for different neutron star surface gravities and atmospheric compositions ( madej et al.2004 ; majczyna et al . 2005 ; and suleimanov et al . 2011 ) . to make the models comparable to the observed evolution of the blackbody normalizations presented in the next section , we plot the color correction factor against the most directly observed quantity , i.e. , the color temperature at infinity . when the color temperature at infinity is less than 2.5 kev , purely helium or low metallicity models show @xmath52 evolution of the color correction factor per kev of color temperature at infinity , while above 2.5 kev , they show an evolution of 12@xmath120% per kev . in contrast , solar metallicity models show a steady increase with color temperature in the @xmath53 kev range . ) which correspond only to neutron stars with radii @xmath54 km . ] we infer the apparent surface area @xmath55 of a neutron star by measuring the x - ray flux @xmath56 and the color temperature @xmath57 at different intervals during the burst , so that @xmath58 here , @xmath59 is the stefan - boltzmann constant , @xmath60 is the distance to the source , and @xmath61 is the color correction factor . the value of the color correction factor is dictated by the predominant source of absorption and emission in the neutron - star atmosphere . it , therefore , depends on the effective temperature , which determines primarily the ionization levels , and on the effective gravitational acceleration , which determines the density profiles of the atmospheric layers . both these quantities decrease as the cooling flux decreases . if we were to assume a constant value for the color correction factor , as it is customary , we would obtain a systematic change in the inferred apparent surface area as the cooling flux of the burst decreases with time . such variations have been discussed in damen et al . ( 1989 ) and in bhattacharyya , miller , & galloway ( 2010 ) . note that this systematic effect can be corrected , in principle , if the data are fit directly with detailed models of neutron - star atmospheres ( see , e.g. , majczyna & madej 2005 ) . a potentially important source of uncertainty in the measurements is introduced by the errors in the absolute flux calibration of the rxte / pca detector . the current calibration of the pca and the cross - calibration between x - ray satellites have been carried out using the crab nebula as a standard candle ( jahoda et al . 2006 ; see also toor & seward 1974 ; kirsch et al . 2005 ; weisskopf et al . the uncertainties in the flux calibration can be due to a potential overall offset between the inferred and the true flux of the crab nebula , which may be as large as 10% ( kirsch et al . 2005 ; weisskopf et al . this can only change the mean value of the inferred apparent area in each source and does not alter the observed dispersion . we will explore this issue as well as uncertainties related to the variability of the crab nebula itself ( wilson - hodge et al . 2011 ) in more detail in paper iii of this series . + figure [ fig : kt_flux_detail ] ( left panels ) show the dependence of the emerging flux on color temperature for all the spectra in the cooling tails of ks 1731@xmath1260 , 4u 1728@xmath134 , and 4u 1636@xmath1536 that we consider to be statistically acceptable ( see 3 ) . we chose these three sources to use as detailed examples because of the high number spectra obtained for each and the fact that they span the range of behavior in cooling tracks that we will discuss below . if the whole neutron star is emitting as a single - temperature blackbody and the color correction factor is independent of color temperature , then @xmath56 should scale as @xmath62 . our aim here is to investigate the systematic uncertainties on the measurement of the apparent surface area in each source at different flux levels . we , therefore , divided the data into a number of flux bins and plotted in the same figure ( right panels ) the distribution of the blackbody normalization values for some representative bins . the blackbody normalization for each spectrum is defined formally as @xmath63 , although in practice this is one of the two measured parameters and the flux is derived from the above definition . according to equation ( [ eq : rapp ] ) , the blackbody normalization is equal to @xmath64 and we report it in units of ( km/10 kpc)@xmath4 . if we do not correct for the dependence of the color correction factor on the flux , we expect the normalization @xmath18 to show a dependence on color temperature that is the mirror symmetric of that shown in figure 2 . the flux - temperature diagrams of ks 1731@xmath1260 , 4u 1728@xmath134 , and 4u 1636@xmath1536 share a number of similarities but are also distinguished by a number of differences . in all three cases , the vast majority of data points lie along a very well defined correlation . this reproducibility of the cooling curves of tens of x - ray bursts per source , combined with the lack of large amplitude flux oscillations during cooling tails of bursts , provides the strongest argument that the thermal emission engulfs the entire neutron - star surface with very small temperature variations at different latitudes and longitudes . in 4u 1728@xmath134 , a deviation of the data points from the @xmath65 correlation is evident at high fluxes and color temperatures ( @xmath66 kev ) , which may be due to the evolution of the color correction factor at high eddington fluxes ( see the discussion in 4 and fig . [ fig : fcolor ] ) . the same deviation is not evident in ks 1731@xmath1260 or 4u 1636@xmath1536 , for which the highest temperatures encountered in the cooling tails were @xmath67 kev . finally , in all three sources , a number of outliers exist at the lowest flux levels , with normalizations that deviate from the above correlation . in ks 1731@xmath1260 and 4u 1636@xmath1536 , the outliers correspond to higher normalization values , whereas in 4u 1728@xmath134 they correspond to lower normalization values with respect to the majority of the data points . any combination of the effects discussed earlier in this and in the previous section may be responsible for the outliers . non - uniform cooling of the neutron star surface will lead to a reduced inferred value for the apparent surface area . reflection of the surface x - ray emission off a geometrically thin accretion disk will cause an increase in the inferred value for the apparent surface area . finally , comptonization of the surface emission in a corona may have either effect , depending on the compton temperature of the radiation . our goal in this work is not to understand the origin of the outliers , but rather to ensure that their presence does not introduce any biases in the measurements of the apparent surface areas inferred from the vast majority of spectra . indeed , including the outliers in a formal fit of the flux - temperature correlation will cause a systematic increase of the apparent surface area with decreasing flux in ks 1731@xmath1260 and in 4u 1636@xmath1536 and a systematic decrease in 4u 1728@xmath134 ( c.f . damen et al . 1990 ; bhattacharyya et al . this issue was largely avoided in our earlier work ( zel et al.2009 ; gver et al . 2010a , 2010b ) by considering only the relatively early time intervals in the cooling tail of each burst , which correspond only to the brightest flux bins . in order to go beyond this limitation here , we consider all spectral data for each burst and employ a bayesian gaussian mixture algorithm , which is a standard procedure for outlier detection in robust statistics ( e.g. , titterington , smith , & makov 1985 ; mclachlan & peel 2000 ; huber & ronchetti 2009 ) . our working hypothesis that the main peak of normalizations in each flux interval corresponds to the signal and the remainder are outliers is shaped by two aspects of the observations . first , at high fluxes , each histogram can be described by a single gaussian with no evidence or room for a second distribution of what we would call outliers . at lower fluxes , when the histograms can be decomposed into two gaussians , the distribution with the highest peak has properties that smoothly connect to those of the single gaussians at higher fluxes . second , the gaussians that we call our signal always contain the majority of the data points compared to the distribution of what we call the outliers . we take these as our criteria for defining our signal . the algorithm , which we describe below in detail , allows us also to measure the degree of intrinsic variation in the apparent surface area for each source that is consistent with the width of the flux - temperature correlation . even though we compare spectra in relatively narrow flux bins , the different observing modes as well as the different number of pcus used in each observation result in a wide range of count rates and , hence , in a wide range of formal errors . this necessitates the use of the bayesian approach we describe below as opposed to , e.g. , performing @xmath33 fits of the distributions over blackbody normalization . consider first the situation in which there are no systematic uncertainties and the inferred blackbody normalization from every spectrum reflects the apparent surface area of the entire neutron star . if this were the case , the intrinsic distribution of blackbody normalizations would be a delta function , while the observed distribution would be broader , with a width equal to the average formal uncertainties of the measurements . in a more realistic situation , there is an intrinsic range of the inferred blackbody normalizations , caused by a combination of the various effects discussed earlier . we assume that , in each flux bin , the intrinsic distribution of normalization values is the gaussian @xmath68\ ; , \label{eq : int_dist}\ ] ] where @xmath69 is its mean and @xmath70 is its standard deviation . the observed distribution of blackbody normalizations will be again broader than the intrinsic distribution because of the formal uncertainties in each measurement . moreover , at low flux levels , a small number of outliers exists , which skews and introduces tails to the observed distribution . our goal here is to quantify the most probable value of the blackbody normalization , i.e. , the parameter @xmath69 in the distribution ( [ eq : int_dist ] ) , as well as the intrinsic range of values in each flux bin , i.e. , the parameter @xmath70 , while excluding the outliers . [ fig:1728_detail ] when the inspection of an observed distribution requires an outlier detection scheme ( as , e.g. , in the lower right panels of fig . [ fig : kt_flux_detail ] ) , we model the distribution of outliers by a similar gaussian , i.e. , @xmath71\;,\ ] ] with a mean @xmath72 and a standard deviation @xmath73 . our model distribution function is , therefore , in general the gaussian mixture @xmath74 where @xmath75 measures the relative fraction of outliers in our sample . we are looking for the parameters in these distributions that maximize the posterior probability @xmath76 , which measures the likelihood that a particular set of parameters is consistent with the data . using bayes theorem , this probability distribution is @xmath77 where @xmath78 is the posterior probability that the particular set of normalization data have been measured for a given set of parameters and the last five terms are the prior probabilities for the model parameters ; the constant @xmath79 ensures that the probability density is normalized . we consider a flat prior probability distribution for the central value of each gaussian in an interval that extends beyond the observed range of normalization values for each source . we also consider a flat prior probability distribution for the standard deviation for each gaussian from @xmath80 ( km/10 kpc)@xmath4 ( which is practically zero ) to a value equal to the width of the range of normalizations . finally , we take a flat prior distribution of @xmath75 between zero and one . our results are extremely weakly dependent on the ranges of parameter values we considered . given a set of @xmath81 observed normalization values @xmath82 and their uncertainties @xmath83 , we calculate the posterior probability of the data given a set of model parameters as @xmath84 \left[p_{\rm int}(a| a_{\rm int } , \sigma_{\rm int } ) + \eta p_{\rm out}(a| a_{\rm out } , \sigma_{\rm out})\right]\;. \label{eq : post_prob_data}\ ] ] here , @xmath82 is the value of each measurement and @xmath83 is its corresponding uncertainty , which we assume to be normally distributed ; @xmath85 is a normalization constant . using this posterior probability distribution , we identify the most probable values of the model parameters , as well as their uncertainties , following standard procedures . figure [ fig:1728_detail ] shows , as an example , the application of the gaussian mixture algorithm to the distribution of blackbody normalizations obtained from fitting the spectra of 4u 1728@xmath134 , when the burst flux was in the range @xmath86 erg @xmath21 s@xmath24 . the histogram shows the distribution of blackbody normalizations measured in this flux interval . the parameters of the gaussian mixture that maximize the posterior probability distribution calculated using equation ( [ eq : post_prob ] ) correspond to an intrinsic distribution with a mean and standard deviation of @xmath87 ( km/10 kpc)@xmath4 and @xmath88 ( km/10 kpc)@xmath4 , respectively , and a distribution of outliers with a mean of @xmath89 ( km/10 kpc)@xmath4 , a standard deviation of @xmath90 ( km/10 kpc)@xmath4 , and a relative normalization of @xmath91 . in order to compare the gaussian mixture with the observed histogram , we first convolve it with a third gaussian distribution with a standard deviation of @xmath92 ( km/10 kpc)@xmath4 that is equal to the average formal errors of all measurements in this flux interval . the result is shown as a red solid line in figure [ fig:1728_detail ] , which demonstrates that our assumption of a gaussian distribution for the majority of the data and for the outliers is consistent with the observations . we can appreciate the importance of excluding the outliers from the sample using the bayesian gaussian mixture algorithm in the following way . if we repeat the above procedure for the same flux interval in 4u 1728@xmath134 but do not allow for the presence of outliers , the parameters of the intrinsic distribution will be @xmath93 ( km/10 kpc)@xmath4 and @xmath94 ( km/10 kpc)@xmath4 , which are significantly different than the results quoted above . moreover , had we assumed that there are no systematic errors but that the underlying distribution was a delta function , the most probable value would have been @xmath95 ( km/10 kpc)@xmath4 . the formal errors in such a measurement would have been substantially smaller compared to the systematic uncertainties . in the following , we show in detail the application of this gaussian mixture algorithm to all the flux intervals for the three sources . 260 , 4u 1728@xmath134 , and 4u 1636@xmath1536 , when the outliers have been removed . each dot represents the most likely centroid of the intrinsic distribution , while each error bar represents its most likely width , as calculated using the procedure outlined in 4.2 and depicted in figure [ fig:1728_detail ] . in each panel , the solid and dashed horizontal lines show the best - fit normalization and its systematic uncertainty inferred using the flux bins that do not correspond to near - eddington fluxes and are denoted by filled circles on the error bars.,title="fig : " ] 260 , 4u 1728@xmath134 , and 4u 1636@xmath1536 , when the outliers have been removed . each dot represents the most likely centroid of the intrinsic distribution , while each error bar represents its most likely width , as calculated using the procedure outlined in 4.2 and depicted in figure [ fig:1728_detail ] . in each panel , the solid and dashed horizontal lines show the best - fit normalization and its systematic uncertainty inferred using the flux bins that do not correspond to near - eddington fluxes and are denoted by filled circles on the error bars.,title="fig : " ] 260 , 4u 1728@xmath134 , and 4u 1636@xmath1536 , when the outliers have been removed . each dot represents the most likely centroid of the intrinsic distribution , while each error bar represents its most likely width , as calculated using the procedure outlined in 4.2 and depicted in figure [ fig:1728_detail ] . in each panel , the solid and dashed horizontal lines show the best - fit normalization and its systematic uncertainty inferred using the flux bins that do not correspond to near - eddington fluxes and are denoted by filled circles on the error bars.,title="fig : " ] [ fig : norm_vs_flux ] lccccccccccc & & + source & 0.51 & 12 & 23 & 34 & 45 & 56 & 67 & 78 & 89 & 910 & average + 4u 1636@xmath1536 & 112.6 & 124.6 & 126.9 & 142.7 & 144.2 & 145.7 & 145.7 & 149.5 & & & 130.7 + & 18.7 & 16.1 & 12.4 & 22.4 & 18.4 & 15.1 & 14.1 & 4.6 & & & 20.9 + 4u 1702@xmath1429 & 151.0 & 167.6 & 180.4 & 184.2 & 185.7 & 171.4 & 151.0 & 119.3 & 98.2 & & 176.6 + & 13.9 & 10.1 & 9.9 & 4.1 & 3.9 & 8.1 & 8.1 & 17.2 & 12.4 & & 11.6 + 4u 1705@xmath144 & 80.2 & 86.9 & 86.9 & 82.4 & & & & & & & 83.9 + & 9.9 & 7.1 & 10.9 & 7.4 & & & & & & & 9.1 + 4u 1724@xmath1307 & 98.7 & 108.3 & 120.4 & 127.4 & 125.4 & 102.8 & & & & & 113.8 + & 7.1 & 4.6 & 12.6 & 3.7 & 19.4 & 15.4 & & & & & 15.4 + 4u 1728@xmath134 & 114.1 & 126.1 & 137.4 & 145.0 & 135.2 & 137.4 & 133.7 & 126.1 & 117.8 & 93.7 & 134.4 + & 15.6 & 13.4 & 10.9 & 8.9 & 18.2 & 18.4 & 16.4 & 13.1 & 15.4 & 13.1 & 14.9 + ks 1731@xmath1260 & 72.6 & 89.9 & 93.0 & 95.2 & 95.2 & & & & & & 96.0 + & 6.1 & 5.4 & 4.1 & 9.6 & 1.9 & & & & & & 7.9 + 4u 1735@xmath144 & 71.6 & 71.6 & 72.6 & & & & & & & & 72.1 + & 4.6 & @xmath96 & @xmath97 & & & & & & & & @xmath98 + 4u 1746@xmath137 & 12.9 & 16.3 & 17.3 & 15.6 & 15.2 & 15.2 & 19.9 & & & & 15.7 + & 1.9 & 0.5 & 0.4 & 2.4 & 2.4 & 1.6 & 6.4 & & & & 2.4 + sax j1748.9@xmath12021 & 92.7 & 87.7 & 91.7 & 87.7 & & & & & & & 89.7 + & 11.9 & 6.6 & 7.6 & 15.4 & & & & & & & 9.6 + sax j1750.8@xmath12900 & 126.9 & 110.8 & 106.8 & 97.2 & 86.2 & & & & & & 93.2 + & 40.5 & 19.9 & 10.1 & 8.1 & 8.6 & & & & & & 9.4 + [ results ] [ cols="<,^,^",options="header " , ] figure [ fig : norm_vs_flux ] shows the result of the outlier detection procedure as applied to the various flux intervals of ks 1731@xmath1260 , 4u 1728@xmath135 , and 4u 1636@xmath1536 ( see also table [ results ] ) . each dot represents the most likely centroid of the intrinsic distribution of blackbody normalizations , while each error bar represents its most likely width . the normalization of the blackbody in the case of 4u 1728@xmath134 shows a strong dependence on flux when the source is emitting at near - eddington rates . this is not seen in ks 1731@xmath1260 and 4u 1636@xmath1536 , for which the color temperature never reached temperatures above 2.5 kev . in fact , in our observed sample , the strong dependence of the normalization on the flux is seen in all three sources with color temperatures in excess of 2.5 kev , namely 4u 1728@xmath134 , 4u 1702@xmath1429 , and 4u 1724@xmath1307 ( see fig . 6 below ) but is absent from the other sources . the evolution of the blackbody normalization at near - eddington fluxes depends on the local gravitational acceleration ( @xmath99 ) on the neutron - star surface and on its composition . therefore , modeling the decline of the blackbody normalization at large fluxes would allow us , in principle , to measure the combination @xmath100 of the mass and radius of the neutron star , as has been attempted by majczyna & madej ( 2005 ) and by suleimanov et al . ( 2011 ) . + + + + + + at low burst fluxes , the observed normalizations show a small albeit statistically significant trend towards lower values . in the case of 4u 1728@xmath134 , where the potential decrease is the largest , the normalization changed from @xmath101 ( km/10 kpc)@xmath4 to @xmath102 ( km/10 kpc)@xmath4 as the flux declined from @xmath103 erg s@xmath24 @xmath21 to @xmath104 erg s@xmath24 @xmath21 . this corresponds to a @xmath105% reduction in the normalization . in other sources , the decline is even smaller . this weak dependence seen in the data argues against the models with solar composition , as shown in figure [ fig : fcolor ] . the decline in the normalization can , in principle , be accounted for with a @xmath106% increase in the color correction factor ( see eq . [ [ eq : rapp ] ] ) towards low temperatures . as can be seen in figure [ fig : fcolor ] , current atmosphere models do not predict such an evolution . however , as we discussed in the beginning of 4 , a number of effects related to the physics of bursts on the neutron star surface ( uneven cooling and burst oscillations ) , as well as the reduced sensitivity of the pca at low energies , are capable of causing the observed trend in the normalization . moreover , the @xmath107 effect in the data we would aim to model is comparable to the uncertainty in inferring theoretically the color correction factor from the atmosphere models and is also comparable to the deviation of the observed and theoretical spectra from blackbodies . the latter concern can be remedied by fitting directly theoretical model spectra to the data , but reducing the theoretical uncertainties present in the models to less than @xmath108 few percent is significantly more challenging . an alternative approach is to allow for a range of values for the color correction factor that span the spread of theoretical uncertainties , metallicities , and fluxes of the cooling tails . in earlier work ( gver et al . 2010a , 2010b ) we allowed for a 7% range in the color correction factor ( from 1.3 to 1.4 ) , which is adequate to account for the evolution in the blackbody normalizations shown by the data . following this approach , we need to identify the range of blackbody normalizations we obtained from the data at different flux intervals as a systematic uncertainty in the measurement . in order to achieve this , we applied the bayesian gaussian mixture algorithm to the combined data set for each source for all flux intervals that correspond to a color temperature @xmath109 kev . this way , we computed the most probable value for the blackbody normalization for each source in a wide flux range as well as the systematic uncertainties in that measurement , which account for the potential decline of the normalization with decreasing flux . in figure [ fig : norm_vs_flux ] , we identify the flux intervals we used for each source with a filled circle in the middle of each error bar . we also depict with a solid line the most probable value of the normalization in this wide flux range and with dashed lines the range of systematic uncertainties . our results are summarized in table 3 . the blackbody normalizations for ks 1731@xmath1260 , 4u 1728@xmath134 , and 4u 1636@xmath1536 are [email protected] ( km/10 kpc)@xmath4 , [email protected] ( km/10 kpc)@xmath4 , and [email protected] ( km/10 kpc)@xmath4 , respectively , with the uncertainties dominated entirely by systematics . figures [ fig : radii1][fig : radii3 ] and tables [ chi2table][results ] show the results obtained after applying the procedure outlined in 4 to seven more sources from table [ sourcestable ] . the last two sources , 4u 0513@xmath140 and aql x-1 show large variations in the cooling tails of individual bursts and are discussed in detail in the appendix . moreover , in the appendix we also discuss a number of bursts from 4u 1702@xmath1429 and a long burst from 4u 1724@xmath1307 , which we did not include in the analysis . as in the case of the three sources discussed in the previous section , the vast majority of the spectra observed during the cooling tails of x - ray bursts are very well described by blackbody functions , with only marginal allowance for systematic variations . in the flux - temperature diagrams , the cooling tails largely follow a well defined and reproducible track . finally , in each flux interval of most sources , the range of blackbody normalizations obtained from a large number of bursts is consistent with a small degree of systematic uncertainties . the measurements of the average blackbody normalizations in all sources are dominated by systematic uncertainties ( the main exception is 4u 1735@xmath144 ) . however , these uncertainties are small , ranging from @xmath44% in the case of 4u 1702@xmath1429 to @xmath110% in the case of 4u 1636@xmath1536 . the apparent radius of each neutron star scales as the square root of the blackbody normalization . as a result , the errors in the spectroscopic determination of neutron - star radii introduced by systematic effects in the cooling tails of x - ray bursts are in the range @xmath0% ( see table [ radii_table ] ) . such small errors by themselves do not preclude distinguishing between different equations of state of neutron - star matter . a total of 46 bursts have been observed from the source 4u 1702@xmath1429 . six of these bursts reach high fluxes and have been categorized as photospheric radius expansion bursts by galloway et al . ( 2008a ) . the remaining 40 bursts typically reach lower fluxes . in the main body of this paper , we focused on the 6 bright bursts from this source and exclude the remaining for two reasons that we explore in this appendix . the left panel of figure [ fig:1702_appendix ] shows the histogram of @xmath33/dof values for the 880 spectra observed during the cooling tails of the 40 bursts and compares them to the expected distribution given the number of degrees of freedom . it is evident that blackbody functions provide statistically unacceptable fits to the majority of the spectra . had we not considered these spectra unacceptable , we would find that , approximately 4 seconds after the bursts start , the inferred blackbody normalizations increase rapidly to large values . the increase in normalization is , in fact , correlated with an increase in the @xmath111/dof values , as shown in the right panel of figure [ fig:1702_appendix ] , rendering them even more suspect . these two arguments strongly suggest that the spectra of the 40 weak x - ray bursts from 4u 1702@xmath1429 are not dominated by the thermal emission from the neutron star . the first x - ray burst observed from 4u 1724@xmath1307 on 1996 november 08 during observation 10090 - 01 - 01 - 02 is very different compared to the other two x - ray bursts that we used in the main text . the first burst is unusually long ( @xmath112 s ) , while the other two bursts are much shorter ( @xmath113 s range , as is typical for the vast majority of the type - i bursts observed ) . moreover , while the spectra of the short bursts are accurately modeled with blackbody functions ( see top panels of fig . [ fig : radii1 ] ) , fitting the spectra during the cooling tail of the long burst results in unacceptably large values of @xmath33/dof and the distritubiton of the resulting @xmath33/dof values do not follow the expected @xmath33 distribution ( see left panel of fig . [ fig:1724_appendix ] ) . adding a certain amount of systematic uncertainty to the data can in principle result in a decrease in the individual @xmath33/dof values ; however , it can not change the fact that the resulting @xmath33/dof distribution does not follow the expected distribution . if we went ahead and further analyzed these statistically unacceptable spectra , we would have obtained values for the blackbody normalization that are larger compared to those of the shorter bursts ( see right panel of fig . [ fig:1724_appendix ] ) . the large values and the seemingly random distribution of @xmath33/dof suggest that the x - ray spectra in the cooling tail of the long burst from 4u 1724@xmath1307 are not dominated by the thermal emission from the neutron star . in fact , int zand & weinberg ( 2010 ) found evidence for atomic edges and a reflection component in the spectra of this burst , which they attributed to the presence of heavy metals in the ashes of previous bursts that were exposed by the long radius - expansion episode . for these reasons , we excluded the long x - ray burst observed from 4u 1724@xmath1307 from our analysis . note that suleimanov et al.(2011 ) chose the spectra from this burst in order to infer the mass and radius of the neutron star in 4u 1724@xmath1307 . had they chosen the other two bursts from the same source , the spectra of which are actually well described by blackbody functions , they would have obtained a radius of the neutron star that is @xmath114% smaller , making their result consistent with the radii inferred for the other sources using this method ( zel et al . 2009 ; gver et al . 2010a , 2010b ) as well as with radii inferred from quiescent neutron stars in globular clusters ( e.g. , webb & barret 2007 ; guillot , rutledge , & brown 2010 ) . six bursts have been observed from the source 4u 0513@xmath140 . while most of the x - ray spectra during the cooling tails of these bursts can be described well by blackbody functions ( left panel of fig . [ fig:0514 ] ) , the flux - temperature diagrams show irregular behaviour for the cooling tails ( right panel of fig . [ fig:0514 ] ) . in particular , two of the bursts show a cooling tail that is reminiscent of the long burst from 4u 1724@xmath1307 discussed above : early in the cooling phase the temperature of the blackbody decreases while the flux remains constant and , when the flux starts decreasing , the cooling occurs with blackbody normalizations that are on average larger compared to the other bursts . it is plausible that the spectra of these bursts from 4u 0513@xmath140 are also affected by the presence of atomic lines , but the relatively low flux of 4u 0513@xmath140 ( which is a factor of @xmath115 smaller compared to that of 4u 1724@xmath1307 at similar temperatures ) still allows us to fit the observed spectra with blackbody functions . we do not consider this source suitable for a radius measurement based on the burst data . the transient source aql x-1 has shown 51 x - ray bursts , from which we extracted 2191 spectra . most of these spectra are well fit by blackbody functions , but their cooling tracks in the flux - temperature diagram depend strongly on the properties of the bursts ( see fig . [ fig : aqlx1 ] ) . in particular , a large number of bursts are relatively short and reach modest fluxes ( typically @xmath116 erg s@xmath24 @xmath21 ) with cooling tracks in the flux - temperature diagram that are reproducible . on the other hand , many bursts , which are relatively longer , reach fluxes that are a factor of @xmath117 higher , are often characterized by radius expansion episodes , and follow a range of cooling tracks in the flux - temperature diagram with blackbody normalizations that are typically larger than those of the other bursts . one plausible exlanation is related to the fraction of the neutron star surface that is engulfed by the thermonuclear flash . in weak flashes , the burning front may not propage across the entire surface and , therefore , only a fraction of the stellar surface contributes to the normalization of the blackbody . on the other hand , in stronger flashes , a larger fraction of the neutron star surface is covered . the source aql x-1 is one of the two main examples ( the other being the source 4u 1636@xmath1536 ) discussed by bhattacharrya et al . ( 2010 ) as evidence for systematic variations in the burning area during thermonuclear x - ray bursts . it is important to emphasize here , however , that the case of aql x-1 is very unusual compared to the remaining sources that we analyzed and is , perhaps , related to the presence of a weak but dynamically important magnetic field in this source , as inferred from the observation of intermittent persistent pulsations ( casella et al . 2008 ) . houck , j. c. , & denicola , l. a. 2000 , in astronomical society of the pacific conference series , vol . 216 , astronomical data analysis software and systems ix , ed . n. manset , c. veillet , & d. crabtree , 591+

we study inhibitory coherence ( i.e. , collective coherence by synaptic inhibition ) in an ensemble of globally - coupled type - i neurons which can fire at arbitrarily low frequencies . no inhibitory coherence is observed in a homogeneous ensemble composed of only subthreshold neurons ( which can not fire spontaneously without noise ) . by increasing the fraction of ( spontaneously firing ) suprathreshold neurons , heterogeneity - induced inhibitory coherence is investigated in a heterogeneous ensemble of subthreshold and suprathreshold neurons . as passes a threshold , suprathreshold neurons begin to synchronize and play the role of coherent inhibitors for the emergence of inhibitory coherence . thus , regularly - oscillating ensemble - averaged global potential appears for . for this coherent case suprathreshold neurons exhibit coherent mixed - mode oscillations with a fast subthreshold ( small - amplitude ) hopping frequency and a lower spiking frequency . by virtue of their coherent inhibition , sparsely synchronized suprathreshold neurons suppress noisy activities of subthreshold neurons . thus , only coherent subthreshold hoppings appear in the individual potentials of subthreshold neurons . we also characterize the inhibitory coherence in terms of the `` statistical - mechanical '' spike - based and correlation - based measures and find that the degree of inhibitory coherence increases with increasing for . finally , effect of sparse randomness of synaptic connectivity on the inhibitory coherence and universality of the heterogeneity - induced inhibitory coherence are briefly discussed .

1110.6565

recently , much attention has been paid to rhythms of the brain @xcite . coherence of neural oscillations may be used for efficient sensory and cognitive processing ( e.g. , feature integration , selective attention , working memory , and decision making ) @xcite . this kind of neural coherence is also correlated with pathological rhythms associated with neural diseases ( e.g. , epileptic seizures and tremors in the parkinson s disease ) @xcite . here , we are interested in these coherent brain rhythms . a brain circuit is composed of a few types of excitatory principal cells and diverse types of inhibitory interneurons . interneuron diversity increases the computational power of principal cells @xcite . effect of chemical synapses on coherent brain rhythms has been much investigated in neural systems composed of excitatory and/or inhibitory neurons @xcite . historically , recurrent excitation between principal cells is the conventional coherence mechanism @xcite . however , when the decay time of the synaptic interaction is enough long , mutual inhibition between interneurons ( rather than excitation ) may synchronize individual neural firings @xcite . by providing a coherent oscillatory output to the principal cells , interneuronal networks play the role of the backbones ( i.e. , pacemakers ) of many brain rhythms such as the thalamocortical spindle rhythms @xcite and the fast gamma rhythms in the hippocampus and the neocortex @xcite . when the feedback between the excitatory and the inhibitory populations is strong , neural coherence occurs via the `` cross - talk '' between the two populations @xcite . in these computational studies of neural coherence , different types of network architectures have been considered @xcite ; all - to - all networks where every neuron is coupled to every other neuron , sparse random networks where synaptic connections are sparse , and complex networks such as small - world networks ( with predominantly local connections and rare long - distance connections ) @xcite and scale - free networks ( with a few percent of hub neurons with an exceptionally large number of connections ) @xcite . neurons in the neural system exhibit a variety of morphological and physiological properties . however , close to threshold , this remarkable richness may be grouped broadly into two basic types of excitability , often referred to as type i and type ii @xcite . when the strength of a constant input current passes a threshold , type - i neurons can fire at arbitrarily low frequencies and they can smoothly encode the strength of the input into the output firing frequency . in contrast , type - ii neurons have a non - zero minimum frequency of firing and they fire in a narrow frequency band which is relatively insensitive to changes in the strength of the applied current . different types of excitability occur because neurons have different bifurcations of resting and spiking states @xcite . for the type i neurons , oscillations emerge via a saddle - node bifurcation on an invariant circle . as the bifurcation parameter ( i.e. , strength of the injected current ) passes a threshold , the stable and the unstable fixed points coalesce and then disappear , leaving a large - amplitude stable periodic orbit . this is a global bifurcation and the frequency of the global loop can be arbitrarily small . on the other hand , for type - ii neurons a transition from a resting state to a periodically spiking state occurs through a hopf bifurcations with a finite non - zero firing frequency . according to their bifurcations , neurons may also be classified into integrators and resonators @xcite . type - i neurons act as integrators without subthreshold oscillations , and they prefer high - frequency input : the higher the frequency of the input , the sooner they fire . in contrast , type - ii neurons exhibit damped subthreshold oscillations and act as resonators : they prefer oscillatory input with the same frequency as that of damped oscillations . according to their excitability type , neurons make distinctly different responses to stimuli which have important implications for their distinct roles in generating population rhythms @xcite . in this paper , we study inhibitory coherence ( i.e. , collective coherence by synaptic inhibition ) in an ensemble of globally - coupled type - i neurons . neural models exhibiting the type-1 excitability include the connor model for the crab leg axons @xcite , the wang - buzsaki model for inhibitory interneurons @xcite , the hindmarsh - rose model @xcite , and the morris - lecar ( ml ) model @xcite under some circumstances . in section [ sec : he ] , we describe the biological conductance - based ml neuron model with voltage - gated ion channels . the ml neurons ( used in our study ) exhibit the type - i excitability , and they interact via inhibitory gabaergic synapses whose activity increases fast and decays slowly . inhibitory coherence ( which is our main concern ) is particularly important because it plays a significant role in integration of sensory and cognitive information ; for example , impaired inhibitory coherence is believed to be associated with schizophrenia and attention deficit disorder @xcite . hence , it is important to understand mechanisms for the emergence of inhibitory coherence . many works exploring mechanisms of neural coherence were done in neural systems composed of spontaneously firing ( i.e. , self - oscillating ) suprathreshold neurons ( above the threshold ) @xcite . for this case , neural coherence occurs via cooperation of regular firings of suprathreshold neurons . unlike the suprathreshold case , subthreshold neurons ( below the threshold ) can not fire spontaneously without noise ; they can fire only with the help of noise . stochastic excitatory coherence ( i.e. , collective coherence between noise - induced spikings by synaptic excitation ) was observed in a population of excitatory subthreshold neurons @xcite . due to the stochastic excitatory coherence , synaptic current , injected into each individual neuron , becomes temporally coherent . hence , temporal coherence resonance of an individual subthreshold neuron in the network may be enhanced . furthermore , stochastic inhibitory coherence ( i.e. , collective coherence between noise - induced spikings by synaptic inhibition ) was also investigated in a population of inhibitory subthreshold ml neurons exhibiting the type - ii excitability @xcite . weak stochastic inhibitory coherence was thus found to appear via cooperation of individual irregular oscillations ( i.e. , a regular small - amplitude ensemble - averaged oscillation emerges from sparsely synchronized neurons discharging irregularly at lower rates than the network oscillation ) . these sparsely synchronized neural oscillations have been intensively investigated in other types of neural networks @xcite and they are believed to be associated with cortical rhythms in cognition [ e.g. ultrafast rhythm ( 100 - 200 hz ) , gamma rhythm ( 30 - 100 hz ) and beta rhythm ( 15 - 30 hz ) ] with irregular and sparse neural discharges @xcite . in contrast to the case of subthreshold type - ii ml neurons , no stochastic inhibitory coherence is observed in a homogeneous population of subthreshold type - i ml neurons . hence , subthreshold type - i integrator neurons ( without subthreshold oscillations ) seem to be much more difficult to synchronize by inhibition than subthreshold type - ii resonator neurons ( exhibiting subthreshold oscillations ) . to take into consideration the effect of ( spontaneously firing ) suprathreshold neurons on the inhibitory coherence , we consider a heterogeneous inhibitory ensemble of subthreshold and suprathreshold type - i ml neurons . heterogeneity ( or diversity ) has been found to make constructive effects on collective coherence in various physical , biological , neural , and social systems @xcite . in section [ sec : ic ] , we investigate heterogeneity - induced inhibitory coherence by increasing the fraction of suprathreshold neurons @xmath0 in the whole population . as @xmath0 passes a threshold value @xmath1 suprathreshold neurons begin to synchronize and they play the role of coherent inhibitors for the emergence of inhibitory coherence in the whole heterogeneous population . thus , for @xmath3 the ensemble - averaged global potential @xmath4 exhibits a regular small - amplitude oscillation . for this coherent case , individual suprathreshold neurons exhibit intermittent spikings phase - locked to @xmath4 at random multiples of the period of @xmath4 . due to the stochastic spike skipping of suprathreshold neurons , the interspike interval ( isi ) histogram has multiple peaks and partial occupation occurs in the raster plot of neural spikes . in addition to the coherent intermittent spikings , coherent subthreshold ( small - amplitude ) hopping oscillations also appear in the individual potentials of suprathreshold neurons . thus , sparsely synchronized suprathreshold neurons exhibit coherent mixed - mode oscillations with two well - separated frequencies , a fast subthreshold hopping frequency imposed by the collective network oscillation and a lower firing frequency of individual suprathreshold neurons . by virtue of their coherent inhibition sparsely synchronized suprathreshold neurons suppress noisy activities of subthreshold neurons . thus , only coherent fast subthreshold hopping oscillations ( without spikings ) appear in the individual potentials of subthreshold neurons . we also characterize this heterogeneity - induced inhibitory coherence in terms of `` statistical - mechanical '' spike - based and correlation - based measures , and find that the degree of inhibitory coherence increases as @xmath0 is increased from @xmath1 . in a real brain , each neuron is coupled to only a certain number of neurons which is much smaller than the total number of neurons . the effect of sparseness of synaptic connectivity on the inhibitory coherence is briefly discussed by varying the average number of synaptic inputs per neuron @xmath5 in a random network . emergence of inhibitory coherence is thus found to persist until @xmath5 is larger than a threshold value @xmath6 . we also confirm the universality of the heterogeneity - induced inhibitory coherence in a population of canonical type - i quadratic integrate - and - fire neurons @xcite . this kind of heterogeneity - induced weak inhibitory coherence might be associated with cortical rhythms with stochastic and sparse neural discharges which contribute to cognitive functions in the cerebral cortex ( e.g. , information integration , working memory , and selective attention ) @xcite . finally , a summary is given in section [ sec : sum ] . in this section we describe the biological neuron model used in our computational study . we consider a heterogeneous inhibitory ensemble of @xmath7 globally - coupled subthreshold and suprathreshold type - i neurons . as an element in our neural system , we choose the conductance - based ml neuron model , originally proposed to describe the time - evolution pattern of the membrane potential for the giant muscle fibers of barnacles @xcite . the population dynamics in this neural network is governed by the following set of differential equations : @xmath8 [ eq : cml1 ] where @xmath9 , \\ w_{\infty}(v)&= & 0.5 \left [ 1+\tanh \left\ { ( v - v_3)/{v_4 } \right\ } \right ] , \\ \tau_{r}(v ) & = & 1/ \cosh \left\ { ( v - v_3)/(2v_4 ) \right\ } , \\ s_{\infty } ( v_i ) & = & 1/[1+e^{-(v_i - v_t)/\delta } ] . \end{aligned}\ ] ] [ eq : cml2 ] here , the state of the @xmath10th neuron at a time @xmath11 ( measured in units of ms ) is characterized by three state variables : the membrane potential @xmath12 ( measured in units of mv ) , the slow recovery variable @xmath13 representing the activation of the @xmath14 current ( i.e. , the fraction of open @xmath14 channels ) , and the synaptic gate variable @xmath15 denoting the fraction of open synaptic ion channels . in eq . ( [ eq : cml1]a ) , @xmath16 represents the capacitance of the membrane of each neuron , and the time evolution of @xmath12 is governed by four kinds of source currents . the total ionic current @xmath17 of the @xmath10th neuron consists of the calcium current @xmath18 , the potassium current @xmath19 , and the leakage current @xmath20 . each ionic current obeys ohm s law . the constants @xmath21 , @xmath22 , and @xmath23 are the maximum conductances for the ion and the leakage channels , and the constants @xmath24 , @xmath25 , and @xmath26 are the reversal potentials at which each current is balanced by the ionic concentration difference across the membrane . since the calcium current @xmath18 changes much faster than the potassium current @xmath19 , the gate variable @xmath27 for the @xmath28 channel is assumed to always take its saturation value @xmath29 . on the other hand , the activation variable @xmath13 for the @xmath30 channel approaches its saturation value @xmath31 with a relaxation time @xmath32 , where @xmath33 has a dimension of ms and @xmath34 is a ( dimensionless ) temperature - like time scale factor . each ml neuron is also stimulated by a dc current @xmath35 and a gaussian white noise @xmath36 [ see the 2nd and 3rd terms in eq . ( [ eq : cml1]a ) ] satisfying @xmath37 and @xmath38 , where @xmath39 denotes the ensemble average . the noise @xmath36 is a parametric one which randomly perturbs the strength of the applied current @xmath35 , and its intensity is controlled by the parameter @xmath40 . depending on the system parameters , the ml neuron may exhibit either type - i or type - ii excitability @xcite . throughout this paper , we consider the case of type - i excitability where @xmath41 @xmath42 @xmath43 @xmath44 and @xmath45 . ( for comparison , a result on the order parameter is given for the type - ii case where the values of the above parameters are the same as those in the type - i case except that @xmath46 , @xmath47 , @xmath48 , and @xmath49 . ) for the type - i case , a transition from a resting state to a spiking state occurs for @xmath50 @xmath51 via a saddle - node bifurcation on an invariant circle @xcite , and a firing begins at arbitrarily low frequency . on the other hand , a type - ii neuron exhibits a jump from a resting state to a spiking state through a subcritical hopf bifurcation for @xmath52 @xmath51 by absorbing an unstable limit cycle born via fold limit cycle bifurcation for @xmath53 @xmath51 @xcite , and hence the firing frequency begins from a non - zero value . here a spread in the value of the dc input current @xmath54 is taken into consideration , and thus for each subthreshold ( suprathreshold ) neuron a value of @xmath35 is randomly chosen with a uniform probability in the range of @xmath55 [ @xmath56 $ ] for the type - i case and in the range of @xmath57 [ @xmath58 $ ] for the type - ii case , where the value of the spread parameter @xmath59 is set as @xmath60 @xmath51 . we consider a heterogeneous inhibitory ensemble of @xmath7 globally - coupled subthreshold and suprathreshold ml neurons where the fraction of suprathreshold neurons is given by @xmath61 ( @xmath62 : number of suprathreshold neurons ) . the last term in eq . ( [ eq : cml1]a ) represents the synaptic coupling between neurons in the network . each neuron is connected to all the other ones through global synaptic couplings . @xmath63 of eq . ( [ eq : cml2]c ) represents such synaptic current injected into the @xmath10th neuron . here the coupling strength is controlled by the parameter @xmath64 and @xmath65 is the synaptic reversal potential . we use @xmath66 mv for the inhibitory synapse . the synaptic gate variable @xmath67 obeys the 1st order kinetics of eq . ( [ eq : cml1]c ) @xcite . here the normalized concentration of synaptic transmitters @xmath68 , activating the synapse , is assumed to be an instantaneous sigmoidal function of the membrane potential with a threshold @xmath69 in eq . ( [ eq : cml2]g ) , where we set @xmath70 mv and @xmath71 mv . the transmitter release occurs when the neuron emits a spike ( i.e. , its potential @xmath72 is larger than @xmath69 ) . for the inhibitory gabaergic synapse ( involving the @xmath73 receptors ) , the synaptic channel opening rate , corresponding to the inverse of the synaptic rise time @xmath74 , is @xmath75 @xmath76 , and the synaptic closing rate @xmath77 , which is the inverse of the synaptic decay time @xmath78 , is @xmath79 @xmath76 @xcite . hence , @xmath80 rises fast and decays slowly . numerical integration of eq . ( [ eq : cml1 ] ) is done using the heun method @xcite ( with the time step @xmath81 ms ) similar to the second - order runge - kutta method , and data for @xmath82 @xmath83 are obtained with the sampling time interval @xmath84 ms . for each realization of the stochastic process in eq . ( [ eq : cml1 ] ) , we choose a random initial point @xmath85 $ ] for the @xmath10th @xmath86 neuron with uniform probability in the range of @xmath87 , @xmath88 , and @xmath89 . in this section , we are concerned about inhibitory coherence in a heterogeneous ensemble of @xmath7 globally - coupled subthreshold and suprathreshold type - i ml neurons . by increasing the fraction of suprathreshold neurons @xmath0 , we investigate the heterogeneity - induced inhibitory coherence . ( hereafter , for convenience we omit the dimensions of @xmath54 , @xmath40 , and @xmath64 . ) we first consider a homogeneous population ( corresponding to the case of @xmath90 ) composed of only subthreshold type - i ml neurons , and study the inhibitory coherence by varying both the coupling strength @xmath64 and the noise intensity @xmath40 . emergence of inhibitory coherence may be well described by the ( ensemble - averaged ) global potential , @xmath91 in the thermodynamic limit @xmath92 , a collective state becomes coherent if @xmath93 @xmath94 is non - stationary ( i.e. , an oscillating global potential @xmath4 appears for a coherent case ) , where the overbar represents the time average . otherwise ( i.e. , when @xmath95 is stationary ) , it becomes incoherent . thus , the mean square deviation of the global potential @xmath4 ( i.e. , time - averaged fluctuations of @xmath4 ) , @xmath96 plays the role of an order parameter used for describing the coherence - incoherence transition @xcite . for the coherent ( incoherent ) state , the order parameter @xmath97 approaches a non - zero ( zero ) limit value as @xmath7 goes to the infinity . versus ( a ) both the coupling strength @xmath64 and the noise intensity @xmath40 and versus ( b ) @xmath40 for @xmath98 @xmath99 in @xmath7 globally - coupled inhibitory subthreshold type - i ml neurons . the value of @xmath35 for each subthreshold type - i neuron is randomly chosen with a uniform probability in the range of @xmath55 where @xmath50 @xmath51 and @xmath60 @xmath51 . ( c ) plots of @xmath97 versus @xmath40 for @xmath100 @xmath99 in @xmath7 globally coupled inhibitory subthreshold type - ii ml neurons . the value of @xmath35 for each subthreshold type - ii neuron is randomly chosen with a uniform probability in the range of @xmath57 where @xmath53 @xmath51 and @xmath60 @xmath51 . ] figure [ fig : order](a ) shows plots of the order parameter versus both the coupling strength @xmath64 and the noise intensity @xmath40 . as @xmath7 is increased , the order parameter tends to decrease , independently of @xmath64 and @xmath40 . an example of the order parameter is shown in fig . [ fig : order](b ) for @xmath98 . for any given @xmath40 , @xmath101 is found to decrease as @xmath7 is increased . hence , only incoherent states exist , irrespectively of @xmath40 . this is in contrast to the case of subthreshold type - ii neurons exhibiting inhibitory coherence . figure [ fig : order](c ) shows plots of the order parameter versus the noise intensity for the type - ii case of @xmath100 . unlike the type - i case coherent states exist in an intermediate range of noise intensity [ @xmath102 where the order parameter approaches a non - zero limit value as @xmath7 increases . to take into consideration the effect of ( spontaneously firing ) suprathreshold neurons on the inhibitory coherence , we consider a heterogeneous population consisting of subthreshold and suprathreshold type - i ml neurons for @xmath98 . for convenience , we set the value of noise intensity as @xmath103 and investigate the heterogeneity effect on the inhibitory coherence by increasing the fraction of suprathreshold neurons @xmath0 . globally - coupled inhibitory type - i ml neurons for @xmath98 @xmath99 and @xmath103 @xmath104 ; @xmath105 in ( b1)-(b5 ) and ( c1)-(c3 ) . the value of @xmath35 for each subthreshold ( suprathreshold ) type - i ml neuron is randomly chosen with a uniform probability in the range of @xmath55 [ @xmath56 $ ] where @xmath50 @xmath51 and @xmath60 @xmath51 . plots of the order parameter @xmath97 versus the fraction of suprathreshold neurons @xmath0 in ( a1 ) the whole population and in the two subpopulations of ( a2 ) the suprathreshold and ( a3 ) the subthreshold neurons . raster plots and time series of the global potential @xmath4 in the whole population for @xmath106 ( b1 ) 0 , ( b2 ) 0.2 , ( b3 ) 0.4 , ( b4 ) 0.6 , and ( b5 ) 1.0 . time series of the subensemble - averaged potentials @xmath107 and @xmath108 in the two subpopulations of the suprathreshold and the subthreshold neurons for @xmath106 ( c1 ) 0.2 , ( c2 ) 0.4 , and ( x c3 ) 0.6 . vertical dashed lines in ( c1)-(c3 ) represent the times at which local minima of @xmath4 appear . ] figure [ fig : ct](a1 ) shows plots of the order parameter @xmath97 versus @xmath0 in the whole population . as @xmath0 passes a threshold value @xmath109 , a transition from an incoherent to a coherent state occurs . as shown in fig . [ fig : ct](a1 ) , it is enough to consider only the case of @xmath105 for the study of inhibitory coherence because the order parameter @xmath97 becomes saturated for @xmath105 . hereafter we set the number of neurons as @xmath105 in all cases except the calculation of the order parameter . for an incoherent case of @xmath90 , the raster plot consists of randomly scattered sparse spikes and the global potential @xmath4 exhibits nearly a stationary irregular oscillation [ see fig . [ fig : ct](b1 ) ] ; the amplitude of @xmath4 decreases with further increase in @xmath7 . however , when passing the threshold @xmath1 partially - occupied `` stripes '' ( composed of spikes and indicating collective coherence ) appear in the raster plot together with regularly - oscillating small - amplitude @xmath4 with frequency @xmath110 @xmath111 hz ) [ see fig . [ fig : ct](b2 ) ] . as @xmath0 is further increased , both the pacing degree of spikes and the amplitude of @xmath4 ( representing the degree of collective coherence ) increase , as shown in figs . [ fig : ct](b3)-[fig : ct](b5 ) where @xmath112 ( b3 ) 14.3 hz , ( b4 ) 13.6 hz , and ( b5 ) 14.2 hz . this kind of weak inhibitory coherence also occurs in each subpopulation of the subthreshold and the suprathreshold neurons . as in the case of the whole population , emergence of inhibitory coherence in the subpopulations may be well described by the subensemble - averaged potentials @xmath107 and @xmath108 , @xmath113 where @xmath62 ( @xmath114 ) is the number of suprathreshold ( subthreshold ) neurons . then the order parameters @xmath115 and @xmath116 , defined by the mean square deviation of @xmath107 and @xmath108 , @xmath117 may be used for describing the coherence - incoherence transitions in the subpopulations of suprathreshold and subthreshold neurons , respectively . plots of @xmath115 and @xmath116 versus @xmath0 are shown in figs . [ fig : ct](a2 ) and [ fig : ct](a3 ) , respectively . coherent transition in each subpopulation occurs at the same threshold value @xmath118 . for the case of coherent states , not only @xmath107 but also @xmath108 exhibits regular oscillations whose amplitudes increase as @xmath0 is increased [ see figs . [ fig : ct](c1)-[fig : ct](c3 ) where vertical dashed lines in @xmath107 and @xmath108 denote the times at which local minima of @xmath4 appear ] . both @xmath107 and @xmath108 are phase - locked to @xmath4 . to further understand the emergence of inhibitory coherence , we examine the individual and the global output signals in the subpopulations of subthreshold and suprathreshold neurons . globally - coupled inhibitory type - i ml neurons for @xmath98 @xmath99 and @xmath103 @xmath104 . the value of @xmath35 for each subthreshold ( suprathreshold ) type - i ml neuron is randomly chosen with a uniform probability in the range of @xmath119 [ @xmath56 $ ] where @xmath50 @xmath51 and @xmath60 @xmath51 . the individual potential @xmath120 of the first neuron and the global potential @xmath107 in the subpopulation of suprathreshold neurons for @xmath106 ( a1 ) 0.1 , ( a2 ) 0.2 , ( a3 ) 0.4 , ( a4 ) 0.6 , and ( a5 ) 1.0 . vertical dashed lines in ( a2)-(a5 ) represent the times at which local minima of @xmath107 appear . ( b ) plot of the average firing probability @xmath121 versus the fraction of suprathreshold neurons @xmath0 . the individual potential @xmath120 of the first neuron and the global potential @xmath108 in the subpopulation of subthreshold neurons for @xmath106 ( c1 ) 0 , ( c2 ) 0.1 , ( c3 ) 0.2 , ( c4 ) 0.4 , and ( c5 ) 0.6 . vertical dashed lines in ( c3)-(c5 ) denote the times at which local minima of @xmath108 appear . isi histograms in the whole population for @xmath0 ( d1 ) 0 , ( d2 ) 0.2 , ( d3 ) 0.4 , ( d4 ) 0.6 , and ( d5 ) 1.0 ; each isi histogram is composed of @xmath122 isis and the bin size for the histogram is 5 ms . vertical dotted lines in ( d2)-(d5 ) denote integer multiples of @xmath123 ( period of @xmath4 ) . ] figures [ fig : coh](a1)-[fig : coh](a5 ) show the time series of the individual potential @xmath120 of the first neuron and the time series of the global potential @xmath107 in the subpopulation of the suprathreshold neurons . @xmath107 exhibits a regular small - amplitude oscillation for a coherent case , while it shows a nearly stationary irregular oscillation for an incoherent case . for the case of coherent states , individual suprathreshold neurons exhibit intermittent spikings phase - locked to @xmath107 at random multiples of the period of @xmath107 [ see figs . [ fig : coh](a2)-[fig : coh](a5 ) where dashed lines denote the times at which local minima of @xmath107 appear ] . this `` stochastic phase locking '' leading to stochastic spike skipping is well shown in the isi histogram with multiple peaks [ see figs . [ fig : coh](d2)-[fig : coh](d5 ) ] . in addition to these coherent intermittent spiking phases , coherent subthreshold ( small - amplitude ) hopping oscillations also appear in the individual potentials of suprathreshold neurons . thus , suprathreshold neurons exhibit coherent mixed - mode oscillations with two well - separated frequency scales , a fast subthreshold hopping frequency @xmath124 imposed by the collective network oscillation with frequency @xmath110 @xmath125 14 hz ) and a lower spiking frequency @xmath126 of individual suprathreshold neurons ; @xmath127 ( a2 ) 3.6 hz , ( a3 ) 2.8 hz , ( a4 ) 2.3 hz , and ( a5 ) 1.8 hz . these sparsely synchronized suprathreshold neurons play the role of coherent inhibitors for the emergence of inhibitory coherence in the whole heterogeneous population , as shown below . on the other hand , for the incoherent case only stochastic intermittent spikings occur without any coherent hoppings , as shown in fig . [ fig : coh](a1 ) . figure [ fig : coh](b ) shows the plot of the average firing probability @xmath121 versus @xmath0 in the subpopulation of subthreshold neurons ( i.e. , time - averaged fraction of firing subthreshold neurons in the subpopulation of subthreshold neurons ) . due to inhibition @xmath121 decreases dramatically with respect to @xmath0 . for @xmath128 , one can disregard spikings of subthreshold neurons because @xmath129 becomes very small . the time series of the individual potential @xmath120 of the first neuron and the time series of the global potential @xmath108 in the subpopulation of the subthreshold neurons are shown in fig . [ fig : coh](c1)-[fig : coh](c5 ) . a regular oscillation with small amplitude occurs in @xmath108 for a coherent case , while @xmath108 exhibits a nearly stationary irregular oscillation for an incoherent case . for the case of coherent states , sparsely synchronized suprathreshold neurons suppress noisy activities of subthreshold neurons by virtue of their coherent inhibition , and then individual subthreshold neurons exhibit only coherent subthreshold hoppings ( without spikings ) , in contrast to the suprathreshold case . figures [ fig : coh](d1)-[fig : coh](d5 ) show the isi histograms in the whole population . ( as shown above , spiking neurons in the whole population are just suprathreshold ones for @xmath130 ) the isi histogram for @xmath90 has a very long tail , and hence the average value @xmath131 ms ) of isis is very large . as @xmath0 passes the threshold @xmath1 , multiple peaks tend to appear at integer multiples of @xmath123 ( period of @xmath4 ) [ i.e. , @xmath132 @xmath133 [ e.g. , see the isi histogram for @xmath134 ; vertical dotted lines in the histogram denote integer multiples of @xmath135 ms ) ] . as @xmath0 is further increased , isi histograms with more distinct multiple peaks appear due to the stochastic spike skipping of the suprathreshold neurons , as shown in figs . [ fig : coh](d3)-[fig : coh](d5 ) where @xmath123 = ( d3 ) 69.9 ms , ( d4 ) 73.3 ms , and ( d5 ) 70.3 ms . the most probable peak appears at @xmath136 , and hence suprathreshold neurons fire mostly in alternate global cycles . we characterize the heterogeneity - induced inhibitory coherence in terms of two kinds of `` statistical - mechanical '' spike - based and correlation - based measures . as shown in figs . [ fig : ct](b1)-[fig : ct](b5 ) , inhibitory coherence may be well visualized in the raster plot of spikes . for a coherent case , the raster plot is composed of partially - occupied stripes ( indicating collective coherence ) . to measure the degree of the collective coherence seen in the raster plot , a new spike - based measure @xmath137 was introduced by considering the occupation pattern and the pacing pattern of neural spikes in the `` stripes '' @xcite . particularly , the pacing degree between spikes is determined in a statistical - mechanical way by quantifying the average contribution of microscopic individual spikes to the global potential @xmath4 . the spiking coherence measure @xmath138 of the @xmath10th stripe is defined by the product of the occupation degree @xmath139 of spikes ( representing the density of the @xmath10th stripe ) and the pacing degree @xmath140 of spikes ( denoting the smearing of the @xmath10th stripe ) : @xmath141 the occupation degree @xmath139 in the @xmath10th stripe is given by the fraction of spiking neurons : @xmath142 where @xmath143 is the number of spiking neurons in the @xmath10th stripe . for the full occupation , @xmath144 , while for the partial occupation @xmath145 . the pacing degree @xmath140 of each microscopic spike in the @xmath10th stripe can be determined in a statistical - mechanical way by taking into consideration its contribution to the macroscopic global potential @xmath4 . each global cycle of @xmath4 begins from its left minimum , passes the central maximum , and ends at the right minimum ; the central maxima coincide with centers of stripes in the raster plot [ see figs . [ fig : ct](b2)-[fig : ct](b5 ) ] . an instantaneous global phase @xmath146 of @xmath4 is introduced via linear interpolation in the two successive subregions forming a global cycle . the global phase @xmath146 between the left minimum ( corresponding to the beginning point of the @xmath10th global cycle ) and the central maximum is given by : @xmath147 and @xmath146 between the central maximum and the right minimum ( corresponding to the beginning point of the @xmath148th cycle ) is given by : @xmath149 where @xmath150 is the beginning time of the @xmath10th global cycle ( i.e. , the time at which the left minimum of @xmath4 appears in the @xmath10th global cycle ) and @xmath151 is the time at which the maximum of @xmath4 appears in the @xmath10th global cycle . then , the contribution of the @xmath152th microscopic spike in the @xmath10th stripe occurring at the time @xmath153 to @xmath4 is given by @xmath154 , where @xmath155 is the global phase at the @xmath152th spiking time [ i.e. , @xmath156 . a microscopic spike makes the most constructive ( in - phase ) contribution to @xmath4 when the corresponding global phase @xmath155 is @xmath157 ( @xmath158 ) , while it makes the most destructive ( anti - phase ) contribution to @xmath4 when @xmath159 is @xmath160 . by averaging the contributions of all microscopic spikes in the @xmath10th stripe to @xmath4 , we obtain the pacing degree of spikes in the @xmath10th stripe , @xmath161 where @xmath162 is the total number of microscopic spikes in the @xmath10th stripe . by averaging @xmath138 of eq . ( [ eq : sm ] ) over a sufficiently large number @xmath163 of stripes , we obtain the spike - based coherence measure @xmath137 : @xmath164 globally - coupled inhibitory type - i ml neurons for @xmath98 @xmath99 and @xmath103 @xmath104 . ( a ) spike - based coherence measure @xmath137 : ( a1 ) plot of the average occupation degree @xmath165 versus the fraction of suprathreshold neurons @xmath0 , ( a2 ) plot of the average pacing degree @xmath166 versus @xmath0 , and ( a3 ) plot of the spiking coherence measure @xmath137 versus @xmath0 . to obtain @xmath165 , @xmath166 , and @xmath137 , we follow the @xmath167 stripes for each @xmath0 . ( b ) correlation - based coherence measure @xmath168 : ( b1 ) plot of @xmath168 versus @xmath0 in the whole population and plots of @xmath169 and @xmath170 versus @xmath0 in the two subpopulations of ( b2 ) suprathreshold and ( b3 ) subthreshold neurons . the number of data used for the calculation of a cross - correlation function for each @xmath0 is @xmath171 . ] by varying @xmath0 , we follow @xmath167 stripes and measure the degree of collective spiking coherence in terms of @xmath165 ( average occupation degree ) , @xmath166 ( average pacing degree ) , and @xmath137 for 13 values of @xmath0 in the coherent regime , and the results are shown in figs . [ fig : cm](a1)-[fig : cm](a3 ) . as @xmath0 is increased , the average occupation degree @xmath165 ( denoting the average density of stripes in the raster plot ) increases slowly , but its values are very small ( @xmath172 ) ; only a fraction ( less than 1/20 ) of the total neurons fire in each stripe [ see figs . [ fig : ct](b2)-[fig : ct](b5 ) ] . this partial occupation results from stochastic spike skipping of individual neurons seen well in the multi - peaked isi histograms [ see figs . [ fig : coh](d2)-[fig : coh](d5 ) ] . on the other hand , the average pacing degree @xmath166 increases rapidly near the threshold @xmath1 , and then it grows slowly . this tendency may be understood from the change in the structure of the isi histograms . as @xmath0 is increased , clear well - separated multiple peaks appear , and hence the average pacing degree of the stripes becomes better with increasing @xmath0 . in most region of the coherent region , the values of @xmath166 are large in contrast to @xmath165 . however , the spiking measure @xmath137 of eq . ( [ eq : cm ] ) ( representing the collective spiking coherence ) is very low due to the partial occupation in the raster plot . for the coherent case , subthreshold neurons exhibit only coherent subthreshold hoppings without spikings . hence , only suprathreshold neurons ( exhibiting coherent intermittent spikings ) make contribution to the spike - based measure @xmath137 . from now on , we use another statistical - mechanical measure based on the ensemble average of cross - correlations between the global potential and the individual potentials @xcite , and measure the degree of the inhibitory coherence in the subpopulations of the subthreshold and the suprathreshold neurons as well as in the whole population . the inhibitory coherence in the whole population is quantified in terms of the coherence measure @xmath168 given by the ensemble average of the global - individual cross - correlations @xmath173 between @xmath4 and @xmath12 at the zero - time lag : @xmath174 here the normalized cross - correlation function @xmath175 between @xmath4 and @xmath12 is given by @xmath176 where @xmath177 is the time lag , @xmath178 , @xmath179 , and the overline denotes the time average . this correlation - based measure @xmath168 can be regarded as a `` statistical - mechanical '' measure because it quantifies the average contribution of ( microscopic ) individual potentials to the ( macroscopic ) global potential . hence , @xmath168 is in contrast to the conventional microscopic measure based on the cross - correlations between the individual potentials . as in the case of the whole population , the degree of inhibitory coherence in each subpopulation of the subthreshold and the suprathreshold neurons may be well quantified in terms of the coherence measures @xmath170 and @xmath169 based on the cross - correlations between the global potentials ( @xmath108 and @xmath107 ) and the individual potentials , @xmath180 where @xmath181 by varying @xmath0 , we measure the degree of inhibitory coherence in terms of the correlation - based measures @xmath168 , @xmath169 , and @xmath170 not only in the whole population , but also in the subpopulations of the subthreshold and the suprathreshold neurons , and the results are shown in figs . [ fig : cm](b1)-[fig : cm](b3 ) . all of the coherence measures increase rapidly near the threshold @xmath1 , and then they grow slowly . the values of these correlation - based measures are very large in contrast to the spiking coherence measure @xmath137 . we also note that the degree of inhibitory coherence in the subpopulation of subthreshold neurons is higher than that in the subpopulation of suprathreshold neurons . this can be understood from the oscillating patterns of the global and the individual potentials . the global potentials @xmath107 and @xmath108 exhibit small regular oscillations [ see figs . [ fig : coh](a2)-[fig : coh](a5 ) and figs . [ fig : coh](c3)-[fig : coh](c5 ) ] . like the case of the global potential , the individual subthreshold neurons exhibit only coherent subthreshold hoppings , in contrast to the case of suprathreshold neurons exhibiting both the coherent intermittent spikings and the coherent hoppings . hence , the cross - correlations between @xmath108 and the individual potentials of subthreshold neurons become higher than those between @xmath107 and the individual potentials of suprathreshold neurons . versus the noise intensity @xmath40 for @xmath182 , ( b1)-(b5 ) raster plots of neural spikes and ( c1)-(c5 ) global potentials @xmath4 for various values of @xmath0 and @xmath40 in a heterogeneous ensemble of @xmath7 globally - coupled subthreshold and suprathreshold type - i ml neurons for @xmath98 @xmath99 ; @xmath105 in ( b1)-(b5 ) and ( c1)-(c5 ) . the value of @xmath35 for each subthreshold ( suprathreshold ) type - i ml neuron is randomly chosen with a uniform probability in the range of @xmath119 [ @xmath56 $ ] where @xmath50 @xmath51 and @xmath60 @xmath51 . ] in the above , we study the heterogeneity - induced inhibitory coherence for a fixed value of @xmath103 where @xmath183 . by varying the noise intensity @xmath40 we investigate the effect of noise on the inhibitory coherence for @xmath98 . for @xmath182 , plots of the order parameter @xmath97 versus @xmath40 are shown in fig . [ fig : noise](a ) . the degree of inhibitory coherence decreases monotonically with increasing @xmath40 from zero , and a transition to an incoherent state occurs when passing a threshold @xmath184 @xmath185 . figures [ fig : noise](b1)-[fig : noise](b5 ) and [ fig : noise](c1)-[fig : noise](c5 ) show the raster plots of spikes and the global potentials @xmath4 for various values of @xmath40 and @xmath0 . for @xmath182 , with increasing @xmath40 the stripes in the raster plot become more smeared and the amplitude of @xmath4 decreases . eventually when passing the threshold @xmath184 , incoherent states appear ( i.e. , the raster plot consists of randomly scattered spikes and @xmath4 exhibits a nearly stationary irregular oscillation ) . hence , as @xmath40 is increased the value of @xmath1 ( threshold value of the fraction of suprathreshold neurons for the emergence of inhibitory coherence ) increases ; @xmath186 0.08 and 0.28 for @xmath187 and 15 , respectively . thus , for @xmath188 no inhibitory coherence emerges , as shown in the case of @xmath189 . for various values of @xmath0 and @xmath5 in a heterogeneous ensemble of @xmath7 @xmath190 randomly - coupled subthreshold and suprathreshold type - i ml neurons for @xmath98 @xmath99 and @xmath103 @xmath104 . ( c ) plots of the order parameter @xmath97 versus the effective average number of synaptic inputs per neuron @xmath191 @xmath192 for @xmath182 . the value of @xmath35 for each subthreshold ( suprathreshold ) type - i ml neuron is randomly chosen with a uniform probability in the range of @xmath119 [ @xmath56 $ ] where @xmath50 @xmath51 and @xmath60 @xmath51 . ] so far , we consider the globally - coupled case . however , in a real brain each neuron is coupled to only a certain number of neurons which is much smaller than the total number of neurons . due to the sparseness of the network architecture , the inhibitory coherence ( seen in the globally - coupled case ) is expected to be reduced or destroyed . it is often assumed in models that the coupling between neurons is random @xcite . we briefly investigate the effect of sparse random connectivity on the inhibitory coherence for @xmath98 and @xmath103 by varying the average number of synaptic inputs per neuron @xmath5 in a heterogeneous ensemble of @xmath7 randomly - coupled subthreshold and suprathreshold type - i ml neurons . figs . [ fig : random](a1)-[fig : random](a5 ) and [ fig : random](b1)-[fig : random](b5 ) show the raster plots of spikes and the global potentials @xmath4 for various values of @xmath5 and @xmath0 when @xmath105 . for @xmath182 , with decreasing @xmath5 the stripes of spikes in the raster plot become more smeared and the amplitude of @xmath4 decreases . eventually , incoherent states appear when passing a threshold @xmath6 ( i.e. , the raster plot is composed of randomly scattered spikes and @xmath4 shows a nearly stationary irregular oscillation ) . hence , as @xmath5 is decreased from @xmath193 ( corresponding to the globally - coupled case ) , a larger fraction of suprathreshold neurons is necessary for the appearance of inhibitory coherence ( e.g. , see the cases of @xmath194800 and 500 ) . thus , for @xmath195 inhibitory coherence disappears , as shown in the case of @xmath196 . as is well known , @xmath5 [ rather than @xmath197 ( i.e. , the connection probability per neuron ) ] plays an appropriate sparseness parameter for the coherent transition because there exists a fixed threshold value @xmath6 for large @xmath7 , independently of @xmath7 @xcite . ( in contrast , the threshold value of @xmath197 depends on @xmath7 . ) when @xmath182 , plots of the order parameter @xmath97 versus the effective average number of synaptic inputs per neuron @xmath191 @xmath198 are shown in fig . [ fig : random](c ) for @xmath199 and 3000 , where the correction term @xmath200 takes into account the finite network size effect @xcite . inhibitory coherence emerges when @xmath191 is larger than a threshold @xmath201 @xmath202 ; for the case of @xmath105 , @xmath203 . finally , we examine the universality of the heterogeneity - induced inhibitory coherence in a population of @xmath7 globally - coupled subthreshold and suprathreshold quadratic integrate - and - fire ( qif ) neurons @xcite . the qif neuron model is canonical in the sense that any type - i excitable systems can be transformed into the form of the qif model by a continuous change of variables . here , we use a qif model with parameters derived from the wang - buzsaki conductance - based model @xcite . these qif neurons interact via the same inhibitory gabaergic synapses as those in the case of the ml neurons . the dynamics of the membrane potential @xmath12 of each qif neuron is governed by the following set of differential equations : @xmath204 with the auxiliary after - spike resetting : if @xmath205 ( threshold potential for spiking ) , @xmath206 ( resetting potential ) , where @xmath11 denotes the time ( measured in units of ms ) , @xmath207 , @xmath208 , @xmath209 , @xmath210 , @xmath211 , and @xmath212 . the @xmath10th neuron is stimulated by the applied current @xmath213 and the white gaussian noise @xmath36 whose intensity is controlled by the parameter @xmath40 . the synaptic current injected into the @xmath10th neuron @xmath63 is given in eq . ( [ eq : syn ] ) where the coupling strength is controlled by the parameter @xmath64 , @xmath214 is the synaptic reversal potential , and the synaptic gate variable @xmath67 obeys the 1st order kinetics of eq . ( [ eq : cml1]c ) along with eq . ( [ eq : cml2]g ) with @xmath215 . for the single qif neuron , a transition from a resting to a spiking state occurs when @xmath213 passes the rheobase @xmath216 via a saddle - node bifurcation on an invariant circle . here a spread in the value of @xmath213 is taken into consideration , and thus for each subthreshold ( suprathreshold ) neuron a value of @xmath213 is randomly chosen with a uniform probability in the range of @xmath217 [ @xmath218 $ ] , where the value of the spread parameter @xmath59 is set as @xmath219 . by increasing the fraction of suprathreshold neurons @xmath0 , we investigate the heterogeneity - induced inhibitory coherence for @xmath220 and @xmath221 . globally - coupled inhibitory qif neurons for @xmath220 @xmath99 and @xmath221 @xmath104 ; @xmath105 in ( b1)-(b5 ) . the value of @xmath222 for each subthreshold ( suprathreshold ) qif neuron is randomly chosen with a uniform probability in the range of @xmath217 [ @xmath218 $ ] where @xmath223 @xmath51 and @xmath224 @xmath51 . plots of the order parameter @xmath97 versus the fraction of suprathreshold neurons @xmath0 in ( a ) . raster plots of neural spikes and time series of the global potential @xmath4 for @xmath106 ( b1 ) 0 , ( b2 ) 0.1 , ( b3 ) 0.3 , ( b4 ) 0.5 and ( b5 ) 1.0 . ] figure [ fig : qif](a ) shows plots of the order parameter @xmath97 versus @xmath0 . as @xmath0 passes a threshold value @xmath225 , a transition from an incoherent to a coherent state occurs like the case of ml neurons [ see fig . [ fig : ct](a1 ) ] . for an incoherent case of @xmath90 , the raster plot consists of randomly scattered sparse spikes and the global potential @xmath4 exhibits a nearly stationary irregular oscillation , as shown in fig . [ fig : qif](b1 ) . however , when passing the threshold @xmath1 partially - occupied `` stripes '' appear in the raster plot along with regularly - oscillating small - amplitude global potential @xmath4 [ see fig . [ fig : qif](b2 ) for @xmath226 . as @xmath0 is further increased , both the pacing degree of spikes and the amplitude of @xmath4 ( representing the degree of inhibitory coherence ) increase , as shown in figs . [ fig : qif](b3)-[fig : qif](b5 ) . we have studied the heterogeneity - induced inhibitory coherence by increasing the fraction of suprathreshold neurons @xmath0 in an ensemble of globally - coupled subthreshold and suprathreshold type - i ml neurons . for @xmath90 no inhibitory coherence has been observed , which implies that subthreshold type - i neurons are difficult to synchronize by synaptic inhibition . however , as @xmath0 passes a threshold value @xmath1 , a coherent transition occurs in the subpopulation of suprathreshold neurons , and these synchronized suprathreshold neurons play the role of coherent inhibitors for the emergence of inhibitory coherence in the whole heterogeneous population . consequently , for @xmath3 a regular population rhythm with small amplitude appears in the ensemble - averaged global potential @xmath4 . for this coherent case , both the coherent intermittent spiking and the coherent subthreshold ( small - amplitude ) hopping phases appear in the individual potentials of suprathreshold neurons . thus , these sparsely synchronized suprathreshold neurons exhibit coherent mixed - mode oscillations with two well - separated frequency scales ( i.e. , fast subthreshold frequency and lower spiking frequency ) . by virtue of their coherent inhibition , sparsely synchronized suprathreshold neurons suppress noisy activities of subthreshold neurons . hence , only the coherent fast subthreshold hopping phase appears in the individual potentials of subthreshold neurons . the inhibitory coherence has been characterized in terms of the `` statistical - mechanical '' coherence measures based on spikes and correlations . the degree of inhibitory coherence was thus found to increase as @xmath0 is increased for @xmath3 . the effect of sparse random synaptic connectivity on the inhibitory coherence has also been investigated , and emergence of inhibitory coherence has thus been found to persist only if @xmath5 is larger than a threshold value @xmath6 . finally , the universality of heterogeneity - induced inhibitory coherence has been confirmed in a population of canonical type - i qif neurons . this kind of heterogeneity - induced weak inhibitory coherence which emerges from sparsely synchronized oscillations of suprathreshold neurons might be associated with cortical rhythms with irregular and sparse neural firings which contribute to cognitive functions such as information integration , working memory , and selective attention . impaired sparse synchronization is believed to be related to mental disorders ( e.g. , schizophrenia and autism ) @xcite . s.y.k . completed the final version of the manuscript during his visit to the jeju national university , and he thanks profs . d. kang and d. kim for their hospitality . this research was supported by the basic science research program through the national research foundation of korea funded by the ministry of education , science and technology ( 2010 - 0015730 ) . g. buzs@xmath227ki , _ rhythms of the brain _ ( oxford university press , new york , 2006 ) . wang , physiol . rev . * 90 * , 1195 ( 2010 ) . c. m. gray , j. comput . neurosci . * 1 * , 11 ( 1994 ) . b. hauschildt , n. b. janson , a. balanov , and e. sch@xmath228ll , phys . e * 74 * , 051906 ( 2006 ) ; p. grosse , m. j. cassidy , and p. brown , clin . neurophysiol . * 113 * , 1523 ( 2002 ) ; p. tass , m. g. rosenblum , j. weule , j. kurths , a. pikovsky , j. volkmann , a. schnitzler , and h. j. freud , phys . rev . 81 * , 3291 ( 1998 ) . wang , in _ encyclopedia of cognitive science _ , edited by l. nadel ( macmillan , london , 2003 ) , pp . 272 - 280 . f. bremer , physiol . rev . * 38 * , 357 ( 1958 ) ; r. d. traub , r. miles , and r. k. s. wong , science * 243 * , 1319 ( 1989 ) . c. van vreewijk , l. f. abbott , and g. b. ermentrout , j. comput . neurosci . * 1 * , 313 ( 1994 ) . d. hansel , g. mato , and c. meunier , neural comput . * 7 * , 307 ( 1995 ) . wang and j. rinzel , neural comput . * 4 * , 84 ( 1992 ) . d. golomb and j. rinzel , physica d * 72 * , 259 ( 1994 ) . wang and g. buzs@xmath227ki , j. neurosci . * 16 * , 6402 ( 1996 ) . j. white , c. c. chow , j. ritt , c. soto - trevino , and n. kopell , j. comput . neurosci . * 5 * , 5 ( 1998 ) . m. a. whittington , r. d. traub , n. kopell , b. ermentrout , and e. h. buhl , int . j. psychophysiol . * 38 * , 315 ( 2000 ) . p. h. e. tiesinga , j .- fellous , j. v. jose , and t. j. sejnowski , hippocampus * 11 * , 251 ( 2001 ) . d. hansel and g. mato , neural comput . * 15 * , 1 ( 2003 ) . c. b@xmath228rgers and n. kopell , neural comput . * 15 * , 509 ( 2003 ) ; _ ibid . _ * 17 * , 557 ( 2005 ) . g. buzs@xmath227ki , c. geisler , d. a. henze , and x .- j . wang , trends in neurosci . * 27 * , 186 ( 2004 ) ; s. yu , d. huang , w. singer , and d. nikolie , cereb . cortex * 18 * , 2891 ( 2008 ) ; q. wang , z. duan , m. perc , and g. chen , epl * 83 * , 50008 ( 2008 ) ; a. roxin , h. riecke , and s. solla , phys . lett . * 92 * , 198101 ( 2004 ) ; h. riecke , a. roxin , s. madrugs , and s. solla , chaos * 17 * , 026110 ( 2007 ) . q. wang , m. perc , z. duan , and g. chen , phys . e * 80 * , 026206 ( 2009 ) ; r. morgan and i. soltesz , proc . usa * 105 * , 6179 ( 2008 ) . a. l. hodgkin , j. physiol . * 107 * , 165 ( 1948 ) . j. rinzel and b. ermentrout , in _ methods in neural modeling : from ions to networks _ , edited by c. koch and i. segev ( mit press , cambridge , 1998 ) , p. 251 . e. m. izhikevich , int . chaos * 10 * , 1171 ( 2000 ) . d. hansel , g. mato , and c. meunier , neural comput . * 7 * , 307 ( 1995 ) . s. marella and g. b. ermentrout , phys . e * 77 * , 041918 ( 2008 ) ; b. ermentrout , neural comput . * 8 * , 979 ( 1996 ) . e. m. izhikevich , ieee trans . neural networks * 10 * , 499 ( 1999 ) . g. mato and i. samengo , neural comput . * 20 * , 2418 ( 2008 ) . n. malik , b. ashok , and j. balakrishnan , prana - j. phys . * 74 * , 189 ( 2010 ) . j. a. connor and c. f. stevens , j. phisiol . * 213 * , 31 ( 1971 ) . j. l. hindmarsh and r. m. rose , proc . london b * 221 * , 87 ( 1984 ) . c. morris and h. lecar , biophys . j. * 35 * , 193 ( 1981 ) . k. m. spencer , p. g. nester , m. a. niznikiewicz , d. f. shenton , and r. w. mccarley , j. neurosci . * 23 * , 7407 ( 2003 ) . p. j. uhlhaas and w. singer , nat . neurosci . * 11 * , 100 ( 2010 ) . y. wang , d. t. w. chik , and z. d. wang , phys . rev . e * 61 * , 740 ( 2000 ) . w. lim and s .- y . kim , j. korean phys . soc . * 51 * , 1427 ( 2007 ) ; int . b * 23 * , 703 ( 2009 ) . w. lim and s .- y . kim [ to appear in the j. comput . ( doi : 10.1007/s10827 - 011 - 0330 - 3 ) ] . n. brunel and v. hakim , chaos * 18 * , 015113 ( 2008 ) ; n. brunel and v. hakim , neural comput . * 11 * , 1621 ( 1999 ) ; n. brunel , j. comput . neurosci . * 8 * , 183 ( 2000 ) ; n. brunel and x .- j . wang , j. neurophysiol . * 90 * , 415 ( 2003 ) ; c. geisler , n. brunel , and x .- j . wang , j. neurophysiol . * 94 * , 4344 ( 2005 ) . c. j. tessone , c. r. mirasso , r. toral , and j. d. gunton , phys . * 97 * , 194101 ( 2006 ) ; m. perc , m. gosak , and s. kralj , soft matter * 4 * , 1861 ( 2008 ) ; c. j. tessone and r. toral , eur . j. phys . b * 71 * , 549 ( 2009 ) ; x. li , j. zhang , and m. small , chaos * 19 * , 013126 ( 2009 ) . g. b. ermentrout and n. kopell , siam j. appl . math . * 46 * , 233 ( 1986 ) ; e. m. izhikevich , ieee trans . on neural networks * 15 * , 1063 ( 2004 ) ; p. e. latham , b. j. richmond , p. g. nelson , and s. nirenberg , j. neurophysiol . * 83 * , 808 ( 2000 ) . m. san miguel and r. toral , in _ instabilities and nonequilibrium structures vi _ , edited by j. martinez , r. tiemann , and e. tirapegui ( kluwer academic publisher , dordrecht , 2000 ) , p. 35 . s. c. manrubia , a. s. mikhailov , and d. h. zanette , _ emergence of dynamical order _ ( world scientific , singapore , 2004 ) . w. lim and s .- y . kim , j. korean phy . * 57 * , 1290 ( 2010 ) . e. barkai , i. kanter , and h. sompolinky , phys . a * 41 * , 590 ( 1990 ) . wang , d. golomb , and j. rinzel , proc . natl . sci . * 92 * , 5577 ( 1995 ) . d. golomb and d. hansel , neural comput . * 12 * , 1095 ( 2000 ) .

consider a large uniformly mixing dynamic population , which has constant birth rate and exponentially distributed lifetimes , with mean population size . a markovian sir ( susceptible infective recovered ) infectious disease , having importation of infectives , taking place in this population is analysed . the main situation treated is where , keeping the basic reproduction number as well as the importation rate of infectives fixed , but assuming that the quotient of the average infectious period and the average lifetime tends to 0 faster than . it is shown that , as , the behaviour of the 3-dimensional process describing the evolution of the fraction of the population that are susceptible , infective and recovered , is encapsulated in a 1-dimensional regenerative process describing the limiting fraction of the population that are susceptible . the process grows deterministically , except at one random time point per regenerative cycle , where it jumps down by a size that is completely determined by the waiting time since the previous jump . properties of the process , including the jump size and stationary distributions , are determined .

1506.02540

the mathematical theory for the spread of infectious diseases has a long history and is by now quite rich ( e.g. , @xcite ) . one of the more common type of disease models is called sir ( susceptible @xmath1 infective @xmath1 recovered ) meaning that individuals are at first susceptible . if infected ( by someone ) they immediately become infectious ( being able to spread the disease onwards ) . after some time an infectious individual recovers , which also means that the individual is immune to further infection from the disease . such models were originally studied for populations assuming homogeneous mixing , but during the last few decades considerable effort has been put into analysing epidemic models in communities which are not homogeneously mixing but instead may be described using some type of social structure , such as a community of households ( e.g. @xcite ) or a random network describing possible contacts ( e.g. @xcite ) . the vast majority of papers devoted to these type of problems assume a fixed community and community structure . in the current paper we treat the situation where the population is dynamic in the sense that people die and new individuals are born , or more precisely immigrate into the population . further , we assume that there is also importation of infectious individuals ( randomly in time according to a homogeneous poisson process ) , implying that the disease never vanishes forever . in order to facilitate analytical progress we consider only the case of a homogeneously mixing community , which in network terminology corresponds to treating the complete network . models for recurrent epidemics go back to the deterministic formulations of @xcite and @xcite . a stochastic treatment was given first in the pioneering work of @xcite , who considered an sir model with importation of both susceptibles and infectives , but without disease - unrelated deaths . an alternative model , with disease - unrelated deaths but no importation of infectives , has been studied extensively ( e.g. @xcite and the references therein ) . interest often centres on the time to extinction of infection and the closely - related problem of the critical community size for an infection to persist in a population . we consider a markovian sir epidemic with demography and importation of infectives , in which infectious individuals infect new individuals at constant rate and the infectious period is exponentially distributed . we study limit properties of the epidemic when the average population size @xmath0 tends to infinity . our focus lies on the case where the limit is taken keeping the basic reproduction number @xmath3 ( i.e. the average number of susceptibles infected by a single infective in an otherwise fully susceptible population of size @xmath0 ) and the immigration rate of infectives fixed , whereas the quotient of the average infectious period and the average lifetime tends to 0 faster than @xmath4 . for many infectious diseases this quotient typically lies between @xmath8 and @xmath9 , hence supporting this asymptotic regime , but in the discussion we treat other asymptotic regimes briefly . under the above asymptotic regime , all epidemic outbreaks are short , having duration that tends to @xmath10 in probability as @xmath11 further , as @xmath11 , epidemic outbreaks are either minor , having size of order @xmath12 , or major , having size of exact order @xmath13 . it follows that , as @xmath11 , the behaviour of the three - dimensional process describing the evolution of the fraction of the population that are susceptible , infective and recovered , is encapsulated in a one - dimensional regenerative process @xmath6 , describing the limiting fraction of the population that are susceptible . during each cycle , the process @xmath7 makes one down jump , corresponding to the occurrence of a major outbreak , and except for this increases deterministically , as minor outbreaks have no effect on @xmath14 in the limit as @xmath11 . ( here , @xmath15 , where , for @xmath16 , @xmath17 with @xmath18 being the number of susceptible individuals in the population at time @xmath19 . ) note that @xmath14 does not converge weakly to @xmath7 in the skorohod topology since the sample paths of @xmath7 are almost surely discontinuous but those of @xmath14 almost surely contain only jumps of size @xmath20 , so are close to being continuous . thus to obtain rigorous convergence results , we consider two processes , @xmath21 and @xmath22 , which coincide with @xmath14 , except during major outbreaks during which they sandwich @xmath14 , and prove that both @xmath21 and @xmath22 converge weakly to @xmath7 in the skorohod topology ( theorem [ wconv ] ) . it then follows that certain functionals of @xmath14 converge weakly to corresponding functionals of @xmath7 ( corollary [ convfunct ] ) . the paper is structured as follows . in section [ modelmain ] , we define the model and the limiting regenerative process , give an intuitive explanation of why @xmath7 approximates @xmath14 for large @xmath0 and present the main convergence results . in section [ s - properties ] , we derive some properties of the limiting regenerative process : the jump size distribution , the associated renewal time distribution and the stationary distribution . in section [ illustrations ] , we present simulations supporting the convergence result and illustrating various features of the limiting process . in section [ proofs ] , we prove the main results . we end in section [ disc ] with a discussion summarising our results and also exploring briefly additional questions , such as other asymptotic regimes . we now define the markovian sir epidemic with demography and importation of infectives ( sir - d - i ) . we consider the process to be indexed by a target population size @xmath0 , which we assume is a strictly positive constant . the population model is an immigration - death process with constant immigration rate and linear death rate . for @xmath16 , let @xmath23 denote the population size at time @xmath19 . then @xmath23 increases at constant rate @xmath24 and decreases at rate @xmath25 . the population size hence fluctuates around @xmath0 , which is assumed to be large . the markovian sir - epidemic on this population is defined as follows . for @xmath16 , let @xmath26 and @xmath27 denote the number of susceptibles , infectives and recovered , respectively , at time @xmath19 , so @xmath28 . we assume that @xmath29 and that @xmath30 as @xmath11 , where @xmath31 $ ] is constant . ( the value of @xmath32 has no effect on the ensuing epidemic . ) . a fraction @xmath33 of all births ( i.e. immigrants ) are infectives and the remaining births are all susceptibles , so births of infectives occur at rate @xmath34 and births of susceptibles occur at rate @xmath35 . while infectious , any given infective infects any given susceptible at rate @xmath36 , independently between each distinct pair of individuals . thus , approximately , each infective makes infectious contacts at the points of a homogeneous poisson process having rate @xmath37 , with contacts being with individuals chosen independently and uniformly from the whole population ; a contact with a susceptible individual results in that individual becoming infected , while a contact with an infectious or removed individual has no effect . each infectious individual recovers and becomes immune at rate @xmath38 , implying that the infectious period is exponentially distributed with rate parameter @xmath38 . more formally , the process @xmath39 is a continuous - time markov chain , with state space @xmath40 and transition intensities given by @xmath41 corresponding to birth of a susceptible , birth of an infective , death of a susceptible , death of an infective , death of a recovered , infection of a susceptible and recovery of an infective , respectively . we study specifically the case where the average population size @xmath0 tends to infinity in such a way that * the total importation rate @xmath34 of infectives tends to a strictly positive constant @xmath42 , so @xmath43 as @xmath2 ; and * the infection and recovery rates satisfy @xmath44 and @xmath45 as @xmath2 . for ease of exposition , we assume that @xmath0 is an integer , so sequences of epidemic processes are indexed by the natural numbers . however , all of the results of the paper are easily generalised to the case of a family of epidemic processes indexed by the positive real numbers . to conclude , the parameters of the model are : @xmath0 , the average population size ; @xmath46 , where @xmath47 is the average lifetime and @xmath24 is the population birth rate ; @xmath37 , the infection rate ; @xmath38 , where @xmath48 is the average length of the infectious period ; and @xmath33 , the fraction of births which are infectious , so @xmath34 is the birth ( or importation ) rate of infectives . let @xmath49 , where @xmath17 is the fraction " of the population that is susceptible at time @xmath19 . the process @xmath6 can be viewed as the limit of @xmath14 as @xmath11 under the above asymptotic regime . it is a markovian regenerative process ( e.g. @xcite , chapter v ) , with renewals occurring whenever @xmath50 . between each renewal @xmath51 increases deterministically according to the differential equation @xmath52 except for one down jump ( from above @xmath53 to below @xmath53 ) . this implies that @xmath54 before the jump ( if @xmath55 denotes the time from the last renewal ) . the random time @xmath56 from a renewal to the jump has distribution specified by @xmath57\quad(t \ge 0),\ ] ] with @xmath58 given by , so @xmath59 the size of the jump is specified by the value @xmath60 of the process just prior to the jump . more precisely , @xmath61 , where for @xmath62 , @xmath63 is the unique strictly positive solution to the equation ( cf . @xcite , equation ( 3.15 ) ) @xmath64 in epidemic theory @xmath63 is known as the relative fraction infected among the initially susceptible of an sir epidemic outbreak in which a fraction @xmath65 are initially susceptible and the rest immune . hence , the size of the down jump is @xmath66 . after the down jump , @xmath51 increases deterministically according to the same differential equation until the next renewal point , so @xmath67\ ] ] and the inter - renewal time is @xmath68 $ ] . illustrations of @xmath7 are given in section [ illustrations ] . we first explain heuristically why @xmath7 can be viewed as the limit of @xmath14 as @xmath11 under that asymptotic regime described in section [ model ] . suppose that @xmath0 is large . then when no infective is present , all that happens is that individuals die and new ones are born at approximately the same rate @xmath24 . recovered ( immune ) individuals that die are replaced by susceptible individuals , so the fraction of susceptibles increases at rate @xmath69 which explains the deterministic growth rate of @xmath7 . after an exponentially distributed holding time , with rate parameter @xmath70 , an infective is born into the community . if the fraction susceptible @xmath71 is below @xmath53 , then the effective reproduction number @xmath72 is strictly less than one , implying that , with probability tending to one as @xmath11 , a large outbreak will not occur , so @xmath71 continues to grow approximately deterministically . if @xmath73 when a new born infective enters the community , then with approximate probability @xmath74 that infective gives rise to a major outbreak that infects order @xmath75 susceptibles ( cf . @xcite , pages 53 and 376 ) , otherwise only a minor outbreak , which infects order @xmath76 susceptibles , occurs and @xmath71 continues to grow approximately deterministically . this explains the distribution for @xmath56 , the time from a renewal until a down jump in @xmath7 , which has time varying intensity given by @xmath42 multiplied by the limiting major outbreak probability ( cf . @xcite ) . if a major outbreak takes place , the size of the outbreak among the susceptibles is given approximately by @xmath77 where @xmath60 denotes the limiting ( as @xmath78 ) fraction susceptible just prior to the outbreak and @xmath79 is defined above ( cf . @xcite , page 60 ) . the duration of such a major outbreak is of order @xmath80 ( cf . @xcite ) which tends to @xmath10 by assumption . thus , if there is a major outbreak it happens momentarily and , in the limit as @xmath11 , the fraction susceptible after the outbreak , @xmath81 , satisfies @xmath82 . although the above heuristic argument makes it plausible that the normalised susceptible process @xmath83 converges to the regenerative process @xmath7 , there are two complicating factors in making the argument fully rigorous . first , as explained in section [ sec_intro ] , it is not true that @xmath84 as @xmath5 , where @xmath85 denotes weak convergence in the space @xmath86 of right - continuous functions @xmath87 having limits from the left ( i.e. cdlg functions ) , endowed with the skorohod metric ( e.g. @xcite , chapter 3 ) . as explained also in section [ sec_intro ] , we overcome this problem by considering two processes , @xmath21 and @xmath22 , which coincide with @xmath14 except during major outbreaks , when they sandwich @xmath14 , and show that @xmath88 and @xmath89 as @xmath11 ; see theorem [ wconv ] . the second complicating factor is that the results referred to above concerning the probability , size and duration of a major outbreak are for an epidemic in a static population , whereas our population is dynamic . the results carry over to our setting because , in the limit as @xmath11 , the time scale of an epidemic outbreak is infinitely faster than that of demographic change , but proofs need to be adapted accordingly . before stating our main theorem , some more notation is required . recall that @xmath90 is the number of infectives at time @xmath19 in the sir - d - i epidemic with average population size @xmath0 and that we consider epidemics with no infective at time @xmath10 , i.e. with @xmath29 . let @xmath91 . for @xmath92 , let @xmath93 and @xmath94 . thus , provided @xmath0 is sufficiently large , the @xmath95th major outbreak starts at approximately time @xmath96 and ends at time @xmath97 . ( the choice of @xmath98 to delineate major outbreaks is essentially arbitrary . our proofs work equally well if @xmath98 is replaced by any function @xmath99 which satisfies @xmath100 and @xmath101 as @xmath11 . ) for @xmath16 , let @xmath102 and @xmath103 the following theorem is proved in section [ proof ] . [ wconv ] suppose that @xmath104 . then , as @xmath11 , @xmath105 where @xmath106 . an immediate consequence of theorem [ wconv ] is that suitable functionals of @xmath14 converge weakly to corresponding functionals of @xmath7 . for @xmath107 , let @xmath108 denote @xmath109 for all @xmath110 . a functional @xmath111 is called monotone if either @xmath112 for all @xmath113 satisfying @xmath114 , or @xmath112 for all @xmath113 satisfying @xmath115 . the following corollary , which can clearly be generalised to suitable non - real - valued functionals , follows immediately from theorem [ wconv ] by using the continuous mapping theorem ( e.g. @xcite ) . for @xmath111 , let @xmath116 . [ convfunct ] suppose that @xmath104 , @xmath111 is monotone and @xmath117 . then @xmath118 where @xmath106 . one functional which satisfies the conditions of corollary [ convfunct ] is the first passage time functional @xmath119 , defined for given @xmath120 by @xmath121 the functional @xmath119 is clearly monotone and @xmath122 , cf . @xcite , page 124 . another functional satisfying the conditions of corollary [ convfunct ] is the occupancy time functional @xmath123 , defined for any given @xmath124 and @xmath125 by @xmath126 this functional is again clearly monotone . the proof that @xmath127 is given at the end of section [ proof ] . we now outline some properties of the regenerative process @xmath7 which can be obtained from renewal and regenerative process theory ( e.g. @xcite , chapters iv and v ) . as described in section [ s(t ) ] the stochastic part of the regenerative process is completely specified by the waiting time @xmath56 until the down jump , but it can be specified equivalently by the jump size @xmath128 . noting that @xmath129 ) , it follows from that @xmath130 whence @xmath131 which can be used to obtain the distribution of the jump size @xmath132 . the jump size is strictly less than @xmath133 , as @xmath134 for all @xmath16 . hence , for @xmath135 , @xmath136 \right ) \quad(\mbox{using}~\eqref{sbeforejump})\nonumber \\ & = 1-\left[\frac{r_0\left(1-x-{\rm e}^{-r_0 x}\right)}{\left(r_0 - 1\right)\left(1-{\rm e}^{-r_0 x}\right)}\right]^\kappa \left[\frac{\left(r_0 - 1\right)x}{1-x-{\rm e}^{-r_0 x}}\right]^{\frac{\kappa}{r_0}}. \ ] ] the lifetime distribution for the renewal process describing successive visits of @xmath7 to @xmath53 may be derived as follows . during a cycle , the regenerative process @xmath7 starts at @xmath53 and grows deterministically , according to , until the time @xmath56 of the down jump . after this down jump it again grows deterministically , according to , until it reaches @xmath53 , when the next renewal occurs . if we change the order of these two parts , the process starts at @xmath81 and grows deterministically until it reaches @xmath60 . the lifetime @xmath137 hence equals the time it takes for the deterministic curve defined by to travel from @xmath81 to @xmath60 this time equals @xmath138 this is a monotonic increasing function of @xmath132 , so the renewal time distribution can be obtained numerically using the expression @xmath139 given by . the stationary distribution of @xmath7 can be obtained using regenerative process theory ( e.g. @xcite , chapter v , section 3 ) . during a regenerative cycle , the process @xmath7 traverses @xmath65 if and only if @xmath65 lies between @xmath81 and @xmath60 . if it does , the density for the time spent there is inversely proportional to the derivative @xmath140 . consequently , if we let @xmath141 denote the density of the stationary distribution of @xmath7 , we have @xmath142)\quad(1-\tau(1)<s<1),\ ] ] where @xmath143 ( @xmath144 $ ] ) is the normalizing constant making this a pdf . if @xmath145 , then @xmath146 $ ] if and only if @xmath147 . if @xmath148 , then @xmath146 $ ] if and only if @xmath149 , where @xmath150 is defined by @xmath151 . it then follows using and that , with @xmath152 , @xmath153^\kappa \left[\frac{\left(r_0 - 1\right)\tilde{s}}{1-\tilde{s}-{\rm e}^{-r_0 } \tilde{s}}\right]^{\frac{\kappa}{r_0}}&\quad \text{if } 1-\tau(1 ) < s<1/r_0,\\ c\left(\frac{1-s}{r_0 - 1}\right)^{\kappa\left(1-\frac{1}{r_0}\right)}r_0^\kappa s^{\frac{\kappa}{r_0 } } & \quad \text{if } 1/r_0 \le s < 1 . \end{array } \right.\ ] ] in the next section the density @xmath141 is calculated numerically and shown to agree with corresponding empirical values from simulations . we now present briefly some numerical and simulation results , which illustrate convergence of the epidemic process as well as properties of the limiting stationary distribution of the fraction susceptible @xmath154 . in figure [ epidsimul ] the epidemic is simulated for 100 years in a population of @xmath155 individuals . in all figures , @xmath156 implying that the effective reproduction number @xmath157 is supercritical as soon as the population fraction susceptible exceeds @xmath158 . the average lifetime is @xmath159 years and @xmath160 , so the average length of the infectious period is about 1 week . in the left panels of figure [ epidsimul ] , @xmath161 , so the rate at which new infectives enter the population ( @xmath42 ) equals 1 per 3.75 years , and in the right panels @xmath162 , so new infectives enter the population at rate @xmath163 per year . the upper panels show the fraction of the population that is susceptible over the @xmath164 period and the lower panels show the corresponding fraction that is infective . observe that when @xmath161 major outbreaks are less frequent but larger than when @xmath162 , and that there are appreciably more minor outbreaks when @xmath165 . note also that epidemics are rarer than the importation rate of infectives suggests , for two reasons . first , major outbreaks can occur only when @xmath166 , and secondly , when @xmath167 is above this threshold , major outbreaks do not occur each time an infective enters the community . in the lower left panel of figure [ epidsimul ] some minor outbreaks caused by importation of infectives can also be seen . + in figure [ limitingsimul ] realisations of the corresponding limiting processes are plotted . the same parameter values are used in both figures . the stochastic features of the epidemic and the limiting process are in agreement , suggesting that the limiting behaviour has kicked in when @xmath155 . note that , unlike in figure [ epidsimul ] , there are no near - vertical lines as outbreaks are now instantaneous . we now illustrate properties of the stationary distribution of the fraction susceptible @xmath154 , both for the epidemic with @xmath168 and @xmath155 , as well as for the limiting process . for the three processes , and for three different values of @xmath169 , we simulate the epidemic and limiting processes for 10,000 years and in figure [ stationary - simul ] we plot bar charts of the relative time spent with specified fraction susceptible . the processes are simulated over a very long time span so that the empirical distribution of the fraction susceptible is close to the corresponding stationary distribution . ( recalling the functional @xmath123 defined at the end of section [ subsecmain ] , note that by standard regenerative process theory , for any fixed @xmath120 , @xmath170 as @xmath171 and , by corollary [ convfunct ] , @xmath172 as @xmath11 . ) the values of @xmath173 and @xmath3 are the same as in figure [ epidsimul ] . ( note that the value of @xmath174 , and hence also @xmath175 @xmath176 , is the same for both values of @xmath0 . ) the chosen values of @xmath169 are @xmath177 and @xmath164 , corresponding to importation of infectious individuals on average one every 75 , 25 and @xmath178 years , respectively . in the plots we have also computed @xmath141 , the stationary distribution of the limiting process , numerically as described in section [ s - properties ] . it is seen that the bar charts from the epidemics resemble the limiting stationary distribution @xmath141 , except when @xmath168 and @xmath165 . when @xmath169 is small , few outbreaks take place , so even if the outbreaks are large , the population fraction of susceptibles is close to 1 most of the time , which explains why the stationary distribution @xmath154 is concentrated at values close to 1 . for moderate values of @xmath169 , the stationary distribution has positive mass for nearly all @xmath65 values between @xmath179 ( the fraction susceptible after a major outbreak starting with the entire population being susceptible ) and 1 . the stationary distribution is seen to be concentrated around @xmath53 when @xmath169 is large , owing to the fact that a new major outbreak occurs quite soon after the population fraction of susceptibles exceeds @xmath53 , with the effect that the size of major outbreaks is generally small . these observations imply that the stationary distribution is _ not _ stochastically decreasing ( nor increasing ) in @xmath169 . let @xmath180 be a probability space on which is defined a homogeneous poisson process @xmath181 on @xmath182 having rate @xmath183 and let @xmath184 denote the times of the points in @xmath181 . for @xmath185 , let @xmath186 denote the point process with points at @xmath187 , where @xmath188 @xmath189 . let @xmath190 denote the epidemic process indexed by @xmath0 . then @xmath186 gives the points in time when infectives immigrate into the population in @xmath190 . we construct @xmath190 @xmath191 and @xmath7 by first conditioning on @xmath181 . the process @xmath7 is constructed as follows . recall the definition of @xmath63 at [ tau ] . between the points of @xmath181 , @xmath51 increases deterministically according to the differential equation . for @xmath92 , @xmath7 has a down jump to @xmath192 $ ] at time @xmath193 with probability @xmath194 ( independently for successive @xmath95 ) , otherwise @xmath7 continues to grow according to . thus , @xmath7 can be described as follows . let @xmath195 be the times of the down jumps of @xmath7 , so these form a subset of the points of @xmath181 . let @xmath196 so , for fixed @xmath197 , the solution of with @xmath198 is @xmath199 . let @xmath200 and suppose that @xmath201 is given . then , for @xmath202 , @xmath203 where , for @xmath92 , the initial value @xmath204 , with @xmath205 . the precise definition of the construction of @xmath190 @xmath191 is not relevant at this stage . we prove theorem [ wconv ] by first proving the corresponding result for processes conditioned on @xmath181 . [ condwconv ] suppose that @xmath104 . then , for @xmath206-almost all @xmath181 , @xmath207 in order to prove lemma [ condwconv ] , we need some more notation and an extra lemma ( lemma [ convd ] , below ) . recall that , for @xmath185 , we assume @xmath29 , that @xmath91 and that , for @xmath92 , @xmath208 and @xmath209 . for @xmath185 , let @xmath210 and , for @xmath92 , let @xmath211 . @xmath212 [ convd ] suppose that @xmath104 . then the following hold for @xmath206-almost all @xmath181 . 1 . for @xmath213,@xmath214,@xmath215,@xmath216 and @xmath217 as @xmath11 . 2 . for @xmath218 , @xmath219 3 . @xmath220 as @xmath221 . see section [ proofconvd ] . first note that since results concerning convergence in distribution in the euclidean space @xmath222 carry over in all essential respects to convergence in distribution in @xmath223 ( see @xcite , page 19 ) , the skorohod representation theorem implies that we may assume that the convergence in lemma [ convd ] holds almost surely . let @xmath224 be the set @xmath225 such that ( i ) for @xmath226 @xmath227 ( ii ) for @xmath202 , @xmath228 and ( iii ) @xmath229 as @xmath221 . then @xmath230 for @xmath206-almost all @xmath181 . for @xmath107 , @xmath231 denotes the distance between @xmath232 and @xmath233 in the skorohod metric ( see @xcite , chapter 3.5 ) . let @xmath181 satisfy @xmath230 . we show that for all @xmath234 , @xmath235 it then follows that , under the skorohod metric , both @xmath236 and @xmath237 converge almost surely to @xmath7 , which implies . by proposition 5.3 on page 119 of @xcite , to show that @xmath238 as @xmath11 it is sufficient to show that for each @xmath239 , there exists a sequence @xmath240 of strictly increasing functions mapping @xmath241 onto @xmath241 so that @xmath242 and @xmath243 for ease of exposition we now suppress dependence on @xmath244 . fix @xmath245 and let @xmath246 , so @xmath247 .. then , there exists @xmath248 such that @xmath249 for all sufficiently large @xmath0 . for such @xmath0 , let @xmath250 be the piecewise - linear function joining the points @xmath251 , with @xmath252 for @xmath253 . similarly , let @xmath254 be the piecewise - linear function joining the points @xmath255 , with @xmath256 for @xmath253 . ( note that @xmath257 with strict inequality for @xmath258 . ) the functions @xmath250 and @xmath254 are strictly increasing and satisfy , since @xmath259 and @xmath260 as @xmath11 ( @xmath261 . thus , to complete the proof we show that @xmath262 and @xmath263 considering first , note that for @xmath264 , since @xmath265 is increasing on @xmath266 $ ] and @xmath267 for all @xmath268 , @xmath269 since @xmath270 , @xmath271 and @xmath272 ( as @xmath260 and @xmath250 is continuous ) , so @xmath273 as @xmath7 is right - continuous . also , for @xmath274 , @xmath275 where @xmath276 and @xmath277 now @xmath278 for @xmath279 , so using , @xmath280 a simple argument using the mean value theorem shows that , for @xmath281 $ ] and @xmath282 , @xmath283 as @xmath250 satisfies and @xmath260 as @xmath11 it then follows using that the first term on the right hand side of tends to @xmath10 as @xmath11 . also , for @xmath284 $ ] and @xmath16 , @xmath285 so the second term on the right hand side of tends to @xmath10 as @xmath11 , since @xmath286 as @xmath11 . thus , @xmath287 as @xmath11 . note that @xmath288 for @xmath289 , so implies that @xmath290 also converges to @xmath10 as @xmath11 , whence @xmath291 combining and yields that , @xmath292 a similar argument to the derivation of yields @xmath293 which together with yields , as required . the proof of is similar to that of and hence omitted . we prove the result for @xmath21 . the proof for @xmath22 is identical . recall that if @xmath294 @xmath191 and @xmath132 are random elements of @xmath86 then @xmath295 as @xmath11 if and only if @xmath296 \to \left[f(x)\right]$ ] as @xmath11 for all bounded , uniformly continuous functions @xmath297 ( see , for example , @xcite , chapter 3 , theorem 3.1 ) . let @xmath297 be any such function . then lemma [ condwconv ] implies that , for @xmath206-almost all @xmath181 , @xmath298 = { \rm e}\left[f(s)|\eta\right].\ ] ] hence , by the dominated convergence theorem , @xmath299 & = & \lim_{n \to \infty}{\rm e}_\eta\left[{\rm e}\left[f(\bar s^{(n)}_-)|\eta\right]\right]\\ & = & { \rm e}_\eta\left[\lim_{n \to \infty}{\rm e}\left[f(\bar s^{(n)}_-)|\eta\right]\right]\\ & = & { \rm e}_\eta\left[{\rm e}\left[f(s)|\eta\right]\right]\\ & = & { \rm e}\left[f(s)\right].\end{aligned}\ ] ] this holds for all bounded , uniformly continuous @xmath297 , so @xmath300 as @xmath11 , as required . we end this subsection by showing that the occupancy time functional @xmath123 , defined at , satisfies @xmath127 . recall that @xmath195 denote the jump times of @xmath7 . let @xmath301 and , for @xmath302 , let @xmath303 . let @xmath304 be the set of @xmath225 such that @xmath305 ( and hence also @xmath306 ) tends to @xmath307 as @xmath221 . then , by lemma [ convd ] ( iii ) , @xmath308 . we show that if @xmath309 @xmath191 and @xmath310 , then @xmath311 , for @xmath312 , whence @xmath127 . suppose that @xmath312 . dropping the explict dependence of @xmath7 on @xmath244 , since @xmath313 , by proposition 5.3 on page 119 of @xcite , there exists a sequence @xmath240 of strictly increasing functions mapping @xmath314 onto @xmath314 such that @xmath315 now @xmath316 where @xmath317 and @xmath318 let @xmath319\cap\left(\left\{t_1,t_2,\cdots\right\}\cup\left\{v_1,v_2,\cdots\right\}\right)$ ] . then @xmath320 has lebesgue measure zero and @xmath321 as @xmath11 , for @xmath322 \setminus d$ ] , since @xmath323 , by the first equation in [ skorconv ] , and @xmath7 is continuous at such @xmath19 . thus @xmath324 by the dominated convergence theorem . a similar argument , using in addition the second equation in [ skorconv ] , shows that @xmath325 . thus , @xmath326 , as required . we prove lemma [ convd ] by splitting the sir - d - i epidemic process @xmath190 into cycles , where now a cycle begins at the end of a major outbreak and finishes at the end of the following major outbreak . thus a cycle consists of two stages : stage 1 , during which the susceptible population grows approximately deterministically until there are at least @xmath98 infectives present ; and stage 2 , comprising the major outbreak caused by these @xmath98 infectives , during which the susceptible population crashes . recall that , as @xmath11 , the point process @xmath186 , describing immigration times of infectives in @xmath190 converges almost surely to the point process @xmath181 governing times when down jumps _ may _ occur in the limiting process @xmath7 . lemma [ minoreps ] considers the initial stage 1 and shows , using birth - and - death processes that sandwich the process of infectives , that for @xmath206-almost all @xmath181 , as @xmath5 , for successive importations of infectives until a major outbreak occurs , the probability a given importation triggers a major outbreak converges to the probability that the corresponding importation results in a down jump in the limiting process @xmath7 . consequently , the time until there are at least @xmath98 infectives in @xmath190 converges weakly to the time of the first down jump in @xmath7 , since @xmath186 converges almost surely to @xmath181 . further , application of the law of large numbers for density dependent population processes ( @xcite , chapter 11 ) shows that up until the first down jump of @xmath7 , the scaled process of susceptibles , @xmath327 , converges weakly in the uniform metric to @xmath7 , since minor epidemics infect order @xmath12 individuals . lemmas [ majoreps0 ] and [ majoreps ] concern the limiting size and duration of a typical major outbreak . lemma [ majoreps0 ] considers outbreaks in which the initial number of infectives is of exact order @xmath0 , for which the above - mentioned law of large numbers is applicable . this is then used to prove lemma [ majoreps ] , which considers major outbreaks triggered by @xmath98 infectives . finally , lemma [ convd ] follows easily by induction using lemmas [ minoreps ] and [ majoreps ] , since @xmath190 is markov . the proof involves extensive use of birth - and - death processes that bound the process of infectives in the epidemic model ( cf . we first give some notation concerning birth - and - death processes and then state a lemma , proved in appendix [ appendix1 ] , concerning properties of sequences of such processes . let @xmath328 denote a linear birth - and - death process , with @xmath329 , birth rate @xmath330 and death rate @xmath331 . for @xmath332 , let @xmath333 , where @xmath334 if @xmath335 for all @xmath336 . ( throughout the paper we adopt the convention that the hitting time of an event is infinite if the event never occurs . ) let @xmath337 denote the duration of @xmath338 . for @xmath339 , let @xmath340 denote the total number of births during @xmath341 $ ] in @xmath338 , and let @xmath342 denote the total progeny of @xmath338 , not including the @xmath95 ancestors . further , for @xmath343 , let @xmath344 . [ bdresults ] suppose that @xmath345 @xmath191 , where @xmath346 is constant and @xmath347 as @xmath11 . 1 . if @xmath348 , then 1 . for all @xmath336 , @xmath349 2 . @xmath350 ; and 3 . for any @xmath351 , @xmath352 2 . if @xmath353 , then 1 . @xmath354 ; 2 . @xmath355 as @xmath11 ; 3 . @xmath356 ; and 4 . for any @xmath351 , @xmath357 before proceeding some more notation is required . for @xmath92 , let @xmath358 be the indicator function of the event that the @xmath95th point in @xmath181 yields a down jump in @xmath7 . for @xmath185 and @xmath92 , let @xmath359 and @xmath360 . [ minoreps ] suppose that @xmath361 as @xmath11 . then the following hold for @xmath206-almost all @xmath181 . 1 . for @xmath92 , @xmath362 2 . for @xmath92 , as @xmath11 , @xmath363 3 . for @xmath92 , as @xmath11 , @xmath364 for ease of presentation we suppress explicit conditioning on @xmath181 in the proof . first note that @xmath365 , since @xmath366 is a realisation of a continuous random variable . assume without loss of generality that there is no recovered individual at time @xmath367 . for @xmath16 , let @xmath368 be the number of susceptibles at time @xmath19 under the assumption that the immigration rate for susceptibles is @xmath369 and the immigration rate for infectives is @xmath10 , and let @xmath370 . then , for any @xmath371 , application of theorem 11.2.1 of @xcite ( using the more general definition of a density dependent family given by equation ( 11.1.13 ) of that book ) yields that , for any @xmath372 , @xmath373 recall that @xmath190 denote the epidemic process with average population size @xmath0 . consider the epidemic initiated by the immigration of an infective at time @xmath374 in @xmath190 and let @xmath375 . for ease of exposition , translate the time axis of @xmath190 so that the origin corresponds to @xmath374 . with this new time origin , @xmath376 can be approximated by a linear birth - and - death process @xmath377 having death rate @xmath378 and ( random ) time - dependent birth rate given by @xmath379 . this approximation ignores depletion in the number of susceptibles owing to infection , so @xmath377 is an upper bound for @xmath376 . let @xmath380 and fix @xmath381 . note that , with the change of origin , @xmath382 . then , using , for any @xmath383 , there exists @xmath384 and @xmath385 such that @xmath386 for @xmath387 and @xmath388 , with probability at least @xmath389 , the process @xmath377 is bounded below and above by the birth - and - death processes @xmath390 and @xmath391 , respectively , where @xmath392 and @xmath393 . further , since @xmath394 , for all sufficiently large @xmath0 , the birth - and - death process @xmath390 is bounded below by the birth - and - death process @xmath395 , where @xmath396 . similarly , for all sufficiently large @xmath0 , the birth - and - death process @xmath391 is bounded above by the birth - and - death process @xmath397 , where @xmath398 . suppose first that @xmath399 . then for all @xmath400 , where @xmath401 , the birth - and - death process @xmath397 is subcritical , so by lemma [ bdresults](a)(i ) , for all @xmath336 , @xmath402 setting @xmath403 shows that , for all sufficiently large @xmath0 , with probability at least @xmath404 , @xmath405 , and hence also @xmath406 , is bounded above by @xmath407 throughout its entire lifetime . thus , @xmath408 by lemma [ bdresults](a)(ii ) . hence , since @xmath409 is arbitrary , @xmath410 let @xmath411 . then it follows using that @xmath412 as @xmath11 . suppose instead that @xmath413 . fix @xmath414 , where @xmath415 , and @xmath383 . then , similar to above , there exists @xmath416 such that , for all sufficiently large @xmath0 , with probability at least @xmath389 , @xmath377 is bounded above and below by @xmath407 @xmath417 , respectively , throughout the interval @xmath418 $ ] . for @xmath343 , let @xmath419 , @xmath420 and @xmath421 . note that the birth - and - death processes @xmath395 and @xmath397 are both supercritical . then , by lemma [ bdresults](b)(ii ) , for all sufficiently large @xmath0 , the process @xmath377 is bounded below and above by @xmath407 @xmath417 , respectively , throughout the interval @xmath422 $ ] . using lemma [ bdresults](b)(i ) , it then follows that @xmath423 and @xmath424 letting both @xmath425 and @xmath426 converge down to @xmath10 in and yields @xmath427 further , using lemma [ bdresults](b)(ii ) , it follows that @xmath428 recall that @xmath377 is an upper bound for @xmath376 . we now show that the probability that the two processes coincide over @xmath429 $ ] converges to one as @xmath78 . in @xmath377 births occur at time - dependent rate @xmath379 , whilst in @xmath376 infections occur at time - dependent rate @xmath430 . now @xmath431 for all @xmath16 , almost surely , so the two processes can be coupled by using an independent sequence @xmath432 of independent and identically distributed random variables that are uniformly distributed on @xmath433 , with the @xmath434th birth in @xmath377 ( which occurs at time @xmath435 say ) yielding an infection in @xmath376 if and only if @xmath436 . for @xmath185 and @xmath336 , let @xmath437 be the total number of births in @xmath377 during @xmath341 $ ] . recall that the probability that @xmath377 is sandwiched between the supercritical birth - and - death processes @xmath407 and @xmath417 throughout @xmath429 $ ] converges to one as @xmath78 . it then follows using lemma [ bdresults](b)(iii ) that @xmath438 also , since @xmath439 as @xmath11 , it follows using that , for any @xmath440 , @xmath441\right)=1.\ ] ] suppose that @xmath442 and , for fixed @xmath443 , @xmath444 for all @xmath445 $ ] . then , @xmath446 , for @xmath447 , so if @xmath448 denotes the probability that the @xmath434th birth in @xmath377 yields an infection in @xmath376 , then @xmath449 whence @xmath450 thus , recalling and , the probability that @xmath376 and @xmath377 coincide over @xmath422 $ ] converges to one as @xmath11 , which , together with , yields @xmath451 and , together with , yields @xmath452 we have thus proved parts ( i ) and ( iii ) for @xmath453 . note that , since @xmath454 , when reverting to the original time axis , the probability that the total number of individuals infected during @xmath455 $ ] in @xmath190 is less than @xmath456 tends to one as @xmath11 , which combined with proves part ( ii ) when @xmath453 . parts ( i ) , ( ii ) and ( iii ) for @xmath457 follow easily by induction since the processes @xmath458 ( @xmath185 ) and @xmath7 are markov . before proceeding we state some well - known facts about the final outcome of the deterministic general epidemic ( e.g. @xcite chapter 1.4 ) . for @xmath16 , let @xmath459 and @xmath460 denote respectively the density of susceptibles and infectives at time @xmath19 , so @xmath461 are determined by the differential equations @xmath462 with initial condition @xmath463 , where @xmath464 and @xmath465 . note that time is scaled so that the recovery rate is @xmath466 . then @xmath459 decreases with @xmath19 , @xmath467 and @xmath468 , where @xmath469 is the unique solution in @xmath433 of @xmath470 note that @xmath471 is continuous in @xmath472 and @xmath473 as @xmath474 , where ( recall ) @xmath475 in the following two lemmas , there is no importation of infectives in @xmath476 , though births of susceptibles still occur at rate @xmath369 . for @xmath16 , let @xmath477 . [ majoreps0 ] suppose that @xmath478 as @xmath11 , where @xmath479 and @xmath480 . let @xmath481 . then , as @xmath11 , 1 . @xmath482 , 2 . @xmath483 . for @xmath185 and @xmath336 , let @xmath484 and @xmath485 . let @xmath486 , where @xmath487 . the process @xmath488 is a continuous - time markov chain with transition intensities @xmath489,\\ q^{(n)}_{(s , i),(s-1,i)}&=n\left[\frac{\mu}{\gamma_n}\frac{s}{n}\right],\\ q^{(n)}_{(s , i),(s-1,i+1)}&=n\left[r_0\frac{s}{n}\frac{i}{n}+\left(\frac{\lambda_n}{\gamma_n}-r_0\right)\frac{s}{n}\frac{i}{n}\right],\\ q^{(n)}_{(s , i),(s , i-1)}&=n\left[\frac{i}{n}+\frac{\mu}{\gamma_n}\frac{i}{n}\right ] , \ ] ] corresponding to a birth of a susceptible , a death of a susceptible , an infection of a susceptible , and a recovery or death of an infective , respectively . the transition intensities are written in the above form to indicate that the family of processes @xmath490 is asymptotically density dependent , as defined by @xcite . let @xmath491 be any compact subset of @xmath492 . recall that @xmath493 and @xmath494 as @xmath11 . hence , as @xmath11 , each of @xmath495 , @xmath496 , @xmath497 and @xmath498 converges to @xmath10 . it follows that the conditions of theorem 3.1 in @xcite are satisfied , whence , for any @xmath372 and any @xmath336 , @xmath499 where @xmath500 is the solution of the deterministic general epidemic having initial condition @xmath463 . write @xmath501 for @xmath469 . there exists @xmath502 such that @xmath503 , since otherwise @xmath504 would be strictly positive . given @xmath505 , choose @xmath506 so that @xmath507 there exists @xmath508 such that @xmath509 and @xmath510 . then implies that @xmath511 so , reverting to the original time scale and letting @xmath512 , @xmath513 observe that , whilst @xmath514 , the process @xmath515 is bounded above by the birth - and - death process @xmath516 , where @xmath517 and @xmath393 . now @xmath518 as @xmath11 , so , for all sufficiently large @xmath0 , @xmath516 is in turn bounded above by @xmath519 , where @xmath520 . recall that @xmath521 and @xmath522 denote the total number of births in and the extinction time of @xmath338 , respectively . then @xmath523=\frac{r_0(s_\infty+\epsilon_0)}{1-r_0(s_\infty+\epsilon_0)},\ ] ] and , recalling , application of the strong law of large numbers yields @xmath524 also , lemma [ bdresults](a)(iii ) implies that @xmath525 recall that @xmath526 denotes the process that describes the number of susceptibles in the absence of any infectives and suppose that @xmath527 . for @xmath16 , let @xmath528 and @xmath529 be the total number of births and deaths , respectively , during @xmath341 $ ] in @xmath530 . using and the fact that @xmath528 has a poisson distribution with mean @xmath531 , there exists @xmath532 such that @xmath533 the processes @xmath534 and @xmath519 can be coupled so that @xmath535 whilst @xmath536 . the first equations in and imply that @xmath537 so implies that , with probability tending to @xmath466 as @xmath11 , the coupling holds thoughout the lifetime of @xmath519 . recall that @xmath481 . the coupling implies that @xmath538 , so part ( ii ) of the lemma follows from , since @xmath539 as @xmath11 . further , @xmath540 is at most the sum of @xmath541 and the number of births in @xmath542 $ ] , and at least the difference between @xmath541 and the sum of the number of susceptible deaths in @xmath542 $ ] and @xmath543 , so , and imply that @xmath544 proving part ( i ) of the lemma , since @xmath505 can be arbitrarily small . [ majoreps ] suppose that @xmath545 @xmath191 and @xmath361 as @xmath11 , where @xmath479 . let @xmath546 , @xmath547 then , as @xmath11 , 1 . @xmath548 , where the function @xmath63 is defined at ; 2 . @xmath483 ; 3 . @xmath549 and @xmath550 . fix @xmath551 such that @xmath552 . then , whilst @xmath553 , @xmath406 is bounded below by the birth - and - death process @xmath554 , where @xmath555 and @xmath393 . now @xmath556 as @xmath11 , so , for all sufficiently large @xmath0 , @xmath554 is in turn bounded below by @xmath557 , where @xmath558 . recall that @xmath190 denotes the epidemic process indexed by @xmath0 . for @xmath336 , let @xmath559 be the total number of infections in @xmath190 during @xmath341 $ ] . let @xmath560 . define @xmath561 as in the proof of lemma [ minoreps ] . for @xmath16 , let @xmath528 and @xmath529 be the total number of births and deaths , respectively , during @xmath341 $ ] in @xmath526 . as at , but note that @xmath562 is different here , for any @xmath440 there exists @xmath563 such that @xmath564 also , since @xmath361 as @xmath11 , @xmath565 observe that , if @xmath566 and @xmath567 , then @xmath568 obtained by making @xmath569 and @xmath570 as large as possible and assuming no susceptible dies during @xmath571 $ ] , and @xmath572 obtained by making @xmath569 as small as possible , @xmath573 as large as possible and assuming no susceptible is born during @xmath571 $ ] . recall that , whilst @xmath553 , @xmath376 is bounded below by the birth - and - death process @xmath557 , so @xmath574 , provided @xmath553 throughout @xmath571 $ ] . now @xmath575 as @xmath11 , by lemma [ bdresults](b)(iv ) , so @xmath576 as @xmath11 , for any @xmath372 . setting @xmath577 in , using , and noting that @xmath578 as n @xmath579 , shows that @xmath580 so @xmath581 further , since for any @xmath582 , @xmath583 as @xmath11 , it follows from - that , for any @xmath372 , @xmath584 so @xmath585 it is straightforward to couple the jump processes of @xmath586 and + @xmath587 to show that @xmath588 , where @xmath589 and @xmath590 denotes stochastically greater than . further , recalling that @xmath557 has the same distribution as @xmath591 , it follows using @xcite , theorem 5.4 , that @xmath592 thus , since @xmath593 as @xmath11 , @xmath594 where @xmath595s_0\theta.\ ] ] a similar argument using an upper bounding birth - and - death process yields that @xmath596 where @xmath597s_0\theta.\ ] ] exploiting the markov property of @xmath476 , - and lemma [ majoreps0](i ) imply that , for any @xmath372 , @xmath598 letting @xmath599 , noting that @xmath600 and using the continuity properties of @xmath501 , yield that , for any @xmath372 , @xmath601 proving part ( i ) of the lemma . part ( ii ) follows immediately using , , and lemma [ majoreps0](ii ) . part ( iii ) is an easy concequence of parts ( i ) and ( ii ) and . the lemma follows easily by induction using lemmas [ minoreps ] and lemmas [ majoreps ] . first note lemma [ minoreps ] ( i ) and ( iii ) imply that @xmath602 as @xmath11 , and lemma [ minoreps ] ( i ) and ( ii ) imply that holds for @xmath603 and @xmath604 as @xmath11 . lemma [ majoreps ] ( ii ) then yields that @xmath605 as @xmath11 , lemma [ majoreps ] ( i ) yields that @xmath606 as @xmath11 , and lemma [ majoreps ] ( iii ) yields that @xmath607 and @xmath608 as @xmath11 . now @xmath476 is markov , so , since @xmath609 as @xmath11 , the above argument can be repeated for @xmath302 . part ( iii ) is immediate , since @xmath610 , where @xmath611 are the times of the points in @xmath181 . in the paper it is proved that for an sir epidemic in a dynamic population ( whose size fluctuates around @xmath0 ) , in which there is importation of infectives at a constant rate , the normalised process of susceptibles converges to a regenerative process @xmath7 as @xmath2 . further , properties of the limiting process @xmath7 are derived . the asymptotic regime considered is for the situation when the rate of importation of infectives @xmath612 and the basic reproduction number @xmath3 remain constant with @xmath0 , whereas the average length of the infectious period @xmath48 converges to 0 faster than @xmath4 ( in most real - life epidemics , the ratio of average infectious period and average lifetime lies between @xmath8 and @xmath9 ) . other asymptotic regimes could of course also be considered . for example , if the importation rate of infectives grows with @xmath0 , then there will always be infectives present in the population resembling an endemic situation . if the duration of an infectious period remains fixed ( or at least grows slower than @xmath98 ) , then the duration of a single outbreak will be long and the typical time horizon will not go beyond the first outbreak . a more complicated and interesting scenario seems to be for the asymptotic situation treated in the current paper , but where the epidemic is initiated with a fraction @xmath53 of the population susceptible and a large enough number of infectives . it then seems as if an endemic equilibrium will stabilize , but determining and proving this rigorously remains an open problem . for large but finite @xmath0 , it is possible for the process to get stuck in an endemic situation near the end of a major outbreak ( with states similar to those just described ) . eventually the epidemic leaves this endemic state and returns to the behaviour of the limiting process . in figure [ endemic - simul ] such a simulation is presented . the parameter values are @xmath613 , @xmath614 , @xmath615 ( so the importation rate of infectives is one per @xmath616 years ) , @xmath156 and @xmath617 ( so the average infectious period is @xmath618 months ) . the left and right plots show the fraction of the population that are susceptible and infective , respectively , as functions of time . a quasi - endemic phase lasts roughly from years @xmath619 to @xmath620 . observe that major outbreaks become smaller prior to the process entering the quasi - endemic phase and fluctuations in the number of infectives increase in amplitude prior to the end of the quasi - endemic phase . beside studying other asymptotic regimes , it could be of interest to increase realism in the model , for example , by relaxing exponential distributions of infectious periods and lifetimes and allowing for a latent state ( cf . @xcite , who consider epidemics with importation of susceptibles only ) or by having some population structure , such as network or households ( see the challenges in @xcite and @xcite ) . suppose that @xmath348 . then , for @xmath336 , @xmath621\\ & = & \exp\left(-(1-a)\beta_n t\right ) \to 0 \quad \mbox{as } n\to \infty,\end{aligned}\ ] ] since @xmath622 as @xmath11 , proving part ( a)(i ) . observe that , for any @xmath95 , @xmath623 , where @xmath624 denotes equal in distribution . it follows that @xmath625 , so @xmath626 since @xmath627 is subcritical , proving part ( a)(ii ) . for any @xmath336 , @xmath628\\ & = & \left \lceil{c n}\right \rceil \exp\left(-(1-a)\beta_n t\right)\\ & \to&0\quad\mbox{as } n \to \infty,\end{aligned}\ ] ] since @xmath347 as @xmath11 , proving part ( a)(iii ) . suppose that @xmath353 . note that , since @xmath629 , coupled realisations of @xmath630 can be obtained by setting @xmath631 @xmath632 . now , see e.g. @xcite , page 112 , there exists a random variable @xmath633 , satisfying @xmath634 if and only if @xmath627 goes extinct , such that @xmath635 note that , for any @xmath636 , @xmath637 . if @xmath634 , then @xmath638 , so @xmath639 as @xmath11 , and @xmath640 , so @xmath641 for all sufficiently large @xmath197 , whence @xmath642 for all sufficiently large @xmath0 . if @xmath643 , then @xmath644 , so @xmath645 . also , for any @xmath336 , @xmath646 now @xmath647 as @xmath11 , since @xmath648 , so , since @xmath643 , implies that @xmath649 for all sufficiently large @xmath0 . this holds for any @xmath336 , so @xmath650 as @xmath11 . part ( b)(i ) follows since @xmath651 and part ( b)(ii ) also follows , indeed we have shown that , under the coupling , @xmath652 as @xmath11 . further , it follows using @xcite , theorem 5.4 , that @xmath653 where @xmath654 is the same random variable as in . if @xmath634 , then @xmath655 , so @xmath656 for all sufficiently large @xmath0 . if @xmath643 , then and imply that @xmath657 , so , since @xmath658 as @xmath11 , @xmath659 , whence @xmath660 for all sufficiently large @xmath0 . part ( b)(iii ) now follows . a similar argument to the above shows that , if @xmath643 , then for any @xmath351 , @xmath661 as @xmath11 . thus @xmath662 as @xmath11 , since @xmath663 is the sum of @xmath664 independent copies of @xmath665 . part ( b)(iv ) follows since @xmath666 . t. b. is supported by riksbankens jubilieumsfond , grant p12 - 0705:1 and p. t. is supported by vetenskapsrdet ( swedish research council ) , project 20105873 + f. b. and p. t. gratefully acknowledge support from the isaac newton institute for mathematical sciences , cambridge , where we held visiting fellowships under the infectious disease dynamics programme and its follow - up meeting , during which this work was initiated . ball , f. , britton , t. , house , t. , isham , v. , mollison , d , pellis , l. and scalia tomba , g. ( 2015 ) seven challenges for metapopulation models of epidemics including households models . _ epidemics _ * 10 * , 6367 .

the plasma membrane of human red blood cells consists of a lipid bilayer attached to a regular network of underlying cytoskeletal polymers . we model this system at a dynamic coarse - grained level , treating the bilayer as an elastic sheet and the cytoskeletal network as a series of phantom entropic springs . in contrast to prior simulation efforts , we explicitly account for dynamics of the cytoskeletal network , both via motion of the protein anchors that attach the cytoskeleton to the bilayer and through breaking and reconnection of individual cytoskeletal filaments . simulation results are explained in the context of a simple mean - field percolation model and comparison is made to experimental measurements of red blood cell fluctuation amplitudes .

0806.3897

membranes are essential components of all biological cells @xcite . in addition to their biological importance , lipid bilayers and biomembranes have also attracted considerable attention from physicists due to their fascinating and unusual properties @xcite . one particularly well studied system is the human red blood cell ( rbc ) membrane . the rbc membrane is a composite structure , consisting of a lipid bilayer adhered to an underlying network of filamentous cytoskeletal proteins ( spectrin ) via integral membrane protein anchors ( see fig . [ fig : mesh ] ) . the spectrin network is observed to be quite regular @xcite , with an approximate hexagonal symmetry extending over the entire cell surface . ( it is worth emphasizing that the spectrin based cytoskeletal network of rbcs is completely different from the three dimensional actin networks common to other types of animal cells @xcite . ) given the apparent simplicity of the rbc membrane , it is tempting to attempt modeling with elementary elastic models . indeed , there is a rich history of elastic rbc models to be found in the literature @xcite . without attempting a full historical review here , we comment that no single elastic model has yet been identified that is capable of reproducing the full set of experimental data available for rbc membranes . for example , while the work of lim , wortis and mukhopadhyay @xcite is capable of capturing the range of observed rbc shapes ( stomatocyte , discocyte , echinocyte and non - main - sequence shapes as well ) seen under various chemically induced stresses ( e.g. ph , salt , atp , etc . ) , this model has not been applied to explain the thermal fluctuation amplitudes observed in rbc membranes . and , while the models of gov , zilman and safran @xcite and fournier , lacoste and raphael @xcite appear to do a good job fitting thermal fluctuation data @xcite , these models do not appear capable of explaining mechanical deformation experiments on rbcs @xcite . the failure of simple models to consistently explain both thermal fluctuations and mechanical deformation experiments has been recognized for years @xcite . one recent model does explain both types of data within a single elastic model @xcite , however this model treats the spectrin network as an incompressible and homogeneous viscoelastic plate coupled to the lipid bilayer . it remains unclear why such an approximation should suffice for the sparse cytoskeletal network present in rbcs . additionally , this model seems too simplistic to capture the full range of shape behaviors explained in reference @xcite . the models mentioned in the preceding paragraph make no mention of the role of energy expenditure in the behavior of rbc membranes . this despite the fact that it is known that rbc membranes possess kinase and phosphatase activities capable of altering the properties of spectrin and other network associated proteins via ( de)phosphorylation @xcite . and , certain measurements of rbc fluctuation amplitudes show a correlation between atp concentration and fluctuation magnitude @xcite . one might argue that the difficulty in fitting all rbc behavior to a single elastic model stems from the fact that a truly comprehensive model must incorporate the effects of energy expenditure by the cell in a realistic fashion . gov and safran @xcite are the first to seriously consider active energy expenditure within the rbc from a theoretical standpoint . they have proposed that atp induced phosphorylation and dephosphorylation of the rbc cytoskeletal network leads to a continual dynamic evolution of the integrity of the spectrin network . while this picture remains hypothetical , without direct proof , it is consistent with the general observations relating atp concentration to membrane fluctuation amplitudes . gov and safran ( gs ) @xcite have used this picture to motivate a simple picture for rbc fluctuations under the presence of atp . local breaking and reforming of the spectrin network is captured via non - thermal forces imparted on an elastic membrane model . this model has provided the first plausible explanation for the viscosity dependence of rbc fluctuations @xcite . while elastic models with proper accounting for non - thermal energetics may eventually prove adequate in describing the long wavelength physics of the rbc , it is clear that wavelengths near or below the spectrin mesh size ( @xmath0 ) must be considered within a more microscopic picture . a recent model by dubus and fournier ( df ) @xcite has extended the traditional elastic modeling of rbc membranes to explicitly include the cytoskeletal network at a molecular level of detail . within this model , the spectrin network is considered as a completely regular hexagonal network of phantom entropic springs attached to a fluid lipid bilayer . over wavelengths significantly longer that the spectrin mesh size , the network so modeled becomes mathematically equivalent to an imposed surface tension on the fluid bilayer . at wavelengths comparable to and shorter than network spacing , the system behaves differently from a simple membrane with applied tension . this model was used to compute the spectrum of thermal fluctuation amplitudes for the rbc membrane , but made no attempt to account for non - thermal consumption of energy and only computed thermal ( non - dynamic ) observables . in this paper , we extend the df entropic spring model of the cytoskeleton meshwork to include dynamic evolution of the entire system . we allow the anchor points between spectrin and membrane to laterally diffuse and we allow for dynamic dissociation and association of spectrin links as a molecular level manifestation of the non - thermal energetic picture proposed by gs ( fig . [ fig : snap ] ) . one important consequence of the gs picture is that sufficiently high atp concentrations lead to an appreciable fraction of dissociated spectrin links at the membrane surface . depending upon the timescales for spectrin ( re)association , the effective long - wavelength elastic properties of the membrane interpolate between two limiting cases . if spectrin ( re)association kinetics are much slower than all other timescales in the problem , the effective tension imposed by the network ( as inferred by out - of - plane bilayer undulations ) is well predicted by a simple percolation - theory argument ( see fig . [ fig : perco ] ) . in the opposite limit of fast spectrin kinetics , the effective tension is well predicted by assuming each link in the network has a reduced spring constant proportional to the probability of the link being intact at steady state . at intermediate rates , simulations are seen to interpolate between the two extreme cases . this paper is organized as follows . in sec . [ sec : model ] , we present our mathematical model for the rbc membrane . in sec . [ sec : simulation ] details of our simulation methods are discussed . in sec . [ sec : complete ] results for a fully intact spectrin meshwork are presented , while in sec . [ sec : broken ] we generalize to the more interesting case of a randomly broken network ( both static and dynamically broken ) . we discuss our results in relation to experiment in sec . [ sec : experiments ] and conclude in sec . [ sec : conclusion ] . we treat the rbc membrane as a helfrich fluid sheet @xcite coupled via mobile anchor points to a network of springs . our hamiltonian is @xmath1 ^ 2\nonumber\\&+ & \sum_{\langle i , j\rangle}\xi_{ij}(t)\left\{\frac{\mu}{2 } ( \mathbf{r}_i-\mathbf{r}_j)^2 + \frac{\mu}{2}[h(\mathbf{r}_i)-h(\mathbf{r}_j)]^2+e\right\}.\nonumber\\\label{eq : h}\end{aligned}\ ] ] the first term ( integral portion ) is the standard bilayer bending energy for a monge gauge sheet assuming small deformations @xcite and bending modulus @xmath2 . @xmath3 is the projected area of the membrane , @xmath4 is the position vector in the @xmath5 plane and and @xmath6 is the local displacement of the membrane away from the flat reference configuration specified by @xmath7 ( see fig . [ fig : snap ] ) . we assume that the lipid bilayer itself has a negligible ( bare ) surface tension . in more general situations , eq . [ eq : h ] is easily modified to account for non - vanishing tension inherent to the lipid bilayer portion of the membrane @xcite . the second term ( sum portion ) accounts for the energy of the 2d cytoskeletal meshwork , modelled as a network of entropic springs ( or in other words , ideal chains of polymers ) with effective spring constant @xmath8 ( @xmath9 with @xmath10 and @xmath11 the kuhn and contour length of the polymer , respectively ) . in a fully intact cytoskeletal meshwork , all spring end points ( nodes ) are restricted to lie on the surface of the bilayer , so their coordinates are specified by @xmath12 and @xmath13 , with indices @xmath14 ( or @xmath15 ) labelling different nodes . the sum is over all _ distinct _ nearest neighbor node pairs ( equivalently , over all polymer springs ) , denoted as @xmath16 . the factor @xmath17 is included to account for the possibly incomplete connectivity of the network . it is equal to 1 when the @xmath18 link is connected and 0 otherwise . the constant @xmath19 reflects all other free energy change associated with connecting a detached filament end to a node besides the elastic energy of the spring ( e.g. , the binding energy to the node ) ; we assume this energy is negative and significantly larger than thermal energy scales to insure stability of the network in the absence of non - thermal energy sources . although our starting point is very similar to the df model , we emphasize a few key differences . in df , the lateral positions of nodes are held fixed in the geometry of minimum energy for a flat bilayer surface ; _ i.e. _ the @xmath20 variables are treated as set constants in df , not as variables capable of influencing the energetics and/or dynamics of the system . this approximation renders the hamiltonian analytically tractable , however it is not immediately clear that such a choice fully captures all relevant physics in this system . for example , with fixed node positions equally spaced on a regular lattice , the `` spring '' contribution to eq . [ eq : h ] amounts to a finite differenced version of the usual surface tension contribution to the helfrich hamiltonian . at long wavelengths , eq . [ eq : h ] with fixed nodes is guaranteed to behave as a helfrich sheet under tension . if the nodes are mobile , as physically expected , the spring network represents a simple manifestation a tethered membrane . such membranes are known to exhibit more complicated fluctuations than expected for helfrich fluid bilayers @xcite . we also emphasize that much of the following work is concerned with membrane dynamics , which was not considered in df . one of the most interesting aspects of our model is the dynamic breaking and reformation of cytoskeletal filaments , which could not be studied with the equilibrium approach adopted by df . dynamics in our system are overdamped , owing to the low reynolds number environment present at cellular length scales @xcite . for bilayer height fluctuations , we have the following langevin type equation of motion , which accounts for hydrodynamic flow in the surrounding cytoplasm @xcite @xmath21.\label{eq : langevin}\ ] ] here , @xmath22 is the diagonal part of the oseen tensor @xcite , given by @xmath23 where @xmath24 is the viscosity of the surrounding fluid . the above integral is taken over the entire @xmath25 plane ; it is assumed that the area of interest , @xmath3 , is embedded within a periodically repeating environment of identical subsystems . @xmath26 is the force per unit area on the membrane , @xmath27 , \label{eq : fr}\end{aligned}\ ] ] where @xmath28 is the dirac delta function , and the sum is over all nearest neighbors of node @xmath14 , denoted as @xmath29 . @xmath30 is a spatially local gaussian white noise satisfying the fluctuation - dissipation relation @xmath31 where @xmath32 is boltzmann s constant and @xmath33 temperature of the system . we have another set of langevin equations describing lateral diffusion of the nodes within the bilayer @xmath34,\label{eq : langevin2}\ ] ] with @xmath35\frac{\partial h(\mathbf{r}_i)}{\partial \mathbf{r}_i}\right\},\nonumber\\ \label{eq : fir}\end{aligned}\ ] ] and @xmath36 where @xmath37 is the lateral diffusion constant of the node across the membrane surface . [ eq : langevin ] and [ eq : langevin2 ] completely specify the dynamics of the lipid bilayer and the attached 2d meshwork . notice that the two sets of langevin equations are coupled ( and must be solved simultaneously ) via the shape of the membrane surface . we note that eq . [ eq : fir ] neglects the purely geometric effect of non - flat membrane geometry on the @xmath38 motion of node points @xcite . this approximation significantly simplifies our modeling and it has recently been demonstrated that such geometric effects are very small for the physical parameters studied herein @xcite . it is convenient to recast eq . ( [ eq : langevin ] ) in a fourier basis @xcite . @xmath39\label{eq : langevink}\ ] ] with @xmath40 for integer @xmath41 and @xmath42 . here for the latter convenience of the simulation , we assume in general a rectangular sample with size @xmath43 in real space , and therefore the two lattice constants in @xmath44 space are different . the quantities @xmath45 , @xmath46 and @xmath47 derive from functions periodic in @xmath48 and @xmath49 , due to the assumed periodicity of the system . the fourier transform pair for an arbitrary function , @xmath50 , with such periodicity is @xmath51 the fourier transformed oseen interaction , @xmath52 in contrast , reflects transformation over the full 2d plane . by construction , the dynamics specified by eq . [ eq : langevink ] reflects an infinite network of periodic membrane images interacting via the long range @xmath53 oseen hydrodynamic kernel . the random forces obey @xmath54 for the dynamics of the breaking and reforming of spectrin springs , we consider a simple two state kinetic model . we define the rate to reconnect a link as @xmath55 , and the rate to disconnect a link as @xmath56 , irrespective of the instantaneous position of the endpoints of the spring . accordingly , the value of each @xmath57 jumps back and forth between 1 and 0 . defining @xmath58 as the steady - state probability of a link to be connected at any moment , we have @xmath59 it should be emphasized that our simple model for the breaking and reformation of spectrin filaments does not obey detailed balance since we do not account for the variations in energetics caused by positions of the network nodes within the kinetic scheme . this approach necessarily corresponds to a non - equilibrium situation , with the dynamics of the spectrin network driving membrane fluctuations in a non - thermal manner . qualitatively , this corresponds to the picture proposed by gs , however the detailed kinetics involved in the spectrin ( re)association process are unknown and likely differ substantially from the picture adopted herein . our simple two - state picture represents one possible manifestation of non - equilibrium driving . the coupled equations implied by eqs . [ eq : langevink ] and [ eq : langevin2 ] are not amenable to analytical solution and will be solved via simulation as detailed below . before proceeding , we note that simulations are run on a discrete square lattice . in other words , ( [ eq : fourierr ] ) is replaced by @xmath60 where @xmath61 is the lattice constant and @xmath62 now only take discrete values on the lattice . correspondingly , in @xmath63 space , the reciprocal lattice ( with lattice constant @xmath64 and @xmath65 ) contains @xmath66 points and the summation in eq . [ eq : fourierk ] is finite . two types of meshworks are considered in our simulations . first , in accordance with the true geometry of rbc membranes @xcite , we consider a hexagonal meshwork ( fig . [ fig : corral]a ) . in such a meshwork , when all links are connected , the average positions of all nodes form a hexagonal lattice ( this lattice of nodes should not be confused with the lattice used to discretize the surface as discussed above ) . we consider a finite sample and assume periodic boundary conditions in a rectangular geometry of size @xmath43 approximately commensurate with the embedded hexagonal meshwork ( see fig . [ fig : corral]a ) . the average distance between the nearest neighbor nodes , @xmath67 , is determined from the average surface density of nodes in a rbc membrane . for theoretical interest , we also consider a square meshwork with square lattice symmetry for the average positions of nodes ( fig . [ fig : corral]b ) . here we simply take a periodic square box , i.e. , @xmath68 . to connect with experiment and prior theoretical work , our simulations are used primarily to calculate the mean square height fluctuation of the membrane surface in @xmath63 space , @xmath69 ) we observe that the relation @xmath70 holds fairly well . this expression corresponds to that expected @xcite for a thermal fluid bilayer sheet with effective bending rigidity @xmath71 and effective surface tension @xmath72 , but we stress that its use in interpreting the simulations is empirical . the composite membrane surfaces studied in this work can not be regarded as fluid - like due to the assumed connectivity of the cytoskeleton matrix . much of our analysis is presented in terms of @xmath72 , which is obtained by fitting the simulated data to eq . [ eq : hk2 ] ( while assuming @xmath73 unless noted otherwise ) . for future reference , we define the free fluctuation spectrum " of the sheet , @xmath74 , as the result anticipated from eq . [ eq : h ] in the absence of any cytoskeletal contributions ( i.e. all @xmath75 ) @xmath76 two simulation methods were used to study the composite membrane system : fourier monte carlo ( fmc ) @xcite and fourier space brownian dynamics ( fsbd ) @xcite . in the limit of infinitely slow breaking and reformation of spectrin filaments ( quenched disorder ) , the @xmath57 variables are static over the course of any finite simulation and the system relaxes to a thermal equilibrium dependent upon the connectivity of the network . in this limit , both fmc and fsbd simulations may be used to calculate thermal averages and both should agree with one another . when spectrin links are breaking and reforming in time following the non - equilibrium scheme presented above , we must run dynamic fsbd calculations . our primary interest in this work is the non - equilibrium case , however fmc calculations were performed both to verify the accuracy of our fsbd simulations ( in the static network limit ) and to examine the scaling of some our equilibrium results with system size ( the fmc method is computationally more efficient than fsbd ) . the fmc scheme @xcite is a standard metropolis algorithm @xcite , which uses fourier modes of the bilayer , @xmath45 , as the bilayer degrees of freedom . there are two kinds of degrees of freedom in our system : membrane shape modes and lateral position of the nodes of the spectrin network . for the former , we attempt mc moves on the fourier modes of the system to enhance computational efficiency relative to naive real space schemes @xcite . while for the latter , we attempt moves that displace the @xmath25 position of the nodes to adjacent sites of the real space lattice inherent to the simulations . note that the shape of the membrane surface is continuously variable in our description , while the lateral position of nodes is discrete . the primary advantage to evolving the shape of the bilayer surface in fourier space is that we may tune the maximal size of attempted mc moves for each mode separately . in the case of a simple fluid bilayer ( without attached cytoskeleton ) under finite surface tension it is clear that a good choice for maximal move sizes is ( see eq . ( 1 ) in ref . @xcite ) @xmath77 where @xmath78 is the maximum attempted jump size of mode @xmath63 . in our case , a similar choice works well only for wavelengths sufficiently long that eq . [ eq : hk2 ] is obeyed , however it is a simple matter to tune each @xmath78 individually to optimize simulation efficiency . in practice , maximal jump sizes were tuned to insure that the acceptance ratio of all trials was approximately @xmath79 . the fsbd method has been fully detailed in our previous work @xcite . here we present only a minimal discussion to introduce the method . in this section we consider the simple case when @xmath80 in eq . ( [ eq : h ] ) is independent of time and postpone the discussion of @xmath81 for the next subsection . integrating eqs . ( [ eq : langevink ] ) and ( [ eq : langevin2 ] ) from @xmath82 to @xmath83 for small @xmath84 , we have @xmath85,\nonumber\\ \mathbf{r}_i(t+\delta t)=\mathbf{r}_i(t)+\frac{d}{k_bt}[\mathbf{f}_i(t)\delta t+\mathbf{r}_i(\delta t)].\label{eq : fsbd}\end{aligned}\ ] ] where @xmath86 the statistical properties of @xmath87 and @xmath88 follow directly those of @xmath89 and @xmath90 , @xmath91 in the simulation , @xmath87 and @xmath88 are drawn from gaussian distributions with means and variances specified above . since @xmath92 must be satisfied to insure the height field of the membrane is real valued , only about half of the @xmath93 need to be generated in each time step ; the remaining follow via complex conjugation . the precise formulation of this statement is somewhat complicated by the finite number of modes in our discrete fourier transform . readers are referred to ref . @xcite for a detailed discussion of how the full set of random forces are to be generated while preserving the real valued nature of @xmath6 . at each time step of the fsbd simulation , the following calculations are performed : \(1 ) take @xmath94 and @xmath95 from the last time step . \(2 ) evaluate @xmath26 and @xmath96 using eqs . ( [ eq : fr ] ) and ( [ eq : fir ] ) . \(3 ) compute @xmath97 by fourier transforming @xmath26 . \(4 ) draw @xmath98 and @xmath88 from the gaussian distributions specified above . \(5 ) compute @xmath99 and @xmath100 using eq . ( [ eq : fsbd ] ) . \(6 ) get @xmath101 through inverse fourier transformation for use in the next iteration . there is one complication in step 2 of the above procedure . while we evolve each @xmath102 as a continuous variable , the height field of the membrane is only specified at points on the real space lattice ( more precisely , it is only readily obtainable via fast fourier transformation at these points ) . to compute the forces for use in eqs . [ eq : fr ] and [ eq : fir ] , both @xmath103 and @xmath104 are approximated by assuming the position of node @xmath14 directly coincides with the nearest real space lattice site . while this approximation could potentially cause problems due to discontinuous jumps in the forces , the surfaces under study are only weakly ruffled . it was verified ( in the thermal case ) that fmc simulations with node positions strictly confined to lattice sites and fsbd simulations as outlined here were in good agreement . in practice , we choose @xmath84 small enough that further reduction has no consequence for the reported results , typically on the order of @xmath105 @xmath8s depending upon choice of lattice size @xmath61 . for the case of a dynamic network , every @xmath17 jumps back and forth between 1 and 0 , according to the two rates @xmath55 and @xmath56 ( see sec . [ sec : model ] ) . we use the stochastic simulation algorithm of gillespie ( kinetic monte carlo ) @xcite to pick random times for breaking and reformation events to occur . let us consider the event of reconnecting a link . if the link is disconnected at time @xmath106 , then the waiting time distribution for link reconnection is @xmath107 @xcite . a random number consistent with this distribution is obtained by applying the following transformation to a uniform random deviate @xmath48 @xcite @xmath108 a similar transformation , replacing @xmath55 with @xmath56 is used to determine the time at which an intact filament breaks apart . the set of times so obtained provides a complete trajectory for the behavior of each @xmath57 for use in eqs . [ eq : fr ] and [ eq : fir ] . the continuously distributed times are rounded off to the nearest time point in the discrete fsbd procedure . we note that our model assumes each link must always connect the same two nodes ( or be broken ) - there is no provision for a spectrin filament to dissociate from one node and reconnect elsewhere . in this section we assume a fully connected cytoskeleton meshwork ( @xmath109 for all links at all times ) . in this limit , there are no non - thermal effects and either kmc or fsbd simulations may be performed to calculate the equilibrium spectrum , @xmath110 . the qualitative features of this fluctuation spectrum have been predicted by fournier et . they argued that the behavior of the composite membrane surface should behave simply in two limits . in the short wavelength limit , the effect of the cytoskeleton might be expected to play a minor role ; neglecting the cytoskeletal terms in eq . [ eq : h ] leads to the prediction @xmath111 . in the long wavelength limit , the cytoskeleton should play an important role , but may be regarded as a continuous medium imparting an effective surface tension to the bilayer ( and possibly modifying the bare bilayer bending rigidity ) as in eq . [ eq : hk2 ] . `` short '' and `` long '' wavelengths referenced above are understood to be interpreted relative to the size of the individual spectrin links ( @xmath67 , see fig [ fig : corral ] ) . the original work of fournier et . al . assumed a sharp transition between these two regimes and fit experimental data with a transition at a wavelength of approximately @xmath112 . the simulations below are in qualitative agreement with this picture , but place the transition wavelength at @xmath67 and predict a finite width to the crossover region . both fmc and fsbd simulations were performed with identical results ( as expected ) . simulations were seeded from an initially flat membrane with the cytoskeletal anchor points arranged in a perfect lattice ( as indicated in fig . [ fig : corral ] ) . the initial configuration was allowed to fully equilibrate before collecting any data for analysis . the real - space lattice constant used in the simulations was @xmath113 with box dimensions of @xmath114 and @xmath115 for the hexagonal network and @xmath116 for the square network ( except where indicated otherwise , these values of @xmath61 and @xmath117,@xmath118 and @xmath119 were used in all reported simulations ) . this corresponds to 96 independent nodes ( 192 triangular corrals within the periodic box ) in the case of six - fold connected anchors and 64 independent nodes ( 64 square corrals within the periodic box ) in the case of the four - fold connected anchors . in preliminary runs , it was verified that neither increasing the sample size ( @xmath117,@xmath118 ) nor decreasing @xmath61 by a factors of @xmath120 significantly altered results ; the values outlined above were thus chosen to insure converged results with minimal computational expense . for convenience , all physical parameters used in the simulations are listed in table [ tab : para ] . our simulation results are shown in fig . [ fig : mcmesh ] . in the long wavelength limit , the results are well fit by eq . [ eq : hk2 ] , assuming @xmath121 and using @xmath122 for the effective tension in the square and hexagonal symmetry simulations , respectively . these tension values are not fit constants , but rather may be inferred from the cytoskeletal contribution to eq . [ eq : h ] . surface tension is an energy per unit area , so we may calculate its value as the ratio of entropic spring ( cytoskeleton ) energy per corral to the area per corral . in the square geometry , there are effectively two springs per corral ( each spring is shared by two adjacent corrals ) and using an idealized zero temperature geometry , a single corral has area @xmath123 and total spring energy of @xmath124 . the reported value for @xmath125 follows immediately . a similar calculation leads to the somewhat larger value of @xmath126 in the hexagonal geometry , reflecting the higher density of springs in this case . the numerical value of @xmath126 so calculated is @xmath127 k@xmath128 t @xmath8m@xmath129 j m@xmath130 , which is close to a theoretical fit @xcite of the experimental result @xcite ( see fig . [ fig : mcmesh ] ) . the fluctuation spectra in fig . [ fig : mcmesh ] indicate that free membrane predictions hold reasonably well out to wavelengths of approximately @xmath67 ( @xmath131 ) , with good convergence to long - wavelength behavior established by @xmath132 ( @xmath133 ) . the intermediate regime between wavelengths of @xmath67 to @xmath132 encompasses the crossover between the two limiting cases . while this fact is unfortunate in light of the experimental data for the rbc @xcite , which displays a crossover at longer wavelengths , the simulation predictions are unambiguous . the experimental data remains a mystery , but we do note that it may be accounted for by adoption of an ad hoc harmonic confining potential @xcite . one of the motivations for this work was to test the performance of the analytical theory developed in df . in particular , we anticipated that the mobility of cytoskeletal anchors would lead to some quantitative deviations from the df theory at short wavelengths . in fact , the mobility of anchor points leads to insignificant changes in the fluctuation spectrum for physical parameters relevant to the rbc ( see fig . [ fig : pineornot ] ) . we also note that the detailed analysis of df does slightly better in reproducing the simulated spectrum than does the adoption of surface tensions implied by eq . [ eq : sigma ] ( see fig . [ fig : kccorrect ] ) . at intermediate wavelengths , the small deviations of @xmath71 away from @xmath2 predicted by df do improve the fit as compared to our naive arguments , however the effect is very slight in comparison to the leading order effect of introducing a finite surface tension . as a final point , we note that the anisotropic nature of the square network over all wavelengths leads to some variance in fluctuations for a given magnitude of @xmath44 depending on the direction taken . in principle , the hexagonal network should not suffer from this effect @xcite , however the underlying square lattice taken for our simulations does introduce some anisotropy to the spectra ; these effects are most severe at short wavelengths . the spread of data points plotted in figs . [ fig : mcmesh ] - [ fig : kccorrect ] reflects this directional dependence in the spectra and should not be taken as evidence of statistical noise . as previously mentioned , statistical errors are less than the size of the symbols used in plotting . we now generalize the results of the previous section to include rbc membranes with randomly ( and dynamically changing ) broken cytoskeletal meshworks . as outlined in section [ sec : model ] , we assume the dynamics of spectrin association and disassociation are governed by simple rate processes . the rate constants for connecting a broken cytoskeletal filament , @xmath55 , and breaking an intact filament , @xmath56 , are assumed independent of the distance between filament end points on the bilayer surface and independent of all other connections within the network . this immediately leads to the conclusion that the probability for a filament to be intact at a given time is @xmath134 . equivalently , @xmath58 is the average percentage of intact filaments in the cytoskeletal network . for the moment , we simply take this picture as a hypothetical model for non - equilibrium dynamics in the rbc membrane . a discussion regarding the connection between this model and experiment will be provided in sec . [ sec : experiments ] . our analysis in this section centers around the calculation of @xmath135 , the effective tension of the composite membrane as inferred from long wavelength fluctuations . as indicated by the notation , this tension depends on the degree of connectivity within the network . not apparent in the notation is the fact that this tension also depends upon the magnitude of the rates @xmath56 and @xmath55 ( and not just the ratio of them in @xmath58 ) due to the non - equilibrium nature of the dynamics . extraction of @xmath135 follows the same general prescription as in the previous section ; the fluctuation spectrum is collected and compared to the empirical result of eq . [ eq : hk2 ] . at the longest wavelengths modeled in the simulations , it is found that eq . [ eq : hk2 ] does a good job of reproducing the simulation data . obvious theoretical estimates for @xmath136 are only available in the limiting cases of very fast spectrin ( re)association kinetics and very slow kinetics . in general , the effective tension as a function of @xmath58 ( and kinetic rates ) must be extracted from simulation . in the limit that spectrin ( re)association rates are much faster than all other time scales in the problem , each cyotoskletal link in the network behaves as an intact link with a diminished spring constant . the numerical value for this weakened spring constant is simply the time average of the spring as it flips between the two possible values of @xmath8 and zero . this picture immediately leads to @xmath137 similar arguments have been invoked previously to suggest a possible atp dependence in the shear modulus of the rbc membrane @xcite . such an atp concentration dependence may provide an explanation for rbc shape changes as a function of atp concentration @xcite . in the opposite regime , when network connectivity kinetics are far slower than any other timescale in the system , the membrane may be regarded as evolving thermally under the influence of quenched network disorder . the behavior of the system in the quenched disorder limit is analogous to the 2d percolation problem @xcite . as such , it is expected that the effective tension within the system must vanish at some finite critical @xmath138 . for values of @xmath58 less than @xmath139 , no global connectivity within the network exists and , consequently , there is no restoring force possible in response to area dilations of the membrane surface . the 2d connectivity percolation limit is known to occur at @xmath140 , where @xmath141 is the bond valence for each node ( i.e. @xmath142 with @xmath143 for the square network and @xmath144 with @xmath145 for the triangular network ) . furthermore , within the approximation of the mean field theory , it is expected that the decrease in @xmath146 as @xmath58 drops from one to @xmath139 is linear . that is @xmath147 for all @xmath148 and @xmath149 for @xmath150 . a brief justification for the percolation theory results quoted above may be found in the appendix . the discussion of `` fast '' and `` slow '' kinetics in the previous paragraph was intentionally left ambiguous , without specification of what time scales these quantities were to be compared with . it seems prudent to avoid a detailed discussion of this issue , due to the many different dynamic scales in this particular problem , however a crude discussion is appropriate . given our focus on long wavelength elastic properties , the most obvious timescale for comparison is the membrane relaxation time for the longest wavelength under observation . for equilibrium membranes at tension @xmath146 and bending rigidity @xmath2 , it is well known @xcite that relaxation of @xmath45 is exponential with a characteristic time @xmath151 this result follows immediately from eq . [ eq : langevink ] . considering only the longest wavelength modes in our simulations ( @xmath152 with @xmath153 for the square and @xmath154 for the hexagonal simulations ) and allowing for tensions between zero and the fully connected network values , we find that @xmath155 falls in the range of @xmath156 - @xmath157 . while these values are only required to hold for homogeneous equilibrium membranes , they were verified to hold reasonably well for the non - equilibrium membranes studied here when @xmath135 ( determined from the fluctuation spectrum ) is naively used in eq . [ eq : taum ] . the relaxation rate associated with the two - state spectrin dynamics is @xmath158 , which provides a time scale @xmath159 . the fast kinetics limit discussed above would be expected to apply for @xmath160 and the slow kinetics limit would apply for @xmath161 . fsbd simulations were run including kmc breaking and reforming events in the spectrin meshwork ( as detailed in sec . [ sec : simulation ] ) . although no equilibrium can be reached in these simulations due to the intrinsically non - equilibrium nature of the simulation , the systems do converge to a steady state regime . @xmath110 as a function of @xmath44 were collected , at steady - state , for both the square and hexagonal systems described in the previous section . three different sets of simulations were run in each geometry corresponding to different spectrin kinetic regimes . specifically , @xmath56 was chosen to assume the values @xmath162 s@xmath163 , @xmath164 s@xmath163 , and @xmath165 s@xmath163 . a variety of different connectivity percentages were simulated , spanning @xmath166 . given values for @xmath56 and @xmath58 , @xmath167 follows immediately allowing for complete specification of the model . the three choices for @xmath56 roughly correspond to the regimes @xmath168 , @xmath169 , and @xmath170 respectively . figures [ fig : squbroke ] and [ fig : tribroke ] display our simulation results , plotted in the form of effective surface tensions as a function of network connectivity . the tensions are calculated as in the lower panel of fig . [ fig : mcmesh ] using the largest wavelengths available in the simulations . in addition to the fsbd / kmc simulations , fmc results are plotted for the quenched disorder case corresponding to @xmath171 . in these simulations , an ensemble of random network connectivities consistent with the prescribed @xmath58 values were run without allowing the network connectivity to evolve . the results for @xmath110 were averaged over the ensemble to obtain @xmath135 . fmc was used in these cases for numerical efficiency and is valid since the evolution under conditions of static network connectivity is purely thermal . the speed of the fmc algorithm allowed a set of simulations to be run for larger membrane sizes in order to clarify the role of finite size effects . the figures clearly display both expected regimes . fast network reorganization yields results in good agreement with eq . [ eq : sigmap1 ] , whereas slow reorganization yields results consistent with eq . [ eq : sigmap2 ] . intermediate kinetic regimes interpolate between these two limiting cases . finite size effects appear not to play any role , except perhaps for the case of @xmath58 values very close to @xmath139 in the quenched disorder case , which is to be expected due to the percolation phase transition at this point . at values of intermediate membrane connectivity , it is possible for the systems to display a range of effective surface tension values , depending upon the relative timescales for spectrin kinetics as compared to membrane relaxation . the implications of this fact will be discussed further in the next section . evidence that rbc membrane shapes and shape fluctuations depend upon non - thermal energy expenditure comes from two different experimental sources . measurement of cell shape @xcite and shape fluctuations @xcite under a variety of mgatp concentrations suggest this fact directly . increases in mgatp concentration lead to enhanced membrane surface fluctuations . less directly , it has been shown that rbc membrane fluctuation amplitudes in the presence of mgatp depend upon the viscosity of the surrounding solvent @xcite . thermal properties of a physical system should not be influenced by transport coefficients , such as the viscosity , but should depend only upon system energetics via the partition function . biochemically , it is clear that the presence of mgatp leads to the phosphorylation of spectrin @xcite , which has been implicated in the observed shape changes and fluctuation amplitudes in rbc membranes under varying mgatp concentrations @xcite . whether or not this phosphorylation event ( and/or similar events in other molecular components of the network ) coincides with breaking of spectrin filaments is less clear , however the hypothesis that this is in fact the case @xcite is compelling and served as one of the major motivations for this study . in what follows , we assume that phosphorylation of spectrin induces individual filaments to dissociate from the network . we stress that this model is hypothetical and , indeed , that some studies refute the correlation between spectrin phosphorylation and rbc shape / elastic properties @xcite . in a naive model for spectrin phosphorylation , we assume the following kinetic equations for spectrin dissociation and association with the network as a whole . @xmath172 in the above , @xmath173 and @xmath174 stand for the associated and dissociated forms of spectrin , respectively , and it is assumed that the reactions proceed via catalysis by kinases and phosphatases . we assume atp concentration is held fixed , either by endogenous biochemistry within the rbc in vivo or by experimental control in vitro . using our previously introduced notation , we write @xmath175=cp$ ] and @xmath176=c(1-p)$ ] , where @xmath177 is the total concentration of spectrin on the cell surface . the steady state condition is formulated as @xmath178 , which , in conjunction with kinetic eqs . [ eq : phosphor ] , leads to @xmath179 here , both kinetic constants have been wrapped up into the single constant @xmath180 . @xmath42 is the concentration of atp in solution . @xmath180 has been defined such that the condition @xmath181 implies @xmath182 ( in this section we assume the hexagonal network due to its direct connection to rbc systems ) , i.e. @xmath181 specifies the concentration of atp necessary to reach the percolation phase transition in systems with quenched disorder . while this model neglects the observed dependence of spectrin dephosphorylation on @xmath42 @xcite , it has the advantage of simplicity in the absence of quantitative kinetic data and follows the line of reasoning previously introduced @xcite , allowing for comparison to earlier work . [ eq : patp ] predicts the connectivity of the network as a function of atp concentration and allows for comparison between the modeling of the previous section and experimental data . experimental data for rbc fluctuations as a function of atp concentration is available only in terms of real space height amplitudes , @xmath6 @xcite ( see fig . [ fig : exp ] ) . following eqs . ( [ eq : fourierk ] ) and ( [ eq : hk2 ] ) ( and for simplicity assuming @xmath183 ) , @xmath184 where the interval between adjacent @xmath185 or @xmath186 is @xmath187 . in the limit of large @xmath119 , the sum @xmath188 can be approximated by the integral @xmath189 , and we have @xmath190 the limiting expressions and simulations of the preceding section provide a route toward estimation of @xmath135 under various rates of spectrin association / dissociation . [ eq : patp ] enables us to translate these results into tensions as a function of atp concentration . the resulting tensions may be used in eq . [ eq : realh ] to calculate the real space membrane fluctuation amplitudes as a function of atp concentration . rbc s have a diameter of 7 microns and we use @xmath191 @xcite in our comparison to experiment . it is also important to note that experimental observations of rbc fluctuations were carried out on cells that were allowed to firmly adhere to [ a ] glass substratum " @xcite under the influence of their natural ( presumably of electrostatic origin ) affinity for such surfaces . the adhesion of the bilayer to an underlying supporting matrix will be assumed to contribute a bare surface tension , @xmath192 to the membrane of the rbc @xcite . that is , we assume @xmath193 . we take @xmath180 and @xmath192 as two fitting parameters to reproduce the experimental data . the resulting fits , assuming the two extreme cases of fast spectrin kinetics ( eq . [ eq : sigmap1 ] assumed ) and slow spectrin kinetics ( eq . [ eq : sigmap2 ] assumed ) are displayed in fig . [ fig : exp ] . the values adopted by the fitting parameters in these two cases are @xmath194 ( both cases ) , and @xmath195 mm for the fast kinetics case and @xmath196 mm for the slow kinetics case . it turns out that the limiting case of slow spectrin kinetics provides the most satisfactory fit of the data , not only in comparison to the fast kinetics case , but also with respect to intermediate kinetic regimes ; the sharp onset of tensionless fluctuations at @xmath180 naturally leads to the observed plateau in the experimental data . the slowest possible time - scale associated with rbc membrane fluctuations in our model ( @xmath155 ) is on the order of a couple of seconds . this assumes a tension - free membrane and @xmath197 m . our data analysis suggests that spectrin breaking / reformation dynamics are substantially slower than this , since the quenched - disorder ( percolation theory ) results provide the best fit to the experimental data . this finding would appear to be consistent with the experimental observation @xcite that establishment of steady - state levels of spectrin phosphorylation occurs on the time scale of minutes to tens of minutes with changes in mgatp concentration . previous theoretical work exploring the consequences of spectrin dissociation on membrane fluctuations @xcite has assumed the time - scale for spectrin kinetics to be on the order of milliseconds . within the context of the present model , rates this fast seem unlikely . experimentally , it is observed that rbc fluctuation amplitudes depend not only upon mgatp concentration , but also upon the viscosity of the solvent surrounding the membrane @xcite ; increasing viscosity above that of aqueous buffer solution leads to decreases in observed fluctuations . membrane fluctuations at thermal equilibrium should yield identical statistics regardless of solvent viscosity ( although the dynamics will certainly differ ) , so these experiments provide additional proof of the non - thermal character of rbc fluctuations . within the present model , eq . ( [ eq : taum ] ) clearly displays the linear relationship between inverse viscosity and the relaxation timescales for individual membrane modes . in the preceding section , it was argued that the ratio between membrane relaxation rates and spectrin kinetic rates determines the magnitude of effective tension that must be used if one desires to fit membrane fluctuations to the functional form of eq . [ eq : hk2 ] . at a given connectivity , @xmath58 , fluctuation amplitudes decrease ( @xmath72 increases ) as the spectrin kinetic cycle increases in speed . alternatively , since it is the ratio of kinetics to membrane relaxation that is relevant , we expect fluctuation amplitudes to decrease as viscosity is raised assuming all other system parameters are held fixed . qualitatively , this trend is in agreement with the experimental results . to obtain quantitative comparison with the variable viscosity data it is necessary to evaluate @xmath198 explicitly from direct simulation , without assuming a single effective tension as in eq . [ eq : hr ] . within our model , fluctuation amplitudes are affected by viscosity changes only by shifting the timescale for membrane dynamics relative to spectrin kinetics . a relative shift in timescale ( at constant @xmath58 ) corresponds to a vertical motion between the limiting regimes plotted in fig . [ fig : tribroke ] . in order for viscosity changes to yield any effect , the timescales for spectrin kinetics and membrane relaxation must be comparable . ( e.g. if spectrin dissociation takes an hour , increasing viscosity by a factor of 10 will not remove you from the limiting `` slow kinetics regime '' . similarly , if dissociation took a nanosecond , no reasonable viscosity change could promote the system out of the `` fast kinetics regime '' . ) and , if these timescales are comparable , differences in the individual relaxation rates at each wavelength will lead to different effective tensions as a function of wavelength . in practice , it is not feasible to run simulations for a full @xmath199 patch with sufficient sampling to obtain reliable results . it is possible to run a @xmath200 patch . the effective tensions as a function of wavevector were extracted from this rectangular geometry and were used in a generalized version of eq . [ eq : hr ] with @xmath201 to obtain @xmath198 for the full @xmath199 system . to simultaneously find agreement with the atp concentration data ( fig . [ fig : exp ] ) and the variable viscosity data it is necessary to assume spectrin rates that are slow relative to membrane relaxation when the solvent is pure buffer solution ( as in the conditions of ref . @xcite and fig . [ fig : exp ] ) , but are poised to move out of this limiting regime with small viscosity increases . by trial and error , it was found that @xmath202s for @xmath203 mm ( the atp concentration conditions of ref . @xcite ) satisfies this condition and yields the best fit to both data sets . other model parameters were taken identical to the slow spectrin kinetics " fit to the data of fig . [ fig : exp ] . in particular , @xmath204 , @xmath196 mm and @xmath205 . in fig . [ fig : exp2 ] we plot real space membrane fluctuations as a function of @xmath206 for viscosity values spanning an order of magnitude . we observe a reduction in fluctuation amplitudes of about 20% over the entire range of viscosities studied , which falls just outside the errors of the experimental data @xcite . it is important to stress that figs . [ fig : exp ] and [ fig : exp2 ] simultaneously fit two very different experiments to a single set of physical parameters . our results agree with both experiments quite well . from a mathematical standpoint , the results of the preceding section could be viewed as a success ; we fit the available experimental data with our theoretical model . however , the physical implications of the derived fit parameters are troubling . most worrisome is the critical concentration of atp for onset of the percolation limit , @xmath196 mm . this implies that at physiological conditions ( @xmath207 mm ) , the spectrin network is 89% dissociated . a similar degree of dissociation ( 83 % ) is predicted for the @xmath203 mm conditions of the variable viscosity experiments described above . as presented , it is clear that our model should not be taken seriously in the limit of slow spectrin kinetics and @xmath208 , since we do not allow the connectivity of the network to evolve away from its starting point ( i.e. a given spectrin tetramer is assumed to always link the same two vertices or be disassociated ) . in a severely compromised network , how could a given filament ever be expected to find its assigned association point following a dissociation event ? as mentioned previously , there does not appear to be definitive experimental data relating atp consumption to spectrin network dissociation . rather , this is a plausible hypothesis advocated in reference @xcite , based upon the limited biochemical data that is available for the various constituents of the rbc cytoskeleton . we can not reconcile this picture with the simulations carried out in this work , primarily for the reason outlined in the previous paragraph . it is possible that atp consumption could lead to a significant change in the elastic properties of individual spectrin filaments without completely dissociating the filament from the network . in such a picture , the two forms of spectrin introduced above , @xmath209 and @xmath210 , would correspond to filaments with weak and strong effective spring constants , respectively ( or different natural lengths , etc . ) . in this context , the analysis we present is sensible since network connectivity is always maintained , but the elastic properties of individual links are changing . the percolation limit in this picture corresponds to the point at which it is impossible to follow a path across the spectrin network without stepping on at least one weak @xmath209 link . although such a picture is completely hypothetical , we note that it is theoretically possible to alter the effective elastic properties of spectrin filaments while maintaining lateral integrity of the network @xcite . the underlying physical picture pursued in this work was motivated by and is similar in spirit to the work of safran and gov @xcite ( i.e. a dynamically breaking and reforming spectrin network ) . however , the simulation model implemented to study this system is more similar to the work of dubus and fournier @xcite ( with the additional possibility of dynamically breaking bonds ) . we have interpreted our simulation results within the context of a simple percolation theory . while this picture shares some similarity with the analysis of ref . @xcite in the context of global effects associated with variable atp concentration , we find differences between the present model and ref . @xcite in the description of local fluctuation dynamics . for example , we find no evidence in our simulations for localized forces of the sort discussed in ref . @xcite . instead , the increased amplitudes of fluctuation at high atp concentration are attributable to global properties of the spectrin network within our model . we also find that the rates associated with spectrin kinetics must be orders of magnitude slower than assumed in ref . @xcite in order to reconcile our simulation model with the experiments of refs . however , it should be pointed out that the particular rates assumed in ref . @xcite were only rough estimates and are not essential to the qualitative predictions of that work @xcite . the starting point for all of our simulations , eq . [ eq : h ] , assumes a central - force network for the cytoskeletal matrix and a phantom entropic spring description for the spectrin filaments . in addition , we do not allow for the possible formation of 5-fold or 7-fold defects @xcite in our network . inclusion of direct interactions between spectrin and the bilayer surface , generalizing the description of the network to deviate from the central - force picture , and/or allowing for defects in network connectivity could all potentially alter the quantitative findings of this work . the present simulation model was adopted for ease of numerical implementation and to facilitate comparison to the existing work of dubus and fournier @xcite . this study provides a detailed analysis of the behavior of a specific model for a composite membrane system ( lipid bilayer plus actively labile cytoskeletal network ) . this model is motivated by our ( incomplete ) picture for the structure and kinetics of the rbc membrane surface and may prove useful in the development of more refined models in the future . in particular , it should be noted that more refined descriptions of the spectrin network have been developed that allow for interactions between the bilayer surface and the filaments and a more complete description of polymer elasticity beyond the simple gaussian - chain model @xcite . future modeling , incorporating some of these approaches for the behavior of the cytoskeleton , would be highly desirable and is currently under investigation . we note in closing that a very recent study by evans et . @xcite directly contradicts the experimental results in refs . this recent series of experiments finds no correlation between atp concentration and rbc membrane fluctuations . given the severe disagreement between experimental results on the rbc system , one has almost complete freedom in interpretation of our simulation results . the picture outlined above suggests that we can obtain quite good agreement between simulation and experiments that do measure atp dependence of the rbc fluctuations . within this picture , it seems necessary to assume that our broken " spectrin links actually correspond to intact links with diminished spring constant and/or a non - zero natural length . another , equally valid , interpretation of our results is that it is impossible to reconcile an actively breaking cytoskeletal meshwork picture with available experimental data due to the high degree of network dissociation predicted under physiological conditions within such a picture . given the uncertain experimental landscape , it seems impossible to make a definitive statement in favor of either viewpoint . we thank golan bel , nir gov and kim parker for useful discussions . this work was supported by the national science foundation ( grant no . che-0321368 and grant no . che-0349196 ) . f. b. is an alfred p. sloan research fellow and a camille dreyfus teacher - scholar . the calculation of an effective tension in the limit of quenched disorder within the spectrin network is closely related to the 2d bond percolation problem @xcite . a mean field treatment for the percolation problem was first developed in the context of electric conductivity @xcite and was later extended to calculate the elastic moduli of networks of hookean springs with finite @xcite and zero natural lengths @xcite . the below discussion summarizes the arguments of ref . @xcite for the readers convenience . let us consider a randomly connected meshwork with only a fraction ( @xmath58 ) of the links intact . each intact link is a spring with spring constant @xmath8 and a natural extension of zero . in the spirit of mean field theory , we assume that the global elastic properties of this imperfect meshwork are well approximated by a complete meshwork comprised of springs with spring constant @xmath211 at every link ( see fig . [ fig : meshforce ] ) even though individual links will be either completely intact or fully broken . our problem is to determine an expression for @xmath211 . for this purpose , let us consider an arbitrary link @xmath212 in the meshwork . the real spring constant , @xmath213 , associated with this link is either @xmath8 if the link is connected , or 0 if it is not . if we apply a force @xmath214 across this link ( see fig . [ fig : meshforce ] ) , the extension @xmath215 is given by @xmath216 where @xmath217 is an effective spring constant reflecting the contribution of all network components except the link @xmath212 . it can be shown that @xcite @xmath218 where @xmath141 is the connectivity of the meshwork , for example , @xmath142 for the square meshwork and @xmath144 for the hexagonal meshwork considered in this work . on the other hand , if we assume the mean spring constant @xmath211 for link @xmath212 , we predict a mean extension , @xmath219 to achieve consistency within the mean field approximation , we must have @xmath220 where the second equality is simply a more explicit version of the first . this equation is readily solved to yield @xmath221 where @xmath222 to calculate @xmath135 , we simply replace @xmath8 in eq . [ eq : sigma ] by @xmath211 from eq . [ eq : mum ] , since the latter is now the average spring constant of the equivalent complete meshwork . taking @xmath142 and @xmath144 for the square and hexagonal meshwork respectively , we arrive at eq . [ eq : sigmap2 ] . one should note that eqs . [ eq : sigmap2 ] and [ eq : pc ] are only results of a mean field approximation . in the vicinity of @xmath139 , it is well known that the correlation effects are strong and that mean field theory breaks down . however , the mean field results are adequate to describe the behavior of @xmath146 outside the immediate vicinity of @xmath139 as evidenced by figs . [ fig : squbroke ] and [ fig : tribroke ] , which is sufficient for the purposes of this work . between the nodes . ( b ) : side view : the protein anchors are confined to the lipid bilayer , but may diffuse laterally . the coils underneath the bilayer are spectrin tetramers . the contour length of the tetramers is approximately @xmath223 , so there is considerable extra length coiled up below the membrane surface . ( c ) : one end of a link is disconnected from the node , possibly with the help of atp.,scaledwidth=45.0% ] '' . all lines together ( solid and dotted ) correspond to a complete hexagonal meshwork . ( a ) : above the percolation threshold , @xmath148 , the connected links have global connectivity , i.e. , they form a cluster extending across the entire cell . the corresponding effective surface tension will always be finite . ( b ) : below the percolation threshold , @xmath150 , there is no global connectivity in the meshwork , i.e. connected links form finite islands of connectivity that do not span the cell . the corresponding effective surface tension becomes zero if the links are breaking and reforming slowly . ( c ) : even at instantaneous connectivities below the percolation threshold , there can be a finite surface tension if spectrin breaking / formation kinetics are sufficiently fast . in this limit , each link can be thought of as a spring with a reduced restoring force since out of plane membrane undulations evolve more slowly than the lifetime of the link.,scaledwidth=50.0% ] , ( normalized by @xmath224 ) versus inverse wavenumber for the our rbc membrane model assuming a perfectly intact cytoskeletal network . the squares and triangles indicate simulation results for square and hexagonal meshworks respectively . the error bars are approximately the same size as the symbols . the solid lines are theoretical results using eq . ( [ eq : sigma ] ) for @xmath72 and @xmath73 in eq . [ eq : hk2 ] . the wavevector @xmath225 is indicated on the horizontal axis to show where crossover from free membrane to long - wavelength behavior might be expected to occur . the circles are experimental results taken from ref . lower : the approach of @xmath146 to its limiting long - wavelength value is displayed . in this case , each point in the spectrum on the left was individually fit to eq . [ eq : hk2 ] by adjusting @xmath72 ( assuming @xmath73 ) . the resulting @xmath72 as a function of @xmath226 are shown . at the longest wavelengths simulated , the inferred tensions converge to the predicted values.,title="fig:",scaledwidth=45.0% ] , ( normalized by @xmath224 ) versus inverse wavenumber for the our rbc membrane model assuming a perfectly intact cytoskeletal network . the squares and triangles indicate simulation results for square and hexagonal meshworks respectively . the error bars are approximately the same size as the symbols . the solid lines are theoretical results using eq . ( [ eq : sigma ] ) for @xmath72 and @xmath73 in eq . [ eq : hk2 ] . the wavevector @xmath225 is indicated on the horizontal axis to show where crossover from free membrane to long - wavelength behavior might be expected to occur . the circles are experimental results taken from ref . @xcite . lower : the approach of @xmath146 to its limiting long - wavelength value is displayed . in this case , each point in the spectrum on the left was individually fit to eq . [ eq : hk2 ] by adjusting @xmath72 ( assuming @xmath73 ) . the resulting @xmath72 as a function of @xmath226 are shown . at the longest wavelengths simulated , the inferred tensions converge to the predicted values.,title="fig:",scaledwidth=45.0% ] for a hexagonal lattice . the circles are the data for the case of immobile nodes placed on a perfect lattice . while the triangles are the data for the case of mobile nodes ( identical data to fig . [ fig : mcmesh ] ) . mobility of the cytoskeleton attachment points does not significantly alter the simulation results.,scaledwidth=50.0% ] . the solid line is the df model prediction . at intermediate wavelengths df does a slightly superior job in fitting the simulations . both approximations fail at small wavelengths and converge to the same behavior at long wavelengths.,scaledwidth=50.0% ] versus @xmath58 is plotted for a square meshwork . the circles , triangles , and diamonds are fsbd / kmc results for @xmath227 ( @xmath168 ) , @xmath228 ( @xmath169 ) and @xmath229 ( @xmath170 ) respectively . the squares and the stars are fmc results ( @xmath230 ) . all simulations use box sizes identical to those introduced in the previous section , except for the stars . stars represent fmc simulations on systems that are twice as large in both linear dimensions . for clarity , only error bars for the circles and the stars are shown , but all simulations have similar errors . the dashed line is the prediction of eq . ( [ eq : sigmap1 ] ) . the solid line is the prediction of eq . ( [ eq : sigmap2]).,scaledwidth=50.0% ] versus @xmath58 is plotted for a hexagonal meshwork . all symbols and lines have the same meaning as in fig . [ fig : squbroke ] , only the meshwork connectivity has been changed . ( the obvious discrepancy between simulation and theory at @xmath231 is due to the rectangular geometry of our simulation box . it is impossible to perfectly fit a portion of a hexagonal network within a rectangle , so the theoretical results that assume perfect hexagonal symmetry are only approximately valid in this case.),scaledwidth=50.0% ] , as a function of atp concentration ( normalized by the zero atp values , @xmath232 ) . the circles with error bars are the experimental data @xcite . the solid line is a theoretical fit assuming @xmath126 in eq . ( [ eq : sigmap2 ] ) , corresponding to the case of @xmath233 ( slow spectrin kinetics ) . the dashed line is a theoretical fit assuming @xmath126 in eq . ( [ eq : sigmap1 ] ) , corresponding to the case of @xmath234 ( fast spectrin kinetics).,scaledwidth=50.0% ] , as a function of inverse viscosity of the solvent . the circles with error bars are the simulation results assuming @xmath235 @xmath8 m , @xmath236 , @xmath196 mm , @xmath203 mm and @xmath202s . the solid line is a fit to our simulation data . the dashed line and the two dashdotted lines are the experimental fitting line and the range of experimental errors taken from fig . 2 of ref .

we calculate the two loop anomalous dimension matrix involving current current operators , qcd penguin operators , and electroweak penguin operators especially relevant for weak non leptonic decays , but also important for decays . the calculation is performed in two schemes for : the dimensional regularization scheme with anticommuting ( ndr ) , and in the t hooft veltman scheme . we demonstrate how a direct calculation of diagrams involving in closed fermion loops can be avoided thus allowing a consistent calculation in the ndr scheme . the compatibility of the results obtained in the two schemes considered is verified and the properties of the resulting matrices are discussed . the two loop corrections are found to be substantial . the two loop anomalous dimension matrix , required for a consistent inclusion of electroweak penguin operators , is presented in a subsequent publication . remarks : below you find the * latex source * of our recent paper . the accompanying figures ( in postscript ) are distributed separately in an uuencoded compressed tar - file as required by the rules for postscript files on this bulletin board . the latex distribution is wrapped up in a single file with `` snip '' lines delimiting the individual files . the corresponding file names are indicated in the line `` > > > < filename < < < '' just above the leading `` snip '' lines . the * full * distribution consists of the following files : latex part ( this file ) : 2ladmqcd.tex : latex source of the paper . artcustom.sty , a4wide.sty , a4.sty : required style files except epsf.sty . epsf.sty is of any use only in connection with the rokicki dvips converter . a4wide.sty uses this particular version of a4.sty ( with the later separately available at this bulletin board although i do nt know whether the two are compatible ) . postscript part ( distributed separately ) : fig1qcd.ps , fig2qcd.ps , fig3qcd.ps , fig4qcd.ps , fig5qcd.ps , fig6qcd.ps , fig7qcd.ps : postscript files read in by `` 2ladmqcd.tex '' via epsf.sty provided with the rokicki dvips dvi - to - postscript converter . if you do nt have the rokicki dvips you should comment out the relevant epsf - commands in `` 2ladmqcd.tex '' . best regards , markus lautenbacher markus e. lautenbacher , | phone : + 49/89/3209 - 2387 technical university munich , | fax : + 49/89/32092296 physics dept .- theo . physics t31,| internet : d-8046 garching , frg . [email protected] > > > 2ladmqcd.tex < < < snip snip snip snip snip snip snip snip

hep-ph9211304

effective low energy hamiltonians for non leptonic weak decays of hadrons are usually written as linear combinations of four quark operators . the coefficients of these operators , the wilson coefficient functions , can be calculated in the renormalization group improved perturbation theory , as long as the normalization scale @xmath7 is not too low . the size of these coefficients depends on @xmath7 through the qcd effective coupling constant and on the anomalous dimensions of the operators in question . since the operators generally mix under renormalization , one deals with anomalous dimension matrices rather than with single anomalous dimensions . in the case of the @xmath2 hamiltonian , relevant for instance for the @xmath8 rule and the ratio @xmath9 , there are ten operators @xmath10 which have to be considered . consequently one deals with a @xmath11 anomalous dimension matrix . the ten operators in question can be divided into three classes : * current current operators @xmath12 originating in the usual w exchange and subsequent qcd corrections , * qcd penguin operators @xmath13 originating in qcd penguin diagrams , and * electroweak penguin operators @xmath14 originating in the photon penguin diagrams . explicit expressions for these operators are given in eq . ( [ eq:2.1 ] ) . because of the presence of electroweak penguin operators a consistent analysis must involve anomalous dimensions resulting from both strong and electromagnetic interactions . working to first order in @xmath15 but to all orders in @xmath16 the following anomalous dimension matrix is needed for the leading and next to leading logarithmic approximation for the wilson coefficient functions , @xmath17 in the leading logarithmic approximation only the one loop matrices @xmath18 and @xmath19 have to be considered . inclusion of next to leading corrections requires the evaluation of the two loop matrices @xmath20 and @xmath21 . in addition some one loop finite terms are needed . what is known in the literature are the matrices @xmath18 and @xmath19 @xcite@xcite and a @xmath22 submatrix of @xmath20 involving the current current operators @xmath23 and @xmath24 @xcite . the purpose of the present and a subsequent paper is to complete the evaluation of the matrix @xmath20 and to calculate @xmath21 . together with certain one loop finite terms this will allow the extension of the phenomenology of non leptonic @xmath2 transitions beyond leading logarithms . the results for the @xmath25 submatrix of @xmath20 involving current current and qcd penguin operators together with some phenomenological implications have already been presented by us in ref . @xcite . in the present paper we give the details of these two loop calculations and generalize them to the full matrix @xmath20 . the matrix @xmath21 is considered in a subsequent publication @xcite . in particular , we present the results for an arbitrary number of colours , @xmath26 , which is useful for the applications of the @xmath27 expansion . the phenomenological implications of these new results will be discussed in ref . @xcite . the calculation of the matrices @xmath20 and @xmath21 is very tedious . it involves a large number of two loop diagrams which make the use of an algebraic computer program almost mandatory .- algebra tracer , written in mathematica @xcite , which we extensively used during the calculation , has been developed by two of us @xcite and is available from the authors . ] moreover one has to deal with several subtleties which are absent in the calculations of @xmath18 and @xmath19 . first of all , the two loop anomalous dimensions depend on the renormalization scheme for operators and the treatment of @xmath5 in @xmath28 space time dimensions . it is known for instance on the basis of ref . @xcite , that the @xmath29 submatrix of @xmath20 involving @xmath30 and @xmath24 calculated in the ndr scheme ( naive dimensional regularization with anticommuting @xmath5 ) differs substantially from the one obtained in the hv scheme ( non anticommuting @xmath5 @xcite ) . the same feature is found for the full matrices @xmath20 and @xmath21 . the important point is that the difference between @xmath31 calculated in two different schemes is on general grounds entirely given in terms of @xmath32 and the finite parts of one loop diagrams calculated in the two schemes in question . this relation is very useful for checking the compatibility of two loop calculations performed in different renormalization schemes and plays an important role in demonstrating the scheme independence of physical quantities . it is given in eq . ( [ eq:2.12 ] ) . another subtle point is the dependence of the two loop anomalous dimensions obtained in the ndr scheme on whether a given operator has been put in colour singlet or colour non singlet form such as @xmath33 in @xmath34 dimensions @xmath24 and @xmath35 would be equivalent by means of fierz transformation . their one loop anomalous dimensions are the same . again the two loop anomalous dimensions for these two forms of operators can on general grounds be related to each other by calculating one loop diagrams . this feature turns out to be fortunate in that it offers one way to deal with dangerous closed fermion loops involving odd numbers of @xmath5 in two loop diagrams . the latter are known to be ambiguous in the case of the ndr scheme . indeed , as we will demonstrate explicitly in section 5 and in ref . @xcite , it is possible by using different forms of the operators and the relation mentioned above , to express all diagrams with closed fermion loops in terms of diagrams not containing such loops . in this way , consistent calculations of @xmath20 and @xmath21 can be performed in the ndr scheme and compared with the one obtained in the hv scheme , free of @xmath5 problems . the following sections constitute the calculation of @xmath20 and a detailed discussion of the subtle points mentioned above . in ref . @xcite , the methods developed here are generalized to include photon exchanges and are used to calculate @xmath21 . our paper is organized as follows : in section 2 , we give a list of the ten operators in question , and we classify the one and two loop diagrams into current current and penguin diagrams . we also discuss the basic formalism necessary for the calculation of @xmath20 . in section 3 , we recall the matrix @xmath18 , and we calculate the finite terms of one loop diagrams in the ndr and hv scheme . in section 4 , the calculations and results for two loop current current diagrams are presented . in section 5 , an analogous presentation is given for two loop penguin diagrams . in section 6 , we combine the results of the previous sections to obtain @xmath20 in ndr and hv schemes . we discuss various properties of this matrix , in particular its large-@xmath26 limit . section 7 contains a brief summary of our paper . in appendices a and b explicit expressions for the elements of the @xmath0 matrices @xmath18 and @xmath20 for an arbitrary number of colours ( @xmath26 ) and flavours ( @xmath36 ) are given . in appendix c the corresponding results for @xmath20 in the case of @xmath37 are presented . the ten operators considered in this paper are given as follows @xmath38 where @xmath39 , @xmath40 denote colour indices ( @xmath41 ) and @xmath42 are quark charges . we have omitted colour indices for the colour singlet operators . @xmath43 refers to @xmath44 . this basis closes under qcd and qed renormalization and is complete if external momenta and masses are neglected . however , at intermediate stages of the calculation , we have to retain operators which vanish on shell . this will be discussed in detail in section 5.2 . at one place , it will become useful to study a second basis in which the first two operators are replaced by their fierz conjugates , @xmath45 with the remaining operators unchanged . in fact , the latter basis is the one used by gilman and wise @xcite . we prefer however to put @xmath24 in the colour singlet form as in eq . ( [ eq:2.1 ] ) , because it is this form in which the operator @xmath24 enters the tree level hamiltonian . let us finally recall that the fierz conjugates of the @xmath46 operators @xmath47 and @xmath48 are given by @xmath49 with similar expressions for @xmath50 and @xmath51 . here the fierz conjugates of the @xmath53 penguin operators @xmath54 and @xmath55 , to be denoted by @xmath56 , can be found in analogy to ( [ eq:2.2 ] ) . in order to calculate the anomalous dimension matrices @xmath18 and @xmath20 , one has to insert the operators of eq . ( [ eq:2.1 ] ) in appropriate four point functions and extract @xmath57 divergences . the precise relation between the @xmath57 divergences in one and two loop diagrams and the one and two loop anomalous dimension matrices will be given in the following subsection . here , let us only recall that insertion of any of the operators of eq . ( [ eq:2.1 ] ) into the diagrams discussed below results in a linear combination of the operators @xmath58 . a row in the anomalous dimension matrix corresponding to the inserted operator can then be obtained from the coefficients in the linear combination in question . there are three basic ways a given operator can be inserted into a four point function . they are shown in fig . 1 , where the cross denotes the interaction described by a current in a given operator of eq . ( [ eq:2.1 ] ) , and the wavy line denotes a gluon . we shall refer to the insertions of fig . [ fig:1](a ) as `` current current '' insertions . the insertions of fig . [ fig:1](b ) and ( c ) will then be called `` penguin insertions '' of type 1 and type 2 respectively . the complete list of the diagrams necessary for one and two loop calculations is given in figs . at the one loop level one has three current current diagrams , fig . 2 , to be denoted as in ref . @xcite by @xmath59 @xmath60 and one penguin diagram of each type , fig . 3 , to be denoted by @xmath61 and @xmath62 for type 1 and type 2 insertion respectively . at the two loop level there are 28 current current diagrams shown in fig . 4 , to be denoted by @xmath63 @xmath64 , and 14 penguin diagrams of each type shown in fig . 5 , to be denoted by @xmath65 @xmath66 and @xmath67 @xmath68 respectively . the diagrams obtained from the ones given here by left - right reflections have not been shown . in the case of current current diagrams also up - down reflections have to be considered . in fig . 6 , we show two penguin diagrams which have no @xmath57 divergence and hence do not contribute to the anomalous dimensions , and in fig . 7 , some examples of penguin diagrams which vanish identically in dimensional regularization are given . the anomalous dimensions of the operators @xmath58 , calculated in the @xmath69 scheme , are obtained from the @xmath57 divergences of the diagrams of figs . 2 5 and from the @xmath57 divergence of the quark wave function renormalization . let us denote by @xmath70 a column vector composed of the operators @xmath58 . then @xmath71 where @xmath72 stands for bare operators . working in @xmath73 dimensions , we can expand @xmath74 in inverse powers of @xmath75 as follows @xmath76 where @xmath77 is the qcd coupling constant . inserting ( [ eq:2.5 ] ) into ( [ eq:2.4 ] ) one derives a useful formula @xmath78 let us next denote by @xmath79 and @xmath80 the renormalized and the bare four quark green functions with operator @xmath70 insertions . strictly speaking @xmath79 and @xmath80 are matrices , because the insertion of a single operator in a given diagram results in a linear combination of operators . at the one loop level @xmath81 is obtained by evaluating the diagrams of figs . 2 and 3 . at the two loop level it is found by evaluating the diagrams of figs . 4 and 5 and subtracting the corresponding two loop counter terms . for the results of tabs . 1 4 , we have made these subtractions diagram by diagram . next @xmath82 where @xmath83 is a matrix which represents the renormalization of the four quark fields . in the case of pure qcd it is a diagonal matrix with all the diagonal elements being equal to @xmath84 , where @xmath85 is the quark wave function renormalization , defined by @xmath86 we expand @xmath81 and @xmath85 in inverse powers of @xmath75 as follows @xmath87 where @xmath88 and @xmath89 \ ; . \label{eq:2.11a}\ ] ] demanding @xmath90 to be finite , we find @xmath91 , and , using ( [ eq:2.6 ] ) , the basic formulae for the one and two loop anomalous dimension matrices , @xmath92 \label{eq:2.10 } \ , , \\ \big({\hat{\gamma}_{\rm s}^{(1)}}\big)_{ij } & = & -\,4 \left[\,2\,a_{2}\,\delta_{ij } + \left(b_{2}\right)_{ij}\,\right ] \label{eq:2.11 } \,.\end{aligned}\ ] ] the two loop anomalous dimension matrices depend on the renormalization scheme and in particular on the treatment of @xmath4 in @xmath93 dimensions . in ref . @xcite , we have derived a relation between @xmath94 and @xmath95 calculated in two different renormalization schemes @xmath96 and @xmath97 , @xmath98 + 2 \,\beta_{0 } \delta{\hat{r}_{\rm s}}\qquad { \rm with } \qquad \delta{\hat{r}_{\rm s}}\ ; = \ ; ( { \hat{r}_{\rm s}})_{b } - ( { \hat{r}_{\rm s}})_{a } \ , , \label{eq:2.12}\ ] ] where @xmath99 is the leading coefficient in the expansion of the @xmath40function @xmath100 and the @xmath101 are defined by @xmath102 \langle \vec{q}^{(0 ) } \rangle \ ; . \label{eq:2.14}\ ] ] here , @xmath103 is a tree level matrix element and @xmath104 denotes renormalized one loop matrix elements calculated in the scheme @xmath105 . the matrices @xmath101 are obtained by calculating the finite terms in the one loop diagrams of figs . 2 and 3 . relation ( [ eq:2.12 ] ) is very useful as it allows to test compatibility of the two loop anomalous dimension matrices calculated in two different renormalization schemes . it also plays a central role in demonstrating the scheme independence of physical quantities @xcite . let us recall the values for @xmath99 , @xmath106 , @xmath107 , and @xmath108 , @xmath109 @xmath110 \ , , \label{eq:2.16}\ ] ] where @xmath111 . here , @xmath26 is the number of colours and @xmath36 the number of active flavours . all four quantities in eqs . ( [ eq:2.15 ] ) and ( [ eq:2.16 ] ) are common to the ndr and hv scheme . moreover @xmath99 and @xmath106 are gauge independent . the coefficients @xmath107 and @xmath108 given here correspond to the feynman gauge . in the course of the discussion of the anomalous dimensions of the penguin operators @xmath47 and @xmath48 it will be useful to have at hand the anomalous dimension of the mass operator @xmath112 , @xmath113 to two loops , the anomalous dimensions @xmath114 and @xmath115 are given by @xmath116 \ , , \label{eq:2.19}\ ] ] for both schemes @xcite . finally let us recall that for the anomalous dimension of the weak current , @xmath117 the one - loop coefficient @xmath118 vanishes in both schemes . however , at the two loop level in the hv scheme , @xmath119 is found to be non vanishing @xcite , @xmath120 in the hv scheme the axial current attains a finite renormalization if one subtracts minimally , as we did , and one gets a mixing between chiral components . this is not so aesthetic and on hindsight it would have been more elegant to redefine the currents non - minimally to retain standard normalization . we do not proceed along this line but in our computation the mixing reappears in the non - vanishing of the two loop coefficient @xmath119 . however , since the one - loop matrix elements @xmath121 also change correspondingly , the final physical quantities are normalized properly and scheme independent in the end . the contributions of diagrams @xmath59 @xmath60 of fig . 2 to the matrix @xmath18 have a very simple structure . the mixing between different operators can be divided into five blocks , each containing two operators , @xmath122 with no mixing between different blocks . moreover , the mixing among @xmath53 operators @xmath123 is described by a universal @xmath124 matrix . ( [ eq:2.10 ] ) , one finds @xmath125_{\rm cc } } \ ; = \ ; \left ( \begin{array}{cc } -\,6/n & 6 \\ 6 & -\,6/n \end{array } \right ) \ , , \label{eq:3.2}\ ] ] with identical results for @xmath126 and @xmath127 . similarly , @xmath128_{\rm cc } } \ ; = \ ; \left ( \begin{array}{cc } 6/n & -\,6 \\ 0 & -\,6 \,(n^{2}-1)/n \end{array } \right ) \ , , \label{eq:3.3}\ ] ] with identical result for @xmath129 . comparing with eq . ( [ eq:2.19 ] ) , we note that @xmath130_{\rm cc } } \ ; = \ ; { \left [ \gamma_{88}^{(0 ) } \right]_{\rm cc } } \ ; = \ ; 2 \,\gamma_{m}^{(0 ) } \ , , \label{eq:3.4}\ ] ] showing that the one loop anomalous dimensions of the operators @xmath47 and @xmath48 are twice the one loop anomalous dimension of the mass operator , provided only current current diagrams @xmath59 @xmath60 are taken into account . it should be pointed out that this result is valid for an arbitrary number of colours @xmath26 , but remains only true in the large-@xmath26 limit when penguin insertions are taken into account @xcite . the contributions of the diagrams @xmath61 and @xmath62 of fig . 3 to the matrix @xmath18 have also a very simple structure : * the insertions of @xmath30 , @xmath131 , and @xmath132 into the diagrams @xmath61 and @xmath62 vanish . this follows either from colour conservation , or flavour conservation , or finally from the dirac structure , as one can easily verify by inspecting the diagrams of fig . thus , the corresponding rows in @xmath133 vanish , @xmath134 * the insertion of any of the remaining operators into the diagrams of fig . 3 results always into a unique linear combination of the qcd penguin operators @xmath135 @xmath47 , multiplied by an overall factor characteristic for the inserted operator . denoting by @xmath136 the row vector in the space ( @xmath30 @xmath55 ) , the non - vanishing elements of @xmath18 , coming from the diagrams of fig . 3 , are as follows , @xmath137 where @xmath138 and @xmath139 denote the number of up and down quark flavours respectively ( @xmath140 ) . in obtaining ( [ eq:3.7 ] ) , eq . ( [ eq:2.10 ] ) has been used with @xmath107 omitted , since this contribution has already been included in ( [ eq:3.2 ] ) and ( [ eq:3.3 ] ) . combining the results of eqs . ( [ eq:3.2 ] ) , ( [ eq:3.3 ] ) , ( [ eq:3.5 ] ) , and ( [ eq:3.7 ] ) , we find the known complete one loop anomalous dimension matrix @xmath18 . for completeness this matrix has been given explicitly in appendix a. it should be stressed , that these results do not depend on the renormalization scheme for @xmath58 and are also valid for @xmath141 . this is no longer the case for the finite terms in the diagrams of figs . 2 and 3 , to which we now turn our attention . the contribution to the matrix @xmath143 defined in eq . ( [ eq:2.14 ] ) , resulting from the diagrams @xmath59 @xmath60 of fig . 2 . * depends on the renormalization scheme for @xmath58 , but * does not depend on whether @xmath58 , or their fierz conjugates @xmath141 , are inserted in these diagrams . the general structure of @xmath144_{\rm cc}}$ ] is the same as in the case of @xmath145_{\rm cc}}$ ] presented in section 3.1 . since @xmath121 is generally gauge dependent and dependent on the infrared structure of the theory we give here only results for @xmath146 which are free from these dependences . the results for @xmath143 in the landau gauge have been given in ref . @xcite . we find @xmath147_{\rm cc } } \ ; = \ ; \left ( \begin{array}{cc } 2 n - 4/n & 2 \\ 2 & 2 n - 4/n \end{array } \right ) \ , , \label{eq:3.8}\ ] ] with identical results for @xmath148 and @xmath149 . similarly , @xmath150_{\rm cc } } \ ; = \ ; \left ( \begin{array}{cc } 2 n - 8/n & 6 \\ 4 & 4 n - 8/n \end{array } \right ) \ , , \label{eq:3.9}\ ] ] with identical result for @xmath151 . turning our attention to the finite parts of the penguin diagrams of fig . 3 , it should be first noted that at the one loop level one does not face the problem of the evaluation of @xmath152 $ ] in @xmath28 dimensions . consequently , the diagram @xmath61 can be calculated in the ndr renormalization scheme without any difficulties . nevertheless , different results are obtained for the ndr and hv scheme . the reason that no problem with closed fermion loops arises at this level is as follows . although at the intermediate state @xmath153 appears it is contracted with @xmath154 and @xmath155 where @xmath156 is the momentum of the gluon . thus actually only the trace @xmath157 has to be calculated , which however is zero . the structure of the results is similar to the ones given in section 3.2 . in addition to the above mentioned scheme dependence , in the case of the ndr , but not the hv scheme , the results for the insertions of the operators @xmath24 , @xmath135 , @xmath158 , @xmath159 , and @xmath55 depend on whether these operators are taken in the basic form of eq . ( [ eq:2.1 ] ) or in the fierz conjugate form @xmath141 . in the case of pure qcd corrections the results for @xmath30 , and for the @xmath160 operators @xmath131 @xmath48 , do not depend on the form used . these properties will play an important role in section 5 . beginning with the basis of eq . ( [ eq:2.1 ] ) , we find that the insertions of @xmath30 , @xmath131 , and @xmath132 into the diagrams @xmath61 and @xmath62 vanish in both renormalization schemes considered . thus the corresponding rows in @xmath161 and @xmath162 vanish , @xmath163 with an identical result for hv scheme . the remaining insertions are non - zero but @xmath164 turns out to vanish for @xmath165 , @xmath166 , @xmath167 and @xmath55 : @xmath168 for the remaining operators we find then @xmath169 where the vector @xmath170 has been defined in ( [ eq:3.6 ] ) . since , as stated above , @xmath171 and @xmath172 insertions give different results in the ndr scheme , we find that the @xmath173 calculated in the bases ( [ eq:2.1 ] ) and ( [ eq:2.2 ] ) differ . one obtains @xmath174 - { \hat{r}_{\rm s , ndr}}[\ref{eq:2.1 } ] \ ; = \ ; \left ( \begin{array}{c } 0 \\ -\frac{1}{3 } p \\ 0 \\ \vdots \\ 0 \end{array } \right ) \ , . \label{eq:3.16}\ ] ] this result will play an important role in section 5 . the generalization of the one loop matrix of eq . ( [ eq:3.2 ] ) to two loop level can be done using the results of ref . @xcite , where the eigenvalues of the @xmath22 submatrix @xmath175 have been calculated in the ndr and hv schemes . since also at two loop level the corresponding submatrix is symmetric with equal entries on the diagonal , one has @xmath176_{\rm cc } } = & { \left [ \gamma_{22}^{(1 ) } \right]_{\rm cc } } = & \frac{1}{2 } \big(\ , { \tilde{\gamma}}{+}^{(1 ) } + { \tilde{\gamma}}{-}^{(1 ) } \,\big ) + 2 \,\gamma_{j}^{(1 ) } \ , , \label{eq:4.1 } \\ { \left [ \gamma_{12}^{(1 ) } \right]_{\rm cc } } = & { \left [ \gamma_{21}^{(1 ) } \right]_{\rm cc } } = & \frac{1}{2 } \big(\ , { \tilde{\gamma}}{+}^{(1 ) } - { \tilde{\gamma}}{-}^{(1 ) } \,\big ) \ , , \label{eq:4.2}\end{aligned}\ ] ] with @xmath177 given by formulae ( [ eq:5.2 ] ) and ( [ eq:5.3 ] ) of ref . @xcite for ndr and hv schemes respectively , and @xmath119 in ( [ eq:2.22 ] ) of the present paper . the details of the calculation of the diagrams @xmath63 @xmath64 of fig . 4 , with @xmath178 insertions , can be found in ref . @xcite , where a table of @xmath179 and @xmath57 singularities in the diagrams @xmath63 @xmath64 has been given . in performing these calculations one has to take care of evanescent operators which , although vanishing in @xmath180 , affect the two loop anomalous dimension matrix of physical operators @xmath58 of eq . ( [ eq:2.1 ] ) . for this reason , the treatment of counter diagrams has to be done with care . the discussion of evanescent operators can be found in ref . @xcite , and a nice presentation is given also in @xcite . here it suffices to give only the projections on the space of physical operators . denoting the evanescent operators generally by @xmath181 , with @xmath39 , @xmath40 , @xmath182 , @xmath183 being dirac indices , the relevant projection for diagrams involving @xmath178 operators used in ref . @xcite is given by @xmath184 where in the hv scheme @xmath185 has to be taken in @xmath34 . using eqs . ( [ eq:4.1 ] ) and ( [ eq:4.2 ] ) , together with the results of ref . @xcite , we find in the case of the ndr scheme @xmath186_{\rm cc } } \ ; = \ ; \left(\begin{array}{cc } a_{\rm ndr } & b_{\rm ndr } \\ b_{\rm ndr } & a_{\rm ndr } \end{array}\right ) \ , , \label{eq:4.4}\ ] ] where @xmath187 with identical results for @xmath148 and @xmath149 . for the hv scheme we find @xmath188_{\rm cc } } \ ; = \ ; \left(\begin{array}{cc } a_{\rm hv } & b_{\rm hv } \\ b_{\rm hv } & a_{\rm hv } \end{array}\right ) \ , , \label{eq:4.7}\ ] ] where @xmath189 with identical results for @xmath148 and @xmath149 . the calculation of the @xmath190 insertions into the diagrams @xmath63 @xmath64 proceeds in analogy with the one for @xmath53 operators . however , the projection on the space of physical operators takes this time the following form , @xmath191 in order to check fierz symmetry properties , it is instructive to calculate simultaneously the insertions of the @xmath192 operators , e.g. @xmath193 of eq . ( [ eq:2.3 ] ) . in this case , the projection on the space of physical operators takes the form @xmath194 the singular terms in the diagrams @xmath63 @xmath64 with @xmath131 and @xmath193 insertions in the ndr and hv scheme are collected in table [ tab:1 ] . the two loop counter terms have been included diagram by diagram . also corrections to counter diagrams resulting from the mixing with evanescent operators have been taken into account . colour factors have been omitted . they can be found in ref . the overall normalization is such that after the inclusion of colour factors separate contributions to @xmath195 of eq . ( [ eq:2.11 ] ) are obtained . in the case of diagrams 29 31 , we have included in table [ tab:1 ] the colour factors present in the gluon self - energy @xmath196 we make the following observations : * the @xmath179 singularities are the same in both schemes as expected , * the @xmath57 singularities differ , which implies different two loop anomalous dimension matrices , * the fierz symmetry in diagrams @xmath63 @xmath64 is preserved in both schemes . the fierz conjugate pairs of diagrams are ( 4,6 ) , ( 7,9 ) , ( 10,12 ) , ( 13,15 ) , ( 16,21 ) , ( 17,20 ) , ( 18,19 ) , ( 22,24 ) , ( 25,27 ) , and ( 29,31 ) . the remaining diagrams are self conjugate . in ref . @xcite , where the operators @xmath197 and @xmath24 have been considered the fierz conjugate diagrams with @xmath197 or @xmath24 insertions had the same singular structure . now , the fierz symmetry relates the @xmath131 insertion of a given diagram to the @xmath193 insertion of the fierz conjugate diagram . as seen in table [ tab:1 ] , this relation is satisfied for all fierz conjugate diagrams : @xmath131 and @xmath193 insertions give the same result . we should , however , emphasize that as in ref . @xcite , also here this structure is only obtained after the evanescent operators have been properly taken into account . otherwise , the fierz symmetry in pairs ( 16,21 ) , ( 17,20 ) , ( 18,19 ) is broken and the results for diagrams 5 and 23 are modified . clearly , the fact that the results in table [ tab:1 ] satisfy all these relations , is a good check of our calculations . including colour factors , summing all diagrams @xmath63 @xmath64 , and using the formula ( [ eq:2.11 ] ) , gives the current current contribution to the two loop anomalous dimension matrix of @xmath190 operators . for the ndr scheme we find @xmath198_{\rm cc } } \ ; = \ ; \left(\begin{array}{cc } c_{\rm ndr } & d_{\rm ndr } \\ e_{\rm ndr } & f_{\rm ndr } \end{array}\right ) \ , , \label{eq:4.13}\ ] ] where @xmath199 with identical results for @xmath151 . here , @xmath115 is the two loop anomalous dimension of the mass operator given in eq . ( [ eq:2.19 ] ) . for the hv scheme we find @xmath200_{\rm cc } } \ ; = \ ; \left(\begin{array}{cc } c_{\rm hv } & d_{\rm hv } \\ e_{\rm hv } & f_{\rm hv } \end{array}\right ) \ , , \label{eq:4.18}\ ] ] where @xmath201 with identical results for @xmath151 . the matrices given in eqs . ( [ eq:4.4 ] ) , ( [ eq:4.7 ] ) , ( [ eq:4.13 ] ) , and ( [ eq:4.18 ] ) give the part of the two loop anomalous dimension matrix coming from the current current diagrams of fig . 4 . let us denote this part by @xmath202_{\rm cc}}$ ] and @xmath203_{\rm cc}}$ ] for ndr and hv schemes respectively . we note , that * these two matrices differ considerably from each other . in particular all diagonal entries in the hv scheme increase as @xmath204 , whereas in the ndr scheme this is only the case for the elements ( 6,6 ) and ( 8,8 ) . * since @xmath205 and @xmath206 grow like @xmath26 for large @xmath26 and @xmath207 and @xmath208 grow like @xmath209 in both schemes considered , we find remembering @xmath210 , that these two elements approach non - vanishing constants in the large-@xmath26 limit . in the ndr scheme all other elements vanish in this limit and this fact will not be changed by including penguin diagrams . in the hv scheme all diagonal elements survive the large-@xmath26 limit . * the simple relation between @xmath205 , @xmath206 and @xmath114 , given in ( [ eq:3.4 ] ) , is violated at two loop level in both schemes . however , in the ndr scheme it is recovered for large @xmath26 @xcite . * the vanishing elements ( 5,6 ) and ( 7,8 ) in eq . ( [ eq:3.3 ] ) become non zero at the two loop level . * using the results of this and the previous section , it is interesting to note that the relation ( [ eq:2.12 ] ) is satisfied even if only current - current contributions to @xmath18 , @xmath143 and @xmath20 are taken into account : @xmath211_{\rm cc } } \ ; = \ ; { \left[{\hat{\gamma}_{\rm s , ndr}^{(1)}}\right]_{\rm cc } } + \left [ \delta{\left[{\hat{r}_{\rm s}}\right]_{\rm cc } } , { \left[{\hat{\gamma}_{\rm s}^{(0)}}\right]_{\rm cc } } \right ] + 2 \,\beta_{0 } \delta{\left[{\hat{r}_{\rm s}}\right]_{\rm cc } } \ , , \label{eq:4.23}\ ] ] where @xmath212_{\rm cc } } = { \left[{\hat{r}_{\rm s , hv}}\right]_{\rm cc } } - { \left[{\hat{r}_{\rm s , ndr}}\right]_{\rm cc } } \label{eq:4.24}\ ] ] this is easy to understand . indeed , the two loop current current diagrams of fig . 4 have subdiagrams of current current type only , and consequently the compatibility of the two calculations should be verified within current current diagrams alone . the fact that this is indeed the case is a good check of our results . the calculation of the penguin diagram contributions to @xmath20 can be considerably simplified by first analyzing the general structure of these contributions . the two loop penguin diagrams are shown in fig . 5 , where two types of insertions of a given operator , b ) and c ) of fig . 1 , have to be considered . yet , as we shall show now , the full matrix @xmath213 can be obtained by calculating just the insertions of the operators @xmath30 , @xmath24 , @xmath50 , and @xmath193 , which moreover receive only contributions from the diagrams @xmath67 @xmath68 , i.e. penguin diagrams without closed fermion loops . we first note that the anomalous dimensions of the @xmath214 penguin operators @xmath135 , @xmath158 , @xmath159 , and @xmath55 can be expressed in terms of the anomalous dimensions of @xmath30 , @xmath24 , @xmath197 , and @xmath35 as follows , @xmath215 where the last terms in ( [ eq:5.1 ] ) ( [ eq:5.4 ] ) are only due to internal down quarks . similarly , the anomalous dimensions of the @xmath46 operators @xmath132 and @xmath48 can be expressed in terms of those of @xmath131 and @xmath47 respectively , @xmath216 we next note that @xmath197 , @xmath35 , @xmath131 , and @xmath47 insertions are non zero only for @xmath65 @xmath66 , whereas @xmath30 , @xmath24 , @xmath50 , and @xmath193 insertions are non zero only for @xmath67 @xmath68 . in the hv scheme , there is no difficulty in evaluating the diagrams @xmath65 @xmath66 which involve closed fermion loops . consequently , in this scheme , eqs . ( [ eq:5.1 ] ) ( [ eq:5.6 ] ) can directly be used to find the matrix @xmath217 . however , in the ndr scheme it is much easier to deal with the diagrams @xmath67 @xmath68 . we will now show that relation ( [ eq:2.12 ] ) can be used to reduce the full calculation to the latter diagrams . let us first observe that eq . ( [ eq:2.12 ] ) can also be used to relate the two loop anomalous dimensions obtained in the bases ( [ eq:2.1 ] ) and ( [ eq:2.2 ] ) . since these two bases differ only in the two first operators , and the current current anomalous dimensions of @xmath175 and @xmath218 are the same , we find @xmath219 } \ ; = \ ; \brackp{{\hat{\gamma}_{\rm s}^{(1)}}[{(\ref{eq:2.1 } ) } ] } + \left [ \widetilde{\delta}{\hat{r}_{\rm s } } , { \hat{\gamma}_{\rm s}^{(0)}}\right ] + 2 \,\beta_{0 } \widetilde{\delta}{\hat{r}_{\rm s}}\ , , \label{eq:5.7}\ ] ] where @xmath220 with @xmath221 given in eq . ( [ eq:3.16 ] ) . ( [ eq:5.7 ] ) and ( [ eq:5.8 ] ) allow to find @xmath222 and @xmath223 , once @xmath224 and @xmath225 are known . in the ndr scheme we then have @xmath226 whereas in the hv scheme , the insertions of @xmath30 and @xmath24 are equal to the insertions of @xmath197 and @xmath35 , respectively . next , we recall the important result of section 3 : the finite pieces in one loop diagrams ( the matrix @xmath143 ) did not depend on whether @xmath131 and @xmath47 or @xmath50 and @xmath193 had been used . since @xmath227 and @xmath228 form closed sets under the renormalization due to current current diagrams , and the insertion of any of these four operators in one loop penguin diagrams give always linear combinations of @xmath135 @xmath47 it is evident that @xmath229 in both hv and ndr schemes . to summarize , we have shown that the full matrix @xmath213 can be obtained by calculating just the insertions of the operators @xmath30 , @xmath24 , @xmath50 , and @xmath193 into the diagrams @xmath67 @xmath68 . so far , we have restricted our discussion to `` on shell '' results . it is well known however , that to renormalize operators one has to include mixing with all operators having the same quantum numbers . in the literature @xcite , it has become standard to divide these into 3 classes i , ii@xmath230 , and ii@xmath231 , according to * class i : gauge invariant operators that do not vanish by virtue of the equations of motion , * class ii@xmath230 : gauge invariant operators that vanish by virtue of the equations of motion , and * class ii@xmath231 : gauge non invariant operators . in our case , the situation is even slightly more complicated , since one has to consider also mixing with evanescent operators , which vanish when restricted to 4 dimensions . the renormalized operators @xmath58 of eq . ( [ eq:2.1 ] ) , corresponding to the unrenormalized operators @xmath232 , which belong to class i , are given by @xmath233 here , @xmath234 are the evanescent operators , @xmath235 are in class ii@xmath231 , and @xmath236 are gauge invariant operators of dimension 6 that just involve two quark fields . there are @xmath237 independent operators of this sort ( up to total derivatives ) , having @xmath238 quantum numbers which arise in our computation up to two loops,\gamma^{\mu}(1-{\gamma_5 } ) \,d \,\right)$ ] does nt enter at two loops . ] @xmath239\big ] \gamma^{\nu}(1-{\gamma_5 } ) \,d \,\right ) \ , , { \nonumber}\\ g_{2 } & = & \left ( \bar s \big\{d^{2},d_{\mu}\big\ } \gamma^{\mu}(1-{\gamma_5 } ) \,d \,\right ) \ , , \label{eq:5.2.2 } \\ g_{3 } & = & \left ( \bar s d_{\mu } d_{\nu } d_{\lambda } s^{\mu\nu\lambda } ( 1-{\gamma_5 } ) \,d \,\right ) \ , , { \nonumber}\end{aligned}\ ] ] where @xmath240 is the covariant derivative and @xmath241 . the operator @xmath242 is proportional to the penguin operator @xmath243 which occurs at one loop , @xmath244 and @xmath245 all belong to class ii@xmath230 . in arbitrary gauges it is known that class i operators generally mix with class ii operators , in particular with those of class ii@xmath231 . one can simplify the work considerably by using the background field gauge @xcite , for which no such mixing with gauge non invariant operators occurs . in our calculation , we therefore used such a gauge , and thus have @xmath246 in eq . ( [ eq:5.2.1 ] ) . however this choice only affects the diagrams with triple gluon vertices , @xmath247 and @xmath248 of fig . in order to include mixing with the operators of class ii@xmath230 , ( [ eq:5.2.2 ] ) , we have to calculate the operator insertions into four quark green functions with arbitrary external momenta . this results in additional momentum dependent operators which reflect the presence of class ii@xmath230 operators . the various tensor products , leaving out the external quark fields , which appear in the calculation are as follows , @xmath249 where @xmath250 , @xmath251 are incoming ( outgoing ) quark momenta and @xmath252 . the operators @xmath58 of eq . ( [ eq:2.1 ] ) all correspond to the structure @xmath253 . the procedure to calculate the diagrams of fig . 5 followed that described in ref . that is , we first classified all independent integrals occuring in the diagrams and their singular parts , computed for arbitrary external momenta . then the contractions with the corresponding @xmath182matrix structures , using the ndr rules and the hv rules , were performed separately . for the ndr scheme this was done by `` hand '' and by using the algebraic computer program tracer @xcite . for the hv scheme we used tracer nearly exclusively for the penguin diagrams . the operator structures @xmath254 and @xmath255 already vanish going on - shell , i.e. using the equations of motion , while the on shell projection of @xmath256 and @xmath257 is given by @xmath258 as a check , the on - shell result was also calculated in the ordinary feynman gauge . since only diagrams @xmath247 and @xmath248 are affected , after using the equations of motion their sum should agree with the result in the background field gauge . up to evanescent contributions this is indeed the case . a second technicality of the calculation that we like to discuss in somewhat more detail is the treatment of @xmath5 . there are basically two diagrams , namely diagrams 3 and 4 of fig . [ fig:5 ] , where one encounters problems with the traces involving @xmath5 . there we meet the ambigous trace expression @xmath259\ , , \label{eq : usig}\ ] ] where @xmath260 is defined as before . all other ambigous expressions can be reduced to this one together with other non ambigous terms . besides the treatment of this problem in ndr as described in the previous section , we can attempt an ansatz for @xmath261 to replace it by @xmath262 with @xmath263 representing the ambiguity which is of @xmath264 . although @xmath263 does appear at intermediate steps of the calculation of diagrams 3 and 4 , after inclusion of evanescent contributions , it cancells out in the final expression . the result thus obtained agrees with the result obtained through the procedure of sect . this gives us confidence that our treatment of ndr is correct . we would like to point out that the replacement of eq . ( [ eq : repl ] ) seems to be related to the procedure recently proposed by krner et.al . we emphasize that in both works the scheme presented is a prescription , which although shown to work upto two loops for certain processes , still needs to be proven consistent to all orders . the singular terms in the diagrams @xmath67 @xmath68 with @xmath24 and @xmath193 insertions in the ndr and hv scheme , are collected in tables [ tab:2][tab:4 ] . loop counter terms have been included diagram by diagram . colour factors have been omitted . the overall normalization is such that after the inclusion of colour factors separate contributions to @xmath195 of eq . ( [ eq:2.11 ] ) are obtained . the singularites in tables [ tab:2][tab:4 ] include also contributions from penguin diagrams obtained from fig . [ fig:5 ] by left - right reflections . it should be stressed that these additional diagrams give generally different singularities than the diagrams given explicitly in fig . [ fig:5 ] . as usual , the result of a given penguin diagram contains two dirac structures , @xmath265 in diagrams @xmath67 @xmath266 the singular factors multiplying @xmath267 and @xmath268 are the same , but they differ in the case of @xmath269 @xmath68 . we observe , that diagrams @xmath67 , @xmath270 , and @xmath266 do not contribute to the anomalous dimensions . the singular terms for @xmath24 and @xmath193 given in tables [ tab:2][tab:4 ] are also valid for @xmath30 and @xmath50 respectively . however , since @xmath30 and @xmath50 are in the colour non singlet forms , whereas @xmath24 and @xmath193 are in the singlet form , the colour factors differ . in fact , it turns out that in the case of the insertion of @xmath30 and @xmath50 only the colour factors in diagrams 3 , 4 , 11 , and 12 are non vanishing , and only these diagrams contribute to @xmath271 and @xmath272 . including colour factors , using the results of tables [ tab:2][tab:4 ] and eq . ( [ eq:2.11 ] ) without @xmath108 ( taken already into account in the current current contributions ) , we obtain the expressions for @xmath224 , @xmath225 , @xmath273 , and @xmath274 , from which the full matrix @xmath213 can be found by means of formulae of sect . 5.1 . in the ndr scheme we have @xmath275 where the vectors are in the @xmath276 space . in the hv scheme we have @xmath277 using next eqs . ( [ eq:5.1 ] ) ( [ eq:5.6 ] ) in conjunction with ( [ eq:5.9 ] ) ( [ eq:5.12 ] ) , we find the full matrices @xmath278 and @xmath217 . we observe that for large @xmath26 the contributions from penguin diagrams to @xmath20 grow at most as @xmath26 . since @xmath279 the full contribution of penguin diagrams to the third term in eq . ( [ eq:1.1 ] ) vanishes in the large-@xmath26 limit . yet for @xmath37 the penguin diagrams play a considerable role in the numerical values for @xmath20 . adding the current - current and penguin contributions to @xmath20 found in sections 4 and 5 respectively , we obtain the complete two loop matrices @xmath280 and @xmath281 . we first note that 48 elements of these matrices vanish in both schemes . these are given by @xmath282_{\rm ij } } \ ; = \ ; { \left [ { \hat{\gamma}_{\rm s , hv } ^{(1)}}\right]_{\rm ij } } \ ; = \ ; 0 \label{eq:6.1}\ ] ] where @xmath283 the remaining 52 elements of each matrix are non vanishing . these entries in the anomalous dimension matrix are given in table [ tab:5 ] for arbitrary number of colours and flavours . for phenomenological applications we need only the results with @xmath37 . we give them in appendix c for an arbitrary number of flavours . it is evident from this table that the two loop @xmath284 corrections to the anomalous dimensions of the operators @xmath285 are substantial although they are quite different for the two schemes considered . using the results of table 5 and the one - loop results for @xmath18 , @xmath142 of section 3 we can test whether the compatibility relation ( [ eq:2.12 ] ) is satisfied . this turns out to be indeed the case which constitutes a test of our calculation . on general grounds , renormalizability relates the singularities of degree @xmath286 in the @xmath287-th order of perturbation theory with the singularities of degree @xmath288 in the @xmath289-th order . this fact allows for an additional check of our computation by considering the corresponding relation between the coefficients of the @xmath179 singularities at two loops and the @xmath57 singularities at one loop . demanding the finiteness of the renormalized green function @xmath90 , and the anomalous dimension matrix @xmath290 , a procedure analogous to the derivation of eqs . ( [ eq:2.10 ] ) and ( [ eq:2.11 ] ) leads to @xmath291\,\hat{1}\ , , \label{eq : test}\ ] ] where @xmath292 is the coefficient of the @xmath179 singularity at @xmath1 of the bare greens function @xmath81 , and @xmath293 is the leading order coefficient of the renormalization group function associated with the gauge parameter . in the feynman gauge it is given by @xmath294 the values for @xmath99 and @xmath107 were already given in eqs . ( [ eq:2.15 ] ) and ( [ eq:2.16 ] ) respectively . however , to perform the test we have to add the contributions of the diagrams of fig.[fig:6 ] to the result which can be obtained by means of tables [ tab:1][tab:4 ] . despite the fact that they have no @xmath57 divergence and hence do not contribute to the anomalous dimension matrix , they contribute to @xmath292 . their contribution to @xmath292 is given by @xmath295 where the term proportional to @xmath99 stems from the renormalization of the gluon propagator and the term proportional to @xmath296 from the vertex renormalization . let us recall that we work in the background field gauge , where the @xmath40function can be obtained directly from the renormalization of the gluon propagator @xcite . putting everything together , we find that the relation of eq . ( [ eq : test ] ) is indeed satisfied . in addition , similar to the test of the compatibility of different schemes , this relation could also be split in a part solely originating from current current diagrams which is satisfied separately , and a relation involving mixing between current current and penguin diagrams . we will however not elaborate on this any further . we have presented the details and the explicit results of the calculation of the two loop anomalous dimension matrix @xmath284 involving current - current , qcd - penguin and electroweak - penguin operators . performing the calculation in two schemes for @xmath5 , ndr and hv , we have verified the compatibility of the anomalous dimension matrices obtained in these two schemes . we have shown how the use of two different operator bases allows to avoid a direct calculation of penguin diagrams with closed fermion loops . this enabled us to find in an unambiguous way the two loop anomalous dimensions in the simplest dimensional regularization scheme , the one with anticommuting @xmath5 . the results of this paper generalize our earlier paper @xcite where electroweak penguin operators have not been taken into account . the two loop anomalous dimensions matrix presented for the first time in ref . @xcite and generalized in section 6 of the present paper will play a central role in any analysis of non - leptonic decays of hadrons which goes beyond the leading logarithmic approximation . a consistent next - to - leading order analysis involving electroweak penguin operators requires the calculation of @xmath21 , i.e. the two loop anomalous dimension matrix @xmath297 . a detailed account of this calculation is given in @xcite . the full renormalization group analysis with the anomalous dimension matrix of eq . ( [ eq:1.1 ] ) will be presented soon in ref . there the numerical values for the wilson coefficient functions @xmath298 of the operators @xmath285 including next - to - leading order corrections can be found . * acknowledgement * we would like to thank many colleagues for a continuous encouragement during this calculations . one of us ( a.j.b . ) would like to thank bill bardeen and david broadhurst for very interesting discussions . m.e.l . is grateful to stefan herrlich for a copy of his program feynd for drawing the feynman diagrams in the figures and to gerhard buchalla for stimulating discussions . 10 and b. w. lee , ( 1974 ) 108 . and l. maiani , ( 1974 ) 351 . , v. i. zakharov , and m. a. shifman , ( 1977 ) 670 . and m. b. wise , ( 1979 ) 2392 . and r. d. peccei , ( 1980 ) 289 . and m. b. wise , ( 1984 ) 245 . and j .- grard , ( 1987 ) 156 . , ( 1987 ) 551 . , ( 1989 ) 33 . and l. randall , ( 1989 ) 221 , erratum * 235b * ( 1990 ) 412 . , a. j. buras , and m. k. harlander , ( 1990 ) 313 . , g. curci , g. martinelli , and s. petrarca , ( 1981 ) 461 . and p. h. weisz , ( 1990 ) 66 . , m. jamin , m. e. lautenbacher , and p. h. weisz , ( 1992 ) 69 ; addendum ibid . _ nucl . _ * b375 * ( 1992 ) 501 . , m. jamin , and m. e. lautenbacher , two loop anomalous dimension matrix for @xmath2 weak non - leptonic decays ii : @xmath299 , . , m. jamin , and m. e. lautenbacher , effective hamiltonians for @xmath2 and @xmath3 non - leptonic decays beyond leading logarithms in the presence for electroweak penguins , . , , addison - wesley publishing company inc . , 1991 . and m. e. lautenbacher , . and m. veltman , ( 1972 ) 189 . and d. maison , ( 1977 ) 11 , 39 , 55 . , ( 1981 ) 384 . , a. j. buras , and j .- grard , ( 1986 ) 133 . , a. j. buras , and j .- grard , ( 1987 ) 787 . and b. grinstein , ( 1991 ) 239 . and e. de rafael , ( 1991 ) 311 . and zuber , ( 1975 ) 467 , 482 , 3159 . and b. w. lee , ( 1976 ) 160 . and j. a. dixon , ( 1978 ) 1113 ( 1983 ) 349 . , ( 1981 ) 189 . , d. kreimer , and k. schilcher , ( 1992 ) 503 . -1 cm @xmath300 tables [ tab:1][tab:4 ] list the @xmath301 and @xmath57singularities of the various feynman diagrams obtained in the ndr and hv scheme . the values given already include diagram by diagram two loop counterterms . colour factors are omitted and a common overall factor @xmath302 is to be understood . in table [ tab:1 ] the column labeled @xmath303 refers to the numbering of current current diagrams in figs . [ fig:2 ] and [ fig:4 ] . similarly , the @xmath170-column in tables [ tab:2][tab:4 ] refers to penguin diagrams of figs . [ fig:5 ] . in addition , in table [ tab:1 ] the column labeled @xmath304 gives the multiplicity of each diagram , which already has been included in the singularities quoted . the singularites in tables [ tab:2][tab:4 ] include also contributions from diagrams obtained from fig . [ fig:5 ] by left - right reflections . . [ tab:1 ] [ cols="^,^,^,^,^,^,^,^ " , ] @xmath305 @xmath306 @xmath307 @xmath308 = 1.7 in figure 1 : : : the three basic ways of inserting a given operator into a four point function : ( a ) current current- , ( b ) type 1 penguin- , ( c ) type 2 penguin - insertion . the curled lines denote gluons . the 4-vertices `` @xmath309 '' denote standard operator insertions . figure 2 : : : one loop current current diagrams contributing to @xmath18 . the meaning of lines and vertices is the same as in fig . 1 . possible left - light or up - down reflected diagrams are not shown . figure 3 : : : one loop type 1 and 2 penguin diagrams contributing to @xmath18 . the meaning of lines and vertices is the same as in fig . 1 . figure 4 : : : two loop current current diagrams contributing to @xmath20 . the meaning of lines and vertices is the same as in fig . 1 . in addition shaded blobs stand for self - energy insertions . possible left - light or up - down reflected diagrams are not shown . figure 5 : : : two loop penguin diagrams contributing to @xmath20 . the curled lines denote gluons . square - vertices stand for type 1 and 2 penguin insertions as of figs . [ fig:1](b ) and ( c ) , respectively . possible left - light reflected diagrams are not shown . figure 6 : : : two loop penguin diagrams having no @xmath57 divergence and hence do not contribute to the anomalous dimension matrix @xmath20 . the meaning of lines and vertices is the same as in fig . 5 . in addition shaded blobs stand for self - energy insertions or vertex corrections . figure 7 : : : two loop penguin diagrams vanishing identically in dimensional regularization . the meaning of lines and vertices is the same as in fig . 5 . in addition shaded blobs stand for self - energy insertions or vertex corrections .

this is a part of a multiwavelength study aimed at use of complementary photometric , polarimetric and spectroscopic data to achieve an understanding of the activity process in late - type stars . here we present the study of fr cnc , a young , active and spotted star . we performed analysis of ( the all sky automated survey ) data for the years 20022008 and amended the value of the rotational period to be 0.826518 d. the amplitude of photometric variations decreased abruptly in the year 2005 , while the mean brightness remained the same , which was interpreted as a quick redistribution of spots . and broad band photometric calibration was performed for 166 stars in fr cnc vicinity . the photometry at terskol observatory shows two brightening episodes , one of which occurred at the same phase as the flare of 2006 november 23 . polarimetric observations indicate the probable presence of a supplementary source of polarization . we monitored fr cnc spectroscopically during the years 20042008 . we concluded that the rv changes can not be explained by the binary nature of fr cnc . we determined the spectral type of fr cnc as k7v . calculated galactic space - velocity components ( ) indicate that fr cnc belongs to the young disc population and might also belong to the ic 2391 moving group . based on li i .8 measurement , we estimated the age of fr cnc to be between 10120 myr . doppler tomography was applied to create a starspot image of fr cnc . we optimized the goodness of fit to the deconvolved profiles for axial inclination , equivalent width and sin , finding sin = km s and . we also generated a synthetic-band lightcurve based on doppler imaging that makes simultaneous use of spectroscopic and photometric data . this synthetic lightcurve displays the same morphology and amplitude as the observed one . the starspot distribution of fr cnc is also of interest since it is one of the latest spectral types to have been imaged . no polar spot was detected on fr cnc . [ firstpage ] stars : activity stars : flare stars : rotation stars : individual : fr cnc .

1111.6595

it is well - known that late - type stars show magnetic activity similar to the activity of our sun , but the physics of ` stellar ' activity is not yet well understood . in addition , the activity level manifested by late - type stars is much higher than that observed for the sun . fr cnc (= bd+16 1753 = mcc 527 = 1es 0829 + 15.9 = 1rxs j083230.9 + 154940 = hip 41889 ) was first mentioned as a probable active star when it was identified as the optical counterpart of a soft x - ray source 1es 0829 + 15.9 in the _ einstein slew survey_. it has @xmath12 mag , spectral type k8v , the x - ray flux of @xmath13 ( @xcite @xcite ; @xcite @xcite ) . lately , this object was rediscovered as an x - ray source 1rxs j083230.9 + 154940 in the _ rosat _ all - sky survey ( rass ) with lower x - ray flux at the level of @xmath14 ( @xcite @xcite ) . the x - ray luminosity of @xmath15 and the ratio of x - ray to bolometric luminosity @xmath16 of @xmath17 ( @xcite @xcite ) indicates that this object has an active corona ( @xcite @xcite ) . in the _ hipparcos _ catalogue this star was mentioned as an unsolved variable star with the identifier hip 41889 and @xmath18 mag amplitude of variability ( @xcite @xcite ) . it was classified as by dra type star ( i.e. its variability caused by rotational modulation of starspots ) and given the name fr cnc by @xcite . for analysis of _ hipparcos _ observations see @xcite . fr cnc ( @xmath19 and @xmath20 ) has @xmath21 marcsec parallax ( @xcite @xcite ) that implies a distance of @xmath22 pc and an absolute magnitude of @xmath23 . the kinematics of fr cnc suggests that it is a very young ( 3555 myr ) main - sequence star and a possible member of the ic 2391 supercluster , as it was shown by @xcite . @xcite concluded that fr cnc is not a binary system , based on two measurements of the rv . the presence of ca ii h & k and h@xmath24 emission lines in the spectra indicates high chromospheric activity in fr cnc ( @xcite @xcite ; @xcite @xcite ) . in quiescent state this object manifests optical variability with the dominant period @xmath25 d due to the presence of starspots and axial rotation ( @xcite @xcite ) . in addition , photometry obtained in 2005 february april with kilodegree extremely little telescope indicates fr cnc optical variability with @xmath26 d period ( @xcite @xcite ) when monitoring the praesepe open cluster for transiting exoplanets . the first ever - detected optical flare of this object was observed during ccd photometry of fr cnc on 2006 november 23 at crimean astrophysical observatory ( ukraine ) with 38-cm cassegrain telescope and described by @xcite . the flare was observed in @xmath27-bands ( see fig . [ ibvs ] ) . the amplitude reached even 1 mag in the @xmath28-band and was decreasing towards the @xmath29-band . the flare energy output in the @xmath28-band was about @xmath30 and flare to quiescent flux ratio was 38.63 per cent . most of the information on the photospheric activity ( e.g. starspots ) of by dra - type stars comes from photometric observations . the mean brightness level is strongly dependent on the percentage of spotted area of the surface , while changes in spot distribution over the surface could result in changes of the amplitude of variability . fr cnc has a short ( for such class of objects ) rotational period of 0.8267 d. as was shown in @xcite , this short rotational period leads us to expect large flare activity of the star . the detection of a flare on fr cnc on 2006 november 23 motivated us to continue photometric monitoring of this object as well as to study its archival _ asas-3 _ observations ( the all sky automated survey ; see @xcite for description of equipment and data pipeline ) . and @xmath31bands as well as the difference _ check star@xmath32comparison star _ ( ` ch ' on the plot)'',width=302 ] fr cnc was observed in @xmath11-band with _ asas-3 _ survey during 2002 december 2008 may ( 6 observational seasons , see fig . [ asaslt ] ) . all the data were split into separate data sets according to the seasonal gaps in observations and folded with the @xmath33 d period ( from @xcite @xcite ) and represented in fig . [ asas_phase ] ( plotted twice for clarity ) . table [ table : asaslog ] represents the log of observations . the initial epoch was common to calculate phases for all 6 phase diagrams and was chosen arbitrary as hjd ( utc ) = 2452635.72669 ( first point in dataset ) . no evidence of flares in the _ asas-3 _ data was found . the vertical dashed line on fig . [ asas_phase ] indicates the phase when the flare on 2006 november 23 occurred . _ asas-3 _ data are not covering flare on 2006 november 23 . nearest _ asas-3 _ observations were done on jd 2453881 and jd 24544091 , what is 181 days before and 29 days after the flare . _ asas-3 _ observations in 2006 - 2007 at a phase @xmath340.88 ( 143 days after the flare ) show that fr cnc was brighter than during the rest of the time . this probably could be related to brightening episode , which was detected at the same phase during terskol observations ( see sect . [ tersk ] ) . profiles of variability as well as the amplitude of variability are different from season to season , while the mean brightness remains constant ( @xmath35 mag within the error limits of @xmath36 mag ) . to illustrate this , we plotted the amplitude and the mean brightness as a function of mean epoch of observations ( fig . [ asas_amplitude ] ) . an abrupt decrease of amplitude in the year 2005 is clearly seen . one of the possible interpretations is continuous spottedness of the star and redistribution of spots / spot groups from season to season : i.e. spots did nt disappear , but distributed more uniformly over fr cnc surface . to support this idea , notice the constancy of the mean brightness level . if spots disappear then increase of brightness and decrease of amplitude could be expected . the data set was searched for periodic variation of brightness ( in order to estimate with better accuracy the known rotational period ) using the period04 package , developed by patrick lenz ( institute of astronomy , university of vienna , austria ; see @xcite @xcite ) . discrete fourier transform ( dft ) algorithm was applied for statistical analysis . julian dates are heliocentrically corrected . the average zero point of @xmath37 mag was subtracted to prevent the appearance of additional features on the periodogram centered at frequency 0.0 . making a periodicity analysis of _ photometry , we found a dominant frequency of @xmath38 @xmath39 , while , as defined by equation [ eq1 ] , @xmath40 ( therefore , period p = @xmath41 d ; see fig . [ periodogram ] ) . the detected periodicity could be interpreted as the rotational period and it is in good agreement with the period founded by @xcite , but with improved accuracy ( due to the longer time span of _ observations ) . no other periodicity of fr cnc brightness modulations was found on the basis of _ asas-3 _ observations . the dft routine was applied separately to each season of observations as well . the obtained periodograms do not reveal any other significant periodicity , but only the same peak as for the periodogram for the whole time - string . the dominant frequency remains constant within the error limits of @xmath42 for all of 6 periodograms . empirical results from observational analysis ( @xcite @xcite ) and numerical simulations ( @xcite @xcite ) have shown that the ratio in amplitude in the periodogram between signal and noise should not be lower than 4.0 to give good confidence in the detected peak . we calculated the s / n - ratio from the periodogram for the determined dominant frequency , so s / n = 6.88 . the spectral window function is important to confirm that the obtained frequencies are real or an artefact of the window function . this was done by assigning 1 to brightness values with the same observation times and checking the resulting diagram . there was no evidence of significant power at the location of the peak ( fig . [ periodogram ] ) . the dominant frequency on the spectral window is @xmath43 with amplitude @xmath44 , which is due to daily gaps in observations . parameter uncertainties were calculated from an error matrix , which is a by - product of non - linear least - squares fitting procedure . other types of uncertainties are those which could be calculated from analytically derived formulae assuming an ideal case . based on some assumptions one can derive a formula for the uncertainties in frequency and signal amplitude at this frequency . see @xcite and @xcite , @xcite for the derivation based on a monoperiodic fit . the determined equation can be applied for each frequency separately : @xmath45 @xmath46 where @xmath47 is the number of time points , @xmath48 is the time length of the data set , @xmath49 denotes the residuals from the fit and _ a _ refers to the signal amplitude at the particular frequency . both analytical and least - squares error matrix calculations give similar results : @xmath50 and @xmath51 . so , on the periodogram ( fig . [ periodogram ] ) we plotted the @xmath52-level ( dashed line ) to show that the detected peak exceeds it significantly . another method for 1@xmath53-level of amplitude calculation was proposed for use by r.a . fisher ( see @xcite @xcite ; @xcite @xcite ) and often called as fisher randomization test . the idea is to take the original light curve and preserving the time column , shuffle the corresponding intensities around . that destroys any coherent signal in the light curve while keeping the time sampling intact . we are left with a shuffled light curve of pure white noise . the next step is to compute a dft of this light curve which will look really noisy . the standard deviation of the average amplitude of such a dft is close to the 1@xmath53 limit . we iterated this 25 times , therefore we use the average of 25 standard deviation values to determine 1@xmath53-level . this pipeline gave us the value of @xmath54 , which is slightly less than @xmath55 from analytical and least - squares error matrix calculations , hence we plotted the bigger value on our periodogram to be confident . .[table : asaslog ] log of _ asas-3_-observations of fr cnc [ cols="^,^,^,^ " , ] clclllcccc number & date & telescope & instrument & detector & spect . range & orders & dispersion & fwhm@xmath56 & s / n + & ( dd / mm / yyyy ) & & & & ( ) & & ( ) & ( / pixel ) & h@xmath24 + 1 & 29/0307/04/2004 & 2.2-m@xmath57 & foces@xmath58 & 2048x2048 24@xmath59 site@xmath60d & 345010700 & 112 & 0.040.13 & 0.080.035 & 40 + 2 & 1113/04/2004 & 2.0-m@xmath61 & hfosc@xmath62 & 2000x4000 site st-002 , grism gr14 & 32706160 & 1 & 3.57 & 7.231.47 & 200 + & & & & 2000x4000 site st-002 , grims gr8 & 58008350 & 1 & 3.23 & 3.663.99 & 200 + 3 & 1621/12/2006 & 2.2-m@xmath57 & foces@xmath58 & 2048x2048 24@xmath59 site@xmath60d & 360010700 & 106 & 0.080.1 & 0.080.04 & 40 + 4 & 2426/02/2007 & 2.2-m@xmath57 & foces@xmath58 & 2048x2048 24@xmath59 site@xmath60d & 360010700 & 106 & 0.040.13 & 0.070.41 & 40 + & 0708/05/2007 & & & & & & & & + 5 & 21/03/2008 & not@xmath63 & fies@xmath64 & 2000x2000 eev42 - 40 & 36207360 & 80 & 0.020.04 & 0.050.11 & 80 + @xmath56 full width at half maximum of the arc comparison lines ; + @xmath57 2.2-m telescope at the german - spanish astronomical observatory ( caha , almera , spain ) ; + @xmath61 2.0-m himalayan chandra telescope at the indian astronomical observatory ( mt . saraswati , hanle , india ) ; + @xmath63 nordic optical telescope ( not ) at the observatorio del roque de los muchachos ( la palma , spain ) ; + @xmath58 the fibre optics cassegrain echelle spectrograph ( foces ) ; + @xmath62 himalaya faint object spectrograph and camera ( hfosc ) ; + @xmath64 the high - resolution fibre - fed echelle spectrograph ( fies ) . + a total of 58 high and low resolution spectra of fr cnc have been obtained and analysed in this work . the spectroscopic data were obtained during five observing runs . details of each observing run are given in table [ tab : obs ] : date , telescope , spectrograph , ccd chip , spectral range covered , number of orders included in each echelle spectrum , range of reciprocal dispersion , spectral resolution ( determined as the full width at half maximum , fwhm , of the arc comparison lines ) and mean s / n in the h@xmath24 line region . the spectra were extracted using the standard reduction procedures in the iraf echelle package ( bias subtraction , flat - field division and optimal extraction of the spectra ) . we obtained the wavelength calibration by taking spectra of a th - ar lamp . finally , we normalized the spectra by a polynomial fit to the observed continuum . stellar parameters of fr cnc are given in table [ tab : par ] and table [ tab : par2 ] . the photometric data ( @xmath65 ) , @xmath66 , projected rotational velocity ( @xmath67 ) , and galactic space - velocity components ( @xmath4 ) have been determined in this paper . the astrometric data ( parallax , @xmath68 ; proper motions , @xmath59@xmath69@xmath70 and @xmath59@xmath71 ) are from _ hipparcos _ and _ tycho-2 _ catalogues ( @xcite @xcite ) . cccccccccccccc & @xmath72 & @xmath73 & @xmath74 & _ v _ sin _ i _ & @xmath7 + & & & ( d ) & ( km @xmath75s@xmath9 ) & @xmath76 + k7v & 10.44 & 1.10 & [email protected] & [email protected] & 55 + cccccccccccccc @xmath68 & @xmath59@xmath69 cos @xmath78 & @xmath59@xmath71 & @xmath79 & @xmath80 & @xmath81 & @xmath82 + ( marcsec ) & ( marcsec yr@xmath9 ) & ( marcsec yr@xmath9 ) & ( km @xmath75s@xmath9 ) & ( km@xmath75s@xmath9 ) & ( km@xmath75s@xmath9 ) & ( km @xmath75s@xmath9 ) + [email protected]@xmath83 & [email protected]@xmath83 & [email protected]@xmath83 & [email protected] & [email protected] & [email protected] & 28.05 + + @xmath84 from @xmath85 and @xmath86 catalogues ( @xcite @xcite ) . + fr cnc is classified as a k8v - star by @xcite , while a multicolour photometric study allowed @xcite to classify it as a k5v - star . they also have obtained the spectral energy distribution ( sed ) of the star matching a @xmath87 of 4250@xmath77250 k and a log @xmath88 of [email protected] , that agrees with a k5v classification . we have compared our high resolution echelle spectra , in several spectral orders free of lines sensitive to chromospheric activity , with spectra of inactive reference stars of different spectral types and luminosity classes , observed during the same observing run . this analysis makes use of a modified version of the program starmod ( jstarmod ) developed at penn state university ( @xcite @xcite ; @xcite @xcite ) . this program constructs a synthesized stellar spectrum from artificially rotationally broadened , radial - velocity shifted , and weighted spectra of appropriate reference stars . for fr cnc , we have obtained the best fit with a k7v reference star , which is in closer agreement with the k8v classification rather with the k5v from @xcite . we have determined the heliocentric radial velocities ( rv hereafter ) by making use of cross - correlation technique ( see e.g. @xcite @xcite ) . the spectra of the target were cross - correlated order by order , using the routine fxcor in iraf , against spectra of rv standards with similar spectral type taken from @xcite . we derived the rv for each order from the position of the peak of the cross - correlation function ( ccf ) and calculated the uncertainties based on the fitted peak height and the antisymmetric noise as described by @xcite . [ fig : perfil ] ( top ) shows , the irregular profiles of the ccf ( double peaks and asymmetries ) can produce significant errors in rv measures . these features show regular variations : a double peak moving on time - scale of the rotational period can be seen . photospheric activity features on the stellar surface that disturb the profile of the photospheric lines could induce variations in the peak of the ccf , but a stellar companion could also produce the double peak effect . we checked if these rv variations could be due to a binary nature of fr cnc . we found no evidence for the existence of a companion by measuring the rv of the peaks and by trying to fit the data to a coherent orbit . we also carried out a line bisector analysis to enable us to ascertain whether the rv variations may be attributed to starspots . the ccf was computed for regions which include the photospheric lines commonly used in the doppler imaging technique , while excluding chromospheric and telluric lines ( @xcite @xcite ) . we computed the bisector and , to quantify the changes in the ccf bisector shape , also the bisector inverse slope ( bis ) . the bis was defined as the difference of the average values of the top and bottom zones ( we avoided wings and core of the ccf profile , due to errors of bisectors measurements , which are large in these zones ) . we studied the bisector variations only for foces04 run , as it was the more suitable data for the study . three cases are reported in bibliography ( see for example @xcite , @xcite , @xcite , etc . ) : a ) anticorrelation , which indicates that the rv variations are due to stellar activity ( by active regions at the stellar surface like spots or plages ) , b ) lack of correlation , which indicates the doppler reflex motion around the center of mass due to other bodies orbiting the star , c ) correlation , which , as pointed out by martinez - fiorenzano et al . ( 2005 ) , indicates that the rv variations are due to light contamination from an unseen stellar companion . as shown in fig . [ fig : bis ] , there is an anticorrelation between bis and rv , with a pearson correlation coefficient ( @xmath89 ) of -0.6851 . this result suggests that the rv variations of fr cnc are due to stellar activity variations ( e.g. spots on photosphere ) and not due to a binary nature . when the spectrum of the standard star was broadened to the same rotational velocity of fr cnc , the profiles of the ccf became smoother and could be fitted with a gaussian profile , ( see fig . [ fig : perfil ] , bottom ) . the irregular profiles of the ccf ( double peaks and asymmetries ) can produce significant errors in rv measurements ( see fig . [ fig : perfil ] ) . these irregularities may come from photospheric activity features on the stellar surface . they can distort the profile of the photospheric lines and induce variations in the peak of the ccf . however , this behavior may be caused by the difference in rotational velocity ( @xmath6 sin @xmath7 ) between the target and standard star ( see e.g. @xcite @xcite ) when the standard is a much slower rotator than the target . the ccf is essentially the broadening function that would be applied to the template spectrum . use of a broadened template removes higher moments introduced from the starspots , enabling a relatively unbiased estimate of rv to be determined ( fig . [ fig : perfil ] ) . a mean rv of 17.8 @xmath77 1.6 is obtained for the 2004 data set and is in good agreement with the rv derived through optimisation of parameters in the doppler imaging process ( section 7 ) , where we find @xmath90 = 18.6 @xmath77 0.6 . therefore all the radial velocities given in this paper have been calculated by cross - correlation with this rotational broadened spectrum of the standard star . in table [ tab : vr ] we list , for each spectrum , the heliocentric radial velocities ( @xmath90 ) with their corresponding errors ( @xmath91 ) obtained as weighted means of individual values deduced for each order in the spectra . we also list data points from @xcite for comparison . those orders which contain chromospheric features and prominent telluric lines have been excluded when determining the mean radial velocity . ) .,width=207 ] by using the program jstarmod ( see sect . [ 51spectral ] ) we have obtained the best fits for each observing run using @xmath92 values of @xmath9335 km s@xmath9 . however , this value depends on the rotation of the standard star which has non - zero rotation . therefore the obtained value can only be used as an approximation . to determine an accurate rotational velocity of this star we made use of the following method ( see @xcite @xcite for details ) . rotational velocities , @xmath94 can be written as follows ( see @xcite and references therein ) : @xmath95 where @xmath96 is a coupling constant which depends on the spectrograph and its configuration . the spectrum of each of these stars was broadened using the program jstarmod from @xmath6 sin@xmath7 = 1 km s@xmath9 up to 50 km s@xmath9 and the respective ccf was calculated . @xmath96 was found for every spectrograph by fitting the relation @xmath97 vs @xmath98 . it is well known that @xmath99 is a function of the broadening mechanisms which are present in the atmosphere of the star , except rotation ( @xcite @xcite ) . since the broadening mechanisms are function of the temperature and gravity , we may expect a dependence of @xmath99 on the temperature . to determine this dependence we use synthetic spectra with no rotational velocity computed using the atlas9 code by ( @xcite @xcite ) adapted to work under a linux platform by @xcite and @xcite . once @xmath96 is determined and @xmath99 calibrated with the color index ( @xmath100 ) , @xmath101 ( width of the ccf of the star when is correlated with itself ) is measured for each star , @xmath6sin@xmath7 can be directly calculated using the above formula ( [ eqrv ] ) . in table [ tab : vr ] we list , for each observing run , the averaged @xmath94 value obtained . from table [ tab : vr ] we estimate uncertainties of 1.9 - 3.2 km s@xmath9 based on the standard deviations as each measurement epoch . it is likely that the @xmath6sin@xmath7 values vary by more than these uncertainties since fr cnc is very active and exhibits starspots ( see section 7 ) that significantly distort the rotationally broadened absorption lines . moreover , while the observations taken in the year 2004 cover a complete rotation cycle , those @xmath6sin@xmath7 measurements at other epochs , only include one to two observations and are likely to yield more biased results ( i.e. depending on the location of starspots at the observation phases ) . in section 7 , we model the line profile using our doppler imaging code , allowing for the presence of spots . the resulting fits are thus likely to give a more accurate rotation velocity , @xmath6sin@xmath7 , for fr cnc . llcccccll run & hjd & & @xmath102 + & ( 2400000 + ) & ( km s@xmath9 ) & ( km s@xmath9 ) + + foces04 & 53098.3713 & 16.95 @xmath77 0.58 & 44.1 @xmath77 1.9 + foces04 & 53099.3411 & 18.42 @xmath77 0.57 & + foces04 & 53099.4230 & 17.34 @xmath77 0.66 & + foces04 & 53099.4546 & 16.74 @xmath77 0.69 & + foces04 & 53100.3156 & 16.62 @xmath77 0.64 & + foces04 & 53100.3917 & 15.36 @xmath77 0.56 & + foces04 & 53101.3264 & 18.70 @xmath77 0.48 & + foces04 & 53101.4495 & 20.97 @xmath77 0.55 & + foces04 & 53102.3306 & 19.30 @xmath77 0.55 & + foces04 & 53102.4753 & 17.68 @xmath77 0.55 & + foces06 & 54086.6770 & 19.85 @xmath77 0.63 & 37.2 @xmath77 - + foces06 & 54088.5918 & 19.62 @xmath77 0.45 & + foces06 & 54091.6192 & 20.45 @xmath77 0.53 & + foces07a & 54156.4482 & 19.98 @xmath77 0.52 & 41.3 @xmath77 2.7 + foces07a & 54158.5625 & 20.41 @xmath77 0.80 & + foces07b & 54228.3461 & - & 43.6 @xmath77 3.2 + foces07b & 54229.3496 & - & + fies08 & 54547.4408 & 17.82 @xmath77 0.74 & - + upgren02 & 51626.695 & 27 @xmath77 2.3 & - + upgren02 & 51626.730 & 24 @xmath77 4.1 & - + we computed the galactic space - velocity components ( @xmath103 , @xmath11 , @xmath104 ) and their associated errors of fr cnc using the procedure described by @xcite modified by lpez - santiago ( see @xcite @xcite ) . this procedure uses j2000 coordinates and takes into account correlation in the measures of _ hipparcos_. we use averaged rv calculated here ( 18.63 @xmath77 0.14 km s@xmath9 ) and the proper motions and parallax from _ hipparcos_. the obtained values of the components with its module @xmath105 and associated errors are given in table [ tab : par2 ] . the velocity components in the ( @xmath103 , @xmath11 ) diagram are clearly within the young disc population boundaries ( eggen 1984a , b , 1989 ; montes et al . 2001a , b ) indicating that the star belongs to the young disc and that it might also belong to the ic 2391 moving group , mentioned previously by pandey et al . ( 2005 ) , but the eggen kinematic criteria ( see @xcite for details ) are negative , showing that fr cnc could be not a member of any moving group ( mg hereafter ) in the young disc area . the classical view of mgs ( e.g. eggen 1984a ) , i.e. they come from the remnant of a star - forming cloud has been discussed in recent years . several studies ( e.g. famaey et al . 2007 , 2008 ; antoja et al . 2008 ; zhao et al . 2009 ) seem to support a dynamic or resonant mechanism origin . while both theories are feasible , we will just take into account that the percentage of contamination of the young disc space velocity area by old field population is high ( see @xcite @xcite and reference therein ) and so age constraints are needed to assess if fr cnc belongs to any of young disc moving groups . li i @xmath56707.8 spectroscopic feature is an important diagnostic tool for assessing the age in late - type stars , since lithium is destroyed easily by thermonuclear reactions in the stellar interior . the spectral region of the resonance doublet of li i at @xmath56708 is covered in all the high resolution observations . we measured the equivalent width ( @xmath106 hereafter ) in seven spectra . due to the small value , we were not able to measure @xmath106s directly , or use spectral subtraction technique . therefore @xmath106s have been obtained using the iraf task sbands , performing an integration within a band of 1.6 centred in the lithium line ( maldonado et al . we have obtained an averaged value of 54 m . in our high resolution spectra the li i line is blended with fe i 6707.4 . to subtract the fe i contribution we used the color - index relation from @xcite ( @xcite ) therefore , we obtain the final value of @xmath106(li i ) to be 34 m . in fig . [ fig : li ] we plot as an example , a spectra formed by co - adding the foces04 run spectra and we indicated the position of the li i line . by comparing this value with stars of similar spectral type ( k5-k7 with v - i = 1.3 ) in other mgs members or clusters of a well - known age ( in the same way as in figure 3 of @xcite @xcite ) , it is in agreement with being a young object between 10 - 120 and compatible with being ic 2391 mg member . ideally , further constraints on fr cnc s age should be calculated to confirm its youth . popular age approximations are for example the x - ray flux - age relation ( @xcite @xcite ; equation a3 ) or the commonly used relation between age and @xmath107 index , that measures chromospheric emission in the cores of the broad chromospheric ca ii h & k lines ( see e.g. @xcite @xcite ; @xcite @xcite ; @xcite @xcite ) . the former relation is valid for stars with spectral types earlier than that of fr cnc and so can not be used . also , using the latter relation , from our ca ii h & k fluxes ( see sect . [ 63 ] and table [ tab : vr ] ) we obtain a log@xmath107 = -3.48 which is outside the validity range of the activity - age relation but in this case it is compatible with the young age . both echelle and long slit spectra analysed in this work allowed us to study the behavior of the different indicators from the ca ii h & k to the ca ii irt lines , which are formed at different atmospheric altitudes . the chromospheric contribution to these features was determined by using the spectral subtraction technique described in detail by @xcite and @xcite . the excess emission @xmath106 of different spectral features were measured in the subtracted spectra . in table [ tab : ew ] we give the @xmath106 for the ca ii h & k , h@xmath108 , h@xmath78 , h@xmath109 , h@xmath110 , h@xmath24 , and ca ii irt ( @xmath5@xmath58498 , 8542 , 8662 ) lines for the echelle spectra . these @xmath106s were converted to an absolute surface fluxes by using the empirical stellar flux scales calibrated by @xcite as a function of the star color index . in our case , we used the @xmath100 index and the corresponding coefficients for ca ii h & k , h@xmath24 and ca ii irt , using for h@xmath108 the same coefficients as for ca ii h & k , and derived the h@xmath78 , h@xmath109 and h@xmath110 coefficients of flux by carrying out an interpolation between the values of ca ii h & k and h@xmath24 . the logarithm of the obtained absolute flux at the stellar surface ( log@xmath111@xmath112 ) in ergs @xmath113 s@xmath9 @xmath9 for the different chromospheric activity indicators is given in table [ tab : fl ] . [ fig : hairt ] shows representative observations in the h@xmath24 , and ca ii irt @xmath5@xmath5@xmath114 , 8542 line regions for high resolution spectra . [ fig : haindia ] shows representative observations in the h@xmath24 for low resolution spectra . [ fig : hairtzoom ] shows a closer view of one spectrum from figs . [ fig : hairt ] and [ fig : haindia ] where the emission can be better seen . ( left side ) and ca ii irt @xmath1158498 , 8542 ( right side ) line regions for foces04 observing run ; clear wide and prominent emission arises over the continuum from h@xmath24 and a clear emission in the core of the absorption line is seen in the ca ii irt lines . , width=302 ] line region for hfosc04 observing run ; clear emission arises over the continuum from h@xmath24 in this low resolution spectra . , width=245 ] and [ fig : haindia ] , showing a zoom of the h@xmath24 ( left side ) and ca ii irt ( center ) line regions for foces04 and h@xmath24 ( right ) line region for hfosc04.,width=226 ] we analysed the h@xmath24 line region for all the spectra . this line in the obtained spectra is always observed in emission above the continuum ( see figs . [ fig : hairt ] and [ fig : haindia ] ) . measuring the @xmath106 of this line , we found that the @xmath106 average of the h@xmath24 emission is quite different in every season , showing significant variability in time - scales of a year . @xmath106(h@xmath24 ) = 3.23 for foces04 run while , @xmath106(h@xmath24 ) = 1.67 , 1.72 and 1.87 for foces06 , foces07a and foces07b respectively , @xmath106(h@xmath24 ) = 1.72 for the only value of fies08 run . for the low resolution spectra we have an average value ( in 37 spectra taken during three consecutive nights ) of @xmath106(h@xmath24 ) = 2.14 in hfosc04 run . [ fig : haindiam11 ] shows the variation of @xmath106 vs phase ( calculated with the photometric period ) in hfosc04 run . different symbols represent different nights . the second night ( triangles ) shows higher values of the h@xmath24 @xmath106s . top of fig . [ fig : hairtfec11 ] , represent the variation of @xmath106 vs phase in the foces04 run . @xmath106 vs phase in hfosc04 run . different symbols represent different nights . the second night is in triangles , showing higher values of the h@xmath24 @xmath106s.,width=124 ] @xmath106 vs phase in foces04 run . bottom : variation of ca ii irt @xmath106 vs phase.,width=275 ] comparing the variations of h@xmath24 @xmath106 between the runs and the variation in each run when possible , we note activity level variations on a month - long timescale , from one year to the next ( see table [ tab : ew ] ) . making the comparison with the photometry , we notice a correspondence between marked variations in the light curve and @xmath106(h@xmath24 ) . in the year 2004 both photometry and spectroscopy show a high level of fr cnc activity , while it is decreasing rapidly in the year 2005 and then remains on that level during our further observations . this can be interpreted as an activity cycle of at least 45 years , similar to sun or other stars activity cycles , but further follow up is needed to confirm this . accurate stellar activity cycle can prove useful for study the dynamo interface and activity cycle - rotation - spectral type mechanisms in the stars ( see e.g. lorente & montesinos 2005 ) . the persistence of h@xmath24 emission indicates that it is a very active by dra system , but the vast range of variability levels make this star unusual and interesting for further study . we can see the absorption of h@xmath110 , h@xmath109 and h@xmath78 balmer lines filled in with emission in the observed spectra . [ fig : hbdg ] plots a representative subtracted spectra of these three lines in different nights of foces04 run . the variation of these lines with rotational phase and from season to season follows the same trend as h@xmath24 variation . we also measured the ratio of excess emission in the h@xmath24 and h@xmath110 lines @xmath116 and the ratio of excess emission @xmath117 with the correction : @xmath118 given by @xcite . this corrects the absolute flux density in these lines for the color difference in the components . we have obtained a mean value of @xmath117 @xmath119 . this value is in the limit between the presence of prominence - like material on or above the stellar surface ( @xcite @xcite and @xcite @xcite ) . , h@xmath109 and h@xmath78 lines in different nights of foces04 run . , width=226 ] the ca ii h & k line region is included in most of the spectra but the efficiency of the spectrograph and the ccd decreases very rapidly due to the position of these lines at the end of the echellogram . therefore , the obtained s / n ratio is very low , and the normalization of the spectra is very difficult . in many cases we could not measure the @xmath106 lines and in other we measured them in the observed spectra as it was not possible to apply the spectral subtraction for this region ( see tables [ tab : ew ] and [ tab : fl ] ) . strong emission in the ca ii h & k is seen despite the low s / n . the three lines of the ca ii ( irt ) are included in all our echelle spectra . in all of them a clear emission arising in the core of the absorption lines is seen ( see fig . [ fig : hairt ] ) . averaged values of emission @xmath106 are @xmath106(ca ii ) = 0.74 , 0.93 and 0.81 for @xmath1158498 , 8542 , and 8662 in foces04 run . ii ) = 0.49 , 0.72 and 0.61 in foces06 run . ii ) = 0.66 , 0.86 and 0.70 in foces07a . @xmath106(ca ii ) = 0.79 , 1.03 and 0.77 in foces07b and @xmath106(ca ii ) = 0.71 , 0.89 and 0.79 in fies08 . the variation of the emission in these lines is significant although not as strong as in h@xmath24 ( see table [ tab : ew ] ) . only for the foces04-run we have enough data points to compare variations between h@xmath24 @xmath106s and ca ii @xmath106s . in other active stars a clear anticorrelation is usually seen ( as their emissions come from different features in the stellar surface ) , see e.g. stars in @xcite , @xcite , @xcite , @xcite etc . , but it is very weak in the case of fr cnc ( see fig . [ fig : hairtfec11 ] ) . in addition , we have calculated the ratio of excess emission @xmath106 , @xmath120 , which is also an indicator of the type of chromospheric structure , which produces the observed emission . in solar plage values of @xmath120 @xmath93 1.53 are measured , while in solar prominence the values are @xmath93 9 , the limit of an optically thin emitting plasma ( @xcite @xcite ) . we have found a @[email protected] , indicating that ca ii irt emission comes from plage - like regions . since fr cnc is a rapid rotator with considerable broadening of spectral lines , we generated an indirect starspot map using the doppler tomography of stars ( dots ) imaging code ( @xcite @xcite ) . in order to detect the line distortion due to starspots in the high resolution spectra , we have applied least - squares deconvolution ( @xcite @xcite ; @xcite @xcite ) to the 43626845 wavelength region of foces04 spectra ( obtained at 2004 march april ) . a single line , free of the effects of rotational line blending and with high s / n is thus derived . deconvolution is carried out using a @xmath48 = 2450 k model line list ( vald ; see @xcite @xcite & @xcite ) which indicates that there are 8345 lines in the selected wavelength region with normalized depths of 0.051.0 . regions around hydrogen balmer lines , the mg triplet and na doublet are excluded from the deconvolution . a single line profile with a mean of @xmath5 = 5460.4 is derived for each observed spectrum . the mean s / n of the input spectra over the entire 4362 - 6845 wavelength region was 21.8 , while the mean deconvolved line profiles possess s / n = 1008 ( indicating a gain of 46.2 ) . it should be noted that while the profile shown in fig . [ fig : perfil ] ( upper panel ) represents the broadening function of fr cnc ( i.e. rotational velocity plus starspot distortions ) , the lsd profiles in section 7 are deconvolved using a linelist rather than a template . as such , the lsd profiles still contain the intrinsic stellar profile , and any distortions due to starspots will appear to possess a lower amplitude when compared with a template - derived ccf . for the imaging procedure , we used the standard star , hd 151877 , to represent the local intensity profile of a slowly rotating star . a two temperature model with @xmath121 = 4250 k and @xmath122 = 3000 k was used . the starspot image therefore represents the spot filling factor . details of the doppler imaging technique can be found in @xcite . we optimized the goodness of fit to the 10 deconvolved profiles for heliocentric rv , axial inclination , equivalent width and @xmath6sin@xmath7 , finding @xmath123 = 18.6 @xmath77 0.6 , @xmath6sin@xmath7 = 46.2 @xmath77 0.8 km s@xmath9 and i = 55 @xmath77 5 . the @xmath6sin@xmath7 value is higher than , but still consistent with , our mean @xmath6sin@xmath7 value derived in section 5.3 . since we here use all rotation phases to derive our best fit doppler image , it is likely that the result is less biased than the previous results in section 5.3 that do not take account of the presence of starspots , which affect the profile shape . starspot distortions in the wings of the profiles for instance , are likely to lead to underestimations of @xmath6sin@xmath7 when using a single spectrum , leading to a systematically lower mean estimation of @xmath6sin@xmath7 . we note however that our result for the 2004 observations is consistent with the value tabulated in table 9 ( i.e. 44.1 @xmath77 1.9 km s@xmath9 ) . [ fig : xxx ] shows the deconvolved profiles ( and phases of observation ) while fig . [ fig : yyy ] shows mercator projections of the starspot image of fr cnc . the surface map in fig . [ fig : yyy ] ( upper panel ) indicates that fr cnc possesses considerable spot coverage as suggested by the time varying distortions in the deconvolved profiles in fig . [ fig : xxx ] ( left panel - spectroscopic data only ) . for comparison with the 2003 - 2004 ( jd = 2452934 - 2453146 in table 1 ) lightcurve presented in fig . 3 ( top right ) , we have generated a synthetic lightcurve based on the spectroscopically derived image in fig . [ fig : yyy ] ( upper panel ) . while we are able to recover the observed lightcurve morphology , we considerably _ underestimate _ the amplitude ( fig . [ fig : lighcid ] , upper panel ) . this is likely due , at least in part , to loss of information in the doppler imaging process ( i.e. of the spectroscopic data ) owing to resolution and noise constraints imposed by the fitting procedure which must optimise both the fit and the image entropy . with higher s / n ratio data and spectral resolution it is probable that more precise fitting would yield spot filling factors indicating cooler spots . in addition , there is a tendency for spots close to the equator ( @xmath12420 ) to be reconstructed at slightly higher latitudes than they are physically situated . this is due to the smaller entropy penalty of placing a smaller spot away from the equator as compared with a larger spot nearer the equator . the resulting spot distribution will naturally lead to predicted lightcurves with underestimated amplitudes since equatorial spots yield higher amplitudes . we find a spot filling factor of 6 per cent from our spectroscopic doppler image . this is at the lower end of the typical values ( up to @xmath125 per cent ) found for active stars by oneal , saar & neff ( 1996 ) , oneal , neff & saar ( 1998 ) and oneal et al . ( 2004 ) using tio indicators . we have therefore reconstructed an image that also makes use of the 2003/2004 @xmath11-band photometric data ( e.g. see barnes et al . it should be noted that while the lightcurve data spans several months , the spectroscopic data were taken over only a few days . since the photometric observations taken within @xmath126 month of the spectroscopic data do not show significant deviation from the photometric dataset as a whole , we made use of all photometric points . the resulting simultaneous fits to the data are shown in fig . [ fig : xxx ] ( right panel ) and fig . [ fig : lighcid ] ( bottom panel ) , while the reconstructed image is shown in fig . [ fig : yyy ] ( bottom panel ) . simultaneous imaging that makes use of spectroscopic and photometric data must strike a balance , such that reasonable fits to both data sets are achieved . we have already outlined ( above ) possible modest biases in an image derived from a spectroscopic fit to finite s / n . lightcurve data alone does not contain sufficient information to make reliable images of single stars . any complex spot structure or spot groups will be rendered simply as a single spot . photometric imaging favours reconstruction at lower latitudes because a smaller spot is more easily able to reproduce the amplitude variation ( and yields a lower image entropy penalty ) in an observed lightcurve than a slightly larger spot at higher latitude . allowing the photometric fit to more closely match the observed lightcurve results in a slightly poorer fit to the spectroscopic data . the fitted lightcurve now closely matches the observed lightcurve , since more large scale ( albeit unresolved ) starspot structure has been recovered . this has necessitated some loss of image details ( fig . [ fig : yyy ] , lower panel ) as compared to the spectroscopic reconstruction alone ( fig . [ fig : yyy ] , upper panel ) . the filling factor is now greater at 9 per cent , adding further weight to the argument that spectroscopy alone is unable to recover all starspot information . indeed , if the results from tio are to be believed , there may be further unresolved global starspot covereage on fr cnc and all other stars that have been subjected to such studies . we also carried out a bisector analysis on the lsd profiles , calculating the bisectors in the same way as in sect . 5.2 . as we have mentioned , the lsd profiles should show a smoothing of the bisectors , reflected in the lower value of the bisector we obtained comparing with the original spectra . as expected , the anticorrelation is still seen , but here we have a strongest pearson correlation coefficient of r = -0.8992 . as might be expected for such an active star , fr cnc exhibits a high degree of spot coverage . no spots are visible in the phase range 0.10.3 . this agrees well with the photometric @xmath11-band light curve plotted for the same period ( fig . [ asas_phase ] , 20032004 ) which shows a maximum at phase @xmath127 . this contrast with a significant degree of spotness at other longitudes is likely responsible for the high degree of photometrical modulation seen in this star . the starspot distribution of fr cnc is also of interest since it is one of the latest spectral types to have been imaged . comparison with the k5v rapid rotator lo peg , @xcite , shows a significantly different spot distribution . while lo peg rotates more rapidly ( @xmath6 sin @xmath7 = 65.9 km s@xmath9 ) , showing predominantly mid - high latitude spot structures and a polar spot , fr cnc shows only mid - latitude starspots . small scale spot structure is expected to be variable on a day - long timescale , while @xcite have shown that the polar spot on the k0v star , ab dor , is variable in extent over periods of years . while g , and k stars such as ab dor often show strong polar caps , no such features were seen on the m1m2 dwarf stars hk aqr and re 1816 + 541 ( @xcite @xcite ) . this change in starspot location ( i.e. with latitude ) may be due to a change from a convective shell - type dynamo to a fully convective dynamo . predominantly mid - high latitude spots are expected to arise from a rapidly rotating solar - like dynamo ( @xcite @xcite ; @xcite @xcite ) while a distributed dynamo may be expected to produce spots at all latitudes . however , only further spectroscopic time series observations of fr cnc , with a higher cadence of observations ( enabling better surface resolution to be obtained ) , would reveal the long term stability of the starspot patterns . cccccccccccc & + run & hjd & & & & & & & + & ( 2400000 + ) & k & h & h@xmath108 & h@xmath78 & h@xmath109 & h@xmath110 & h@xmath24 & @xmath58498 & @xmath58542 & @xmath58662 + foces04 & 53098.3713 & @xmath84 & @xmath84 & @xmath84 & @xmath84 & 0.42 & 1.13 & 2.83 & 0.69 & 0.97 & 0.76 + foces04 & 53099.3411 & 2.46 & @xmath84 & @xmath84 & 0.62 & 1.10 & 1.32 & 3.50 & 0.73 & 0.92 & 0.71 + foces04 & 53099.4230 & @xmath84 & @xmath84 & @xmath84 & @xmath84 & 0.40 & 1.44 & 3.80 & 0.77 & 0.94 & 0.81 + foces04 & 53099.4546 & @xmath84 & @xmath84 & @xmath84 & @xmath84 & 0.46 & 1.10 & 3.15 & 0.91 & 0.97 & 0.75 + foces04 & 53100.3156 & 1.54 & 2.74 & 1.70 & 0.21 & 0.67 & 1.30 & 3.12 & 0.73 & 0.99 & 2.18 + foces04 & 53100.3917 & 1.58@xmath128 & 1.47 & 0.73 & 0.25 & 0.30 & 1.03 & 2.84 & 0.71 & 0.95 & 0.71 + foces04 & 53101.3264 & @xmath84 & @xmath84 & @xmath84 & 0.40 & 0.76 & 1.08 & 2.83 & 0.67 & 0.91 & 0.73 + foces04 & 53101.4495 & @xmath84 & @xmath84 & @xmath84 & @xmath84 & 0.27 & 1.23 & 3.15 & 0.71 & 0.86 & 0.70 + foces04 & 53102.3306 & 2.12 & 1.74 & 1.02 & @xmath84 & 0.77 & 1.21 & 3.78 & 0.67 & 0.92 & 0.73 + foces04 & 53102.4753 & @xmath84 & @xmath84 & @xmath84 & @xmath84 & @xmath84 & 1.40 & 3.34 & 0.77 & 0.89 & 0.74 + foces06 & 54086.6770 & 6.58@xmath129 & 6.82@xmath129 & 0.00@xmath129 & 0.36 & 0.30 & 0.74 & 1.49 & 0.29 & 0.63 & 0.58 + foces06 & 54088.5918 & 3.72@xmath129 & 4.30@xmath129 & 1.00@xmath129 & 0.26 & 0.73 & 0.50 & 1.93 & 0.59 & 0.53 & 0.66 + foces06 & 54091.6192 & 5.84@xmath129 & 5.39@xmath129 & 0.90@xmath129 & 0.37 & 0.36 & 0.58 & 1.60 & 0.60 & 0.99 & 0.58 + foces07a & 54156.4482 & 9.69@xmath129 & 5.97@xmath129 & 1.40@xmath129 & 0.32 & 0.24 & 0.74 & 1.80 & 0.68 & 0.86 & 0.71 + foces07a & 54158.5625 & 9.58@xmath129 & 3.78@xmath129 & 0.48@xmath129 & - & 0.13 & 0.76 & 1.65 & 0.64 & 0.86 & 0.69 + foces07b & 54228.3461 & - & - & - & - & - & - & 1.87 & 0.79 & 1.03 & 0.77 + foces07b & 54229.3496 & - & - & - & - & - & - & 1.72 & 0.71 & 0.89 & 0.79 + + @xmath128 values measured with low s / n . + @xmath84 values not measured due to the low s / n . + @xmath129 values measured in the observed spectra ( spectral subtraction is not applied ) . + cccccccccccc + run & hjd & & & & & & & + & ( 2400000 + ) & k & h & h@xmath108 & h@xmath78 & h@xmath109 & h@xmath110 & h@xmath24 & @xmath58498 & @xmath58542 & @xmath58662 + foces04 & 53098.3713 & @xmath84 & @xmath84 & @xmath84 & @xmath84 & 5.88 & 6.35 & 6.91 & 6.24 & 6.38 & 6.28 + foces04 & 53099.3411 & 6.62 & @xmath84 & @xmath84 & 6.03 & 6.30 & 6.42 & 7.00 & 6.26 & 6.36 & 6.25 + foces04 & 53099.4230 & @xmath84 & @xmath84 & @xmath84 & @xmath84 & 5.86 & 6.46 & 7.04 & 6.28 & 6.37 & 6.31 + foces04 & 53099.4546 & @xmath84 & @xmath84 & @xmath84 & @xmath84 & 5.92 & 6.34 & 6.95 & 6.36 & 6.38 & 6.27 + foces04 & 53100.3156 & 6.41 & 6.66 & 6.46 & 5.56 & 6.08 & 6.41 & 6.95 & 6.26 & 6.39 & 6.74 + foces04 & 53100.3917 & 6.42@xmath128 & 6.39 & 6.09 & 5.63 & 5.73 & 6.31 & 6.91 & 6.25 & 6.37 & 6.25 + foces04 & 53101.3264 & @xmath84 & @xmath84 & @xmath84 & 5.84 & 6.14 & 6.33 & 6.91 & 6.22 & 6.36 & 6.26 + foces04 & 53101.4495 & @xmath84 & @xmath84 & @xmath84 & @xmath84 & 5.69 & 6.39 & 6.95 & 6.25 & 6.33 & 6.24 + foces04 & 53102.3306 & 6.55 & 6.47 & 6.23 & @xmath84 & 6.14 & 6.38 & 7.03 & 6.22 & 6.36 & 6.26 + foces04 & 53102.4753 & @xmath84 & @xmath84 & @xmath84 & @xmath84 & @xmath84 & 6.45 & 6.98 & 6.28 & 6.35 & 6.27 + foces06 & 54086.6770 & 6.74@xmath129 & 6.75@xmath129 & - & 5.49 & 5.45 & 5.92 & 6.46 & 5.76 & 6.09 & 6.06 + foces06 & 54088.5918 & 6.49@xmath129 & 6.55@xmath129 & 5.92 & 5.36 & 5.84 & 5.75 & 6.57 & 6.06 & 6.02 & 6.11 + foces06 & 54091.6192 & 6.68@xmath129 & 6.65@xmath129 & 5.87 & 5.50 & 5.53 & 5.81 & 6.49 & 6.07 & 6.29 & 6.05 + foces07a & 54156.4482 & 6.91@xmath129 & 6.70@xmath129 & 6.06 & 5.45 & 5.35 & 5.91 & 6.54 & 6.13 & 6.23 & 6.15 + foces07a & 54158.5625 & 6.90@xmath129 & 6.50@xmath129 & 5.60 & - & 5.10 & 5.93 & 6.50 & 6.10 & 6.22 & 6.13 + foces07b & 54228.3461 & - & - & - & - & - & - & 6.56 & 6.19 & 6.31 & 6.18 + foces07b & 54229.3496 & - & - & - & - & - & - & 6.52 & 6.14 & 6.24 & 6.19 + + notes as in previous table . we have carried out a photometric , polarimetric and spectroscopic study of fr cnc . optical @xmath28-band photometry was carried out during 2007 march 2008 february at terskol branch of the astronomy institute ( russia ) . there are no peculiar features in the 2008 light curve , while in the 2007 photometry two brightening episodes were detected . one of them occurred at the same phase as the flare of november 23 , 2006 ( phase = @xmath130 ) and probably indicates that both of these events ( i.e. the flare on november 23 , 2006 and the photometric brightening episode ) originated from the same long - living active region on fr cnc . the non - detection of any other flares in our photometry except 2006 november 23 implies that fr cnc has a low frequency of flares . we analysed _ photometry obtained in 20022008 in @xmath11-band . no evidence of flares in _ _ data were found . the profiles of variability are different from season to season . the mean magnitude in @xmath11-band remained the same ( @xmath131 mag ) during 20022008 , while the amplitude decreased abruptly in 2005 . the proposed interpretation is a redistribution of spots / spot groups over the surface of the star , while the total percentage of the spotted area was assumed to be constant within the error limits . a detailed periodogram study of the _ asas-3 _ photometric data enabled us to derive a more accurate value for the period of fr cnc . we find that @xmath132 d. in addition , we also presented @xmath133 photometric calibration of 166 stars in fr cnc vicinity , whose @xmath11-magnitude is in the range of @xmath134@xmath135 mag . the @xmath3 broad - band polarimetric observations of fr cnc have been obtained at aries in nainital ( india ) at manora peak . the observed polarization in @xmath28-band is well matched with the theoretical values expected for zeeman polarization model . however , the observed polarization in @xmath11 and @xmath136 bands slightly exceeds the theoretical values and thompson and rayleigh scattering from inhomogeneous regions are not enough to explain the observed polarization excess . therefore the excess of linear polarization should come from an additional source of polarization . taking into account that we conclude that fr cnc is not a binary , the mechanism which can produce additional linear polarization is probably scattering in circumstellar material distributed in an asymmetric geometry ( e.g. see @xcite , @xcite ) . a total of 58 spectra of fr cnc , which have been obtained in 20042008 , were analysed in this work . based on our spectroscopic observations , fr cnc was classified as k7v star . rv analysis supports the single nature of fr cnc . anticorrelation between bis and rv also indicates that the rv variations are due to stellar activity variations and not due to a secondary companion . the kinematics study , based on obtained galactic space - velocity components ( @xmath4 ) of fr cnc , shows that this star clearly lies in the young disc population velocity space and might also belong to ic 2391 moving group , although the eggen kinematic criteria shows that fr cnc may not be a member of any mg in the young disc area . the li i @xmath56707.8 averaged @xmath106 measured is 34 m , giving the spectral type of fr cnc , it is in agreement with being a young object between 10120 myr . the h@xmath24 line was always observed above the continuum in all the obtained spectra . measuring the @xmath106 of this line , we found that the h@xmath24 emission @xmath106 average in every season is quite different . in 2004 , as with the photometry , spectroscopic indicators of chromospheric activity show a high level of activity which decreased in 2005 . the ca ii ( irt ) is included in our echelle spectra . from the ratio of excess emission @xmath106 we found that in fr cnc , ca ii emission comes from plage - like regions . we noticed that fr cnc can show an activity cycle of 4 - 5 years , although further follow up will confirm this periodicity . since fr cnc is a rapid rotator , we generated an indirect starspot map using the doppler tomography of stars imaging code . from it we derive @xmath6 sin @xmath7 = @xmath137 km s@xmath9 and @xmath138 . fr cnc belongs to one of the latest spectral types to have been imaged with the doppler tomography . we independently estimated a rotational velocity of fr cnc during our observations using queloz method ( sect . 5.3 ) and by the d.i . fits ( sect . although they are differences in the results of the two methods , they are consistent . in table [ tab : par ] we have put as the rotational velocity of fr cnc , the value obtained in the d.i . as it is likely more accurate . despite the short rotation period and its late spectral type , fr cnc shows very few flare events . it shows high level of activity as it is a young star , but an unusually short variability due to the redistribution of activity features on the stellar surface . while this variability is reflected in the changes of the amplitude of brightness , the mean brightness permanently is nearly constant , indicating that the percentage of stellar surface covered by spots is also constant . the spots location is also unusual , not showing a polar spot like other f k stars do but a distribution more resembling those seen in m1m2 dwarfs . although this may be indicative of a distributed dynamo , the mid - high latitude spot locations are more suggestive of an interface dynamo under the action of rapid rotation . we can only speculate as to whether fr cnc is representative of a regime in which a convective - shell - type dynamo gives way to a fully convective dynamo . polarimetric observations of the magnetic field by @xcite ( @xcite ) and @xcite ( @xcite ) for example suggest that this occurs at a later spectral type of m4 , whereas other chromospheric indicators show no obvious changes until later m spectral types ( e.g. @xcite @xcite ) . further spectroscopy with a higher cadence would enable more detailed maps to be derived , with multiple epochs enabling the evolution of starspots to be investigated . a. golovin is thankful to dr . anju mukadam and dr . paula szkody for useful discussions concerning fisher randomization test in periodogram analysis , to dr . andrs holl ( konkoly observatory , budapest , hungary ) for valuable comments on votable format and help with converting photometric sequence to it in order to align on dss image in aladin , to nick malygin and dr . ludmila pakuliak for useful discussions and proof - reading of the manuscript . . m.c.glvez-ortiz acknowledges the financial support from the european commission in the form of marie curie intra european fellowship ( pief - ga-2008 - 220679 ) and the partial support by the spanish micinn under the consolider - ingenio 2010 program grant csd2006 - 00070 : first science with the gtc ( http://www.iac.es/consolider-ingenio-gtc ) . glvez - ortiz and j. barnes also has received support from ropacs during this research , a marie curie initial training network funded by the european commission s seventh framework programme . this work was partly supported by the spanish ministerio de ciencia e innovacion ( micinn ) , programa nacional de astronomia y astrofisica under grant aya2008 - 00695 , and under grant aya2008 - 06423-c03 - 03 , and the comunidad de madrid under pricit project s2009/esp-1496 ( astromadrid ) . a. golovin is thankful to universidad complutense de madrid for hospitality and for all the efforts and the help during his visit to spain in 2008 july and 2009 february . kurucz r. l. 1993 , in astronomical society of the pacific conference series , vol . 44 , iau colloq . 138 : peculiar versus normal phenomena in a - type and related stars , ed . dworetsky m. m. , castelli f. , faraggiana r. , 87 .

we present analysis of the asteroid surface density distribution of main belt asteroids ( mean perihelion au ) in five ecliptic latitude fields , , derived from deep _ large binocular telescope _ ( lbt )band ( 85% completeness limit mag ) and _ spitzer space telescope _ irac 8.0 ( 80% completeness limitjy ) fields enabling us to probe the 0.51.0 km diameter asteroid population . we discovered 58 new asteroids in the optical survey as well as 41 new bodies in the _ spitzer _ fields . the derived power law slopes of the number of asteroids per square degree are similar within each ecliptic latitude bin with a mean value of . for the 23 known asteroids detected in all four irac channels mean albedos range from to . no low albedo asteroids ( 0.1 ) were detected in the _ spitzer _ fls fields , whereas in the swire fields they are frequent . the swire data clearly samples asteroids in the middle and outer belts providing the first estimates of these km - sized asteroids albedos . our observed asteroid number densities at optical wavelengths are generally consistent with those derived from the standard asteroid model within the ecliptic plane . however , we find an over density at in our optical fields , while the infrared number densities are under dense by factors of 2 to 3 at all ecliptic latitudes .

0903.4479

the present main belt asteroid size distribution provides important constraints on models of the original size distribution of planetesimals and their collisional evolution . the size - frequency distribution , main belt asteroid scale height , and albedo variations extant in a large , well - sampled population are also essential to model the size distribution of the near earth asteroid ( nea ) population and to determine the evolution of main belt objects as compared to those bodies injected into earth crossing orbits ( e.g. , * ? ? ? most main belt asteroids are found between 2.2 and 3.4 au from the sun and at ecliptic latitudes nominally less than 20 . the size - frequency distribution of main belt asteroids with diameters @xmath12 km is essentially unknown due to observational flux limits . for asteroids with diameters @xmath13 km , the derived size - frequency distribution estimates range from @xmath14 km - sized objects ( optical sloan digital sky survey , sdss ; * ? ? ? * ; * ? ? ? * ) to @xmath15 km - sized objects ( derived from surveys with the _ infrared space observatory _ , iso ; * ) within the main belt . recently , @xcite have derived a size - frequency estimate for the @xmath13 km - sized bodies for two fields at @xmath16 and @xmath17 ecliptic latitude from 24 photometry obtained with the nasa _ spitzer space telescope _ ( _ spitzer _ , * ? ? ? * ; * ? ? ? * ) that concurs with values derived from analysis of the sloan fields . while the infrared ( ir ) and optical methodologies yield comparable size estimates , these surveys are potentially biased as they are unable to detect ( either through intrinsic sensitivity limits or survey design ) fainter objects within the main belt . thus , the size - frequency distribution may not be complete due to the preferential selection of a cross section of the asteroid population with high albedo values or large sizes . @xcite have modeled the collisional evolution of the size distribution of the main belt using the current size - frequency distributions within the main belt as well as the size - frequency distribution of neas to constrain the primordial asteroid distribution . their best - fit models , based on the current size - frequency distribution , predict @xmath18 asteroids within the 0.3 to 1 km size range . collisional models by @xcite predict @xmath19 bodies in a similar size range within the main belt . however , the validity of the population extrapolation is based on an accurate understanding of the current size - frequency distribution necessitating deep observations to provide accurate number counts at both optical wavelengths ( higher albedo objects ) and ir ( larger dark asteroids ) wavelengths . the collisional evolution model by @xcite suggests that the primordial main belt contained @xmath20 to 250 times the current population of objects with diameters , @xmath21 km . the @xcite model produces a parent population of similar magnitude to that required by interpretation of the meteoritic record . for instance @xcite argues that the surface density within the main belt must be at least 100 times higher than that currently observed to be consistent with the meteoritic record . the total mass of known asteroids in the main belt is @xmath22 . however , both the meteoritic record and collisional models require a primordial main belt mass of @xmath23 to @xmath24 . the significant discrepancy between the mass of the primordial main belt and the current mass , primarily derived from large ( @xmath25 km ) sized asteroids , may be resolved by extending the asteroid number census to properly encompass smaller sizes while including populations found at higher ecliptic latitudes . here we present new observational results to constrain the size - frequency distribution of main belt asteroids and their ecliptic scale - height derived from a deep optical survey of select ecliptic fields obtained with the _ large binocular telescope _ ( lbt ) as well as near- and mid - ir fields observed with _ spitzer _ drawn from the data archive . section [ obs ] describes our observations and reduction techniques , [ disc ] discusses the newly discovered asteroids and our derivation albedo , size - frequency distribution , and scale - height , while [ concl ] summarizes our conclusions . optical observations were obtained at the large binocular telescope ( lbt ; * ? ? ? * ) facility of the mt . graham international observatory with the blue channel of the large binocular camera ( lbc ; * ? ? ? * ; * ? ? ? * ) and a single 8.4 m mirror on various nights during 2007 january 16 through 24 ut as part of science demonstration time ( sdt ) activities . the lbc is a wide - field imager incorporating four @xmath26 pixel ccd detectors with a @xmath27 field of view ( fov ) and a 0.23 per pixel platescale . four fields at ecliptic latitudes of 0 , 5 , 10 , and 15 were observed using a series of 4 min exposures in the v - band ( @xmath28 ; @xmath29 ) under non photometric conditions with seeing between @xmath30@xmath31 . fields were selected to be at solar elongations near 160 , such that sufficient asteroid motion on the plane of the sky could be detected within the @xmath32 3 hr observational baseline . observations of each field consisted of 2 pointings , with one pointing offset by @xmath33 , resulting in a total areal coverage per field of 0.2835 square degrees . table [ table : lbtobs ] provides complete observational details . the data were reduced using standard iraf routines in the _ mscred _ package . images were trimmed , bias - subtracted , and flat - fielded using median sky flats created with twilight flats and additional data obtained during 2007 january sdt time . astrometric solutions for each data frame were generated using iraf routines , in conjunction with the stellar positions obtained from the usno a2.0 astrometric catalog @xcite . the solutions had a positional accuracy of @xmath34 for each field and the relative position of detected asteroids are referenced to this grid . asteroid detection was performed using a three color method . images with astrometric solutions were displayed in image display tool ds9 @xcite , with each individual epoch ( table [ table : lbtobs ] ) loaded into a different color table . asteroids were identified by pairs or triplets of individually colored sources . from these identifications , absolute pixel coordinates for each target were obtained , which were subsequently used for photometric measurements . due to the conditions under which the data was obtained , a high signal to noise of 10 is required for any asteroid detection . secondary confirmation of this asteroid detection technique was performed by registering individual epoch frames by the world coordinate system ( wcs ) and then subtracting , leaving positive and negative asteroid pairs . while this latter technique would be preferred over the three color technique , the seeing varied by 1.3 to 3.0 over each night causing the wcs subtraction technique to be less reliable . photometry was performed using a circular aperture with a radius of 10 pixels for each asteroid coordinate . the photometry was calibrated from stellar sources in the fields using published @xmath2band magnitudes taken from the tycho and usno a2.0/yb6 catalogs ( mean photometric errors of the order @xmath10 0.1 mag ) . forty stars ( 10 stars per chip ) were identified by their astrometric positions and were used to find a photometric offset between the measured and reported absolute values from tycho / usno a2.0 . this zero - point photometric offset , which includes the mean photometric error , was then used to calibrate the asteroid photometry in each pointing as described in table [ table : lbtobs ] , column [ 7 ] . we detected 62 asteroids in the lbc @xmath2band images , of which only four are previously known objects with orbital determinations . the 58 newly discovered asteroids detected in this survey were observed in either two or three epochs . due to the field overlap between pointings , some asteroids were detected six times , allowing for both precise astrometry and photometry . table [ table : as_numsum ] summarizes the number of asteroids found in each field as well as the extrapolated number counts per square degree at each latitude and the limiting magnitudes for each field . the extrapolated optical number counts were obtained by multiplying the asteroid number counts in a field by the number 0.2835 square degree fields necessary to tile a 1 square degree field assuming a uniform asteroid density at each latitude . asteroid photometry and number counts in select ecliptic fields discussed in [ disc ] were derived from new ground - based optical observations and point - sources extractions obtained from fields observed with the infrared camera ( irac ; * ? ? ? * ) and the multiband infrared photometer for _ spitzer _ ( mips ; * ? ? ? * ) as part of various ir survey programs retrieved in the _ spitzer _ public archive . post - pipeline ( pipeline v15 ) basic calibrated data ( bcds ) of selected _ spitzer _ irac fields were downloaded from the public archive and utilized for asteroid detection . fields were selected from the ecliptic plane component of the _ spitzer _ first look survey ( fls , program identification ( pid ) 98 ; * ? ? ? * ) , and swire xmm - lss fields ( pid 181 ; * ? ? ? irac 8.0 data is preferable for asteroid detection due to the detector platescale ( @xmath35 ) and detector sensitivity . the 5@xmath36 detection limit for a 500 sec irac 8.0 observation is 27 @xmath37jy , enabling detection of small main belt asteroids with high signal - to - noise ratios ( snr ) in short integration times . for example , the 8.0 flux from a 1 km diameter main belt asteroid radiating as a blackbody with an orbital semi - major axis of 2.5 au ( 3.2 au ) viewed at opposition is @xmath38jy ( @xmath39jy ) . once an asteroid candidate is identified at 8.0 its sky coordinate can then be used to examine pixels at the same location in the irac 3.6 , 4.5 , and 5.8 images . although the flux densities for small - sized asteroids near the outer belt edge are not easily detectable with irac in single frames at shorter wavelengths , their motion will separate them from the confusion caused by faint extragalactic sources or point spread function ( psf ) smearing of the telescope . due to limits on observation durations ( driven by on - board data storage issues and downlink frequency ) , the _ spitzer _ deep large area surveys must consist of multiple epochs of data with multiple pointings . for asteroid detections this is advantageous as the observation duration limit for all irac observations is 6 hrs which allows observations to be repeated with a cadence of a few hours . assuming circular keplerian orbits within the asteroid belt , the angular velocity @xmath40 ( rad s@xmath41 ) , or rate of motion on the sky of an object near opposition is @xmath42 where @xmath43 , @xmath44 is the mass of the sun , and @xmath45 is the orbital distance in meters . for inner main belt asteroids with a semi - major axis of 2.2 au , the expected rate of motion is 45 per hr , while the expected rate of motion for outer main belt asteroids is 24 per hr . however , the actual rates and direction of motion on the sky may significantly differ from these values ( e.g. , near quadrature , when the fls fields were observed at solar elongation @xmath46 115 ) . nevertheless , movement on the sky of many tens of arcsec per hr enables one to detect asteroids in the field on timescales of hours . the fls consisted of irac observations of two 0.13 square degree fields ( 10@xmath47 48 ) centered on a solar elongation of 115 as seen from _ spitzer _ , at ecliptic latitudes @xmath48 and @xmath49 on 2004 january 21 ut . in j2000 coordinates , this corresponds to field centers of r.a . @xmath50 , decl . @xmath512153.8 for the @xmath52 field , and ra @xmath53 , decl . @xmath541318.7 for the @xmath55 field as viewed by _ spitzer_. to detect asteroid motion , each irac field was observed at three epochs separated by 70 min . for each of the latitude fields , we colored each epoch and coadded the irac 8.0 mosaics of the three epochs using a fixed stellar wcs to create a composite red , green , blue ( rgb ) image , which was then examined . in this rgb image , fixed sources appear white , and moving targets were seen as red - green - blue quasi - linear sequences of sources . the typical snr for the detected sources was @xmath56 . using the rgb technique with the irac 8.0 data , we were able to identify 18 sources in the @xmath57 field , and 16 sources in the @xmath55 field . the faintest of these sources have 8.0 fluxes of @xmath58jy and were detected with a snr @xmath59 in both fields . for comparison , @xcite identified 15 known ( six of which we also recover ) and 19 unknown asteroids their analysis of the fls fields , reporting 8.0 fluxes ranging from 5.6 to 0.09 mjy . the swire program was designed to explore large scale structure in the universe with the irac and mips cameras @xcite . this survey consisted of 6 fields selected to overlap those observed at depth in by the _ hubble space telescope _ and the _ chandra x - ray observatory_. two epochs of observations were obtained for each of the six fields to preen moving targets from final mosaics enabling high fidelity point source catalogs to be generated . for the purposes of our asteroid survey we used irac data obtained between 2004 july 23 and 2004 july 28 in the swire xmm - lss field , located at @xmath60 and encompassing an area of 9.1 square degrees . due to the large area covered by this _ survey and duration limitations of single observations , the irac data consists of 16 individual pointings of 0.44 square degrees in two epochs . the two epochs of each pointing are offset by 3.75 hrs , thus significant asteroid motion can be detected . asteroid detection in the swire data sets was also performed by coloring and coadding irac 8.0 mosaics of each epoch to create a composite red - blue ( rb ) image . in this rb image , fixed sources appear purple , and moving targets were seen as red - blue quasi - linear sequences of sources . we identify a total of 46 sources in the field . the faintest sources have 8.0 fluxes of @xmath61jy and were detected with a snr @xmath62 in both fields . of the 46 sources detected , only 14 asteroids have ground based detections and orbital elements . fluxes in all four irac channels were obtained for all asteroids in the fls and swire datasets . pixel coordinates for each asteroid were recorded at the time of detection and were used for photometry . though the irac cameras have an offset between the 3.6/5.8 and the 4.5/8.0 fovs , pipeline processed mosaics in each channel are aligned such that a given sky coordinate corresponds to the same pixel coordinate in all four channels . this allows use of one set of pixel coordinates for all sources in all 4 channels . source photometry was performed using the iraf task _ phot _ and a three - pixel radius circular aperture to measure the asteroid flux and a concentric four - pixel wide sky annulus to determine the median sky contribution to the total flux measured in the integrated object aperture . final fluxes and flux uncertainties were derived using the aperture corrections contained in the irac data handbook @xcite . due to the offset in the irac channel fovs , some asteroids are only detected in the 4.6 and 8.0 data . the number of asteroids detected in all four irac bands is 40 in the swire data and 24 in the fls fields . the mean ( median ) 8.0 asteroid fluxes are @xmath63jy ( @xmath64jy ) , @xmath65jy ( @xmath66jy ) , and @xmath67jy ( @xmath68jy ) for _ spitzer _ fls @xmath16 , fls @xmath69 , and swire fields respectively . the total number of asteroids detected in the _ spitzer _ fields are summarized in table [ table : as_numsum ] . detailed physical parameters for the newly discovered asteroids detected at optical ( 58 ) and ir ( 41 ) wavelengths are presented in on - line machine - readable catalogs . illustrative catalog layouts are shown in table [ table : machine - optical ] and table [ table : machine - spitz ] . the optical catalog , table [ table : machine - optical ] , is similar to the standard minor planet center submission format . in our table [ table : machine - optical ] the corresponding column format is used : [ 1 ] the asteroid name , [ 2 ] a true / false=1/0 flag indicating if asteroid is a new discovery , [ 3][5 ] are respectively the year , month and ut date of observation , [ 6][11 ] are the right ascension and declination ( j2000.0 ) of the detected target , [ 12 ] and [ 13 ] are the observed asteroid magnitude and associated uncertainty , while [ 14 ] is the filter used for the observations . the ir machine - readable catalog , table [ table : machine - spitz ] , is similar in column format where : [ 1 ] is the asteroid name , [ 2 ] is the _ spitzer _ program i d number , [ 3 ] is the astronomical observation request ( aor ) key , [ 4][9 ] is the year , ut date and time at the start of an observation , [ 5 ] and [ 6 ] are the right ascension and declination in degrees , [ 10][13 ] are the right ascension and declination and ecliptic latitude and longitude in sexigesimal degrees , while [ 14 ] through [ 21 ] are the observed fluxes and flux errors in @xmath37jy measured for each asteroid in the four irac channels . note that the fluxes in the table are not color - corrected . the distribution of asteroids within the main belt provides a means to ascertain the mass distribution of material within the main belt and provides constraints for theoretical models developed to describe planet formation from protoplanetary disks encompassing our own proto - sun , as well as those extant around other young stellar systems . by probing the asteroid distribution in reflected optical light and in the thermal ir in coordinated observational programs , a more complete picture of the number of asteroids and sizes and albedos can be obtained . current models of solar system dynamics ( e.g. , * ? ? ? * ) use the distribution of absolute magnitudes of asteroids combined with a single geometric albedo to estimate the size - frequency distribution of asteroids as observationally determined diameters and albedos of main belt asteroids only exist for @xmath32 2400 objects . however , new survey data now available from fields observed with the _ spitzer _ can be used to determine asteroid albedos and diameters for sizes under 1 km , enabling a more robust determination of albedo trends and size as a function of latitude for a more representative population of objects ( e.g. , a broader sampling of the total distribution ) . our study describes the first results for main belt asteroids with diameters of 0.5 km . to test completeness of our optical and _ spitzer _ data sets , we used the iraf task _ mkobjects _ to create synthetic point sources in data frames . to assess completeness in the optical data , synthetic point sources were added with the same seeing values measured in each frame and with coordinates that varied to mimic asteroid motion . to model the _ spitzer _ images , synthetic objects were randomly placed into an image with motions equivalent to the median rates of a body in keplerian orbit at the corresponding ecliptic latitude . synthetic asteroids were then recovered using the same rgb technique originally used to detected asteroids in the observational images . our completeness values are not only sensitive to source fluxes , but also to motion . for valid completeness detection , an asteroid must be detected in at least 2 of the 3 epochs of data . in our optical data , the 85@xmath70 moving target completeness limit in all fields is @xmath3 . because the moving target completeness samples the completeness due to ( 1 ) a limiting flux and ( 2 ) motion , a point source completeness test was also performed . this exercise shows that the point source completeness at @xmath3 is 90@xmath70 . for all asteroids counts at magnitudes greater than 21.3 , the difference in the two completeness values of 5@xmath70 , is applied by assuming that the number of asteroids not detected due to motion is the same at all magnitudes . these completeness values have been used to obtain corrected asteroid number counts which are reported in column [ 3 ] in table [ table : as_numsum ] . figure [ fig : lbt_histdist_4panel_fig1a ] shows the number counts per square degree as a function of magnitude in each optical field . the _ spitzer _ irac 8.0 analysis , under the constraints that the synthetic asteroid rate of motion was comparable to the median rates of motion in a field at a given ecliptic latitude combined with the requirement of a minimum of 2 detections per synthetic asteroid yields an 80@xmath70 completeness limit for the fls and swire data for @xmath71 @xmath37jy . our derived completeness value differs somewhat from that reported for the fls by @xcite who cite 90@xmath70 completeness at 100 @xmath37jy based on fixed target detections . with the completeness limit at 103 @xmath37jy , we calculate that the number of asteroids in the fls @xmath16 and @xmath69 fields are @xmath72 and @xmath73 per square degree respectively , and @xmath74 per square degree for the @xmath75 swire field . figure [ fig : spitzer_histdist_3panel_fig2 ] are histogram plots of the surface density of asteroids as a function of 8.0 flux in the _ spitzer _ fls and swire fields . comparison of these two major data sets in different wavelengths is complicated in part by the potential for a disparate probe of asteroid sizes . using the completeness limits in the @xmath76band and at irac 8.0 we can estimate the minimum radii of asteroids detectable in our survey from @xmath77 and @xmath78 \label{equation : eq - fthermal}\ ] ] where @xmath79^{\frac{1}{4 } } ( cos \phi)^{\frac{1}{4}}\ , ( cos \theta)^{\frac{1}{4}}.\ ] ] the terms @xmath80 and @xmath81 are the observed @xmath76band and irac 8.0 band fluxes ( w m@xmath82 hz ) respectively . variables in the equation for reflected flux are : @xmath83 is the frequency of observation , in this case @xmath76band , @xmath84 and @xmath85 are the radius ( m ) and temperature ( k ) of the sun , @xmath86 is the asteroid heliocentric distance in meters , @xmath87 is the asteroid diameter in meters and @xmath88 is the asteroid geocentric distance , @xmath89 is the geometric albedo of the asteroid in @xmath76band , r is the relative reflectance of the asteroid as measured with respect to the @xmath76band reflectance and @xmath90 is the observed phase angle in radians . other variables in the equation for thermal flux are : @xmath91 is the asteroid-_spitzer _ separation ( au ) , @xmath92 is the asteroid emissivity which is assumed to be 0.9 , a is the bond albedo , @xmath93 is the solar constant , @xmath94 is the asteroid heliocentric distance in au , and @xmath95 is the beaming parameter . adopting mean values for the known asteroids in our datasets @xmath96 au , @xmath97 au ( optical ) and @xmath98 au , @xmath99 au , ( _ spitzer _ asteroids ; table [ table : stmres ] ) , assuming @xmath100 with @xmath101 and @xmath102 , the minimum asteroid radii detected in our data are @xmath103 m in the optical and @xmath104 m in the irac 8.0 channel . although _ spitzer _ is far more efficient than optical imaging surveys at detecting smaller asteroids , asteroids identified from these two data sets comprise a set of bodies with similar size ranges within the main belt . the ir and optical data can be used to obtain the power law size - frequency distribution at 3 similar latitudes . in both the optical and the mid - ir , flux can be used as a proxy for size in obtaining the size - frequency distribution assuming that the bond albedo is well behaved as a function of wavelength . by adopting a power law distribution for the surface density distribution @xmath105 @xmath106 , we use the completeness corrected counts to measure the power law slopes summarized in table [ table : oir - slopes ] . the derived optical and ir slopes are similar at all latitudes within the uncertainties , with a mean value of @xmath107 for the surface density distribution of asteroids . however , the formal errors suggest that the optical data provides a better constraint on @xmath108 . our derived @xmath108-values ( table [ table : oir - slopes ] ) for the _ spitzer _ fields differs somewhat from those cited by @xcite . the differences can be attributed to how the asteroid photometry was conducted ; @xcite employed psf - fitting ( using , at that time , a poorly determined irac psf ) , whereas we resort to aperture photometry to measure the flux density . in addition , differences in completeness also lead to variances in our results as opposed to those presented in @xcite . we cite 80% completeness at 100 @xmath37jys for moving - objects , while @xcite quote 90% completeness at the same flux level for point sources . as the derived slopes of the surface density distributions are similar at optical wavelengths , we employed a kolmogorov - smirnov ( ks ) test using the mean rates of motion of the asteroids at @xmath16 and @xmath109 ecliptic latitude to discriminate whether these bodies detected in the optical fields are drawn from the same size and magnitude population distributions . unlike a @xmath110 test , data is not binned in the ks test and the ks test can be used on data sets of different lengths . due to the low number of asteroids detected in the 10@xmath111 and 15@xmath111 , binning of the data was rejected due to the associated loss of information and arbitrariness of defining bin sizes . the derived probability is 96.88% , strongly suggesting that the two fields represent the same distance ( rate ) population . similarly , a ks test to determine the likelihood that the asteroids represent the same size distribution yields only a 41.32% probability that the objects at @xmath109 represent the same size distribution as those detected at @xmath16 ecliptic latitude . this statistical inference suggests that the size distribution of asteroids at @xmath16 and @xmath109 are dissimilar ; however , the veracity of this conclusion is limited by small number detection statistics of our data . table [ table : ks - mmm ] summarizes our probability analysis of the magnitude distribution ( e.g. , asteroid diameters ) as a function of ecliptic latitude compared to the @xmath16 field . lastly , although the ks test indicates that the @xmath16 and @xmath109 optical fields are likely not the same population , they do appear to follow a distribution in which large , thus bright , objects are rare , whereas small and thus faint , objects are more numerous . our observed asteroid frequency distribution can be juxtaposed with current models describing the evolution of the asteroid populations in the solar system , such as the statistical asteroid model ( sam ) developed by @xcite . the sam uses a set of 8603 asteroids with absolute magnitudes less than 15.75 ( asteroids with diameters of @xmath112 2.5 km ) to determine a size - frequency distribution of asteroids which is assumed to be smooth to diameters of 1 km . this model also incorporates an albedo distribution derived from 15 dynamical families and 3 `` background '' populations based on 1980 asteroids which have diameter and albedo determinations from msx @xcite and iras @xcite . a number of asteroids are excluded from this model , including those with inclinations greater than @xmath113 , those with eccentricities @xmath114 0.3 . the asteroid number counts observed with both _ spitzer _ and the lbt are less than those predicted by the sam . for a limiting flux of 0.06 mjy at 8 the sam predicts @xmath115 asteroids at @xmath16 latitude and @xmath116 asteroids at @xmath69 latitude . one reason the sam may overestimate the number of asteroids is the assumed faint flux limit of @xcite which corresponds to an effective diameter of 0.6 km . this flux limit is fainter than the completeness for both fields , thus the sam over predicts the number of small asteroids which would be observed by _ spitzer_. the overestimate of the number of asteroids may also be due to a fundamental assumption of the sam the power law slope of the size - frequency distribution is continuous to diameters less than 1 km . this is not observed in the optical . the size - frequency distribution appears to fall off around 3 km @xcite ; however , the reasons for this fall off are not well understood . two possible reasons for the fall off in observed data are : 1 ) the bright limiting magnitude of most asteroid surveys excludes detections of asteroids smaller than 1 km in diameter , or 2 ) the asteroids smaller than 1 km may not be solid , but may instead be easily disruptible piles of rubble . if the predicted number of asteroids is scaled from the sam with a model that yields the number of asteroids with diameters @xmath114 0.6 km being 2.4 times the number of asteroids with diameters @xmath114 1.0 km , the sam still predicts @xmath117 asteroids in the @xmath16 latitude field and @xmath118 asteroids in the @xmath69 latitude field . the estimate at the high latitude is nearly in agreement with our fls results ; however , the sam still overestimates the number of asteroids in the @xmath16 field . the latter discrepancy may due to asteroids with high albedos ( @xmath114 0.25 ) which are not detected in the irac channels because they are too cool and thus their thermal fluxes are below the detection thresholds in this data . for the lbt data which has limiting diameters of @xmath32 2 km , the number counts are within the uncertainty of the sam prediction . for the sdss survey which has a point source completeness limit of @xmath119 , the sam predicts @xmath120 asteroids at @xmath16 latitude and @xmath121 asteroids at @xmath69 latitude . this completeness limit is fainter than ours by 0.9 mag ; however , the model values are nearly in agreement with the asteroid number counts at these two latitudes . unfortunately , the sam does not predict asteroid counts at higher latitudes ( @xmath122 ) and the current version of the sam only uses asteroids with inclinations less than @xmath123 to make estimates on number counts . a subset of our newly detected asteroids have large inclinations . in fact this limitation of the sam complicates derivation of the size - frequency distribution and physical characteristics of well known and well characterized targets such as pallas and asteroid dynamical groups such as hungarias and phocaeas @xcite . direct comparison of mean and median irac 8.0 fluxes ( [ saf ] ) of asteroids in the two fls fields and the swire field , reveals that the swire asteroids are , on average across the population , brighter than those asteroids at @xmath16 and @xmath69 latitude detected in the fls fields , by a factor of two . this variance in brightness could be explained as a difference in asteroid diameter assuming that the populations have comparable albedos . we explored this possibility by estimating the diameters and deriving albedos for all asteroids which were detected in all four irac bands using the near earth asteroid thermal model ( neatm ; * ? ? ? * ; * ? ? ? * ) . in this model , the thermal flux is dependent upon the sub - solar temperature and the temperature distribution of the surface of the asteroid . the latter is solely dependent on the albedo and @xmath95 as model variables . thus , our implementation of neatm follows the @xcite and @xcite prescription whereby the model fits the ir fluxes and the optical absolute magnitude of asteroids ( @xmath124 ) by varying the geometric albedo ( through a @xmath110-minimization ) until a best fit is found . the _ spitzer _ fluxes were first color - corrected according to the prescription described in the irac data handbook @xcite , resulting in the measured fluxes being divided by 1.1717 and 1.1215 at 5.8 and 8.0 , respectively . the asteroid diameter is then calculated by the relation of @xcite which uses only the geometric albedo and the absolute magnitude as input values . because irac asteroid spectral energy distributions are composites of both thermal emission and reflected solar light ( e.g. , * ? ? ? * ) , we derived albedo and diameter estimates of the asteroids using only data from irac channels 3 and 4 . from the channel 3 and 4 fits , we find that the percentage of reflected solar light in our channel 1 photometry to be @xmath46 56% and @xmath46 14% in the channel 2 photometry . our modeling results for _ known _ asteroids detected in our irac fields are summarized in table [ table : stmres ] . the mean diameters of asteroids in swire is 5.91 km while the mean asteroid diameter in the fls is 2.28 km . analysis of the mean geometric albedos and derived sizes for these objects suggests they divide between the two populations . the mean geometric albedo for asteroids in the fls fields is @xmath125 as opposed to @xmath126 for asteroids in the swire fields . no low albedo asteroids ( @xmath9 @xmath10 0.1 ) were detected in the _ spitzer _ fls fields , whereas in the swire fields they are frequent . evidently the surface composition of the asteroids in the swire fields contains more carbonaceous material than the asteroids in the fls , as carbonaceous chondrites have significantly lower albedos in the 0.03 to 0.11 range @xcite . however , the marked dichotomy in derived albedos is most likely a result of the two _ spitzer _ surveys sampling different asteroid taxonomy classes within the general asteroid belt . an extensive analysis of @xmath127 objects in the sloan digital sky survey moving object catalog ( sdss moc ) 4 by @xcite demonstrates that there is a strong correlation between sdss color , asteroid taxonomy ( c- , s- , v - type ) , and orbital elements . in fig . [ fig : sdssmoc4_fig3 ] we plot the proper orbital elements @xmath128 ( orbital semi - major axis ) vs. @xmath129 ( orbital inclination ) of our swire ( green squares ) and fls ( red triangles ) detected asteroids overlaid on 207,942 numbered asteroids ( black dots ) with orbital elements in the astorb files @xcite . the location of the major kirkwood gaps as defined by @xcite are indicated , and the colored symbols used to identify the _ spitzer_-detected asteroids are coded by derived albedo . open symbols are for _ spitzer _ asteroids with derived albedos , @xmath130 , while filled symbols identify asteroids with @xmath131 . the majority ( @xmath132% ) of the low albedo asteroids detected in the _ spitzer _ data reside at high inclination in the outer belt . primitive asteroids , with c - type colors and generally low albedos ( @xmath133 ) , are seen to dominate the asteroid population in the outer belt , while s - type asteroids with @xmath9 ranging from @xmath134 to 0.2 tend to be more dominant in the middle and inner asteroid belt @xcite . our swire data clearly samples asteroids in the middle and outer belts , and in fact , for the first time , we are able to estimate the albedos of km - sized asteroids in the outer belt ( table [ table : stmres ] ) . the low albedos ( @xmath130 ) of the @xmath135 km - sized outer belt swire asteroids are similar to the albedos derived for jupiter - family comets , @xmath136 @xcite . recent modeling by @xcite investigating the capture likelihood of cometary planetesimals into the asteroid belt ( leading the establishment of a d - type population ) suggests that the inner edge of the d - type population is near @xmath137 au . whether the population of low albedo asteroids discovered in our swire fields discussed here are organic - rich , primitive objects can only be verified with follow - up spectroscopy . from flux measurements in all 4 irac channels , we can also begin to constrain the colors of asteroids . this is beneficial for both the understanding of bulk asteroid albedo variations as a function of size and other orbital parameters such as orbital semi - major axis or inclination and for the future exclusion of asteroids from galactic or extragalactic catalogs . figure [ fig : color_fig4 ] is an adaptation of the work by @xcite which shows the irac colors of fixed astronomical objects where we have now included asteroid color data derived from the our fls and swire field photometry . asteroids mainly reside in one isolated locus in color space . the size of this locus is dependent upon both temperature and albedo . the shortest wavelength observations with irac at 3.6 trace the fraction of reflected incident solar flux , thus traces albedo , while the 8.0 to 4.5 ratio traces the thermal emission and thus the albedo and distance of an asteroid . our combined optical survey of @xmath138 square degrees and _ spitzer _ irac 8.0 fields encompassing from 0.26 to @xmath139 square degrees of sky at five latitudes perpendicular to the ecliptic plane resulted in detection of 118 main belt asteroids , of which 91 are newly identified objects either at optical or infrared wavelengths . the optical and mid - ir asteroid counts and fluxes demonstrate that the slope of the size - frequency distribution is consistent with that of a similar size population at 0@xmath111 and 15@xmath111 ecliptic latitude . the derived power law slopes of the asteroid surface density distribution are similar at within each @xmath5 ecliptic latitude bin with a mean slope of @xmath140 . the observed asteroid number densities at optical wavelengths are generally consistent with those derived from the standard asteroid model within the ecliptic plane . however , we find an over density at @xmath11 in our optical fields , while the infrared number densities are under dense by factors of 2 to 3 at all ecliptic latitudes . for the 28 known asteroids detected in all four irac channels , mean albedos range from @xmath7 to @xmath8 , statistically suggesting that these are two different populations . our swire data clearly samples low albedo asteroids in the middle and outer belts , and in fact , for the first time , we are able to estimate the albedos of km - sized asteroids in the outer belt . in order to increase our number statistics and lower the errors , we are completing the data - mining of additional , extant multi - epoch survey _ spitzer _ irac and mips dataset and are undertaking complimentary ground based optical observations with the lbt and the 2.3-m bok telescope at kitt peak . these data will provide a survey area more than 3 times larger than that described in this manuscript and will also expand analysis of asteroid number counts to ecliptic latitudes of 20 . the errors in the slope of the size - frequency distribution can not be decreased by greater number statistics alone . the use of fluxes as proxies for asteroid sizes does not account for variations in albedo ; therefore , additional mid - ir data is required to obtain albedos and sizes for a large number of asteroids . we in the process of completing the data - mining of additional , extant multi - epoch survey _ irac and mips dataset while undertaking complimentary ground based optical observations with the lbt and the 2.3-m bok telescope at kitt peak in order to increase our number statistics and lower the completeness statistics uncertainties . these data will provide a ecliptic survey area more than 3 times larger than that described in this manuscript and will also expand analysis of asteroid number counts to ecliptic latitudes of 20 . these data will extend the number of asteroids with accurate sizes to diameters much smaller than the 10 km iras detection limit enabling reliable number counts for asteroids with diameters @xmath141 km and assessment of whether the fall off in the size - frequency distribution of main belt asteroids distribution with diameters @xmath141 km is a real signature , or an artifact introduced by the limiting magnitudes of current asteroids surveys . this work is based on observations made with the _ spitzer _ space telescope , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa . support for this work was provided by nasa through an award issued by jpl / caltech . support for this work also was provided by nasa through contracts 1263741 , 1256406 , and 1215746 issued by jpl / caltech to the university of minnesota . e.r . and c.e.w . also acknowledge support from the national science foundation grant ast-0706980 . this work is also based on data acquired using the large binocular telescope ( lbt ) . the lbt is an international collaboration among institutions in the united states , italy and germany . lbt corporation partners are : the university of arizona on behalf of the arizona university system ; istituto nazionale di astrophysica , italy ; lbt beteiligungsgesellschaft , germany , representing the max - planck society , the astrophysical institute potsdam , and heidelberg university ; the ohio state university , and the research corporation , on behalf of the university of notre dame , university of minnesota , and university of virginia . the authors also acknowledge the detailed and very helpful comments of anonymous referees whose insights greatly improved the manuscript . fowler , j. w. , & chillemi , j. r. 1992 , in iras minor planet survey , eds . e. f. tedesco ( phillips lab . tech . pl - tr-92 - 2049 ) ( hanscom afb , ma : phillips lab . geophys . , air force mater . command ) , 17 veeder , g. j. , & tedesco , e. f.1992 , in iras minor planet survey , eds . e. f. tedesco ( phillips lab . tech . rep . pl - tr-92 - 2049 ) ( hanscom afb , ma : phillips lab . geophys . , air force mater . command ) , 107 lcccccc + epoch 1 & 6:14:21.68 & + 23:25:15.49 & jan 16 & 07:55:20 & 2.13 & [email protected] + epoch 2 & 6:14:21.68 & + 23:25:15.49 & jan 16 & 08:00:01 & 2.74 & [email protected] + epoch 3 & 6:14:21.68 & + 23:25:15.49 & jan 16 & 09:51:30 & 2.52 & [email protected] + + epoch 1 & 6:14:22.33 & + 23:40:18.72 & jan 16 & 08:01:21 & 2.19 & [email protected] + epoch 2 & 6:14:22.33 & + 23:40:18.72 & jan 16 & 09:13:54 & 2.67 & [email protected] + epoch 3 & 6:14:22.33 & + 23:40:18.72 & jan 16 & 09:57:22 & 2.67 & [email protected] + + + epoch 1 & 6:14:56.16 & + 28:25:17.94 & jan 25 & 02:49:41 & 1.50 & [email protected] + epoch 2 & 6:14:56.16 & + 28:25:17.94 & jan 25 & 03:49:01 & 1.64 & [email protected] + epoch 3 & 6:14:56.16 & + 28:25:17.94 & jan 25 & 06:39:50 & 2.20 & [email protected] + + epoch 1 & 6:14:56.78 & + 28:40:19.50 & jan 25 & 02:55:22 & 1.50 & [email protected] + epoch 2 & 6:14:56.78 & + 28:40:19.50 & jan 25 & 03:54:47 & 1.75 & [email protected] + epoch 3 & 6:14:56.78 & + 28:40:19.50 & jan 25 & 06:45:37 & 1.71 & [email protected] + + + epoch 1 & 6:15:33.49 & + 33:25:15.14 & jan 24 & 02:17:27 & 1.44 & [email protected] + epoch 2 & 6:15:33.49 & + 33:25:15.14 & jan 24 & 04:34:49 & 2.12 & [email protected] + epoch 3 & 6:15:33.49 & + 33:25:15.14 & jan 24 & 08:10:00 & 2.64 & [email protected] + + epoch 1 & 6:15:34.08 & + 33:40:19.96 & jan 24 & 02:23:01 & 1.44 & [email protected] + epoch 2 & 6:15:34.08 & + 33:40:19.96 & jan 24 & 04:40:57 & 2.16 & [email protected] + epoch 3 & 6:15:34.08 & + 33:40:19.96 & jan 24 & 08:15:42 & 2.65 & [email protected] + + + epoch 1 & 6:16:17.02 & + 38:25:09.47 & jan 24 & 02:05:07 & 1.35 & [email protected] + epoch 2 & 6:16:17.02 & + 38:25:09.47 & jan 24 & 04:13:06 & 1.79 & [email protected] + epoch 3 & 6:16:17.02 & + 38:25:09.47 & jan 24 & 07:47:27 & 2.62 & [email protected] + + epoch 1 & 6:16:15.22 & + 38:40:11.62 & jan 24 & 02:10:49 & 1.28 & [email protected] + epoch 2 & 6:16:15.22 & + 38:40:11.62 & jan 24 & 04:19:06 & 2.25 & [email protected] + epoch 3 & 6:16:15.22 & + 38:40:11.62 & jan 24 & 07:54:28 & 3.08 & [email protected] + lcccc @xmath16 & 24 & 85 & 91@xmath14210 & 22.3 + @xmath69 & 21 & 74 & 80@xmath1429 & 22.0 + @xmath17 & 11 & 39 & 41@xmath1426 & 22.2 + @xmath109 & 6 & 21 & 23@xmath1425 & 22.2 + @xmath16 & 24 & 85 & 91@xmath14210 & 25.7 @xmath37jy + @xmath69 & 21 & 74 & 80@xmath1429 & 25.7 @xmath37jy + @xmath75 & 11 & 39 & 41@xmath1426 & 35.4 @xmath37jy + lccccccccccccr sdt01&1&2007&01&16.33147&06&14&03.68&+23&21&03.0&19.72 & 0.03&v + sdt01&1&2007&01&16.38195&06&14&01.21&+23&21&08.7&20.34&0.04&v + sdt01&1&2007&01&16.41215&06&13&59.70&+23&21&12.1&20.16&0.04&v + sdt02&1&2007&01&16.33147&06&13&56.23&+23&15&27.2&20.94&0.13&v + sdt02&1&2007&01&16.38195&06&13&53.82&+23&15&26.8&21.54&0.10&v + sdt02&1&2007&01&16.41215&06&13&52.35&+23&15&27.0&21.46&0.04&v + @xmath143 & + lcccccccccccccccccccc 2001qy160 & 98 & 6095104 & 2004&01&21&02&35 & 25.728&181.0084 & -0.0412 & 180.9416 & 0.3633&89.4&8.1 & 18.9&14.4 & 95.7&45.7&939.9&179.4 + 0h & 98&6095104&2004&01&21&02&35&25.728&181.0051 & -0.1183&180.9693&0.2913&2.3&2.2 & 7.7 & 6.9 & 65.3 & 35.5 & 548.8&132.8 + 2004bs160 & 98 & 6095104 & 2004 & 01 & 21 & 02 & 35 & 25.728 & 181.0004 & -0.1816 & 180.9902 & 0.2313 & 3.7 & 3.6 & 1.7 & 1.7 & 52.5 & 52.2 & 103.2 & 49.2 + 2000ax136 & 98 & 6095104 & 2004 & 01 & 21 & 02 & 35 & 25.728 & 180.9245 & -0.1815 & 180.9204 & 0.2012 & 4.8 & 4.6 & 103.7 & 47.2 & 240.9 & 79.9 & 2526.7 & 305.6 + @xmath143 & + lcccr 2004th222 & 18.5@xmath142 14.1 & 53.6@xmath142 30.6 & 241.0@xmath142 80.3 & 1682.6@xmath142 246.3 + 2005yv181 & 6.0@xmath142 5.6 & 16.23@xmath142 12.8 & 128.4@xmath142 55.2 & 933.8@xmath142 179.0 + 2005yv50 & 16.7@xmath142 14.2 & 15.7@xmath142 12.4 & 64.5@xmath142 36.0 & 570.6@xmath142 136.2 + 1999wa1 & 45.1@xmath142 27.1 & 113.4@xmath142 50.0 & 938.5@xmath142 172.8 & 7314.0@xmath142 532.0 + 2004to8 & 109.2@xmath142 48.7 & 259.0@xmath142 82.6 & 1684.4@xmath142 237.5 & 10301.9@xmath142 634.3 + 2006ap1 & 5.8@xmath142 5.4 & 32.8@xmath142 21.7 & 263.8@xmath142 84.7 & 1677.9@xmath142 245.7 + 2005wx162 & [email protected] & 38.0@xmath142 24.1 & 311.5@xmath142 93.5 & 2287.0@xmath142 290.30 + 1999vy171 & 75.7@xmath142 38.5 & 383.9@xmath142 103.8 & 2867.7@xmath142 314.2 & 18500.5@xmath142 856.9 + 1999jz7 & 4.0@xmath142 4.9 & 25.5@xmath142 18.1 & 196.8@xmath142 71.2 & 1512.8@xmath142 232.3 + 2000vp1 & 15.9@xmath142 12.7 & 25.0@xmath142 17.8 & 113.9@xmath142 51.2 & 775.1@xmath142 161.3 + 2005yh117 & 6.7@xmath142 6.2 & 48.0@xmath142 28.4 & 470.0@xmath142 118.0 & 3570.0@xmath142 366.2 + 2002fe15 & 12.0@xmath142 10.1 & 47.9@xmath142 28.4 & 355.7@xmath142 100.8 & 4308.9@xmath142 403.4 + 2005yx74 & 83.1@xmath142 42.7 & 118.5@xmath142 52.0 & 538.9@xmath142 127.5 & 4465.2@xmath142 411.6 + 2002cn96 & 15.5@xmath142 12.3 & 65.0@xmath142 35.3 & 584.1@xmath142 133.3 & 3106.3@xmath142 343.6 + 2001qy160 & 8.9@xmath142 7.8 & 31.8@xmath142 22.6 & 101.3@xmath142 47.5 & 905.4@xmath142 175.8 + 2000ax136 & [email protected] & 217.6@xmath142 74.2 & 1228.2@xmath142 200.3 & 8993.1@xmath142 593.3 + 2004bh160 & 9.1@xmath142 9.3 & 11.4@xmath142 10.9 & 11.5@xmath142 11.7 & 73.6@xmath142 39.6 + 2001qd49 & 7.1@xmath142 7.4 & 11.2@xmath142 12.2 & 16.1@xmath142 16.0 & 109.9@xmath142 51.3 + 2002wl7 & 18.2@xmath142 13.9 & 50.0@xmath142 29.1 & 497.6@xmath142 121.4 & 3239.4@xmath142 347.8 + 1999fz24 & 41.7@xmath142 25.9 & 45.2@xmath142 27.3 & 211.0@xmath142 74.1 & 2116.3@xmath142 277.6 + 2001rp137 & 8.3@xmath142 7.6 & 14.3@xmath142 14.2 & 34.9@xmath142 24.1 & 273.5@xmath142 89.2 + 2004eu91 & 9.3@xmath142 9.2 & 7.2@xmath142 6.8 & 35.7@xmath142 26.5 & 116.5@xmath142 53.3 + 2004ca105 & 61.7@xmath142 35.8 & 54.6@xmath142 33.7 & 59.9@xmath142 34.8 & 487.4@xmath142 124.5 + 2002rg106 & 12.2@xmath142 10.4 & 5.3@xmath142 5.3 & 22.9@xmath142 17.6 & 228.3@xmath142 80.5 + 2004bu99 & 6.3@xmath142 5.8 & 7.9@xmath142 7.4 & 40.3@xmath142 25.9 & 398.9@xmath142 110.9 + 2004bs160 & 8.1@xmath142 8.6 & 3.1@xmath142 3.1 & 21.4@xmath142 30.7 & 101.9@xmath142 49.0 + 2001sb182 & 14.1@xmath142 11.9 & 11.8@xmath142 11.0 & 32.6@xmath142 22.6 & 184.0@xmath142 70.1 + 2004fg8 & 6.1@xmath142 7.2 & 7.7@xmath142 7.0 & 25.5@xmath142 19.2 & 185.5@xmath142 70.7 + 2004 th222 & 2.08 & 1.79&29.4&15.4 & 2.83 & 0.14 & 1.69 & 0.0002 + 2005 yv181 & 3.27&3.09&18.20&15.5 & 3.91 & 0.07 & 0.79 & 0.0157 + 2005 yv50 & 2.38&2.14&25.36&16.0 & 2.45 & 0.12 & 1.66 & 0.0087 + 1999 wa1 & 2.82&2.60&21.25&12.7 & 9.16 & 0.18 & 1.01 & 0.0225 + 2004 to8 & 2.66 & 2.44 & 22.57&14.5 & 9.02 & 0.03 & 1.04&0.5826 + 2006 ap1 & 2.91&2.71&20.53&15.5 & 3.59 & 0.09 & 0.77&0.0035 + 2005 wx162&2.83 & 2.61&21.20&15.3 & 5.10 & 0.05 & 1.03&0.0141 + 1999 vy171 & 2.56&2.34&23.48&14.0 & 10.92 & 0.04 & 1.04&0.6644 + 1999 jz7 & 2.33&2.08&26.02&14.0 & 3.39 & 0.39 & 1.35 & 0.0308 + 2000 vp1 & 2.43&2.21&24.83&15.7 & 1.99 & 0.23 & 1.03&0.0143 + 2005 yh117 & 2.91 & 2.73&20.57&15.4 & 6.73 & 0.03 & 1.01 & 0.0161 + 2002 fe15 & 3.39&3.25&17.54&14.5 & 12.41 & 0.02 & 1.05&0.3090 + 2005 yx74 & 2.85&2.64&20.99&14.8&5.35 & 0.08 & 1.05&0.0859 + 2002 cn96 & 2.95&2.74&20.29&14.8&5.88&0.06&0.87&0.9232 + 2001 qy160 & 2.96 & 2.39 & 17.74 & 15.0 & 3.26 & 0.17 & 1.02&0.0241 + 2000 ax136 & 2.40 & 1.80&22.02 & 13.1 & 6.89&0.21&1.33 & 0.0002 + 2004 bh160 & 2.42&1.82&21.86 & 18.4 & 0.51 & 0.30 & 1.03 & 0.0001 + 2001 qd49 & 2.77 & 2.19&18.98 & 17.5 & 0.94 & 0.20 & 1.00 & 0.0048 + 2002 wl7 & 2.69 & 2.11 & 19.55 & 15.3 & 4.48 & 0.07 & 1.05 & 0.1406 + 1999 fz24 & 3.10 & 2.53 & 16.93 & 13.8 & 5.81 & 0.16 & 1.03 & 0.1115 + 2001 rp137 & 2.93 & 2.36&17.90 & 16.7 & 1.72 & 0.13 & 1.02 & 0.0012 + 2004 eu91 & 2.40 & 1.80&22.04 & 18.1 & 0.51 & 0.39 & 0.76 & 0.1778 + 2004 ca105 & 2.80&2.22&18.77 & 14.7 & 2.27 & 0.44 & 1.03 & 0.0062 + 2002 rg106 & 2.69 & 2.10 & 19.58 & 16.1 & 1.30 & 0.38 & 1.045 & 0.108 + 2004 bu99 & 2.17 & 1.54 & 24.53 & 17.4 & 1.84 & 0.06 & 2.50 & 0.0008 + 2004bs160 & 2.41 & 1.81 & 21.99 & 18.6 & 0.57 & 0.20 & 1.01 & 0.0178 + 2001sb182 & 3.02 & 2.46 & 17.36 & 15.6 & 1.39 & 0.53 & 0.75 & 0.0636 + 2004fg8 & 2.01 & 1.37 & 26.61 & 18.5 & 0.74 & 0.13&1.90&0.0001 + lccc lbt & @xmath76 & @xmath16 & @xmath144 + lbt & @xmath76 & @xmath69 & @xmath145 + lbt & @xmath76 & @xmath109 & @xmath146 + fls & irac 8 & @xmath16 & @xmath147 + fls & irac 8 & @xmath69 & @xmath148 + swire & irac 8 & @xmath75 & @xmath149 + ecliptic latitude ( [ a ] , upper ) and @xmath69 ecliptic latitude ( [ b ] , lower ) . [ fig : lbt_histdist_4panel_fig1a],title="fig : " ] ecliptic latitude ( [ a ] , upper ) and @xmath69 ecliptic latitude ( [ b ] , lower ) . [ fig : lbt_histdist_4panel_fig1a],title="fig : " ] ecliptic latitude ( [ c ] , upper ) and @xmath109 ecliptic latitude ( [ d ] , lower ) . [ fig : lbt_histdist_4panel_fig1b],title="fig : " ] ecliptic latitude ( [ c ] , upper ) and @xmath109 ecliptic latitude ( [ d ] , lower ) . [ fig : lbt_histdist_4panel_fig1b],title="fig : " ] ecliptic latitude ( upper ) , at @xmath69 ecliptic latitude ( middle ) , and at @xmath150 ecliptic latitude for the swire field ( bottom ) . [ fig : spitzer_histdist_3panel_fig2],title="fig : " ] ecliptic latitude ( upper ) , at @xmath69 ecliptic latitude ( middle ) , and at @xmath150 ecliptic latitude for the swire field ( bottom ) . [ fig : spitzer_histdist_3panel_fig2],title="fig : " ] ecliptic latitude ( upper ) , at @xmath69 ecliptic latitude ( middle ) , and at @xmath150 ecliptic latitude for the swire field ( bottom ) . [ fig : spitzer_histdist_3panel_fig2],title="fig : " ]

we report on the star formation histories and extinction in the central kpc region of a sample of starburst galaxies that have similar far infrared ( fir ) , 10 m and k - band luminosities as those of the archetype starburst m82 . our study is based on new optical spectra and previously published k - band photometric data , both sampling the same area around the nucleus . model starburst spectra were synthesized as a combination of stellar populations of distinct ages formed over the hubble time , and were fitted to the observed optical spectra and k - band flux . the model is able to reproduce simultaneously the equivalent widths of emission and absorption lines , the continuum fluxes between 35007000 , the k - band and the fir flux . we require a minimum of 3 populations ( 1 ) a young population of age myr , with its corresponding nebular emission , ( 2 ) an intermediate - age population ( age 500 myr ) , and ( 3 ) an old population that forms part of the underlying disk or / and bulge population . the birthrate parameter , which is defined as the ratio of the current star formation rate to the average past rate , is found to be in the range 1 to 12 . the contribution of the old population to the k - band luminosity depends on the birthrate parameter and remains above 60% in the majority of the sample galaxies . even in the blue band , the intermediate age and old populations contribute more than 40% of the total flux in all the cases . a relatively high contribution from the old stars to the k - band nuclear flux is also apparent from the strength of the 4000 break and the k line . the extinction of the old population is found to be around half of that of the young population . the contribution to the continuum from the relatively old stars has the effect of diluting the emission equivalent widths below the values expected for young bursts . the mean dilution factors are found to be 5 and 3 for the h and h lines respectively .

astro-ph0309339

starburst galaxies harbor thousands of massive stars , that are responsible for their typical nebular spectrum and high far infrared ( fir ) emission ( weedman et al . 1981 ) . massive stars are short - lived , typically less than 10 million years ( myr ) , consequently , most of the derived properties of the starburst regions pertain to the last generation of massive stars . on the other hand , starburst galaxies have potential to form stars for periods much longer than this . for example m82 , the prototype of a starburst , has sufficient amount of gas in order to sustain star formation for a few hundreds of million years ( walter , weiss & scoville 2002 ) . but , does the star formation really continue for periods much longer than the typical age of massive stars ? by their very nature , studies based on emission lines are unable to answer this question . we need techniques that use tracers of lower mass stars , which live for several tens or even a few hundreds of million years . the optical spectrum of these older stellar systems are characterized by prominent balmer absorption lines . indeed , kim et al . ( 1995 ) found such absorption lines to be common in starburst spectra . in recent years , there have been attempts to use the information contained in the absorption lines to derive the properties of one or more assumed older bursts in individual galaxies , e.g. ngc7714 ( lancon et al . 2001 ) and ngc7679 ( gu et al . 2001 ) . stellar populations of a few hundreds of million years have been inferred in these studies . however , these stellar features are not being routinely used to derive properties such as the age and mass of the older stellar populations of a representative sample of galaxies . this is probably because of difficulties in analyzing the absorption features in the presence of nebular emission . recently , poggianti , bressan & franceschini ( 2001 ) have developed a technique based on synthetic spectra that allows investigation of many previous bursts simultaneously . their technique is capable of handling balmer emission lines even in the presence of an underlying absorption line , and vice versa , and hence is ideal for an analysis of the complete star formation history of starburst regions . the star formation history of a starburst nucleus depends on the spatial scale under investigation . observations of nearby starburst galaxies with the hubble space telescope have resolved starburst nuclei into several tens , or in some cases , a few hundreds of compact clusters known as super star clusters ( sscs ) ( oconnell et al . sscs show a considerable range in their colors , suggesting a spread in their ages . typically sscs are found in a region of a few hundred parsecs within the starburst nucleus . in m82 , regions with strong balmer absorption lines are found as far as 1 kpc from the currently active starburst site ( oconnell & mangano 1978 ) . if we were to place m82 at the distance of arp220 , an ultra luminous far infrared galaxy , all the regions discussed above would be inside an aperture of 3 . thus ground - based spectra of distant starburst nuclei would include not only the present burst , but also the outlying older bursts , if any . therefore , we need to study the star formation history of nearby starburst nuclei over a spatial scale of @xmath51kpc in order to compare the results with those of distant galaxies . star - forming regions are always associated with dust , which has the effect of reddening the observed spectrum . the amount of reddening is traditionally determined by comparing the observed balmer emission line ratios with theoretical values ( osterbrock 1989 ) , and making use of the galactic reddening curve ( cardelli et al . an alternative method to find the reddening is to compare the observed ultraviolet continuum slope of the star - forming regions with that expected for young starbursts from synthetic models ( fanelli , oconnel & thuan 1988 ) . using the above two techniques , it is now well established that the reddening suffered by the emission lines in starburst regions is significantly higher than the reddening suffered by the continuum ( fanelli et al . 1988 ; calzetti , kinney & storchi - bergmann 1994 ) . the difference in the reddening is understood to be due to the escape of stars from the parent molecular clouds as they get older ( calzetti et al . 1994 ; silva et al . 1998 ) . calzetti et al . ( 1994 ) also found that the mean attenuation curve towards the starburst regions is more gray as compared to that of the milky way . charlot & fall ( 2000 ) investigated these issues in detail and concluded that a key ingredient that is responsible for a differential reddening between lines and continuum , and a gray reddening curve , is the finite lifetime of the stellar birth clouds . granato et al . ( 2001 ) found that the starburst spectra could be modeled by even using the galactic reddening curve , when each stellar population constituting the starburst is allowed to have its own extinction . it is important to note that the low observed strength of the 2200ultraviolet bump in starburst spectra could also be produced by the models of granato et al . in this work , we study the star formation histories of the central kpc region of a sample of starburst galaxies , focusing on determining the masses of all the stellar populations that contribute to the optical spectrum . the relative contributions of the young and old populations to the k - band flux are also determined . we use the entire optical spectrum , not just a few emission and absorption lines , so that our observables are sensitive not only to the young burst , but also to all the previous bursts . the starburst model of poggianti et al . ( 2001 ) is used . the extinction of each population is a free parameter to be evaluated by the fitting procedure . in sec . 2 , we describe the sample and observations . the starburst model is explained in sec . the star formation history of the sample galaxies is discussed in sec . 4 . concluding remarks are given in sec . 5 devereux ( 1989 ) defined a complete sample of _ nearby starburst galaxies _ , that includes all nearby , 16@xmath1 d(mpc ) @xmath140 ( hubble constant of 75 kms@xmath6mpc@xmath6 ) , non - seyfert 1 galaxies with central 10@xmath0 m luminosity @xmath7 l@xmath8 , declination @xmath9 , right ascension @xmath10 , and galactic latitude @xmath11 . the sample contains 20 galaxies , whose k - band and fir luminosities are very similar to those of the prototypical starburst m82 . the fir luminosities , as estimated from 60 and 100@xmath0 m infrared astronomical satellite ( iras ) fluxes , lie between @xmath12@xmath13 l@xmath8 , which are intermediate between those of normal galaxies and ultra luminous far infrared galaxies . we carried out long - slit optical spectrophotometry of 12 of these galaxies at the guillermo haro astrophysical observatory , cananea , mexico , during two observing runs in 1999 february and december . a grating with 150 linesmm@xmath6 was used , which resulted in a resolution of about 10 , and a spectral coverage @xmath1435006800 . the spectral and spatial samplings were 3.2pix@xmath6 and 0.46pix@xmath6 respectively . slit widths were typically of @xmath15 , centered on the starburst nucleus . the slit was aligned along the east - west direction for the sample galaxies . the objects were observed as close to the meridian as possible , with maximum airmass of @xmath16 . two spectra of 30 minutes each were taken for each galaxy . the instrumental response was calibrated by the observation of the standard stars hr1544 , hr5501 , bd+40 4032 , feige 15 and feige 34 . each frame was bias - corrected and divided by a normalized flat - field using various tasks in the iraf package . two wavelength - calibrated frames of the same galaxy were averaged , in the process removing cosmic ray events . sky spectra were extracted from the object - free regions of the long - slit and were subtracted from each starburst spectrum . the spectra of the sample starbursts were extracted over a slit length of @xmath17 , which corresponds to physical scales of 0.77 to 1.76 kpc . the average signal - to - noise ratios of the extracted spectra are @xmath1450 and 20 around 5500 and 3700 respectively . we estimate that the spectrophotometry is accurate to better than 10% over the entire wavelength range covered . rlcccrrcccc 470 & sb & 30.50 & 10.13 & 14.65 & 16.62 & 37.73 & 0.39 & 0.83 & 2.31 & 3.95 + 1022 & sba & 18.50 & 10.13 & 14.15 & 28.82 & 100.84 & 0.31 & 0.70 & 3.32 & 2.91 + 2273 & sba & 28.40 & 10.03 & 13.95 & 6.84 & 16.42 & 0.35 & 1.02 & 3.74 & 0.20 + 2750 & sabc & 38.40 & 10.17 & 14.15 & 7.12 & 24.47 & 0.19 & 0.39 & 2.84 & 2.37 + 2782 & saba & 37.30 & 10.39 & 13.92 & 10.84 & 19.19 & 0.26 & 0.56 & 2.04 & 1.49 + 3504 & sabab & 26.50 & 10.47 & 13.10 & 11.04 & 23.50 & 0.24 & 0.60 & 3.30 & 2.40 + 4102 & sabb & 17.00 & 10.44 & 13.00 & 25.00 & 56.75 & 0.49 & 1.03 & 5.25 & 0.43 + 4194 & ibm & 39.10 & 10.82 & 13.55 & 18.63 & 44.27 & 0.24 & 0.48 & 3.02 & 1.17 + 4273 & sbc & 35.10 & 10.49 & 14.10 & 17.53 & 67.90 & 0.21 & 0.66 & 2.69 & 2.96 + 4385 & sb0 & 33.50 & 10.01 & 14.45 & 8.77 & 17.82 & 0.16 & 0.30 & 1.17 & 2.40 + 4818 & sabab & 21.50 & 10.26 & 13.41 & 13.77 & 78.52 & 0.55 & 1.05 & 4.65 & 3.51 + 4984 & sab0 & 21.30 & 10.01 & 12.80 & 4.19 & 26.42 & 0.59 & 1.00 & 3.53 & 1.98 + [ table1 ] the general properties of the observed starburst nuclei are presented in table [ table1 ] . the hubble types and distances to the galaxies are given in columns 2 and 3 respectively . the distances are based on the recessional velocities in the nearby galaxy catalog ( tully 1988 ) . column 4 contains the fir luminosity ( l@xmath18 ) , which was calculated from the 60 and 100 @xmath0 m iras fluxes ( helou et al . iras beams , in general , enclose the entire galaxy , and hence a part of the iras flux could originate in regions outside the nucleus . however recent high resolution mid - infrared images of starburst galaxies obtained with the infrared satellite observatory ( iso ) show that most of the iras flux originates within a spatial scale of around 1 kpc ( foerster - schreiber et al . 2003 ) , which is the slit - length used for the extraction of spectra of our sample galaxies . we estimated the v - band magnitudes in an aperture of @xmath19 ( @xmath20 ) using the compilation of multi - aperture photometric data by prugniel & heraudeau ( 1998 ) . these magnitudes are tabulated in column 5 . v - band magnitudes were also obtained from our slit spectra ( @xmath21 ) . the two magnitudes were measured with the specific purpose of providing two approximate limits on the ratio between fir and v - band luminosity ( fir / v ) . the fir / v ratios using the aperture and slit v - band fluxes are given in columns 6 and 7 respectively ( fir / v = l@xmath18/(l@xmath22 ) , where l@xmath23 is the monochromatic luminosity in the @xmath24-band ) . for compact sources , the real values are expected to be close to fir/@xmath21 , whereas for extended sources , they may be as low as fir/@xmath20 . the k - band flux inside our slit ( see section 3.3 ) , normalized to the slit v - band flux is given in column 8 . column 9 contains the b@xmath25v color obtained from the extracted spectra . the visual extinction ( a@xmath26 ) and the gas metallicity @xmath27 are given in columns 10 and 11 respectively . a@xmath26 was obtained by comparing the ratio of h@xmath3 to h@xmath4 to its case b value , making use of the galactic extinction curve ( osterbrock 1989 ) . however , the measurement of a@xmath26 in the presence of underlying absorption is not trivial ( rosa - gonzlez , terlevich & terlevich 2002 ) . the measured values are somewhat sensitive to the assumed values of the equivalent widths of the underlying absorption lines , and this can increase the uncertainty in @xmath27 . mccall , rybski , & shields ( 1985 ) demonstrated that the stellar absorption correction and the extinction for each galaxy could be estimated if the h@xmath28 line is detected in emission . they recommended a mean absorption correction of 2 to the equivalent widths of the first three lines in balmer series . however the h@xmath28 emission line is not seen in five of the sample galaxies . hence we can not follow the method used by mccall et al . ( 1985 ) uniformly for all the galaxies . alternatively , we followed an iterative method to obtain the best estimate of a@xmath26 . first , we used the observed spectra with a 2 correction to the balmer equivalent widths , to make a preliminary estimation of the nebular abundances . then the entire spectral fitting analysis , described in detail in section 3.1 , was carried out using these abundances . we then measured the visual extinction on the resulting young population model component . this component is not affected by the underlying line absorption arising from the intermediate - age populations . the abundance measurement was refined using this new value of the extinction . the strong line method was used to derive the oxygen abundance from the observed nebular lines using the calibration of mccall et al . the oxygen abundance was converted to metallicity in solar units by assuming that @xmath29 for the sun ( holweger 2001 ) . in principle , the analysis of the equivalent width ( ew ) of the hydrogen emission lines ( e.g. h@xmath4 ) provides information on the age of the most recent burst of star formation ( leitherer et al . 1999 ) , and the balmer decrement gives the extinction values . however , as discussed in the introduction , a typical slit may enclose several stellar populations of different ages and a complex distribution of dust . thus in practice , the ew method is heavily affected by the uncertainty concerning the contributions of intermediate age and old populations to the continuum ( and even line absorption ) below the emission lines . on the other hand , the observed spectral energy distributions ( seds ) contain much more information than what can be derived from a simple application of the balmer decrement method and equivalent width analysis . the physical insight can be even higher if one uses the iras fluxes and k band photometry . the fir emission provides a strong constraint on the current star formation rate ( sfr ) , though the difference in aperture sizes between the iras data and our spectra may constitute a serious problem . as for the nuclear k - band luminosity , there is a debate as to whether it originates from the old population or is provided by red supergiant stars associated with the recent star formation history ( see section 4.2 ) . in order to obtain a complete picture of star formation and extinction in starburst galaxies , we have analyzed the seds by means of the population synthesis technique . the technique adopted to model the observed spectra is an improved version of that described by poggianti et al . the star formation history over the hubble time is broken up into several episodes . each episode is described by a burst of star formation , whose rate is constant over a certain period of time . evolution of the spectra within a single episode is assumed to be negligible and hence each episode is represented by a sed of a simple stellar population ( ssp ) . a ssp is characterized by an age and metallicity . the ages of ssps are suitably selected to provide the main spectral characteristics seen in starburst spectra ( emission and absorption features and the overall continuum ) and are combined together with appropriate intensity and extinction . we used nine ssps , eight of them formed over the last 1 gyr , while the 9@xmath30 one represents the underlying old population . ages of the first eight ssps are 2 , 7 , 10 , 50 , 100 , 300 , 500 and 1000 myr . the duration of each star formation episode is chosen in such a way that the sed remains representative of all the ages those are included by the episode . the resulting durations are 4 , 4 , 7 , 60 , 125 , 200 , 350 and 1250 myr respectively . the only place where we would be using these durations is in the conversion between the stellar mass and the star formation rate . henceforth , episodes of star formation corresponding to the first three ssps will be referred to as the `` young population '' , those for the next four ssps as the `` intermediate age population '' , and the populations older than @xmath51 gyr as the `` old population '' . the `` old population '' is meant to represent the sum of the bulge and disk components that enter the slit area . for each ssp we adopt a uniform screen attenuation , using the standard extinction law for the diffuse medium in our galaxy ( r@xmath31a@xmath32/e@xmath33 , cardelli et al . while a more complex picture of the extinction can not be excluded , it was comprehensively shown by poggianti et al . ( 2001 ) that the characteristics of the emerging spectrum require a significant amount of _ foreground _ dust ( screen model ) . indeed in the case of a uniform mixture of dust and stars , increasing the obscuration does not yield a corresponding increase in the _ reddening _ of the spectrum : the latter saturates to a value e@xmath34 , which is too low to account for the observed emission line ratios ( see also calzetti et al . the extinction value a@xmath32 is allowed to vary for the different stellar populations and the extinguished spectral energy distributions of all the single generations are added up to give the total integrated spectrum . allowing the extinction of each population to vary results in a degeneracy between extinction and star formation rate , in such a way that there can be populations with very high sfr , but completely extinguished , thus contributing insignificantly to the total sed . if the extinguished population is old , then it will not even contribute to the fir emission . this degeneracy can increase somewhat arbitrarily the predicted mass to light ratio of the system . in order to avoid such extinguished populations , we automatically tested and excluded from the solution those intermediate and old populations which contribute less than 1% to the total flux at all wavelengths . although dusty young populations may not contribute significantly to the optical bands , they are important contributors to the fir emission , and hence the populations younger than 10 myr are retained irrespective of their contribution to the total flux at any wavelength . in this way , the algorithm naturally selects not more than four ssps . this is consistent with the time resolution of a general population synthesis technique , when one takes into account observational errors and uncertainties of the models . our ssps have a salpeter - like initial mass function ( imf ) n(m)dm @xmath35 dm with masses between 0.15 and 120m@xmath8 . the stellar spectral atlas is that of jacoby , hunter & christian ( 1984 ) , extended in the uv and near infrared with kurucz ( 1993 ) models as described in bressan , granato & silva ( 1998 ) . the young population spectrum includes the nebular spectrum , which is calculated by means of case b photoionization models computed with cloudy ( ferland 1996 ) with the following parameters : mass of the ionizing cluster 10@xmath36m@xmath8 , electron number density n=100@xmath37 , inner radius 15 pc and metallicity rescaled to that of the ssp . since our main interest is to model the hydrogen spectral features ( emission and absorption lines ) , changing these parameters would not affect our conclusions significantly . the gas metallicity of the majority of our sample of starburst galaxies , where the metallicity could be reliably measured , is super - solar . possible exceptions are ngc2273 and ngc4102 . the former galaxy is known to have a weak seyfert component ( devereux 1989 ) , which is possibly contaminating the [ ] @xmath385007 line , thus resulting in an under - estimation of its metallicity . ngc4102 is the most extinguished galaxy of the sample , and its low metallicity could be due to a possible over - correction to the extinction of the [ ] @xmath39 line . thus in general , the young populations are metal rich . however the metallicity of the old stars must be lower than that ; in the galactic bulge the average value is around half solar ( sadler , rich & terndrup 1996 ) . thus , except for ngc2273 , we have adopted z=2.5z@xmath8 for the intermediate - young populations and z=0.5 z@xmath8 for the old population . for ngc2273 , we have adopted z = z@xmath8 for the young stellar populations . the results do not depend critically on the adopted metallicity . to evaluate the best - fit model we minimize a merit function ( mf ) which was constructed from the ews of h@xmath40 , h@xmath4 and h@xmath3 and the relative intensities of the continuum flux in 12 almost featureless narrow spectral windows ( 35803680 , 38453880 , 39003920 , 40004060 , 41404300 , 44004600 , 47004820 , 50805180 , 54005700 , 59506200 , 63506450@xmath41 and 66306660 ) . a mf is obtained from the differences between modeled and observed values at the selected spectral windows / features , following the definition : @xmath42 where @xmath43 and @xmath44 are respectively the model and observed fluxes ( or ews ) , and @xmath45 and @xmath46 are the corresponding errors and weights . the errors in the spectral window were estimated as the rms fluctuations of the observed flux in the corresponding wavelength interval , while errors of ews are expressed as a percentage value above a minimum value of 0.1 , 0.1 and 1 , respectively for h@xmath40 , h@xmath4 and h@xmath3 . the minimization procedure applies the algorithm of adaptive simulated annealing ( ingber 1989 ) for a continuum multi - dimensional parameter space , to a merit function . it is important to stress that the algorithm of adaptive simulated annealing is able to avoid local minima and provides convergence to the global minimum . however , the uncertainties of the synthesis technique combined with the observational errors do not allow us to distinguish sub - populations within the same broad age range : young , intermediate - age , and old . for the same reason equally good fits can be obtained by making use of ssps of reasonably different metal content ( i.e. solar or half - solar for the intermediate age and old populations ) . additional constraints that may also be considered , such as the observed ratio between the fir to v - band luminosity and the k to v - band luminosity , are discussed below . all objects have published iras fluxes at 12 , 25 , 60 and 100 @xmath0 m ( devereux 1987 ) so that in principle we might constrain the total light that is absorbed in the model by comparison with the observed total infrared luminosity . by analogy with the observational quantity fir / v , defined in the previous section ( see table [ table1 ] ) , we compute a theoretical quantity , fir / v , representing the ratio between the absorbed light and the quantity f@xmath47(5550)@xmath485550 ( ergs@xmath6 ) . the absorbed light is easily obtained as the difference between the attenuated and the intrinsic light of the stellar populations that provide the best fit . however it is not easy to compare with the corresponding observational quantity , because only a fraction of the radiation emitted in the fir could be sampled by our slit , whose aperture is smaller than the iras beam and covers only the central region ( @xmath491 kpc ) of the galaxy . as discussed in the previous section , aperture problems limit the use of the fir / v values . we thus decided to consider the fir / v ratio only as an output of the model , to be compared with the values provided in table [ table1 ] . it is worth noting that a model that fits the optical data provides only a lower limit to the expected fir emission ( for the assumed extinction law ) , because of the possible presence of heavily attenuated young populations contributing to the fir , but not to the optical emission . for extinction curves that are flatter than the galactic curve , the model fir luminosities would be smaller than the values reported in this study . for all our program galaxies , devereux ( 1989 ) has provided multi - aperture k - photometry at 3.6 , 5.3 , 7.2 and 9.3 diameters . the origin of the k - band luminosity in nuclear starbursts is one of the key aspects investigated by devereux . both the old and intermediate , and even the young populations , can provide a significant contribution to the k - band luminosity . these three populations have observationally distinguishable stellar features in the optical spectrum , especially when looking at the balmer discontinuity and the prominence of hydrogen and metallic absorption lines . with our synthesis models we can thus check the consistency of the contribution of the different populations in the near infrared and in the optical , simultaneously . in order to homogenize the data by devereux with ours , we have calculated the fraction of the k - band flux that falls within our slit aperture . to this purpose we have computed the surface brightness profile that fits the multi - aperture data . the data are consistent with exponential luminosity profiles , with scale lengths between 11.5 . this suggests that the source of the k - band luminosity is the stellar disk rather than a very compact source . the maximum luminosity allowed for a compact source is @xmath50% of that observed within the 3.6 aperture . from the fitted surface brightness profile , we calculated the k - band flux that falls within our slit aperture , and then computed the corresponding ratios with the visual flux of our spectra , f@xmath51/f@xmath32 . the resulting values are tabulated in column 8 of table [ table1 ] . the observed spectra of the sample galaxies are shown in figure 1 . these spectra indicate a wide range of optical properties in spite of having nearly similar nir and fir luminosities . spectra are arranged according to their colors in order to easily visualize the gradual change from blue ( ngc4385 ) to red ( ngc4818 ) starbursts . the most noticeable characteristic of these spectra is that the h@xmath3 and h@xmath4 lines are seen in emission , while the higher order lines of the balmer series are seen in absorption . this characteristic behaviour was also present in the sample of very luminous infrared galaxies analyzed by poggianti et al . the transition from emission to absorption happens at h@xmath40 , with 11 of the 12 galaxies having the absorption feature stronger than the emission line . the only exception is the wolf - rayet galaxy ngc4385 ( mrk52 ) . besides , several metallic lines , notably the h & k lines are also seen in absorption , even in ngc4385 . the simultaneous presence of absorption and emission lines is clearly an indication that more than one population contributes to the observed spectrum . to disentangle the contribution of the different stellar components , and also to investigate the role of dust in reddening the spectrum , especially in red starbursts , we have applied the model described in the previous section to our starburst spectra . in order to highlight the success of the method , we summarize here the salient characteristics of the different stellar populations . the balmer emission lines suggest the presence of a young population rich in ob stars . the intermediate age stellar populations are characterized by the prominence of hydrogen absorption lines and by a strong gradient around 3700 . they do not show the discontinuity at 4000 ( bruzual 1983 ) , which is instead a characteristic feature of the old populations . between 3700 and 4000 the spectral shape of the intermediate age stellar populations is thus very different from that of the old populations . at wavelengths longer than 4000 , the continuum of the intermediate age populations is characterized by the absence of typical absorption bands ( e.g. mg and g - band ) shown by the old populations ( age @xmath52 1 gyr ) . thus , a significant contribution of the old populations not only changes the equivalent width of the hydrogen lines ( emission or absorption ) but also modifies the shape of the continuum between 3700 and 4000 and introduces visible features in the optical region . another noticeable characteristic of the old populations is the depth of the k line at @xmath53 3934 . this line is almost absent in young and intermediate age populations . it becomes prominent at ages greater than a few gyrs and thus it is a strong indicator of the dominance of the old populations . on the contrary the h line is blended with the h@xmath54 line and may appear strong even for intermediate age populations . in the present modeling , we have not used the k line as a constraint , because of lack of information on its behaviour at different metallicities . indeed the atmosphere models used to compute ssps are based on solar metallicity stars and only the indirect effect of metallicity on the turnoff of the isochrone can be modeled . recent advances in atmosphere models will allow the inclusion of this constraint in the near future ( rodriguez - merino et al . however , though not included as a constraint , we may notice that the k and h+h@xmath54 absorption lines are fairly well reproduced by the models , providing a further strong support to the validity of the solution . in summary , when a good fit to the data is reached , the solution is unique as far as the three broad age ranges are considered . more specifically , while almost equally good solutions can be obtained by interchanging the different populations within a given episode ( young , intermediate or old ) , it is not possible to find a consistent solution by suppressing or adding an entire episode of star formation . rcccccccccrcr 470 & 1.78&0.73&4.46&0.66 & 1.91 & 0.49&300&0.39 & 0.40&21.88 & 0.37 & 4.35 + 1022 & 0.95&1.09&2.04&0.62 & 0.39 & 0.28&50 & 0.12 & 0.25&27.83 & 0.33 & 1.21 + 2273 & 3.77&1.28&0.15&0.00 & 0.34 & 0.00&50 & 1.62 & 0.55&22.86 & 0.45 & 16.26 + 2750 & 2.95&0.98&1.90&0.00 & 0.35 & 0.00&50 & 0.52 & 0.33&17.00 & 0.23 & 5.25 + 2782 & 4.26&0.65&3.48&0.00 & 0.52 & 0.00&50 & 2.07 & 0.51&12.49 & 0.32 & 20.73 + 3504 & 8.24&1.14&2.62&0.00 & 0.30 & 0.00&100&1.06 & 0.44&34.08 & 0.29 & 10.65 + 4102 & 7.25&1.78&0.00&2.53 & 0.10 & 0.00&50 & 1.01 & 0.66&100.18&0.70 & 10.10 + 4194&18.32&1.05&5.49&0.00&0.45 & 0.00&300&1.51 & 0.53&47.42&0.29 & 15.34 + 4273 & 0.58&0.92&0.19&0.00&0.09 & 0.00&50 & 0.18 & 0.41&15.90 & 0.28 & 1.79 + 4385 & 1.22&0.35&1.95&0.18&0.32 & 0.04&50 & 0.21 & 0.13 & 9.86 & 0.16 & 2.16 + 4818 & 5.08&1.59&0.08&0.00&0.02 & 0.00&300&0.43 & 0.68&154.38&0.72 & 4.34 + 4984 & 0.77&1.27&0.99&0.37&0.24 & 0.15&300&0.70 & 0.68&18.25 & 0.57 & 7.09 + [ table2 ] for purpose of conciseness , we show in figure [ f2 ] only three model fits , while the results for all the galaxies are presented in table 2 . ngc4385 and ngc4984 are selected as representative blue and red starbursts , respectively . ngc1022 represents an intermediate case . for each galaxy , the bottom panel of the figure shows the best fit model ( lines extending shortward of 3500 ) superimposed on the observed spectrum . the spectra are normalized to their @xmath24-band luminosity . the residuals ( observed @xmath25 model ) and the estimated errors at the 14 fitted points are shown below the corresponding features . the residuals on the h@xmath3 , h@xmath4 and h@xmath40 line bands are identified by circled error bars . the histogram in the top panel represents the star formation history ( in m@xmath8yr@xmath6 ) with the relevant sfrs on the left axis . the filled square at each histogram position represents the color excess @xmath55 for each population . the @xmath55 scale is shown on the right axis . models reproduce the observed spectra fairly well , as can be judged by the low residuals at the fitted continuum bands . the fits were even able to reproduce the strength of some of the spectral features , such as the h and k absorption lines , that were not included in the merit function . the relatively large residuals and errors on h@xmath4 and h@xmath40 lines are due to their small equivalent widths caused by the stellar absorption lines almost wiping out the nebular emission lines . observed ( left column ) and model ews of h@xmath3 , h@xmath4 and h@xmath40 are written on the plots . we did not try to fit the emission lines of [ ] and [ ] , because their intensities depend strongly on the physical conditions in addition to the strength of the stellar populations . for each starburst , the model has found four populations two young populations , one intermediate and one old population , each with its distinct mass ( or sfr ) and extinction . two young populations may represent either two short - duration bursts , one older than the other , a continuous star formation lasting for more than 4 myr , or two bursts of nearly the same age , but with vastly differing extinction . the results of the fit for all the galaxies are presented in table 2 . columns 210 contain the model derived sfr and the reddening @xmath55 for the four selected populations . for the intermediate - age population , we also tabulate the age of the best - fitted ssp in column 8 . model derived ratios , fir / v and f@xmath51/f@xmath32 are given in columns 11 and 12 respectively . the last column contains the total mass sampled by the slit with the adopted imf . a comparison of the computed fir / v values and the observed values ( columns 6 and 7 of table 1 ) , indicates that the two values are in the same range , taking into account the differences in the optical and fir apertures , discussed in section 3.2 . it may be recalled that in very luminous infrared galaxies , which on average are 10 times more luminous than our galaxies , observed values are systematically higher than the model values , suggesting a significant amount of star formation hidden from the optical view ( poggianti et al . the fact that the observed values are comparable with the predictions of the model in our sample of galaxies , implies that most of the fir emission detected by iras originates in the starburst nucleus and that the starburst regions do not contain a significant amount of optically hidden star - forming regions . an insight into the relative importance of the young , intermediate - age and old populations in the overall history of nuclear star - formation can be obtained by analyzing the spectra of individual populations . in figure [ mod ] , we plot the best - fit model spectrum along with the spectra of individual populations for the three representative galaxies . observed k - band photometry is also displayed in these plots . note that the plotted spectra are attenuated by the extinction values obtained for each population . in the following sub - sections , we discuss in detail the mass and extinction of the different populations . in the presence of selective extinction , deriving masses of different populations is not straight forward . while the intrinsic mass - to - light ratio is the basic ingredient of any population synthesis technique , the option to vary attenuation for each population , introduces a new degree of freedom . unreasonable high masses can be obtained by letting the attenuation to increase with the strength of an intermediate or old stellar population . the algorithm described in section [ models ] automatically copes with this degeneracy by selecting the minimum mass - to - light ratio solution . mass , rather than sfr , is the fundamental output of the model . the sfr of each population , tabulated in table 2 , is obtained by dividing the mass by the assumed duration of star formation ( see section 3.1 ) . the derived mass of each population is plotted as a bar histogram in the upper panel of figure [ hmas ] . for each galaxy , the histograms from left to right denote the masses of young , intermediate - age and old populations . it can be seen that the mass of the young population is comparable to that of the intermediate - age population in most cases , whereas the mass of the old population is significantly higher . the mass of the young population is plotted against the old stellar mass in figure [ momy ] . the dotted lines in this plot delimit the zone where the young stellar mass is 0.1% and 1% of the old stellar mass . all the points lie between these two limits . this indicates that the recent starburst contributes a tiny amount to the overall visible mass of the nucleus . it is interesting to note that corresponding fraction is 5% in blue compact galaxies ( prez - gonzlez et al . the inferred total stellar masses of the starburst regions are in the range of masses that are usually derived for the galactic bulges . the ratio between the current star formation rate ( in the last 8 myr ) and the average past star formation rate , sfr/@xmath2sfr@xmath52 , often referred to as the birthrate parameter ( kennicutt , tamblyn & congdon 1994 ) , is a useful parameter to measure the strength of the most recent burst . the value of this ratio is less than 0.1 for normal early - type disk galaxies , which are the hubble types that dominate our sample . our sample of starbursts have birthrate parameter in the range 112 , with an average value of @[email protected] . this reiterates that the galaxies of our sample are indeed starburst galaxies . all starburst regions are found to have an intermediate - age stellar population . unfortunately , it is difficult to distinguish differences in spectral features of an evolving starburst between @xmath515 and 500 myr . an evolved spectra of an intense short - duration burst will be indistinguishable from a lower intensity continuous star formation process . there can be three possible origins for the intermediate - age population : ( 1 ) it represents stars formed earlier in the current star formation episode , ( 2 ) it represents the exponential tail of the disk star formation history , ( 3 ) it represents an episode of star formation similar to the present one , but that took place around a few hundred million years ago . in the following paragraphs we discuss these possibilities in more detail . in majority of our sample galaxies , the best - fit ssp that represents the intermediate age population has an age of 50 myr ( see column 8 in table 2 ) . considering that the sed almost remains unchanged between 1575 myr , the intermediate age population could be as young as 15 myr . case ( 1 ) would then imply that the star formation is going on un - interrupted for at least 15 myr . the histograms of figure 2 , suggests that there is a gap in star formation between 8 and 15 myr . this gap corresponds to the populations of red supergiants , which are also represented by the 7 myr ssp . hence equally good solutions could have been obtained by extending the duration of the young burst upto 15 myr . the average present sfrs , in that case , would be a factor of two lower than those presented in table 2 . if the sfr remains constant , then the mass of the processed stars scales proportionately with the duration of the burst . we have seen in the previous section that the mass of the intermediate age population is comparable to that of the young population . this implies that the duration of the intermediate age population is comparable to that of the young population ( @xmath57 myr ) . the duration would be even smaller if the sfrs were higher at earlier times . hence the maximum duration of the present episode for a continuous star formation model is @xmath530 myr . if case ( 2 ) above is correct , then the masses of the populations would be proportional to the life - time of the stars constituting the populations . this would imply that the mass of the intermediate - age population is around 1% of that of the old population . star formation rate in the disks of early - type galaxies decreases exponentially with time and the expected mass would be even less . the derived masses of the intermediate - age populations are indeed @xmath58% of that of the old population , and hence we can not rule out the possibility ( 2 ) . the third possibility , case ( 3 ) , is also plausible because the mass of the intermediate - age population is comparable with that of the present burst . perhaps a way to distinguish the case ( 2 ) from the other two cases is to look at the spatial distributions of the stellar populations . in case ( 2 ) , stellar populations are expected to be uniformly distributed whereas in cases ( 1 ) and ( 3 ) old star clusters are expected . the fact that sscs are frequently encountered in starburst regions and that some of them are of intermediate age and spatially separated from the present active knot , makes us believe that case ( 3 ) is the most likely possibility . it is also possible that all the three processes of formation of intermediate - age population are at work in starburst regions . the relative importance of the three processes might vary from one galaxy to another . the origin of the k - band luminosity in nuclear starbursts has been the subject of debate ever since the first starburst sample was defined by balzano ( 1983 ) . she found that the near infrared colors as well as the k - band luminosities , were unaffected by the starburst activity . this can be seen in table 3 , where we have tabulated balzano s h@xmath25k and j@xmath25k colors of starburst and bright ngc galaxies in columns 2 and 3 respectively . nir colors of starburst nuclei are indistinguishable from those of the nuclei of normal galaxies . cizdziel et al . ( 1985 ) supported balzano s findings , by analyzing the k - band growth curves of starburst nuclei , which indicated a lack of central concentration . on the other - hand , rieke et al . ( 1980 , 1993 ) interpreted the entire k - band luminosity in the archetype starburst m82 , as due to the red supergiants associated with the present starburst activity . our sample of starburst galaxies is well - suited to resolve the above controversy . devereux ( 1989 ) , from where our sample was chosen , discussed the existence of a correlation between the k - band and the 10@xmath0 m luminosities , for galaxies that are luminous at 10@xmath0 m . since 10@xmath0 m emission is related to the recent star formation , this correlation implies that the k - band emission is also related to the recent star formation . on the contrary , by means of our detailed population synthesis technique , we find that , in the same galaxies showing the correlation between 10@xmath0 m and k - band luminosity , a major fraction of the k - band emission comes from the old population . a strong contribution to the k - band luminosity from the young and/or intermediate age stellar populations is excluded by the overall fit of the broad continuum features and inspection of some metallic lines ( e.g. k ) of the optical spectrum . to better clarify this point , we plot in figure [ lksb ] , the fractional contribution of the old stars to the total k - band luminosity against the birthrate parameter . we also show a comparison with theoretical predictions based on simple stellar population , assuming that the old stars have an average age of 10gyr . calculations were done for two ages of the young component : 10myr and 5myr , which represent phases of starburst with and without the red supergiants respectively . the metallicities for the young and old components were assumed to be the same . the age of the young component and the metallicity for the different plotted lines are : a ) 10myr and z=2.5z@xmath8 ( solid line ) ; b ) 10myr and z = z@xmath8/5 ( dotted line ) ; c ) 5myr and z=2.5z@xmath8 ( short dashed line ) ; d ) 5myr and z = z@xmath8/5 ( long dashed line ) , respectively . as expected , the contribution of the old population decreases with increasing birthrate , but only for a birthrate parameter larger than about 7 this contribution falls below 50% of the total light . we have only two cases where this contribution is less than 50% , ngc470 ( 42% ) and ngc1022 ( 38% ) . ngc1022 is among the galaxies with the lowest sfrs in our sample ( 1m@xmath8yr@xmath6 ) , and its high birthrate is due to the low mass of the old population . this is to stress once more that the relative contributions of the young and the old populations to the k - band depends on the birthrate parameter rather than the present sfr . the figure also illustrates the dependence of the contribution of the old population , on the age of the dominant young population and the metallicity . for a given birthrate , the contribution of the old population is higher for lower age and lower metallicity of the young population . it can be inferred from the theoretical curves in the plot that for young bursts which are yet to evolve into the red supergiant phase , more than 80% of the k - band flux originates from old stars , unless the birthrate parameter is much higher than the plotted range . notice that a lower contribution from the old population in ngc1022 may also be inferred from the relatively low strength of the k line ( fig . 1 ) . further insight about the nature of the k - band emission may be obtained by analysing the radial intensity profiles . as discussed in sec . 3.3 , the devereux multi - aperture k - photometry , is consistent with exponential luminosity profiles ( scale - lengths between 11.5 ) , with the compact nucleus not contributing more than 50% within the 3.6 aperture . the contribution of the compact nucleus would be even less if the real intensity profiles of program galaxies are @xmath59 bulge profiles rather than the exponential ones that we have assumed . this disagrees with devereux claim of `` central concentration '' of k - band flux , in these galaxies . we however find an evidence of systematic reddening of v@xmath25k color towards the center , which is expected in all disk galaxies due to the flattening of the disk at shorter wavelengths due to optical depth effects in _ dusty _ disks ( burstein , haynes & faber 1991 ; evans 1994 ) . in the light of this new analysis , we may ask what is the origin of the correlation found between the k - band luminosity and the 10@xmath0 m luminosity , in these galaxies . figure [ momy ] helps us to clarify this point . in this figure , the mass of the young component , which is directly proportional to the star formation rate , is compared with the mass of the old component , essentially the total mass of the bulge . there is a trend for the more massive galaxies to host a stronger nuclear starburst . thus , the relation between the k - band and the 10@xmath0 m luminosity can be essentially ascribed to the underlying relation between the total galaxy mass ( hence the mass of the old component and the k - band luminosity ) and the gas mass ( hence star formation rate and 10@xmath0 m luminosity ) . this explanation is also supported by the fact that devereux s m82-like starburst sample galaxies are earlier than sb ( and hence relatively bigger ) , while the mean hubble type of balzano s sample is sc . on the other hand the observed lack of high 10@xmath0 m emission in faint k - band sources is basically due to the lower gas mass ( not fraction ) of low mass galaxies . thus the m82-like starburst sample of devereux contains early type star - forming galaxies . ccccc @xmath60 & 0.32 @xmath560.12 & 0.31@xmath61 & 0.34 @xmath560.05 & 0.51 @xmath560.17 + @xmath62 & 1.02 @xmath560.12 & 1.04@xmath61 & 1.08 @xmath560.08 & 1.05 @xmath560.14 + [ table4 ] interestingly , the @xmath62 and @xmath60 colors of the models that best fit the observed spectra agree excellently with those reported for balzano s sample of starburst galaxies ( balzano 1983 ) . this is demonstrated in table 3 . model colors of the total and only the young burst are given in column 4 and 5 respectively . it can be seen that the observed @xmath60 colors of starburst galaxies are indistinguishable from those of non - star - forming ngc galaxies . thus most of the k - band luminosity of our starburst nuclei , seems to originate from old disk / bulge stars . however , we expect that in starbursts with a birthrate parameter that exceeds the average value found in our sample ( about 5 ) , the young stars begin to dominate the nir fluxes . this may be the case of high redshift starbursts with similar sfrs but with a significantly lower amount of old stars . one of the aspects where our approach to spectral decomposition differs from the rest ( e.g. lancon et al . 2001 ) is in the treatment of extinction . in our starburst models , the extinction of each population is a free parameter , whose value is chosen to be consistent with both the continuum shape and the balmer line ratios . in addition , the derived extinction does not depend on the extent of contamination by the underlying absorption , as the old populations are simultaneously determined in the models . the dots above the histograms in figure 2 indicate the reddening value of each population . the reddening values , @xmath55 , of each population are also given in table 2 . it can be seen that the reddening of the young populations ( age @xmath18 myr ) is always higher than that of intermediate - age and old populations . poggianti et al . ( 2001 ) also found a similar result for a sample of very luminous infrared galaxies ( vlirgs ) . the reason for the higher extinction of the young population is well - known the young population is still embedded in its parent molecular clouds , which are known to be dusty ( silva et al . 1998 ) . among the two young populations , the extinction of the youngest population is systematically higher . this may suggest that stars are already escaping from their parent molecular clouds at around 4 myr . the ratio of the extinctions of the old to the young population is plotted against the extinction of the young population in figure [ ebv ] . the latter value is calculated from the summed model spectra of the two youngest populations ( see description of table 1 in section 2 for more details ) . the dotted line in the figure represents the average value of the ratio for the sample galaxies , which is equal to @xmath63 . note that the rms of the ratio is only 0.13 mag , in spite of a range of more than 4 mag for the extinction of the young population . thus the observed extinction towards the old stars is proportional to the extinction towards the young stars . such a correlation is not expected if all the dust that causes the extinction of the old stars is mixed with the stars . moreover the @xmath55 of the old population in all cases , except ngc4385 , is higher than 0.18 mag , the limiting reddening for a uniform mixture of dust and stars ( poggianti et al . all these arguments suggest that the same clouds that cause the extinction of the young population , are the principal contributors to the extinction of the old stars . this could be understood under a geometry in which the young star - forming regions are surrounded by molecular clouds , and the majority of the old ( bulge / disk ) stars are distributed uniformly around these clouds . only the light from the old stars that are behind the cloud would suffer high extinction , whereas the light from the rest of the stars would have a maximum reddening of @xmath55=0.18 mag . the net extinction of the old population happens to be 43% of that of the young population . galaxies with a ratio considerably lower than the mean ratio ( e.g. ngc1022 ) may be those where the young star - forming region is located at the far - side of the bulge / disk , and vice versa . analyzing a sample of starburst nuclei , calzetti et al . ( 1994 ) noticed that the continuum near h@xmath3 and h@xmath4 suffered about half the attenuation as suffered by the nebular lines . our result would be identical to that obtained by calzetti et al . , if the near ultraviolet and optical continua are predominantly contributed by the old stars , i.e. stars that do not supply ionizing photons . we have demonstrated that this is indeed the case in previous subsections . hence our detailed analysis reinforces the interpretation put forward by calzetti et al . ( 1994 ) that the reason for a lower continuum extinction as compared to the extinction of nebular lines is the relatively lower extinction towards the older stars as compared to extinction towards the young massive stars . our analysis also suggests , for instance , that the extinction ratio determined from the optical spectrum can not be simply extrapolated to the ultraviolet , because of the different contribution of the old and the young populations at these wavelengths ( see e.g. panuzzo et al . 2003 , for a thorough discussion of this problem ) . in most cases , we find the reddening of the intermediate - age population to be lower than that of the old population . is it a real effect or is it due to some inherent limitation of the model used ? in order to address this question , we go back to figure 3 , where the typical contributions of the three populations are plotted . it can be seen that the wavelength range where the intermediate - age population contributes , if at all , covers less than 500 between the blue limit ( @xmath64 ) and h@xmath40 . in this range , there are only two galaxies ( ngc470 and ngc1022 ) , in which the intermediate - age population is the dominant contributor . significantly , the derived extinctions of the intermediate and old populations compare well in these two cases . in the rest of the galaxies , the available baseline is too small to reliably estimate the reddening . the amount of intermediate age population in most galaxies is basically determined from the observed strength of the h@xmath40 line . as our criterion was to choose models that have minimum extinction ( or equivalently minimum mass - to - light ratio ) , the models are able to give only a lower limit to the extinction of the intermediate - age population . a fair estimate of the mass of the intermediate - age population may be obtained by assuming that it has the same reddening as that of the old population . we find that this mostly affects those cases where the mass of the intermediate - age population is significantly less than that of the young population . in such cases , the masses estimated this way turn out to be almost equal to those of the young population . in rest of the cases , the masses increase by around 40% . the flux ratios and equivalent widths of balmer emission lines are often used as diagnostics of extinction and age of the most recent burst in starburst regions . the existence of previous generations of stars , if not taken into account , can seriously affect the determination of both the extinction and age of the young burst . older populations affect in two ways ( a ) the intermediate - age populations have strong balmer absorption lines in their spectra , which can decrease the flux of the emission lines , ( b ) old population can contribute significantly to the continuum , which can lower the emission equivalent width . the latter effect is often referred to as dilution . for an accurate determination of the extinction value it is necessary to correct for the contamination from underlying stellar absorption lines , and for the age determination it is necessary to know the dilution factor , @xmath65 , which is the ratio of the observed continuum flux to that of the young burst . the following equations are used to calculate the true emission line flux , @xmath66 and equivalent width , @xmath67 , from the observed equivalent width , @xmath68 , and continuum flux , @xmath69 : @xmath70 @xmath71 @xmath72 is the correction term , which mccall et al . ( 1985 ) found to be 2 for the prominent balmer lines . mccall et al . determined this value empirically for a large number of regions , by forcing the extinction corrected ratios of the first three balmer lines to their photo - ionization values . on the other hand , veilleux et al . ( 1995 ) measured the absorption ews of h@xmath4 using simultaneous fitting of the emission and absorption lines for a large sample of infrared luminous galaxies . the measured values ranged from 13 , with a median value of 2 . the decomposition of the observed spectra into the spectra of individual populations , allows us to make an independent determination of @xmath72 . @xmath72 depends on the equivalent width and the relative continuum contribution of intermediate age and old populations . we summed up the spectra of intermediate age and old stellar populations to get a pure absorption - line dominated spectrum for each of our starburst nucleus . we then measured the ew and continuum fluxes for the first four lines of the balmer series in both the total and absorption - line dominated spectrum . @xmath72 and the dilution factor @xmath65 are then simply given by @xmath73 and @xmath74 where @xmath75 , and @xmath76 are the equivalent width and the continuum flux of the old population alone . table 4 contains the statistical properties of @xmath72 and @xmath65 for the four balmer lines . the mean underlying absorption correction to h@xmath3 is @xmath77 , whereas for the other three lines , it is close to @xmath78 . the errors on the ews are the rms deviations for the sample , and not errors in the measurements . hence the difference in the mean values for h@xmath3and the rest of the lines is significant , although the average values are within the quoted errors . on the other hand , it is a practice to use a correction of 2 for all the lines , following mccall et al . it may be recalled that mccall et al . had forced the correction to be the same for all the balmer lines . in this work , we show that this assumption is not strictly true , and the average value of @xmath79 obtained by mccall et al . is a compromise value for all the lines . we emphasize that in a given galaxy correction to h@xmath3 is always smaller than that for the other three lines . the average value of correction could be sample - dependent . this is because the exact numerical value of the correction depends on the relative contribution of the intermediate age population , which could be different in different star - forming regions . for instance , kennicutt ( 1992 ) found a mean value as high as 5 for the integrated spectra of galaxies . we believe the average values obtained by us are representative of starburst nuclei of nearby galaxies and hence we recommend the use of these values , for a reliable estimation of a@xmath26 and metallicity from nebular lines in starburst nuclei . the mean dilution factors of the h@xmath3 and h@xmath4 ews are @xmath80 and @xmath81 , respectively for our sample of starburst nuclei . observed ews of starburst nuclei are generally lower than those expected for young bursts ( e.g. veilleux et al . 1995 ) . the analysis carried out in our work supports the idea that an underlying old population is the primary cause for the observed low values of the balmer ews in starburst nuclei . lcccccccc average & 1.57 & 2.48 & 2.32 & 2.49 & 4.7 & 3.2 & 2.7 & 2.4 + rms & 0.26 & 0.60 & 0.74 & 0.91 & 1.9 & 1.2 & 0.9 & 0.8 + min & 1.15 & 1.48 & 1.20 & 1.22 & 2.3 & 1.6 & 1.4 & 1.3 + max & 2.11 & 3.33 & 3.28 & 3.81 & 22.5 & 31.3 & 40.2 & 46.2 + we have analyzed optical spectra of a sample of starburst galaxies that have their far infrared , 10@xmath0 m , and k - band luminosities similar to those of the prototype starburst m82 . we have adopted a theoretical approach that allows us to resolve their star formation history into three major episodes , the on - going star formation ( @xmath82 myr ) , the intermediate age population ( 50500 myr ) and the old populations ( older than 1 gyr ) . the extinction of these three main populations is a free parameter and is fixed , together with the star formation history , by the simultaneous fitting of the ews of the main hydrogen lines and a number of narrow bands in the continuum . these data are complemented by the k - band flux , estimated for the area within our slit from existing aperture photometry . the fir emission could also be used to constrain the model , but its uncertain contribution within our slit prevented us to use it as a firm constraint . all galaxies have clear signatures of the above three main episodes of star formation , with birthrate parameter ( current sfr over average past value ) , between 1 and 12 . the mass of the intermediate population is generally comparable to the mass formed during the current burst and , together , they provide only a small contribution to the total mass sampled by our spectra . the spectra sample approximately the central 1 kpc region , and the mass within this region turns out to be between 10@xmath83m@xmath8 and 2@xmath4810@xmath84m@xmath8 , with an average value @xmath498@xmath4810@xmath83m@xmath8 . these values are within the mass range found for bulges of spiral galaxies . there is a clear tendency for the most massive bulges to host a stronger starburst activity . the mass of the young component increases linearly with that of the old bulge : m@[email protected]@xmath86 . the relation between the mass of the young and old component is the most likely cause of the correlation reported by devereux ( 1989 ) between the 10@xmath0 m luminosity , which is proportional to the present sfr ( i.e. the mas of the young component ) and the k - band luminosity , which is proportional to the old component . this correlation led devereux to conclude that the k - band luminosity is mainly produced by red supergiant stars of the young component . on the contrary , our population synthesis technique allows us to exclude this possibility . in the majority of the cases , the old population contributes more than 60% of the flux to the k - band , and along with the intermediate - age population it contributes more than 40% even to the blue continuum . test cases show that it is not possible to reproduce the broad features of the continuum and in particular the depth of the k line by suppressing the older population , even by assuming a higher metal content . hence , in most of the m82-like starburst galaxies , the k - band luminosity can be used as an estimate of stellar mass . however , when the birthrate parameter exceeds the average value of our sample galaxies ( @xmath495 ) , the young populations start contributing a significant fraction to the nir luminosity . this should be taken into account when trying to determine , from the restframe optical - nir luminosity , the mass of the old component in high redshift galaxies , where at a given sfr the birthrate parameter is likely to be larger . the regions with on - going star formation have systematically higher extinction than the older populations : a@xmath26(old)/a@xmath26(young)@xmath87 . the fact that the continuum of starburst nuclei is mainly contributed by the old population , makes this result identical to that obtained by calzetti et al . ( 1994 ) , who found a@xmath26(continuum)/a@xmath26(emission line)=0.50 . it was not possible to reliably determine the extinction of the intermediate - age population due to the small base wavelength over which this population dominates , if at all , in the sample nuclei . the decomposition of the observed spectra into spectra of individual populations , allowed us to determine the underlying absorption ews for balmer lines . the resulting mean value for h@xmath3 is @xmath77 , and around @xmath78 for the rest of the balmer lines , where the errors represent the real variations from galaxy to galaxy and are not due to measurement errors . hence in a given galaxy correction to h@xmath3 is always smaller than that for the other three lines . these values are improvements over the values suggested by mccall et al . ( 1985 ) , who obtained the correction by forcing it to be the same for the first three balmer lines . as expected , the value of @xmath79 that was obtained by mccall et al . is intermediate between that of h@xmath3 and the rest of the balmer lines . the continuum from the old population plays an important role in reducing ( or diluting ) the observed emission ews below the values expected for a purely young burst . we find that the mean dilution factors for h@xmath3 and h@xmath4 are @xmath80 and @xmath81 , respectively . we thank bianca poggianti , pasquale panuzzo , alberto franceschini and roberto terlevich for thoughtful discussions . we also thank the referee for several useful comments , which have enabled us to improve the original manuscript . this work was partly supported by the conacyt projects 39714-f , 36547-e and j37680-e . a. b. thanks inaoe for its kind hospitality during his visits . this research has made use of the nasa / ipac extragalactic database , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . granato , g. l. , silva , l. , bressan , a. , lacey , c. g. , baugh , c. m. , cole , s. , & frenk , c. s. 2001 , astrophysics and space science supplement , 277 , 79 gu , q. s. , huang , j. h. , de diego , j. a. , dultzin - hacyan , d. , lei , s. j. & bentez , e. 2001 , a&a , 374 , 932

the influence of nuclear isospin mixing on parity - violating elastic electron scattering is studied for the even - even , nucleic , mg ,si , ands . their ground - state wave functions have been obtained using a self - consistent axially - symmetric mean - field approximation with density - dependent effective two - body skyrme interactions . some differences from previous shell - model calculations appear for the isovector coulomb form factors which play a role in determining the parity - violating asymmetry . to gain an understanding of how these differences arise , the results have been expanded in a spherical harmonic oscillator basis . results are obtained not only within the plane - wave born approximation , but also using the distorted - wave born approximation for comparison with potential future experimental studies of parity - violating electron scattering . to this end , for each nucleus the focus is placed on kinematic ranges where the signal ( isospin - mixing effects on the parity - violating asymmetry ) and the experimental figure - of - merit are maximized . strangeness contributions to the asymmetry are also briefly discussed , since they and the isospin mixing contributions may play comparable roles for the nuclei being studied at the low momentum transfers of interest in the present work .

0806.0552

one motivation for studying parity - violating ( pv ) electron scattering from nuclei is to use it as a tool to extract information on the weak neutral current ( wnc ) and thereby to test the validity of the standard model in the low - energy regime @xcite ( see also @xcite ) . this possibility lies in the fact that the pv asymmetry acquires a very simple , model - independent expression in terms of basic coupling constants , with nuclear structure effects cancelling out , if certain conditions are met . the assumptions required to arrive to such a simple expression are : ( 1 ) that the focus is placed on elastic scattering from spin - zero nuclear targets , for then only coulomb - type monopole form factors enter ; ( 2 ) that strangeness content in the wnc can be neglected , for then only isoscalar and isovector matrix elements occur , with no third type of contribution ; and ( 3 ) that the nuclear ground states have isospin zero , permitting only a single coulomb monopole matrix element to occur , namely , the isoscalar one . in fact , while restriction ( 1 ) can be met with a wide range of even - even nuclei , restrictions ( 2 ) and ( 3 ) are not completely attainable . strangeness content in the wnc has been extensively studied in previous work ( see , for instance , the review article @xcite ) and is not the primary focus of the present study . instead , the present work is aimed at the last issue of potential isospin mixing in nuclear ground states and how this affects the pv asymmetry , albeit taking into account present uncertainties in the strangeness content of the nucleon . isospin mixing constituted part of the goal of a previous study by donnelly , dubach and sick ( dds ) @xcite where the effect of isospin mixing in nuclear ground states on pv electron scattering was first studied . the approach taken there was to use a simple two - level model in which the @xmath5 ground state and an excited @xmath6 state , both with angular momentum / parity 0@xmath7 , were admixed with a mixing parameter @xmath8 . the deviation in the pv asymmetry due to isospin mixing was found to be proportional to this mixing parameter @xmath8 and to the ratio between the inelastic isovector and the elastic isoscalar coulomb monopole form factors . in dds a shell - model approach with only one active shell in a spherical harmonic oscillator basis was used to obtain the spin - isospin reduced coulomb matrix elements ( isoscalar and isovector ) for @xmath1c and @xmath3si . coulomb distortion effects on the electron waves were also neglected in that earlier work . isospin dependence in pv electron scattering is crucial in attempting to determine the precision up to which the standard model constants can be deduced and to what extent strangeness effects in the wnc can be studied . at the same time , this dependence can be exploited to provide information about the spatial distribution of neutrons in the nuclear ground state . indeed , this idea was proposed as part of the original study by dds , namely that a measurement of the electron - scattering pv - asymmetry can provide a direct measurement of the fourier transform of the neutron density . in a subsequent study @xcite the idea in dds was extended to allow for coulomb distortion of the electron wave function , something also done in the present work . neutron densities are less well known than charge densities ( studied using parity - conserving electron scattering ) as most of this information comes from hadronic probes where the reaction mechanism involved is more difficult to interpret than is the case for semi - leptonic electroweak processes . thus , an electroweak measurement of the neutron density can serve to calibrate the measurements made with hadronic probes . in fact , the idea proposed by dds is now being realized in the parity radius experiment ( prex ) at jefferson laboratory which has the goal of measuring the neutron radius of @xmath9pb using pv elastic electron scattering . such a measurement has implications for astrophysics ( including the structure of neutron stars ) , for atomic parity non - conservation studies , for the structure of neutron - rich nuclei , and for determinations of neutron skin thicknesses of nuclei . in this work we extend the previous study undertaken by dds to examine the effects on the pv asymmetry induced by isospin admixtures in the nuclear ground states of n = z nuclei occurring when coulomb interactions between nucleons are taken into account and pairing correlations are included . the nuclear structure is described within a self - consistent axially - symmetric hartree - fock formalism with skyrme forces ; in fact , three different forces are used to explore the sensitivity of the pv asymmetry deviation to the nuclear dynamics . several additional extensions beyond the work of dds are made , namely , inclusion of the spin - orbit correction to the coulomb monopole operators , use of modern electroweak form factors for protons and neutrons , inclusion of strangeness contributions in the nucleon form factors , and , as noted above , incorporation of the effects caused by the coulomb distortion of the electron wave functions . the present study includes four n = z isotopes , namely , @xmath1c , @xmath2 mg , @xmath3si , and @xmath4s . with respect to nuclear structure issues , the main difference between the treatment in dds and here is that in dds coulomb effects in the nuclear ground state are considered perturbatively to find the admixture of giant isovector monopole strength ( @xmath10 , @xmath6 ) , as explained at length in @xcite . here , within the self - consistent mean - field approach with two - body density - dependent effective interactions , the collective isospin mixing effect of the coulomb force is included non - perturbatively in the isospin - non - conserving hartree - fock ( hf ) mean field , along with other collective effects such as pairing and deformation . as a result , the hf+bcs ground state is made up of quasiparticles with rather complex admixtures of harmonic oscillator wave functions in many different major shells . in discussing the results one sees that there are limited kinematical regions where the pv asymmetry is measurable at the levels required , _ i.e. , _ as characterized by the figure - of - merit ( fom ) which typically peaks at momentum transfers around 0.5 @xmath11 . for the region extending up to a little beyond 1 @xmath11 the effects of isospin mixing on the pv asymmetry will be seen later to be characteristically at the level of a few percent for light nuclei , up to more than 10% for the heavier nuclei considered . these are certainly within the scope of existing and future measurements of pv electron scattering . for example , the happex - he measurements at jlab @xcite have already been performed at low @xmath12 , _ i.e. , _ for kinematics similar to those of relevance in the present work . these yielded pv asymmetries at the 4% level ( which are statistically , not systematically limited at present ) , while the future prex measurements on pb @xcite aim for 3% precision in the pv asymmetry . in fact , in recent work ( see @xcite and references therein ) it has been noted that interpretations of results from experiments like happex and g0 are reaching the point where they are limited by effects from isospin mixing . for heavier nuclei , as discussed in the present work , the effects from isospin mixing will be seen to be even larger than for the nucleon or few - body nuclei . at higher momentum transfers the effects of isospin mixing are still larger ; however , the fom is smaller in this region and , as a consequence , our attention in the present work is focused only on the lower momentum transfer region . as far as other contributions are concerned , after discussing our basic results , and thereby setting the scale of the isospin mixing , we return briefly to show the size of potential strangeness effects and of the expected influence of meson - exchange currents at these lower momentum transfers . other effects such as parity admixtures in the nuclear ground state , dispersion corrections , or radiative corrections are expected to provide negligible modifications to the asymmetry @xcite and are not considered in this work . the outline of the paper is the following . in sect . ii we start by introducing the formalism necessary to describe pv in polarized elastic electron scattering from spin - zero n = z nuclei . the correction to the asymmetry due to isospin mixing is isolated and analyzed in terms of the various ingredients involved , such as the roles played by the nucleon form factors and by potential strangeness contributions . there we also discuss the effects expected from electron distortions . in sect . iii we present our results on the pv asymmetry for the four nuclei under study . first we explore the influence on the results of the various ingredients in the formalism ( different pairing gaps , different skyrme interactions , the spin - orbit correction ) and go on to compare the present work with past studies of isospin - mixing effects in pv electron scattering . following this , the principal results of the study are presented for the four chosen nuclei and finally , in sect . iv , we summarize the main conclusions of our work . polarized single - arm electron scattering from unpolarized nuclei can be used to study parity violation , since both electromagnetic ( em ) and weak interactions contribute to the process via @xmath13 and @xmath14 exchange , respectively . the pv asymmetry is given by @xcite @xmath15 where @xmath16 is the cross section for electrons longitudinally polarized parallel ( antiparallel ) to their momentum . keeping only the square of the photon - exchange amplitude for the spin - averaged em cross section and using the interference between the @xmath13 and @xmath17 amplitudes in the cross section difference , in plane wave born approximation ( pwba ) the asymmetry @xmath18 in the standard model can be written as @xmath19 where @xmath20 and @xmath21 are the fermi and fine - structure coupling constants , respectively , and @xmath22 is the four - momentum transfer in the scattering process . @xmath23 is the pv response and @xmath24 is the em form factor , both containing the dependence of the asymmetry on the nuclear structure . in the case of elastic electron scattering between @xmath25 states , only the coulomb - type monopole operators can induce the transition , @xmath26 where @xmath27 , @xmath28 is a kinematical factor that cancels out in the ratio , and @xmath29 is the em(wnc ) monopole coulomb form factor . then one has @xmath30 if we now consider only @xmath0 nuclei and assume that they are isospin eigenstates with isospin zero in their ground states , then only isoscalar matrix elements contribute and the weak and em form factors become proportional : @xmath31 accordingly , the pv asymmetry does not depend on the form factors : @xmath32 a_a \beta ^{(0)}_v \cong 3.22 \times 10^{-6 } |q^2|\ ; \text{(in fm}^{-2}\text { ) } \ , , \label{referencevalue}\ ] ] where , within the standard model , @xmath33 , @xmath34 being the weak mixing angle . the actual pv asymmetry deviates from this constant value by a correction @xmath35 , where @xmath36 or equivalently ( for @xmath37=0@xmath7 nuclei ) @xmath38 and it accounts , in particular , for the effects of nuclear isospin mixing and strangeness content in the pv asymmetry . the ratio between the wnc and em form factors can be written as @xmath39 where the subscript in parenthesis indicates isoscalar ( @xmath5 ) and isovector ( @xmath6 ) parts . the coulomb isoscalar ( @xmath5 ) and isovector ( @xmath6 ) multipole operators can be written in terms of the contributions of order 0 and 1 in @xmath40 ( subscripts ( 0 ) and ( 1 ) ) @xcite : @xmath41 the nucleon form factors @xmath42 and @xmath43 are included in the definitions of the two contributions as follows : @xmath44 \theta^{m_j}_j(q\textbf{x})\ , , \label{operatormso}\end{aligned}\ ] ] where the kinematical factors @xmath45 and @xmath46 contain the energy @xmath47 and momentum @xmath12 transferred to the nucleus by the electron : @xmath48 and @xmath49 are the electric ( magnetic ) nucleon form factors discussed later . the basic multipole operators @xmath50 and @xmath51 are respectively the standard zeroth - order coulomb operator and the spin - orbit first - order correction , and are defined as : @xmath52 \ , .\end{aligned}\ ] ] by using the relations between form factors in isospin space ( @xmath53 ) and form factors in charge space ( proton : p and neutron : n ) , @xmath54 one can write the ratio in eq . ( [ fft01 ] ) as @xmath55 where the operators with tildes , @xmath56 , have the same structure as in eqs . ( [ operatorm ] , [ operatorm2 ] ) , but contain the wnc nucleon form factors @xmath57 , @xmath58 to be defined in the next subsection . in the present study we are interested in the @xmath59 ( @xmath60=0 ) multipole . the charge operator @xmath61 matrix elements in a spherical harmonic oscillator ( s.h.o . ) basis are given by @xmath62 for @xmath59 the spin - orbit operator @xmath63 is just proportional to @xmath64 and one has in the same s.h.o . basis : @xmath65 \:\bigg\langle n'l'j ' \bigg| \frac{j_1(qr)}{qr } \bigg| nlj \bigg\rangle \ : \delta_{l'l } \:\delta_{j'j } = \\\nonumber & & -\frac{1}{\sqrt{4\pi } } \:\frac{1}{2 } \left[j(j+1)-l(l+1)-\frac{3}{4}\right ] \ : \int \frac{j_1(qr)}{qr } \ : r_{nl}(r)r_{n'l}(r ) \ : r^2 \ : dr \ , .\end{aligned}\ ] ] in both expressions @xmath66 is a spherical bessel function of order @xmath67 . the matrix elements of the charge monopole operators ( em and wnc ) between two s.h.o . wave functions are therefore @xmath68 \langle n'lj | \theta^0_0(q\textbf{x } ) | nlj \rangle \,\end{aligned}\ ] ] @xmath69 \langle n'lj | \theta^0_0(q\textbf{x } ) | nlj \rangle \ , , \end{aligned}\ ] ] where the proton ( @xmath70 or @xmath71 ) and neutron ( @xmath72 or @xmath73 ) form factors @xmath20 and @xmath74 will be defined in the next subsection . note that one should not confuse @xmath75 as a superscript on @xmath76 and @xmath77 meaning neutron with @xmath75 as a subscript meaning the quantum number labeling the single - particle basis states the context will also make the distinction clear . the expressions above are used once the s.h.o . expansion of the hf single - particle wave functions is performed ( see below ) . the total monopole matrix elements also contain a center - of - mass correction factor to account for the fact that the hf single - particle wave functions are referred to the center of the mean - field potential , not to the nuclear center of mass , as they should be to avoid the spurious movement of the nucleus as a whole . for a s.h.o . potential , this correction factor takes the form @xmath78 , with @xmath12 the momentum transfer ( in @xmath11 ) , @xmath79 the oscillator parameter ( in fm ) and @xmath80 the total number of nucleons . this is usually employed even when not using a s.h.o . basis ; the correction cancels out when one constructs the ratios of form factors appearing in the asymmetry @xmath18 . finally , let us conclude this section with a brief discussion of potential meson - exchange current effects on the pv asymmetry ( see also @xcite ) . such contributions enter in very different ways in cross sections and in the asymmetry . in the former they typically provide effects at the few percent level , whereas for the latter the situation is not so obvious . first , were there to be no significant isospin mixing and strangeness contributions , _ i.e. , _ with a purely isoscalar , no - strangeness ground state , then the mec effects would cancel in the ratio that gives the pv asymmetry , and , in fact , the result would be completely unaffected by any aspect of hadronic structure . this is an old result @xcite from the earliest studies of elastic pv electron scattering from nuclei . secondly , with isospin mixing the situation is more complicated , since now there are both isoscalar and isovector electromagnetic current matrix elements involved and mec effects are different for the two . however , two arguments suggest that these contributions are not significant for the present study at low momentum transfers : the multipoles involved are c0 ( isoscalar and isovector ) where mec effects , being dominantly transverse , are suppressed anyway and the effect is one of modifying the predicted isovector effect from occurring as a purely one - body matrix element to one where both one - body and typically few percent two - body contributions occur . this would modify the isospin - mixing pv results by a multiplicative ( not additive ) factor of typically a few percent , which is not significant for the present study . thirdly , with strangeness also present , now in the nucleon itself and also in mec contributions , additional effects may occur . a study was done @xcite for the important case of @xmath81he and , based on the results presented there , it may be that at higher values of @xmath12 ( beyond roughly 3 @xmath11 ) such effects could become significant . on the other hand , in the cited study such effects were shown to be negligible for the range of momentum transfers of interest in the present work and accordingly here we also neglect any contributions of this type . the electric nucleon form factor can be expressed either in terms of protons and neutrons or in terms of isoscalar and isovector parts , @xmath82 where @xmath83 is 1 for protons and @xmath84 for neutrons . in analogy , in the standard model , the wnc nucleon form factor is given by @xmath85 where the strangeness contribution to the form factor is isoscalar . this can also be written in terms of proton and neutron form factors , @xmath86 where @xmath87 a similar result applies to the nucleon magnetic form factors by simply substituting g@xmath88 by g@xmath89 in the previous expressions . the nucleon form factors @xmath42 and @xmath43 , from which @xmath90 and @xmath91 are obtained , have been computed using the parametrization by hhler @xcite . within the standard model we have , @xmath92 the strangeness form factor @xmath93 has been parametrized according to @xmath94 ( see , for instance , @xcite ) with @xmath95 the parameters @xmath96 and @xmath97 are constrained by pv electron scattering measurements on hydrogen , deuterium and helium-4 ; the values chosen as representative are discussed below and in sect . iii . the strangeness term in the wnc form factors in eq . ( [ wncff ] ) can be considered separately , giving rise to a decomposition of the pv asymmetry deviation , @xmath98 , where the isospin mixing term @xmath99 is proportional to @xmath100 and @xmath101 is proportional to @xmath102 . the isospin mixing piece is computed considering @xmath103 and @xmath104 in the wnc form factors , whereas the strangeness term is computed considering @xmath105 , @xmath106 , @xmath107 and @xmath108 in the wnc form factors . we evaluate the strangeness contribution neglecting the small differences between neutron and proton densities , assuming g@xmath109=0 , and neglecting the small spin - orbit contribution so only @xmath110 enters and not @xmath111 . this gives @xmath112 once we have introduced the nucleon form factors , it is interesting to notice that if one neglects the spin - orbit correction to the coulomb operator , eq . ( [ operatormso ] ) , one obtains the more intuitive expression @xmath113 with @xmath114 where @xmath115 are the ground - state radial densities for protons and neutrons with @xmath116 as above . if one further neglects the electric neutron form factor @xmath117 and the strangeness form factors @xmath110 , one arrives at the simple expression @xmath118 or equivalently @xmath119 which are similar to those used in dds . as a general consideration , one faces a compromise between optimizing the pv signal ( _ i.e. , _ deviations in the asymmetry ) and the figure - of - merit , discussed below . since the asymmetry generally increases with @xmath120 , while the cross section decreases , this leads to the search for a well defined region of optimal kinematics . to measure properly effects in the pv asymmetry due to isospin mixing , the deviation should be greater than or of the order of a few percent of the reference value @xmath121 given in eq . ( [ referencevalue ] ) . this condition will be fulfilled in general for some intervals of @xmath12 , but not for others . at the same time the relative error of the asymmetry should be kept as small as possible . this relative error depends on technical characteristics of the experimental setting ( detector solid angle , beam luminosity and running time ) gathered in @xmath122 , as well as on intrinsic properties of the target , the projectile and their kinematics gathered in the so called figure - of - merit ( fom ) @xmath123 ( see @xcite ) : @xmath124 clearly , in a study of the present type where the focus is placed on theoretical aspects of the hadronic and nuclear many - body physics in the problem , it is not appropriate to go into further detail on specific experimental conditions , _ i.e. , _ on the specifics of @xmath122 ( see , however , @xcite for some general discussions of experimental aspects of pv electron scattering ) . presenting the fom , as we do in the following section , provides a sufficient measure of the `` doability '' of pv elastic scattering asymmetry measurements whose goal is to explore isospin mixing in the nuclear ground state . in particular , the fom is proportional to the asymmetry itself squared and to the differential cross section of the scattering ( essentially the parity - conserving cross section ) : @xmath125 and therefore shows a diffraction pattern , but with the same negative - slope trend as the differential cross section when the momentum transfer increases , which favors low momentum transfer experiments . to generate the ground - state wave function in the hf+bcs approximation we use a skyrme type density - dependent nucleon - nucleon interaction ( sly4 force @xcite ) , allowing for axially - symmetric deformation . first , the hf equations are solved to generate the self - consistent hf mean field , and pairing correlations are taken into account at each iteration solving the bcs equations to generate the single quasiparticle wave functions . the deformed hf+bcs calculation gives a set of single - particle levels , occupation numbers , and wave functions @xmath126 . the latter are expanded in a deformed harmonic oscillator basis @xmath127 which we also expand in a s.h.o . basis @xmath128 to facilitate the comparison with previous work based on the shell - model approach , @xmath129 where a truncation of 11 major shells @xmath130 has been used in both the deformed ( @xmath131 ) and the spherical ( @xmath132 ) expansions . each hf single - particle state @xmath133 has a parity and an angular momentum projection @xmath134 which are shared by any of the basis wave functions taking part in its expansion ( @xmath135 , @xmath136 ) . the hf+bcs outputs give for every state @xmath133 the energy , the occupation probability @xmath137 and the coefficients @xmath138 . the coefficients in the s.h.o . basis expansion @xmath139 are obtained as @xmath140 with @xmath141 and @xmath142 where the @xmath143 are associated legendre polynomials and the functions @xmath144 are defined in terms of hermite and generalized laguerre polynomials that contain , respectively , the cylindrical @xmath145 and @xmath146 dependence of the eigenstates @xmath147 of the deformed harmonic oscillator potential @xcite . the contribution of each pair of s.h.o . states to the total charge monopole form factors ( em or wnc ) can be analyzed in terms of two pieces as follows : @xmath148 @xmath149 applicable to protons and neutrons ( @xmath150 ) separately . the first factor in the above equations is the matrix element of the charge monopole operator ( em or wnc ) between two s.h.o . states as defined in eq . ( [ ff_basis_em ] ) or ( [ ff_basis_wnc ] ) , which is momentum - dependent and whose @xmath151-dependence is only due to the spin - orbit correction ( eq . ( [ m_spinorbit ] ) ) . this matrix element vanishes unless both basis functions have the same @xmath152 quantum numbers . the nuclear ground - state structure information is contained in the second factor , which is the spherical part of the hf+bcs ground - state density matrix ( for protons or neutrons ) in the s.h.o . basis : @xmath153 it contains the coefficients of pairs of spherical harmonic oscillator components differing at most in the radial quantum number @xmath75 , as well as the occupation probabilities @xmath154 of each of the hf single - particle states @xmath133 . these contributions are added up so that the calculated density refers to the whole hf+bcs ground state of the nucleus . the spherical density matrix elements contain information on the nuclear ground - state structure of the target isotopes . we note that the total hf+bcs density matrix may also contain a non - spherical part , with matrix elements @xmath155 , which for deformed nuclei is nonzero and which does not contribute to the charge monopole form factors . this ensures that only the spherically symmetric ( @xmath59 ) part of the neutron and proton densities contributes and that there is no need to make any further angular momentum projection in the deformed nuclei @xcite . the analysis of the quantities defined above is especially appropriate when comparing results of our hf+bcs calculations with former shell - model results , since each quasiparticle state can always be expressed as a combination of s.h.o . basis states with different radial quantum numbers @xmath75 . this fact allows for nonzero spherical density matrix elements @xmath156 with different @xmath75 and @xmath157 , off - diagonal in the spherical harmonic oscillator basis . when the coulomb interaction is included in the generation of the self - consistent mean field , the spherical density matrix in eq . ( [ density ] ) is slightly different for protons and neutrons . this amounts to saying that the self - consistent ground - state mean field is not a pure @xmath5 isospin state , but contains isospin admixtures , mainly @xmath6 @xcite , which contribute to the pv @xmath158 amplitude . in other words , from the proton and neutron densities @xmath159 , @xmath150 , one may construct an isoscalar ground - state density @xmath160 and an isovector ground - state density @xmath161 that contribute respectively to the isoscalar and isovector parts of the pv amplitude . another aspect that must be addressed is the coulomb distortion of the incoming and outgoing electron wave functions . while we are dealing with relatively light nuclei for which coulomb distortion effects are generally small ( at least where the fom is significant ) , it is important to perform calculations which fully take into account coulomb distortion to obtain realistic predictions that can serve as a reference for future experiments . it is also interesting to compare these results with the pwba calculations described in previous sections . to this end , we follow the standard treatment of coulomb distortion for elastic pv electron scattering within a partial wave formalism . we solve the dirac equation for massless electrons in the coulomb potential generated by the nucleus , closely following the work of refs . @xcite to obtain the distorted wave ( dw ) results . in fig . [ xs ] we compare our unpolarized elastic electron scattering cross sections with experimental data for @xmath1c @xcite , for @xmath2 mg @xcite , for @xmath3si @xcite and for @xmath4s @xcite . for ease of presentation , the cross sections have been transported to a common energy of 400 mev . our calculations of cross sections have been obtained from the hf+bcs ground - state densities as described in the previous section . in obtaining these results , no experimental information on , for instance , charge radii , has been employed or fit . in spite of this fact , very good agreement is found up to a scattering angle of 60@xmath162 , _ i.e. , _ a transfer momentum up to around 2 @xmath11 , which is the region of interest according to the results on fom to be shown below . to consider the distorted - wave correction in the pv asymmetries , eq . ( [ asymmetry_sigmas ] ) has been used , together with eq . ( [ asymmetryratio ] ) for the asymmetry deviations . the effects of coulomb distortion and nuclear isospin mixing can be analyzed separately by considering different ingredients in the calculation of the asymmetries in eq . ( [ asymmetryratio ] ) . in fig . [ asym_gamma_dw_pw ] we show the typical distortion effects for 1 gev electrons in the case of @xmath3si . the left - hand panel shows isospin - mixing ( superscript i ) asymmetries in pwba and when distortions of the electron wave function are taken into account , where the smoothing of the pwba divergences appears as the most obvious effect of distortion . one must be aware of the fact that without isospin effects and distortion of the electrons , the wnc and em form factors in eq . ( [ hadronicratio ] ) are exactly proportional , and thus the asymmetry follows a very simple @xmath120 dependence ( shown by the long dashed line in the left - hand panel of the figure ) , even at the zeros of the form factor corresponding to the diffraction minima . when isospin effects are considered , the diffraction minima for the wnc and em form factors occur at slightly different values of @xmath12 and thus the asymmetry shows extreme variations at the approximate locations of these diffraction minima , as can be seen in the left - hand panel of fig . [ asym_gamma_dw_pw ] ( dashed curve ) . the main effect of the distortion of the electron waves for these light systems is to fill in or to smooth out the diffraction minima , thus severely reducing the amplitude changes of the asymmetry at these diffraction minima , as can also be seen ( solid curves ) in fig . [ asym_gamma_dw_pw ] . furthermore , the distorting potential introduces asymmetry deviations even when no isospin mixing effects are present in the calculation of the nuclear structure . in summary , when nuclear isospin mixing is not considered , _ i.e. , _ fixing the same proton and neutron densities , the pwba result shows simply a @xmath120 behaviour , while the results that take into account the distortion of the electron wave function deviate smoothly from this with a dip where the diffraction minimum occurs . when isospin mixing is included , the asymmetry is increased with respect to the non - isospin - mixing case as @xmath12 approaches the value at the diffraction minima , and it is reduced after the diffraction minima . this behavior is seen both in the pwba and in the dw calculations . in the same fig . [ asym_gamma_dw_pw ] , the plots in the right - hand panel show asymmetry deviations @xmath163 due to the nuclear isospin mixing obtained replacing @xmath164 in eq . ( [ asymmetryratio ] ) by the ratios @xmath165 and @xmath166 . in addition we show the pure effect of distortion ( ignoring nuclear isospin mixing ) obtained from the ratio @xmath167 and the combined effect of distortion and isospin mixing yielded by using the ratio @xmath168 . as one can see in the figure , the effect of distortion is to smooth out the divergences appearing at the position of diffraction minima in pwba , but anywhere else one sees that the deviations in the asymmetry introduced by isospin mixing effects are very similar both in pwba and when the distortions are fully taken into account . analogous results are obtained for @xmath1c , @xmath2 mg and @xmath4s . one should notice that , since we are plotting the absolute value of the asymmetry deviation on a logarithmic scale , zeros of this function appear as downward divergences which remain in the dw calculation . in the next section we focus the discussion on the results corresponding to the asymmetry deviation . this is defined as the ratio whose numerator is the difference between the asymmetry with isospin mixing and without , and whose denominator is simply the asymmetry with no isospin mixing , where all asymmetries are from full distorted - wave calculations . that is , one has @xmath169 this deviation may be directly compared with experiment . it can be seen that this generalized @xmath170 yields results that are similar to the pwba prediction , except for the regions deep in the minima of the cross section . taking into account that @xmath171 can be computed accurately in a model - independent way ( tables can be produced for each nucleus and different electron energies without difficulty ) , and the only inputs needed are the ( experimental ) charge distributions , which are well known for the nuclei studied here , we conclude that the distortion effects will not prevent the determination of isospin mixing effects , provided that the data are compared with a full dw calculation . indeed , the organization of the asymmetry data according to @xmath170 introduced in the equation above may be enough to permit a direct comparison with the simple pwba predictions for @xmath35 . it is worth pointing out that the filling of the minima of the cross section is important for the determination of the asymmetry and of the deviations from the non - isospin - mixing prediction , as it is precisely the region near the minima where the isospin mixing effects on the asymmetry are also more evident . the filling of the minima of the form factor and cross section induced by coulomb distortion will help to make these details of the asymmetry more easily measurable , as the cross section will be much larger in the region of the minima than predicted with the pwba calculation . four even - even , n = z isotopes have been studied in this work , namely @xmath1c , @xmath2 mg , @xmath3si and @xmath4s , which for each element are the most abundant isotopes ( 79@xmath172 abundance for @xmath2 mg , and higher than 90@xmath172 for the other isotopes ) . furthermore , all targets have reasonably large excitation energies of the first ( 2@xmath7 ) excited state , and all but sulfur are suitable in elemental form for high - current electron scattering experiments , as required for a measurement of the pv asymmetry . in our calculations the ground - state deformation is obtained self - consistently . for each nucleus the calculations have been performed using the optimum values of the axially - symmetric harmonic oscillator parameters @xmath79 and @xmath12 @xcite , which define the average length and the axes ratio of the oscillator well . the self - consistent intrinsic proton quadrupole moments are shown in table [ tabledef ] for three different sets of pairing gaps : @xmath173 = 0 mev ( absence of pairing ) ; @xmath173 = 1 mev ; and @xmath173 as obtained from experimental mass differences @xcite of neighboring nuclei through a symmetric five - term formula . these values of the pairing gaps are practically the same for protons and neutrons , namely 4.5 mev for @xmath1c , 3.13 mev for @xmath2 mg , 2.88 mev for @xmath3si , and 2.17 mev for @xmath4s . since these isotopes are relatively light , these pairing gaps are too large because shell effects are not taken into account in the mass formula . the larger pairing gaps tend to give smaller self - consistent deformations . for @xmath1c and @xmath4s we obtain spherical equilibrium shapes in agreement with their vibrational experimental spectra . table [ tabledef ] also includes experimental intrinsic charge quadrupole moments @xcite . for @xmath2 mg and @xmath3si , whose experimental spectra show a rotational band that confirms their deformed shape , the self - consistent values of the quadrupole moment are in better agreement with experiment for the smaller pairing gap values . for this reason , in what follows we focus on results obtained for @xmath173 = 1 mev . values of @xmath174 @xmath175 and @xmath176= 52.15 @xmath175 are reported for @xmath1c and @xmath4s respectively in @xcite . however , in these two isotopes a ground - state rotational band is by no means recognizable @xcite , implying that their intrinsic quadrupole ground - state values should be zero and that the ones appearing in @xcite do actually correspond to their vibrational 2@xmath7 states . the interpretation of @xmath1c in terms of a 3@xmath21 structure is gaining more and more experimental support @xcite . obviously this picture is not reducible to a single determinantal function . however , we note the fact that the hf+bcs energy which we obtain as a function of quadrupole deformation shows a shallow minimum around @xmath22 = 0 for @xmath1c , and this is consistent with that picture . in any case , the final results for pv asymmetries are not very sensitive to the nuclear deformation . this was tested by computing the pv asymmetry for two different shapes of @xmath4s , spherical ( @xmath177 0 @xmath175 , which is the self - consistent deformation ) and prolate ( @xmath177 50 @xmath175 , which has been fixed by means of a quadrupole constraint in the hf calculation ) . the results for the two cases were found to be extremely similar , which shows that our predictions are quite independent of the nuclear shape as long as the corresponding self - consistent densities are properly used in the calculations . in the figures to follow we present the results of isospin mixing for the four isotopes under study , using a sly4 hf+bcs mean field with @xmath173=1 mev . for comparison , although this is not the primary focus of the present study , we also show results with / without strangeness present , and accordingly a few words are in order concerning what should be regarded to be a sensible value for @xmath178 . the experimental value for @xmath96 in eq . ( [ str_para_1 ] ) is still evolving , given the appearance of new experiments , more detailed analyses of the combined experimental results ( see _ e.g. , _ @xcite ) and considerations of complications such as isospin mixing in the nucleon @xcite and @xmath81he @xcite and the role of two - photon exchange @xcite . presently , from pv electron scattering measurements involving @xmath179h and @xmath81he , the value of @xmath96 is consistent with zero , and accordingly in the figures we show isospin - mixing results computed with @xmath180 . the strongest constraint on electric strangeness to date comes from the happex - he experiment at jlab @xcite which yielded the result @xmath181 at @xmath1821.4 @xmath11 ( corresponding to @xmath183 ( gev / c)@xmath184 ) , which translates into a range for the electric strangeness parameter used in our parametrization ( see eqs . ( [ str_para_1 ] ) and ( [ str_para_2 ] ) ) of @xmath185 ; to be conservative we have simply added statistical and systematic errors here . accordingly , in the present study , when referencing our new results to existing estimates of electric strangeness we have adopted the two limit values @xmath186 as a rough measure when strangeness is present , and @xmath180 when it is not all three of them are essentially consistent with existing knowledge on electric strangeness . of course , if ( when ) the amount of electric strangeness is better known from pv electron scattering studies of @xmath179h and @xmath81he the present analysis can also easily be updated , both via more refined parametrizations of @xmath110 and in concert with the isospin mixing effects that provide the main focus of the present work . with reference to the results shown below , one should note that the happex - he pv measurements have been performed in comparable low-@xmath12 kinematical regions . for direct comparison with @xmath163 we use the @xmath178 of eq . ( [ gamma_s ] ) . note that when @xmath96 is negative the effects of isospin mixing and strangeness tend to cancel in the low-@xmath12 region ( @xmath187 1 @xmath11 ) , whereas the opposite would be true were @xmath96 to be positive , which is not ruled out at present . clearly , in the case where no significant electric strangeness is found in the nucleon the curves with isospin mixing and @xmath188 are the appropriate ones . it should also be noted that the strangeness contribution is not very sensitive to details of the nuclear structure ( in particular , the approximation of eq . ( [ gamma_s ] ) , which is the one plotted in the graphs , is completely independent of the nuclear structure ) , whereas the isospin contribution does depend on the details of the nuclear structure . for magnetic strangeness ( see eqs . ( [ str_para_1 ] ) and ( [ str_para_2 ] ) ) following @xcite we have assumed that @xmath189 , although the effects in this case are very small and choosing any value that is consistent with present experimental knowledge would lead to negligible differences . we now begin our discussions of some of the main results of our work , focusing here on an exploration of the basic assumptions made in the present approach , using the case of @xmath3si to illustrate the various effects under study . in fig . [ gamma_iso_strange_si28 ] the left - hand panel shows the total pv asymmetry deviation @xmath35 using the form factor ratio in eq . ( [ ffpn ] ) , together with the isospin mixing contribution @xmath163 and the strangeness contribution @xmath178 discussed above ( as well as below , where the full set of nuclei are inter - related ) . discussion of the right - hand panel , which shows the fom as well as a comparison of plane- and distorted - wave results , is also postponed to later where results for the full set of nuclei are presented . at that point the `` doability '' of measurements of elastic pv electron scattering is discussed in a little more detail . however , before proceeding with those general results , let us comment on some of the specifics , using the case of @xmath3si as prototypical . first , the effect of using different values for the pairing gap in obtaining the asymmetry deviation is shown in fig . [ gamma_pairing_si28 ] , where one can see that the results change very little , even for the larger pairing gap values , as long as the proton and neutron gap values are the same . in conclusion , our results on the pv asymmetry , as shown in the figure , are very stable against changes in the proton and neutron pairing gaps , provided both have the same value . if the proton and neutron pairing gaps were to be notably different , as in one of the examples of fig . [ gamma_pairing_si28 ] , this would lead to a very different structure of the proton and neutron distributions which translates into a different shape for the pv asymmetry deviation . secondly , let us discuss the sensitivity of the results to the use of different skyrme interactions . in fig . [ gamma_forces_si28 ] results for @xmath3si are shown for three different choices , sly4 , sk3 and sg2 . sly4 force @xcite is an example of a recent parametrization and sk3 @xcite is an example of a simple and old force . @xcite has been also widely used in the literature and provides a good description of bulk and spin - isospin nuclear properties as well . clearly in the low-@xmath12 region of interest in the present work there is very little sensitivity to the choice of force . the small spread in @xmath35 seen in the figure may be taken as a sort of theory uncertainty with which other competing effects can be compared . thirdly , consider the spin - orbit contribution which was introduced as a correction in eq . ( [ operatorm ] ) . while expected to be a small correction , it is important to make sure that this contribution does not confuse the interpretation of the pv asymmetry in terms of isospin mixing . typically when treating parity - conserving elastic electron scattering the spin - orbit contribution is dominated by isoscalar effects and is known to provide a small correction . however , since the isospin mixing involves a delicate interplay between isoscalar and isovector matrix elements and since the isovector spin - orbit contribution in particular may be sizable ( see eq . ( [ operatormso ] ) ) : the isovector / isoscalar form factor ratio there is @xmath190/[2 g_m^{(0)}-g_e^{(0 ) } ] \cong 11 $ ] ) , this contribution has also been included in the present work for completeness . results obtained for the pv asymmetry deviation in @xmath3si performed with and without the spin - orbit correction are almost indistinguishable for momentum transfers lower than about 3.5 @xmath11 , except near the diffraction minima where the spin - orbit contribution becomes ( fractionally ) significant and in the dips where @xmath191 is small anyway . one concludes that , at least in those regions away from the peaks seen in @xmath191 , one should not expect the spin - orbit contributions to lead to confusion when attempting to extract the isospin - mixing behaviour of the pv asymmetry . fourthly , a comparison between previous shell - model calculations and our results is now in order . by examining , for instance , figs . 3 or 7 in dds together with the present results we immediately see that the effects of isospin mixing are considerably larger here . in the shell - model calculation of dds , the value of the diagonal density matrix elements for those states below the active shell was the maximum possible occupation of the states in the isoscalar case and zero in the isovector case . for the states in the active shell , only the sum of the density matrix elements was fixed , but not the individual values . all the off - diagonal density matrix elements vanished in the shell - model calculation , but not in the hf calculation for the reasons mentioned above . these off - diagonal matrix elements are the main contributors to the total isovector form factors obtained in the hf case . in particular , it is of interest to trace back which contributions of the isovector spherical density matrix elements ( eq . ( [ density ] ) ) contain the largest major shell mixings . to illustrate this analysis we show in table [ multipolar ] the spherical isovector density matrix elements ( diagonal and off - diagonal ) for different ( @xmath192 ) contributions of @xmath3si . it is found that the off - diagonal isovector density between the harmonic oscillator levels 0d@xmath193 and 1d@xmath193 , _ i.e. , _ @xmath194 , has the largest value , followed by @xmath195 mixings in p@xmath196 and p@xmath197 ( @xmath198 and @xmath199 respectively ) . all of these are normalized to the value of the largest element , @xmath200 . the relative weight of each off - diagonal contribution to the total isovector form factor can change as the momentum transfer varies , and therefore it is necessary in this case to compare the quantities @xmath201 . these are also shown in the table for @xmath12 = 0.1 , 0.5 and 1 @xmath11 . since the coulomb monopole operator between two s.h.o . wave functions @xmath202 is proportional to @xmath12 for off - diagonal contributions , they tend to vanish as @xmath203 . but for @xmath204 , although still relatively small , off - diagonal contributions recover their prominent status determined by the densities . now let us turn to general results at low momentum transfers shown in figs . [ gamma_iso_strange_si28 ] and [ gamma_iso_strange_c12][comp_gammai_dw ] for the full set of four nuclei . in all cases the left - hand panels show results for the total pv asymmetry deviation @xmath191 in pwba for three values of electric strangeness : results with no electric strangeness ( solid curve ) , with @xmath205 ( labeled @xmath206 ) and with @xmath207 ( labeled @xmath208 ) . the strangeness contribution itself ( with @xmath209 ) is shown for reference as a dotted curve . in the right - hand panels the @xmath191 obtained in plane - wave and distorted - wave cases are compared ( solid dark and light lines , respectively ) and the fom obtained either at a fixed scattering angle of @xmath210 = 10@xmath211 ( in pwba ) or at fixed incident energy @xmath212 gev ( in dw ) ( solid and dashed lines , respectively ) is presented . the former gives a region around the peak where the fom reaches reasonable values for the pv deviation to be measured , whereas the latter just shows that the smaller the scattering angle , the larger the fom . when @xmath213 the fom with fixed scattering angle behaves as @xmath214 , whereas the fom with fixed incident energy behaves as a constant , this different behavior for @xmath213 being apparent in the figures . to set the scale of `` doability '' we refer back to our earlier discussions : the ability to measure the pv asymmetry is characterized ( at least ) by the fom and at constant angle this increases in going from light nuclei such as @xmath81he to a case such as @xmath4s . using basic input concerning practical beam currents , electron polarizations , detector solid angles and acceptable running times , in past work ( see dds and also the discussions in sec . 3.5.2 of @xcite ) , the lower limit was typically chosen to be 0.01 , _ i.e. , _ giving rise at least to a 1@xmath172 change with respect to the reference ( standard model ) value of the total pv asymmetry . in other words , if something affects the pv asymmetry at ( say ) the few percent level it can be regarded to be accessible with present technology . indeed , for elastic pv scattering from he 4% has already been reached @xcite and upcoming experiments such as that on pb aim for even higher precision . said another way , the peak value of the fom increases with nuclear species faster than @xmath80 , while the practical luminosity decreases roughly as @xmath215 , making the peak fom @xmath216 increase by roughly a factor of two in going from @xmath81he ( see @xcite ) to @xmath4s , and thus the `` doability '' actually increases in going from light to heavier nuclei . accordingly , in the present study in the figures we have shown where @xmath191 falls below the 1% level ( namely , a bit more challenging than measurements that have already been performed ) to indicate the rough division between what is presently within reach for pv electron scattering and what will have to wait for future advances in technology ; specifically , in the figures we have indicated with shading where the effects fall below 0.01 and are therefore likely to be inaccessible for the foreseeable future . table [ kinematicschoice ] shows the intervals of momentum transfer ( in @xmath11 ) appropriate for the measurement of the isospin - mixing part of the pv asymmetry deviations ( or , in other words , when @xmath180 ) . these intervals have been chosen so that the deviations are higher than 0.01 and at the same time the fom at fixed angle is no more than one order of magnitude lower than the maximum value reached at the peak . they are equivalently given in terms of incident electron energies ( in mev ) . in the case of fixed incident energy , the intervals transform into scattering angle ranges , also shown in the table ( in degrees ) for @xmath212 gev . these angles are never smaller than 5@xmath211 , which can be considered an experimental lower bound . as a rough estimate , one can say that in the upper half of the intervals the value of the fom is very similar no matter what the combination of incident energy and scattering angle used to get the correct transfer momentum . note , furthermore , that the fom is actually only the naive measure ( i.e. , it characterizes @xmath217 , as discussed above ) , whereas , with the isospin - mixing effects increasing with @xmath12 , a somewhat large value of momentum transfer will actually be the optimal one for experimental study . that is , a compromise must be reached to choose a momentum transfer range where both the pv asymmetry deviation and the fom are as large as possible . given the structure of the fom at fixed angle , decreasing with the momentum transfer due to its cross section dependence , the region of interest corresponds to the first bump of the fom graph , _ i.e. , _ between 0 and 1.5 @xmath11 . within this transfer momentum range , one has to find a more restrictive region where the pv asymmetry deviation is large enough to be measurable . the present work is not intended to be a detailed study of experimental possibilities and so we only use the fom for general guidance . in summary , everything above 10@xmath218 in the figures can be considered to be measurable , while the shaded region below this value will be experimentally more challenging . in figs . [ gamma_iso_strange_si28 ] and [ gamma_iso_strange_c12][gamma_iso_strange_s32 ] we show in the left - hand panel the pv asymmetry deviations for the three choices of electric strangeness , @xmath219 0 and + 1.5 , corresponding roughly to the presently known range of acceptable values . clearly for positive versus negative electric strangeness the interplay between strangeness and isospin mixing can be quite different : for negative @xmath110 the two effects tend to cancel , whereas for positive @xmath110 they interfere constructively . in the former case this makes the predicted @xmath191 small , but still potentially accessible , especially in the region at or slightly above @xmath220 @xmath11 . in contrast , for positive electric strangeness the asymmetry deviation typically rises above the 10% value in regions where the fom is peaking clearly a significant modification of the standard model pv asymmetry . one presumes that the @xmath81he case is the best suited to studies of electric strangeness , since isospin mixing is predicted to be much smaller then ; however , the interplay of isospin mixing and strangeness seen here for the four nuclei under investigation suggests that these cases could provide information not only about the former effect , but also the latter . in figs . [ gamma_iso_strange_si28 ] and [ gamma_iso_strange_c12][gamma_iso_strange_s32 ] we show in the right - hand panel and in a reduced transfer momentum range the isospin - mixing contribution to the asymmetry deviation computed within dw from eq . ( [ asymmetryratio_dw ] ) and once again in pwba . the main effect of the dw calculation is to smooth the divergence of the pwba asymmetry deviation by the appearance of a double - bumped structure . out of the region of the peaks , the differences between pwba and dw are seen to be negligible . finally , in order to emphasize how the isospin contribution to the pv asymmetry changes as the atomic number increases , we show results from a fully distorted calculation of it in fig . [ comp_gammai_dw ] for the four nuclei under study . we restrict ourselves to the momentum transfer region of most interest for future experiments , as suggested in table [ kinematicschoice ] , and the asymmetry deviation is shown in a non - logarithmic scale . as can be seen in the figure , the isospin contribution in this region of momentum transfer becomes higher as the atomic number increases and typically @xmath191 goes from a few percent for carbon to well over 10% for the heavier cases . it should be noted that this figure shows effects from isospin mixing alone : if strangeness contributions of order those discussed above were to be present then the total would be modified . for instance , if positive electric strangeness contributions consistent with present experimental limits were present in addition to the isospin - mixing effects discussed in the present work , then the total asymmetry deviations would be roughly twice those seen in the figure , namely , comparatively large effects would be observed . in the present work a new study has been undertaken of the effects expected from isospin - mixing in nuclear ground - state wave functions on elastic parity - violating electron scattering at momentum transfers extending up to about 1.5 @xmath11 . four n = z 0@xmath7 nuclei have been considered , @xmath1c , @xmath2 mg , @xmath3si , and @xmath4s , each expected to be very close to eigenstates of isospin with @xmath5 in their ground states . however , as first discussed in @xcite ( dds ) , the coulomb interaction occurs asymmetrically between pp and pn / nn nucleon pairs in the nucleus , thereby giving rise to small isospin mixing and thus the nuclear ground states considered here have small admixtures with t@xmath2210 . while such effects are essentially negligible in the parity - conserving cross section , they can play a measurable role in the parity - violating asymmetry , and accordingly , whether the focus is placed on isospin mixing itself or on how these effects may confuse interpretations of the pv asymmetry in terms of standard model tests or with respect to strangeness content in the weak neutral current , it is important to evaluate their influence . in the older work of dds a limited - model - space shell model was employed to estimate the isospin mixing and the mixing via the coulomb interaction was handled perturbatively via a simple two - level approximation . furthermore , in the study of dds the idea of using pv elastic electron scattering to determine ground - state neutron distributions as in the case of @xmath9pb ( which forms the basis of the prex experiment ) was put forward ; here the focus has been limited to a few special n = z nuclei and n@xmath221z cases such as lead have not been re - considered . in the present work a self - consistent axially - symmetric mean - field approximation with density - dependent effective two - body skyrme interactions , including coulomb interactions between pp pairs , has been used in direct determinations of the ground state wave functions . the small differences between the proton and neutron density distributions thereby obtained yield both isoscalar and isovector ground - state coulomb monopole matrix elements and produce modifications in the pv asymmetry from the model - independent result obtained in the absence of isospin - mixing and strangeness contributions . additionally , the effects of pairing in this mean - field approximation have also been investigated , as have effects from strangeness contributions in the single - nucleon form factors and from subtle spin - orbit contributions in the coulomb monopole operators . * in the present work one observes considerably larger effects from isospin mixing than were found in dds , especially since here important matrix elements both diagonal and off - diagonal are naturally included , whereas in the earlier study the restriction of the shell - model space to a single major shell yielded a special constraint on the @xmath12-dependences of the isovector matrix elements . * specific influences of nuclear dynamics ( different forces , different pairing gaps ) and of subtleties in the current operators ( spin - orbit effects ) were investigated and seen not to affect the pv asymmetry at low-@xmath12 significantly . * results using either plane or distorted electron waves were obtained and , for the relatively light nuclei considered in the present work , their differences were seen to be small at low momentum transfers except in the vicinity of a diffraction minimum where the cross section is also small . * kinematic ranges where potential future measurements might be undertaken are discussed by studying both the deviations in the pv asymmetry ( the differences seen with / without isospin mixing ) and the experimental figure - of - merit . we have shown that the isospin mixing effects considered in this work will have a measurable effect on the asymmetries , even for the light nuclei considered . * furthermore , in going from the lightest n = z nuclei to heavier cases one sees that the asymmetry deviations increase making the isospin - mixing effects all the more evident . * in exploring the interplay between the isospin - mixing effects and those effects that may arise from electric strangeness contributions one sees an interesting constructive / destructive interference scenario : with positive electric strangeness the two contributions add at low momentum transfers , whereas when negative they subtract . presently the sign of the electric strangeness form factor is unknown ( the form factor is , in fact , consistent with zero ) and so such interferences may provide information on isospin mixing , namely the main focus of the present work , but also on strangeness . this work was supported by ministerio de ciencia e innovacin ( spain ) under contracts no . fis2005 - 00640 and no . fis2008 - 01301 . thanks ministerio de ciencia e innovacin ( spain ) for financial support . j.m.u . acknowledges support from intas open call grant no 05 - 1000008 - 8272 , and ministerio de ciencia e innovacin ( spain ) under grants fpa-2007 - 62616 and fpa-2006 - 07393 , and ucm and comunidad de madrid under grant grupo de fsica nuclear ( 910059 ) . this work was also supported in part ( twd ) by the u.s . department of energy under contract no . de - fg02 - 94er40818 . twd also wishes to thank ucm - gruposantander for financial support at the universidad complutense de madrid . ccccc isotope & @xmath223 = 0 & @xmath223 = 1 mev & @xmath223 from mass diff . & @xcite @xmath1c & @xmath2240 & @xmath2240 & @xmath2240 & - @xmath2 mg & 56.67 & 54.70 & 38.54 & 58.1 @xmath3si & -45.35 & -43.41 & -28.62 & -57.75 @xmath4s & @xmath2240 & @xmath2240 & @xmath2240 & - cccccccc & & & & @xmath225 & @xmath226 & @xmath227 & @xmath228 & @xmath229 & & @xmath230 = 0.1 @xmath231 & @xmath230 = 0.5 @xmath231 & @xmath230 = 1 @xmath231 & & 0 & 0 & 0.058 & * 0.208 * & 0.229 & 0.145 & & 0 & 1 & 0.168 & 0.005 & 0.136 & * 0.346 * 0 & 1/2 & 0 & 2 & 0.052 & 0.000 & 0.005 & 0.048 & & 1 & 1 & 0.028 & 0.100 & 0.079 & 0.024 & & 1 & 2 & * 0.203 * & 0.011 & * 0.244 * & 0.304 & & 0 & 0 & 0.093 & * 0.335 * & 0.309 & 0.078 & & 0 & 1 & * 0.349 * & 0.013 & * 0.329 * & * 0.555 * 1 & 1/2 & 0 & 2 & 0.048 & 0.000 & 0.006 & 0.048 & & 1 & 1 & 0.093 & 0.330 & 0.217 & 0.047 & & 1 & 2 & 0.018 & 0.001 & 0.023 & 0.018 & & 0 & 0 & 0.184 & * 0.662 * & * 0.610 * & 0.016 & & 0 & 1 & * 0.628 * & 0.024 & 0.592 & * 1.000 * 1 & 3/2 & 0 & 2 & 0.137 & 0.000 & 0.018 & 0.137 & & 1 & 1 & 0.186 & 0.659 & 0.434 & 0.094 & & 1 & 2 & 0.031 & 0.002 & 0.040 & 0.032 & & 0 & 0 & 0.044 & * 0.156 * & * 0.119 * & 0.007 & & 0 & 1 & * 0.061 * & 0.003 & 0.061 & * 0.060 * 2 & 3/2 & 0 & 2 & 0.008 & 0.000 & 0.001 & 0.007 & & 1 & 1 & 0.001 & 0.004 & 0.002 & 0.000 & & 1 & 2 & 0.002 & 0.000 & 0.002 & 0.001 & & 0 & 0 & 0.142 & 0.506 & 0.386 & 0.024 & & 0 & 1 & * 1.000 * & 0.045 & * 1.000 * & * 0.986 * 2 & 5/2 & 0 & 2 & 0.048 & 0.000 & 0.008 & 0.044 & & 1 & 1 & 0.284 & * 1.000 * & 0.541 & 0.075 & & 1 & 2 & 0.379 & 0.003 & 0.049 & 0.021 ccisotope & for @xmath163 @xmath1c & 0.74 @xmath232 q @xmath232 1.42 & 838 @xmath232 @xmath233 @xmath232 1607 & 8.38 @xmath232 @xmath210 @xmath232 16.07 @xmath2 mg & 0.51 @xmath232 q @xmath232 1.10 & 577 @xmath232 @xmath233 @xmath232 1245 & 5.77 @xmath232 @xmath210 @xmath232 12.45 @xmath3si & 0.48 @xmath232 q @xmath232 1.08 & 543 @xmath232 @xmath233 @xmath232 1222 & 5.43 @xmath232 @xmath210 @xmath232 12.22 @xmath4s & 0.45 @xmath232 q @xmath232 1.05 & 509 @xmath232 @xmath233 @xmath232 1188 & 5.09 @xmath232 @xmath210 @xmath232 11.88

multilevel monte carlo ( mlmc ) and unbiased estimators recently proposed by mcleish ( _ monte carlo methods appl . _ , 2011 ) and rhee and glynn ( _ oper . res . _ , 2015 ) are closely related . this connection is elaborated by presenting a new general class of unbiased estimators , which admits previous debiasing schemes as special cases . new lower variance estimators are proposed , which are stratified versions of earlier unbiased schemes . under general conditions , essentially when mlmc admits the canonical square root monte carlo error rate , the proposed new schemes are shown to be asymptotically as efficient as mlmc , both in terms of variance and cost . the experiments demonstrate that the variance reduction provided by the new schemes can be substantial .

1512.01022

multilevel monte carlo ( mlmc ) methods pioneered by @xcite and @xcite are now standard for estimation of expectations of functionals of processes defined by stochastic differential equations ( sdes ) . while the mlmc techniques have origins in the integral operators @xcite and sdes @xcite , they can be applied also in other application domains , where estimates with gradually increasing accuracy are available ; see the recent review by @xcite and references therein . more recently , so - called debiasing techniques @xcite have attracted a lot of research activity ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , although similar ideas have been suggested much earlier in more specific contexts ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? these techniques are based on similar ideas as mlmc , but instead of optimal allocation of computational resources for minimising the error , the primary focus is in providing unbiased estimators . monte carlo inference is straightforward with independent unbiased samples , allowing to construct confidence intervals in a reliable way @xcite . debiasing techniques may also be employed within a stochastic approximation algorithm @xcite . in particular , in a stochastic gradient descent type algorithm @xcite relevant for instance in maximum likelihood inference @xcite , unbiased gradient estimate implies pure martingale noise , which is supported by a well - established theory ( e.g. * ? ? ? * ; * ? ? ? the debiasing techniques involve balancing with cost and variance , which often boils down to similar methods and conditions as those that are used with mlmc . the connection between mlmc and debiasing techniques has been pointed out earlier at least by @xcite , @xcite and @xcite , but this connection has not been fully explored yet . the purpose of this paper is to further clarify the connection of mlmc and debiasing techniques , within a general framework for unbiased estimators . many techniques for unbiased estimation other than those considered here have been suggested in the literature . for instance , there is a whole body of literature for ` perfect sampling ' by markov chains @xcite or with certain classes of sde models @xcite ; see also the recent monograph by @xcite and references therein . perfect sampling can be used to construct unbiased estimators , but the problem is generally more prestigious and often harder to implement . see also the recent article by @xcite for discussion of other related unbiased estimation techniques . the rest of the paper is organised as follows . the multilevel monte carlo and the previous debiasing methods are presented briefly in sections [ sec : mlmc ] and [ sec : unbiased ] , respectively . section [ sec : general ] introduces a new general unbiased scheme with an explicit expression for its variance ( theorem [ thm : general ] ) . the unbiased estimators suggested by @xcite and @xcite are reformulated as specific instances of this scheme , as well as an obvious ` hybrid ' scheme with mlmc and unbiased components ( example [ ex : hybrid - mlmc ] ) . new unbiased estimators are suggested in section [ sec : new ] . two of the new schemes , termed stratified and residual sampling estimators , have provably lower variance than simple averages of independent unbiased estimates ( proposition [ prop : consistency ] ) . because stratification is a well - known variance reduction technique , these estimators may be well - known , but they do not seem to be recorded in the literature yet . the first main finding of this paper is theorem [ thm : residual - eff ] which shows that the asymptotic variance of two of new schemes are asymptotically equal to that of mlmc under general conditions . this result suggests that unbiasedness can often be achieved with asymptotically negligible additional variance . the expected cost of the methods is discussed in section [ sec : cost ] . the new schemes appear even more appealing after seeing that the expected cost of mlmc and the unbiased schemes are also asymptotically equivalent ( proposition [ prop : mlmc - expected - cost ] ) and therefore the efficiency of an estimator with new schemes can be made arbitrarily close to mlmc ( corollary [ cor : batch - complexity ] ) . the limiting variance formulation in theorem [ thm : residual - eff ] leads into an easily applicable optimisation criteria for the sampling distribution related to the new estimators . section [ sec : stopping ] presents a further generalisation of the unbiased scheme , which accomodates further conditioning and dependent randomisation schemes based on stopping times . numerical experiments in section [ sec : numerical ] show how the efficiency bounds predicted by theory are attained in four examples , three of which were also studied by @xcite . the paper is concluded by a discussion about the implications of the findings in section [ sec : conclusions ] . some practical guidelines and possible future topics are discussed as well . the starting point of mlmc is the existence of estimators @xmath0 , which are increasingly accurate approximations of a given ` target ' random variable @xmath1 , whose expectation is of interest . that is , @xmath2 are random variables with @xmath3 , and the task is to provide a numerical approximation of @xmath4 . simulation cost of a single realisation of @xmath5 increases in @xmath6 , which calls for optimisation of computational resources . such a scenario arises for instance in the context of sde models , commonly applied in option pricing . in such an application , we might have @xmath7 , where @xmath8 is a given ` payoff ' function , and @xmath9 corresponds to the terminal value of the process @xmath10}$ ] solving the sde @xmath11 , \qquad x_0\equiv x_0 , \label{eq : sde}\ ] ] where @xmath12 stands for the standard brownian motion , the functions @xmath13 are the drift and diffusion parameters of the model , and @xmath14 is the initial value . unless @xmath15 and @xmath16 are of certain specific form , the random variable @xmath9 can not be simulated exactly , but admits easily implementable approximations based on time - discretisation of . the commonly applied dyadic uniform meshes are defined as follows @xmath17 the _ milstein _ discretisation ( e.g. * ? ? ? * ) corresponding to @xmath18 is defined by letting @xmath19 and iteratively calculating @xmath20,\end{aligned}\ ] ] for @xmath21 , where @xmath22 and @xmath23 are independent @xmath24 random variables . the final value of such an iteration provides an approximation of @xmath9 , so we may set @xmath25 for @xmath26 . the cost of simulating a single realisation of @xmath5 is of order @xmath27 , and under certain , fairly general , conditions on @xmath28 and @xmath29 , the mean square error @xmath30 with @xmath31 ( e.g. * ? ? ? the mlmc is an efficient way to use such approximations in order to approximate @xmath4 . it is based on the following , seemingly trivial observation @xmath32 which suggests that one may construct an estimate of @xmath33 using a separate monte carlo approximation of each term @xmath34 : @xmath35 where @xmath36 and @xmath37 are independent random variables with @xmath38 . it is direct to check that then @xmath39 . there are two key requirements which make this useful : a. the random variables @xmath37 are usually independent realisations of @xmath40 , where the approximations @xmath5 and @xmath41 are _ dependent _ , such that @xmath42 is typically small when @xmath6 is large . in the context of the sde example discussed above , such a coupling arises naturally when using a common brownian path in the discretisations leading to @xmath5 and @xmath41 . b. it allows to _ optimise _ the computational effort devoted to each ` level ' to estimate @xmath34 . the key benefit is that fewer samples @xmath43 are often necessary with higher @xmath6 , which leads to lower overall cost . theorem 1 of @xcite quoted below gives the complexity of mlmc under common exponential framework . it assumes that the expected cost ( computational complexity ) of each term @xmath37 is @xmath44 , so that the expected cost of @xmath45 is @xmath46 . [ thm : giles ] suppose @xmath47 are independent , and there exist positive @xmath48 with @xmath49 , such that for all @xmath50 , @xmath51 then , there exists a positive constant @xmath52 such that for any @xmath53 there are values @xmath54 and @xmath55 such that the mlmc estimator satisfies the following mean square error ( mse ) bound @xmath56 and satisfies the expected cost bound @xmath57 the sde application discussed above often satisfies the conditions of theorem [ thm : giles ] with @xmath58 @xcite . in a multidimensional sde setting , the antithetic truncated milstein @xcite often lead to @xmath58 as well . this highlights why mlmc has become so popular it is often possible to attain a ` canonical ' rate @xmath59 , equivalent with a square root error rate @xmath60 , despite the bias , using simple standard discretisation methods . crude monte carlo , that is , taking a fixed level @xmath54 and averaging independent realisations of @xmath61 , leads to worse rates . this canonical square root error rate can be attained with similar assumptions using the debiasing techniques which are discussed next . as with mlmc , assume that @xmath2 and @xmath1 are integrable random variables with @xmath62 . taking @xmath63 , @xmath64 the fundamental observation behind the unbiased schemes is that one may employ randomisation to pick a finite number of terms of the series to construct estimates with expectation @xmath4 . the two results below due to @xcite propose two such estimators , along with expressions for their second moments . proofs of theorems [ thm : single - term ] and [ thm : sum ] are shown to follow as a consequence of theorem [ thm : general ] in section [ sec : general ] ( examples [ ex : single ] and [ ex : sum ] ) . [ cond : single - term ] suppose @xmath65 are uncorrelated random variables with @xmath66 , and @xmath67 is a probability distribution such that @xmath68 for all @xmath69 and @xmath70 [ thm : single - term ] suppose @xmath65 and @xmath67 satisfy condition [ cond : single - term ] , and @xmath71 is independent random variable with @xmath72 , then the _ single - term estimator _ @xmath73 @xcite first suggested the following estimators , but with different conditions and different expression for the variance . [ cond : sum ] suppose @xmath65 are random variables with @xmath66 , and @xmath67 is a probability distribution such that @xmath74 for all @xmath69 and either one of the following hold : a. [ item : coupled - sum ] @xmath40 and @xmath75 b. [ item : independent - sum ] @xmath65 are uncorrelated and @xmath76 [ thm : sum ] suppose @xmath65 and @xmath67 satisfy condition [ cond : sum ] and @xmath71 is independent with @xmath72 , then the _ sum estimator _ @xmath77 satisfies @xmath78 . in case of condition [ cond : sum ] , @xmath79 and , in case of condition [ cond : sum ] , @xmath80 we follow @xcite and call the estimator of theorem [ thm : sum ] _ coupled sum _ in case of condition [ cond : sum ] and _ independent sum _ in case of condition [ cond : sum ] . the consistency assumption in the latter case is slightly milder than that of @xcite , where condition [ cond : sum ] is assumed also in case of uncorrelated @xmath65 . let us briefly return to the case where @xmath81 and @xmath82 corresponds to the solution of a time - discretised sde with a mesh of @xmath27 points as discussed in section [ sec : mlmc ] . many approximation schemes admit a weak rate @xmath83 , which implies that @xmath84 ( cf . * ? ? ? if additionally @xmath85 with @xmath86 , often satisfied for instance by the milstein scheme @xcite or in the multivariate case by the antithetic truncated milstein scheme @xcite , then taking @xmath87 , where @xmath88 satisfies conditions [ cond : single - term ] and [ cond : sum ] . because the cost of level @xmath6 is of order @xmath27 , this sampling scheme admits a finite expected cost ( see section [ sec : cost ] for details ) . this shows that unbiased estimators with finite variance and finite expected cost can be obtained with the same conditions under which mlmc admits the canonical error rate . in case of correlated @xmath65 in condition [ cond : sum ] , a slightly more stringent condition about the ` strong rate ' @xmath30 with @xmath31 is required . @xcite pointed out that averaging @xmath89 independent single term estimators @xmath90 introduced in theorem [ thm : single - term ] corresponds to an estimator of the form @xmath91 where @xmath92 is the number of samples from level @xmath6 , and the expectation of @xmath92 is @xmath93 . this shows a close connection with mlmc , which corresponds taking @xmath94 ( up to some level @xmath54 ) . inspired by this remark , and the techniques by @xcite , consider the following general unbiased scheme , with an expression for its variance . [ thm : general ] suppose @xmath2 and @xmath1 are integrable random variables such that @xmath3 , and let @xmath65 be square integrable random variables such that @xmath95 for all @xmath50 , with @xmath96 . assume @xmath97 for @xmath98 are independent realisations of the process @xmath65 . suppose @xmath92 are non - negative square integrable random integers independent of @xmath99 and with @xmath100 for all @xmath50 . define for @xmath101 @xmath102 if there exist a strictly increasing sequence of positive integers @xmath103 such that @xmath104 and @xmath105 almost surely , then the estimator @xmath106 is unbiased @xmath107 and @xmath108 . the proof of theorem [ thm : general ] is given in appendix [ app : proof - general ] . consider first a simple example of theorem [ thm : general ] , where @xmath92 are taken independent . let @xmath109 be independent with @xmath110 and @xmath111 , and suppose @xmath112 , and let @xmath65 be uncorrelated . then , @xmath113 if @xmath114 a.s . and @xmath115/\mu_i<\infty$ ] , then the corresponding estimator is unbiased with variance @xmath116 . in particular , taking @xmath92 poisson with intensity @xmath93 where @xmath117 , then @xmath118 by the borel - cantelli lemma and @xmath119 . the averages of single term estimators and sum estimators introduced in theorems [ thm : single - term ] and [ thm : sum ] correspond to certain dependence structures of @xmath120 , as we shall see next . all details in the following examples are given , as the new schemes introduced in section [ sec : new ] will be based on examples [ ex : single ] and [ ex : sum ] below . [ ex : single ] suppose condition [ cond : single - term ] holds and @xmath121 are independent . define @xmath122 and call the estimator @xmath123 . it is easy to see that this corresponds to an average of @xmath89 independent single term estimators @xmath90 . note that @xmath109 follow a multinomial distribution with parameters @xmath89 and @xmath67 . we have @xmath124 and @xmath125 = n p_i { \mathbbm{i}\left\{i = k\right\ } } + n(n-1 ) p_i p_k,\ ] ] so @xmath126 . by the independence of @xmath65 , @xmath127 by assumption @xmath128 , and therefore also @xmath62 , which imply @xmath129 . the variance satisfies @xmath130 so @xmath131 . the above also proves theorem [ thm : single - term ] with @xmath132 . [ ex : sum ] suppose condition [ cond : sum ] holds and define @xmath133 , where @xmath134 are independent with @xmath135 . this estimator , which we denote @xmath136 , corresponds to average of @xmath89 independent sum estimators @xmath137 . we have @xmath138 , and for all @xmath139 , @xmath140 so @xmath141 . we also have @xmath142 because @xmath143 for @xmath144 , so for @xmath145 , @xmath146 if condition [ cond : sum ] holds , we obtain @xmath147 because @xmath148 . let @xmath149 be from lemma [ lem : subsequence ] in appendix [ app : single - sum ] . then for any @xmath150 @xmath151 which converges to zero as @xmath152 . the variance satisfies by dominated convergence @xmath153 we may conclude that the variance of the estimator is @xmath154 . suppose then that condition [ cond : sum ] holds . it is straightforward to check that then @xmath155 which satisfies @xmath129 by assumption , and by dominated convergence @xmath156 taking @xmath132 concludes also the proof of theorem [ thm : sum ] . the last example is an obvious ` hybrid ' scheme involving mlmc and an ` unbiased tail ' scheme , which also falls into the framework of theorem [ thm : general ] . [ ex : hybrid - mlmc ] assume that @xmath157 and that @xmath158 for @xmath159 , and @xmath160 are positive with @xmath161 , let @xmath134 be independent with @xmath162 , and define @xmath163 for @xmath164 . the estimator can be written as @xmath165 the first term coincides with an mlmc estimator with @xmath54 levels , with expectation @xmath33 , and the second term is an average of single term estimators of @xmath166 , with @xmath167 samples . note that we could have used any unbiased scheme in the latter part , provided it satisfies the conditions in theorem [ thm : general ] . the new sampling schemes discussed next in section [ sec : new ] provide a different view on the balancing ; see in particular theorem [ thm : residual - eff ] . let us now turn into new practically interesting estimators which correspond to specific choices in theorem [ thm : general ] . the estimators are based on stratification , a classical variance reduction technique in survey sampling ( e.g. * ? ? ? * ) , which have also been widely used in monte carlo ; see for instance @xcite . the following lemma states classical results on stratification ( cf . * ? ? ? * for proof ) . [ lem : strat ] let @xmath168 be an integrable random variable and let @xmath169 be exhaustive disjoint events with @xmath170 , and let @xmath171 . assume @xmath172 are random variables with conditional laws @xmath173 . the stratified estimator @xmath174 satisfies a. [ item : strat - pessimistic ] unbiasedness @xmath175 and @xmath176 . b. [ item : prop - alloc ] if @xmath177 are independent and @xmath178 ( proportional allocation ) , then @xmath179 and @xmath180 . in what follows , assume @xmath67 are positive and such that @xmath181 , and denote @xmath182 . let us first consider uniform stratification schemes . for that purpose , recall the definition of the generalised inverse distribution function @xmath183 corresponding to @xmath67 , @xmath184 the well - known inverse distribution function method states that a uniform @xmath185 is transformed by @xmath186 into @xmath187 with @xmath72 . [ def : stratified ] let @xmath188 , and assume @xmath189 are independent @xmath190 random variables for @xmath191 . let @xmath192 . a. [ item : strat - single ] define @xmath193 , then the estimator @xmath194 defined as in theorem [ thm : general ] is the _ uniformly stratified single term estimator_. b. [ item : strat - sum ] define @xmath133 , then the estimator @xmath195 defined as in theorem [ thm : general ] is the _ uniformly stratified sum estimator_. [ def : systematic ] let @xmath196 and define @xmath197 for all @xmath191 , where @xmath198 , and let @xmath199 . define then @xmath92 and @xmath200 as in definition [ def : stratified ] and , respectively , and the corresponding _ systematic sampling single - term estimator _ @xmath201 and _ systematic sampling sum estimator _ @xmath202 . the consistency and a variance bound for the uniformly stratified and systematic sampling estimators are stated in proposition [ prop : consistency ] , after introducing another slightly different stratification scheme . [ def : residual ] let @xmath188 , define @xmath203 and let @xmath204 . define the ` residual ' probability distribution @xmath205 as @xmath206 , and let @xmath207 be independent random variables such that @xmath208 for @xmath209 . a. define @xmath210 where @xmath211 , then the estimator @xmath212 defined as in theorem [ thm : general ] is the _ residual sampling single term estimator_. b. define @xmath213 , where @xmath214 and @xmath215 , then the estimator @xmath216 defined as in theorem [ thm : general ] is the _ residual sampling sum estimator_. [ prop : consistency ] the estimators in definitions [ def : stratified ] , [ def : systematic ] and [ def : residual ] satisfy : a. assume condition [ cond : single - term ] , then @xmath217 b. assume condition [ cond : sum ] , then @xmath218 proposition [ prop : consistency ] follows as a consequence of lemma [ lem : strat ] ; the detailed proof is given in appendix [ app : proof - prop - consistency ] . instead of using @xmath219 in systematic sampling , we could use instead any sequence @xmath220 $ ] , and let @xmath221 , where @xmath185 . this would not be stratification , but it is direct to check the estimators still retain the same expectation , and also the same pessimistic variance bound . for instance , using a low - discrepancy sequence @xmath222 would correspond to randomised quasi - monte carlo ( e.g. * ? ? ? * ) . let us next consider pseudo - code implementations of the stratified and systematic sampling estimators . algorithm [ alg : stratified ] describes an efficient implementation which produces @xmath223 as in definition [ def : stratified ] . the function defined as but with the lines [ strat - for ] and [ strat - draw ] interchanged produces @xmath223 following definition [ def : systematic ] . @xmath224 ; @xmath225 [ strat - for ] draw independent @xmath226 [ strat - draw ] @xmath227 @xmath228 ; @xmath229 @xmath230 @xmath231 algorithm [ alg : single - term ] calculates one realisation of @xmath232 or @xmath233 corresponding to the sde setting discussed in section [ sec : mlmc ] , when called with @xmath234 and @xmath235 or @xmath236 , respectively . similarly , the independent sum and coupled sum estimators @xmath237 or @xmath238 are simulated if @xmath239 or @xmath240 , respectively . algorithm [ alg : single - term ] could be applied similarly with @xmath241 leading to @xmath242 and @xmath243 , by defining such that @xmath244 satisfy @xmath245 and drawing independent @xmath246 . @xmath247 ; @xmath248 @xmath249@xmath250 @xmath251^{-1}$ ] @xmath252 [ line : tailprob ] simulate brownian path @xmath253 at mesh @xmath254 . @xmath255 [ for - single - coupled ] simulate brownian path @xmath253 at mesh @xmath256 . @xmath257 where @xmath258 is calculated using @xmath253 . @xmath259 @xmath260 @xmath261 proposition [ prop : consistency ] states that all the new estimators are unbiased , and that the uniformly stratified and residual sampling estimators can not be worse than averages of independent single term and sum estimators , in terms of variance . in fact , they often have a strictly lower variance , but it is generally difficult to quantify the benefit . the following theorem [ thm : residual - eff ] is often more useful , indicating that residual sampling attains asymptotically the efficiency of mlmc under general conditions . let us first formulate ` idealised ' mlmc strategies based on limiting allocations @xmath67 . [ def : mlmc ] assume @xmath67 are non - negative with @xmath262 and define @xmath263 . for any @xmath188 , define @xmath203 , @xmath264 , @xmath265 and @xmath266 and denote @xmath267 practical implementations of mlmc are often based on application of stopping rules , which may determine @xmath43 and @xmath268 during simulation . definition [ def : mlmc ] can be therefore viewed as ` idealised ' version of mlmc , where ( nearly ) optimal allocation strategy @xmath67 is known beforehand , and @xmath43 and @xmath268 are determined in terms of a single ` running - time ' parameter @xmath89 . [ thm : residual - eff ] below , @xmath269 ( resp . @xmath270 ) stands for either @xmath212 or @xmath194 ( resp . @xmath271 or @xmath272 ) . a. [ item : residual - eff - single ] assume condition [ cond : single - term ] then @xmath273 b. [ item : residual - eff - sum ] assume condition [ cond : sum ] , then @xmath274 and if @xmath275 , then @xmath276 . c. [ item : eff - independent - sum ] assume condition [ cond : sum ] , then @xmath277 the proof of theorem [ thm : residual - eff ] is given in appendix [ app : proof - eff ] . [ rem : eff - diff ] the limiting variances in theorem [ thm : residual - eff ] can be significantly lower than the upper bounds given in proposition [ prop : consistency ] , which correspond to averaging independent unbiased estimates . indeed , @xmath278 , so @xmath279.\ ] ] likewise , in case of condition [ cond : sum ] , @xmath280 , so @xmath281 note that @xmath282 for all @xmath6 , so all the terms in the sum are positive . let us next consider the cost of simulating estimators @xmath261 of the form given in theorem [ thm : general ] . assume each @xmath37 has random cost @xmath283 , such that the total cost of @xmath261 is @xmath284 . assume also that @xmath285 are independent of @xmath109 and @xmath286 have common mean @xmath287 . the expected cost can be written down as follows . @xmath288 the following records the immediate fact that the estimators introduced in section [ sec : new ] have the same expected cost as the simple averages of independent estimates . [ prop : cost - new ] denote the cost of the estimator @xmath269 by @xmath289 , where ` @xmath290 ' is a place holder . then , a. @xmath291 , b. @xmath292 . the following proposition records that the mlmc estimators introduced in definition [ def : mlmc ] have asymptotically equivalent cost than the corresponding unbiased estimators with same limiting allocation strategies [ prop : mlmc - expected - cost ] the expected cost of mlmc estimators in theorem [ thm : residual - eff ] satisfy @xmath293 the expected cost of @xmath294 is @xmath295 clearly @xmath296 , @xmath297 and @xmath298 as @xmath299 , so if @xmath300 , the result follows by dominated convergence ; otherwise one can use fatou s lemma . the proof with @xmath301 is identical . @xcite formulate an asymptotic efficiency principle , which states that if @xmath302 are i.i.d . estimators with finite variance @xmath303 and with expected cost @xmath304 , then the average of such estimators has asymptotic relative efficiency @xmath305^{-1}$ ] . considering this notion of efficiency , let @xmath188 , and consider the following estimator , which is the average of @xmath54 independent realisations of the estimators above , @xmath306 where @xmath307 are independent realisations of @xmath269 , where ` @xmath290 ' is a place holder . by letting @xmath308 , one may consider the asymptotic efficiency of this estimator as in the following result , stating that stratified and residual sampling procedures always improves upon averages of independent single term and sum estimators , and can be made arbitrarily close to mlmc in asymptotic efficiency . [ cor : batch - complexity ] suppose @xmath188 is fixed , then the following asymptotic efficiency holds when @xmath308 . a. if condition [ cond : single - term ] holds and @xmath309 , then the asymptotic efficiency of both @xmath310 and @xmath311 is no worse than that of @xmath312 , and can be made arbitrarily close to that of @xmath313 by choosing @xmath89 sufficiently large . b. if condition [ cond : sum ] holds and @xmath314 , then the asymptotic efficiency of both @xmath315 and @xmath316 is no worse than that of @xmath317 , and can be made arbitrarily close to that of @xmath318 by choosing @xmath89 sufficiently large . strictly speaking , the use of asymptotic efficiency principle requires that the multilevel estimators @xmath317 would be guaranteed to satisfy a functional central limit theorem , which is out of the scope of this work ; see , however , the recent work of @xcite in this direction in a different setting . if we assume @xmath89 is taken sufficiently large so that the limits in theorem [ thm : residual - eff ] are approximately attained , the asymptotic efficiency principle suggests a rule for tuning the distribution @xmath319 for stratified and residual sampling distributions : the distribution @xmath67 which minimises the asymptitic inverse relative efficiency ( ire ) @xmath320 or @xmath321 maximises the efficiency . for the single term estimator , condition [ cond : single - term ] , this leads to @xmath322 the solution to is proportional to @xmath323 , if @xmath324 ( * ? ? ? * proposition 1 ) . this is straightforward to implement in practice , because reliable estimates of @xmath325 are easily available . a straightforward practical procedure , also implemented in the experiments below , is to use the ` empirical optimal distribution ' @xmath326 to define directly the first @xmath54 probabilities @xmath327 and a tail probability @xmath328 . the tail probabilities @xmath329 for @xmath330 are then defined to follow a parametric distribution which guarantees a finite variance based on theory . in the experiments , the tail distribution was chosen to be geometric . in case of the independent sum estimator , condition [ cond : sum ] , similar minimisation with @xmath331 in place of @xmath332 yields the asymptotically optimal distribution @xmath333 . however , contrary to the single - term estimator , @xmath334 must additionally be non - increasing . because is invariant under multiplicative constants on @xmath67 , the independent sum estimator can never outperform the single term estimator , in terms of asymptotic ire . the coupled sum estimator , condition [ cond : sum ] , leads to the optimisation problem @xmath335 this is more involved for two reasons . estimation of the terms @xmath336 is not straightforward , but in practice , a reasonable approximation may be obtained by approximating @xmath1 with @xmath61 , where @xmath54 is large . the optimisation is more challenging as well , and in fact , the optimal strategy may require considering ` subsequence ' estimators , employing level differences @xmath337 , for some subsequence @xmath338 . this leads to a combinatorial problem , which is discussed by @xcite , who also describe a dynamic programming algorithm which finds such optimal @xmath339 , up to an index @xmath54 . in some applications it is not possible to choose @xmath67 which yield both finite variance and finite expected cost . then , it is not possible to attain a canonical square root convergence rate , but @xcite suggest another approach : choose @xmath67 that ensure finite variance , but which imply infinite expected cost . using a result due to @xcite , they deduce complexity results for unbiased estimators which are close to what are possible with mlmc . however , quantifying the efficiency in the present setting is not straightforward , so this is left for future work . the general unbiased scheme proposed in theorem [ thm : general ] is based on independent randomisation , that is , @xmath109 are assumed independent of @xmath47 . such a scheme is often appropriate in practice , but it is also possible to think of cases where @xmath109 could depend on @xmath47 , for instance in a stopping time fashion . it is also possible to retain unbiasedness while replacing @xmath340 in the estimator by a conditional expectation . we consider below a scheme which accomodates both of these generalisations , while retaining unbiasedness . [ cond : stopping - dependence ] suppose @xmath341 , @xmath342 and @xmath1 are integrable random variables , @xmath343 and @xmath344 , @xmath345 are @xmath16-algebras , and @xmath109 are non - negative integer - valued random variables , such that @xmath346 almost surely for all @xmath50 . [ cond : stopping - dependence2 ] suppose condition [ cond : stopping - dependence ] holds , and for all @xmath347 : a. [ item : general - consistency ] @xmath348 almost surely . b. [ item : general - bound ] there exists random variables @xmath349 such that @xmath350 and with @xmath351 . c. [ item : general - summability ] @xmath352 and @xmath353 almost surely . [ thm : dependent - randomisation ] assume condition [ cond : stopping - dependence ] holds and denote @xmath354 and @xmath355 whenever well - defined . a. [ item : stopping - gen ] if condition [ cond : stopping - dependence2 ] and hold , then @xmath356 . if also condition [ cond : stopping - dependence2 ] holds , then @xmath107 . b. [ item : stopping - indep ] suppose the assumptions of theorem [ thm : general ] hold and @xmath99 are independent of @xmath345 . if @xmath357 < \infty \end{aligned}\ ] ] for all @xmath358 and @xmath104 for some subsequence @xmath103 , then @xmath107 and @xmath359 . proof of theorem [ thm : dependent - randomisation ] is given in appendix [ app : proof - dependent - randomisation ] . theorem [ thm : dependent - randomisation ] is a generalisation of theorem [ thm : general ] , and leads to new , potentially interesting estimators . for instance , let @xmath193 where @xmath360 are positive random integers independent of @xmath361 . if we take @xmath362 for all @xmath363 , then @xmath364 . this can be viewed as a residual sampling scheme applied with the ( random ) probability distribution @xmath365 . analogously , the residual sampling scheme may be viewed through such conditioning , where @xmath366 are deterministic . it is unclear whether theorem [ thm : dependent - randomisation ] allows estimators that have practical appeal , such as greater efficiency compared with the estimators introduced in section [ sec : new ] . theorem [ thm : dependent - randomisation ] allows @xmath367 to depend on @xmath99 in a stopping time fashion , analogous to wald s identity . suppose condition [ cond : stopping - dependence ] holds and @xmath99 are as in theorem [ thm : general ] . a. if @xmath92 are stopping times with respect to @xmath47 , that is , @xmath368 and @xmath369 for all @xmath347 , then taking @xmath370 trivial , condition [ cond : stopping - dependence2 ] holds because @xmath371 = { \mathbb{e}}\delta_i { \mathbb{p } } ( j\le n_i \mid { \mathcal{f}}_{i-1}),\ ] ] and , similarly condition [ cond : stopping - dependence2 ] with @xmath372 . b. if condition [ cond : single - term ] holds for some probability distribution @xmath67 , then @xmath373 . the stopping time formulation may prove theoretically useful , and suggests a possibility to be used in conjunction with stopping rules developed in the mlmc context ( e.g. * ? ? ? however , the practical relevance of such an approach may be limited , because the expectation of a stopping time is often unavailable . it appears also difficult to derive useful explicit variance expressions when @xmath374 depend on @xmath47 . the performance of the estimators was studied with four sde examples , where expectations of final - value functionals are estimated . three of the models were the same as those used by @xcite . the first is a _ geometric brownian motion _ ( gbm ) @xmath375,\qquad x_0\equiv 1,\ ] ] with @xmath376 and @xmath377 , where @xmath12 is the standard brownian motion . the target functional is the final value european option @xmath378 , with approximate expected value @xmath379 @xcite . the second model was a _ cox - ingersoll - ross _ ( cir ) model @xmath380,\qquad x_0\equiv 0.04,\ ] ] with @xmath381 , @xmath382 and @xmath383 , and with final value european option @xmath384 . according to extensive simulations , the expected value was found to be approximately @xmath385 . the third model is a bivariate _ model @xmath386 , \qquad \begin{array}{rl } s_0\equiv 1 \\ v_0\equiv 0 , \end{array}\ ] ] with @xmath376 , @xmath381 , @xmath382 , @xmath383 , and where @xmath387 are coordinates of a correlated brownian motion with coefficient @xmath388 . the functional @xmath389 , with expected value @xmath390 @xcite . the fourth model is an artificial model termed _ modified gbm _ , which has the same volatility term as gbm but a time - dependent drift : @xmath391 , \qquad x_0 \equiv 1,\ ] ] and @xmath392 . the target functional is the mean , which was found to have an approximate expected value @xmath393 . this last model is intended to have bigger @xmath394 than the previous examples , highlighting the differences between the new estimators over averages of independent estimators bigger ; see remark [ rem : eff - diff ] . the implementation developed for the tests is available at https://bitbucket.org/mvihola/unbiased-mlmc . in all but the heston model , the standard milstein scheme described in section [ sec : mlmc ] was employed . the antithetic truncated milstein scheme proposed by @xcite was applied with the heston model . we considered single term , independent sum and coupled sum estimators . with the first two , the distributions @xmath67 were set to approximately optimal in each model as discussed in section [ sec : cost ] : @xmath327 were set empirically to minimise variance , and the tail probability @xmath395 was calculated from empirical data . the tail distribution @xmath332 for @xmath164 was geometric , @xmath396 , with the parameter @xmath397 set in all cases to @xmath398 . this provided a very good fit with the empirical optimal distribution @xmath399 with @xmath400 in case of gbm and @xmath401 with all other examples ; see also theoretical results on the milstein scheme @xcite and the antithetic truncated milstein @xcite . in the coupled sum estimator , the differences @xmath402 were all based on a single brownian trajectory . the optimisation was based on estimators of @xmath403 , where @xmath61 corresponded to discretisation with a mesh of @xmath404 . an optimal subsequences were found using the dynamic programming algorithm of @xcite . figure [ fig : res - all ] shows results based on averages of @xmath405 independent runs of each algorithm in each model . replications of estimators with @xmath406 , @xmath407 , @xmath405 and @xmath408 . ] the graphs in figure [ fig : res - all ] show the estimated inverse relative efficiency ( ire ) of the methods : the average cost multiplied with average square deviation from the ground truth values as given above . according to the theory , this quantity is constant with independent averages , and with residual and stratified sampling , it converges to the same limit as the corresponding mlmc estimator . table [ tab : res ] shows the corresponding numerical values with @xmath409 . .estimated ires in the experiments with @xmath409 . [ cols= " < , < , < , < , < , < , < , < , < , < , < , < , < " , ] [ tab : res ] the experiments appear to align well with the theoretical findings . in all but the last example , the mlmc estimator admitted the best ire with small @xmath89 , and the new estimators appear to admit performance in between the independent and the mlmc case . the performance of the new schemes come close to mlmc performance as @xmath89 increases , as verified by the differences with @xmath409 reported in table [ tab : res ] , which are all relatively small , except for the sum estimators in the modified gbm example indicating still some discrepancy . note that the mlmc implemented in the experiments is the ` idealised ' version involving same pre - determined allocation strategy as unbiased estimators ( description [ def : mlmc ] ) . this explains the discrepancy between the mlmc ire reported here and by @xcite , who employ the original version of the mlmc @xcite . the findings of @xcite suggest that unbiased estimators applied with stopping rules may sometimes improve upon the original mlmc . the differences in performance of single term and independent sum estimators with gbm , cir and heston examples are all relatively small , as @xmath89 increases , the ires of the new single term and sum estimators become negligible , as anticipated by the theory . the coupled sum estimator appears to admit greater efficiency with cir and heston examples , but the differences between independent average estimators and the new estimators are small . the modified gbm example demonstrates that the new estimators can be significantly more efficient . the numerical values shown in table [ tab : res ] indicate a 1316 fold increase in terms of relative efficiency with the new single term estimators and similar performance with mlmc . the increase is 1760 fold with sum estimators . in both cases , the stratified and systematic sampling estimators appear to perform slightly better than the residual sampling estimator . this paper presented a general framework for unbiased estimation , which admits previous debiasing schemes as special cases , and accomodates new , lower variance unbiased estimators . the proposed stratified sampling and residual sampling schemes are promising classes of estimators , as they enjoy good theoretical behaviour they can not only improve on averages of independent estimates ( proposition [ prop : consistency ] and proposition [ prop : cost - new ] ) , but also can have a significant gain in efficiency as illustrated in the experiments . indeed , the stratified and residual sampling estimators can be made arbitrarily close to mlmc in efficiency under general assumptions ( theorem [ thm : residual - eff ] and proposition [ prop : mlmc - expected - cost ] ) , highlighting the close connection between the mlmc and the debiasing schemes , and showing that unbiasedness may be achieved with virtually no sacrifice on efficiency . unbiasedness is an important quality of estimators , when employed as part of stochastic optimisation algorithms @xcite , or in a ` compound sampling ' context @xcite . it also enables rigorous stopping rules , which can lead to benefits over mlmc stopping rules @xcite . while the theory presented in this paper does not give guarantees on the limiting efficiency of the systematic sampling , it is expected to behave often similar to stratified and residual sampling schemes , as illustrated by the experiments . however , as stratified and residual sampling enjoy good theoretical properties , and because systematic sampling appears to perform comparatively in practice , stratified or residual sampling are recommended as safer alternatives . the empirical evidence from the numerical experiments in section [ sec : numerical ] suggests that stratified sampling might sometimes perform slightly better than residual sampling . this , together with the straightforward implementation of stratified sampling ( algorithm [ alg : stratified ] ) , makes it appealing for practical purposes . averages of independent realisations of the single term and the independent sum estimators may have different efficiencies in general . in case of residual and stratified sampling , the optimally tuned estimators often coincide in asymptotic efficiency . in general , the single term estimator always dominates the independent sum estimator in terms asymptotic efficiency , rendering the single term estimator preferable over the independent sum estimator . based on the experiments , the optimally tuned coupled sum estimator may sometimes lead into significant performance gains . this suggests also that dependent level estimators might be worth considering in the mlmc context , where independent level estimators are currently widely employed . despite the potential performance gain of the coupled estimators , the single term estimator ( resp . standard independent level mlmc ) may still often be preferable , because of its simplicity of implementation , and because of simpler and more robust criterion for tuning @xmath67 ( resp . @xmath410 ) . the simplicity of the optimisation criterion suggests , as noted by @xcite , an algorithm which automatically tunes the distribution @xmath67 during repeated simulation of @xmath194 or @xmath212 . @xcite suggest methods for finding optimal discretisation hierarchies in the mlmc context , which could also be explored in the debiasing context . the mlmc literature provides many other potential further research topics . it is yet unclear how well the unbiased estimators can compete with mlmc in scenarios with slower than canonical rate of convergence . this is touched by @xcite , but quantifying the effect of the new estimators needs further arguments . it is possible to employ essentially all techniques developed in the context of mlmc with debiasing . these include , for instance , quasi - monte carlo @xcite , adaptive time steps @xcite or adaptive importance sampling scheme based on a drifted brownian motion @xcite . this research was supported by the academy of finland ( grants 274740 and 284513 ) , and computational resources were provided by csc , the it center for science , finland . the author wishes to thank christel geiss , stefan geiss , peter glynn , chang - han rhee , raul tempone and anni toivola for useful discussions and remarks , and the associate editor and the reviewers of _ operations research _ for helpful comments , and for the suggestion to consider dependent randomisation , which led to the developments in section [ sec : stopping ] . 42 [ 1]#1 [ 1]`#1 ` urlstyle [ 1]doi : # 1 s. agapiou , g. o. roberts , and s. j. vollmer . unbiased monte carlo : posterior estimation for intractable / infinite - dimensional models . preprint arxiv:1411.7713 , 2014 . m. b. alaya and a. kebaier . central limit theorem for the multilevel monte carlo euler method . _ , 250 ( 1):0 211234 , 2015 . a. beskos and g. o. roberts . exact simulation of diffusions . probab . _ , 150 ( 4):0 24222444 , 2005 . v. s. borkar . _ stochastic approximation : a dynamical systems viewpoint_. cambridge university press , 2008 . isbn 978 - 0 - 521 - 51592 - 4 . n. collier , a .- haji - ali , f. nobile , e. von schwerin , and r. tempone . a continuation multilevel monte carlo algorithm . _ bit _ , 550 ( 2):0 399432 , 2015 . b. delyon , m. lavielle , and e. moulines . convergence of a stochastic approximation version of the em algorithm . _ , 270 ( 1):0 94128 , 1999 . s. dereich and t. mueller - gronbach . general multilevel adaptations for stochastic approximation algorithms . preprint arxiv:1506.05482 , 2015 . j. dick , f. y. kuo , and i. h. sloan . high - dimensional integration : the quasi - monte carlo way . _ acta numer . _ , 22:0 133288 , 2013 . r. douc , o. capp , and e. moulines . comparison of resampling schemes for particle filtering . in _ proc . image and signal processing and analysis , 2005 . _ , pages 6469 , 2005 . w. feller . a limit theoerm for random variables with infinite moments . _ amer . j. math . _ , 680 ( 2):0 257262 , 1946 . m. b. giles . multilevel monte carlo path simulation . _ , 560 ( 3):0 607617 , 2008 . m. b. giles . multilevel monte carlo methods . _ acta numer . _ , 24:0 259328 , 2015 . m. b. giles and l. szpruch . antithetic multilevel monte carlo estimation for multi - dimensional sdes without lvy area simulation _ , 240 ( 4):0 15851620 , 2014 . m. b. giles and b. j. waterhouse . multilevel quasi - monte carlo path simulation . in _ advanced financial modelling _ , radon series on computational and applied mathematics , pages 165181 . de gruyter , 2009 . p. glasserman . _ monte carlo methods in financial engineering _ , volume 53 . springer , 2003 . p. glynn . randomized estimators for time integrals . technical report , dtic document , 1983 . p. w. glynn and c .- h . exact estimation for markov chain equilibrium expectations . _ j. appl . _ , 51a:0 377389 , 2014 . p. w. glynn and w. whitt . the asymptotic efficiency of simulation estimators . _ , 400 ( 3):0 505520 , 1992 . p. w. glynn and w. whitt . the asymptotic validity of sequential stopping rules for stochastic simulations . _ , 20 ( 1):0 180198 , 1992 . a .- haji - ali , f. nobile , e. von schwerin , and r. tempone . optimization of mesh hierarchies in multilevel monte carlo samplers . partial differ . equ . _ , 40 ( 1):0 76112 , 2016 . m. h. hansen , w. n. hurwitz , and w. g. madow . _ sample survey methods and theory _ , volume 1 methods and applications . wiley , 1953 . s. heinrich . multilevel monte carlo methods . in _ large - scale scientific computing _ , pages 5867 . springer , 2001 . h. hoel , e. von schwerin , a. szepessy , and r. tempone . adaptive multilevel monte carlo simulation . in _ numerical analysis of multiscale computations _ , pages 217234 . springer , 2012 . m. l. huber . _ perfect simulation_. chapman hall / crc , 2015 . p. e. jacob and a. h. thiery . on nonnegative unbiased estimators . _ , 430 ( 2):0 769784 , 2015 . a. jentzen , p. e. kloeden , and a. neuenkirch . pathwise approximation of stochastic differential equations on domains : higher order convergence rates without global lipschitz coefficients . _ , 1120 ( 1):0 4164 , 2009 . c. kahl and p. jckel . fast stong approximation monte carlo schemes for stochastic volatility models . , 60 ( 6):0 513536 , 2006 . a. kebaier and j. lelong . coupling importance sampling and multilevel monte carlo using sample average approximation . preprint arxiv:1510.03590 , 2015 . p. e. kloeden and e. platen . _ numerical solution of stochastic differential equations_. springer , 1992 . h. j. kushner and g. g. yin . _ stochastic approximation and recursive algorithms and applications_. number 35 in applications of mathematics : stochastic modelling and applied probability . springer - verlag , @xmath411 edition , 2003 . isbn 0 - 387 - 00894 - 2 . lyne , m. girolami , y. atchade , h. strathmann , and d. simpson . on russian roulette estimates for bayesian inference with doubly - intractable likelihoods . _ _ , 300 ( 4):0 443467 , 2015 . d. mcleish . a general method for debiasing a monte carlo estimator . _ monte carlo methods appl . _ , 170 ( 4):0 301315 , 2011 . j. g. propp and d. b. wilson . exact sampling with coupled markov chains and applications to statistical mechanics . _ random structures algorithms _ , 90 ( 1 - 2):0 223252 , 1996 . rhee and p. w. glynn . a new approach to unbiased estimation for sde s . in _ proceedings of the winter simulation conference _ , page 17 . winter simulation conference , 2012 . rhee and p. w. glynn . unbiased estimation with square root convergence for sde models . res . _ , 630 ( 5):0 10261043 , 2015 . h. robbins and s. monro . a stochastic approximation method . _ the annals of mathematical statistics _ , 22:0 400407 , 1951 . t. rychlik . unbiased nonparametric estimation of the derivative of the mean . _ statist . _ , 100 ( 4):0 329333 , 1990 . t. rychlik . a class of unbiased kernel estimates of a probability density function . _ appl . math . ( warsaw ) _ , 220 ( 4):0 485497 , 1995 . h. strathmann , d. sejdinovic , and m. girolami . unbiased bayes for big data : paths of partial posteriors . preprint arxiv:1501.03326 , 2015 . m. vihola . unbiased estimators and multilevel monte carlo . preprint arxiv:1512.01022v2 , 2015 . m. vihola , j. helske , and j. franks . importance sampling type correction of markov chain monte carlo and exact approximations . preprint arxiv:1609.02541 , 2016 . c. walter . point process - based monte carlo estimation . _ statist . the following lemma is due to an argument by @xcite . [ lem : subsequence ] suppose @xmath412 as @xmath413 . then , there exists strictly increasing @xmath149 such that for all @xmath150 , @xmath414 let @xmath415 . if @xmath416 for infinitely many @xmath6 , choose them as @xmath103 otherwise , let @xmath417 be such that @xmath418 for all @xmath419 , and define recursively @xmath420 . now , for any @xmath421 and any @xmath422 , @xmath423 define @xmath424 and for @xmath425 @xmath426 then by dominated convergence @xmath427 = \sum_{i=1}^m { \mathbb{e}}\delta_i = { \mathbb{e}}y_m,\ ] ] because @xmath428 . for @xmath429 , by dominated convergence , @xmath430 \label{eq : single - level - diffs } \\ & = \frac{1}{{\mathbb{e}}n_i { \mathbb{e}}n_k } { \mathbb{e}}\bigg[\sum_{j,\ell=1}^\infty \delta_i^{(j)}\delta_k^{(\ell ) } { \mathbbm{i}\left\{j\le n_i\right\}}{\mathbbm{i}\left\{\ell\le n_k\right\ } } \bigg]\nonumber\\ & = \frac{1}{{\mathbb{e}}n_i { \mathbb{e}}n_k } \bigg({\mathbb{e}}(\delta_i \delta_k ) { \mathbb{e}}\bigg[\sum_{j=1}^\infty { \mathbbm{i}\left\{j\le n_i\right\}}{\mathbbm{i}\left\{j\le n_k\right\}}\bigg ] \nonumber\\ & \phantom{=\frac{1}{{\mathbb{e}}n_i { \mathbb{e}}n_k}\bigg(}+ { \mathbb{e}}\delta_i { \mathbb{e}}\delta_k { \mathbb{e}}\bigg[\sum_{\stackrel{j,\ell=1}{j\neq \ell}}^\infty { \mathbbm{i}\left\{j\le n_i\right\}}{\mathbbm{i}\left\{\ell\le n_k\right\ } } \bigg]\bigg ) \nonumber\\ & = \frac{\big({\mathbb{e}}(\delta_i\delta_k ) - { \mathbb{e}}\delta_i{\mathbb{e}}\delta_k\big){\mathbb{e}}(n_i\wedge n_k ) + { \mathbb{e}}\delta_i{\mathbb{e}}\delta_k{\mathbb{e}}[n_i n_k]}{{\mathbb{e}}n_i { \mathbb{e}}n_k},\nonumber\end{aligned}\ ] ] because @xmath431 . we deduce that for @xmath432 , @xmath433 = v_{\ell , m } + ( { \mathbb{e}}y_m - { \mathbb{e}}y_\ell)^2 . \label{eq : full - level - diffs}\end{aligned}\ ] ] therefore , by assumption @xmath434 is cauchy in @xmath435 . because @xmath436 a.s . , we have @xmath437 a.s . , and therefore @xmath438 in @xmath435 . we deduce that @xmath439 . similarly we find the expression @xmath440 $ ] . consider , and let @xmath92 correspond to @xmath212 . we have for any @xmath139 @xmath463 so @xmath464 . in case of @xmath194 , it is not difficult to check that the number of strata @xmath465 partially overlapping @xmath6 , that is , such that @xmath466 but @xmath467 , is at most two , so @xmath468 , and consequently @xmath469 . denote @xmath470 for @xmath425 , where @xmath109 correspond either to the residual sampling or stratified sampling scheme . we have ( cf . example [ ex : single ] ) @xmath471 because @xmath472 as @xmath299 , we deduce that @xmath473 . on the other hand , @xmath474 is no greater than the corresponding variance of single term estimators , and therefore for @xmath475 , @xmath476 we deduce that for any @xmath477 , due to independence , @xmath478 both terms on the right can be made arbitrarily small by choosing @xmath479 large enough and letting @xmath299 . an easy calculation shows that @xmath480 where the first term converges to @xmath481 as @xmath299 . for the latter , observe that @xmath482 and @xmath483 as @xmath299 , so by dominated convergence the last sum in vanishes . let us then turn into . let us calculate first @xmath484 and because @xmath485 we obtain @xmath486 note that @xmath487 , so @xmath488 the first sum in therefore converges to @xmath489 as @xmath299 and the latter sum in converges to zero as above . consider then @xmath216 . as above , @xmath490 . in case of @xmath195 , at most one stratum @xmath491 contains some @xmath492 and @xmath6 , so @xmath493 . therefore , @xmath494 for both schemes , and the variance of @xmath495 admits the bound @xmath496 the second term on the left equals @xmath497 , and clearly @xmath498 as @xmath299 . now , take @xmath103 from lemma [ lem : subsequence ] , then stratification implies that for @xmath150 , @xmath499 we conclude as above by writing for any @xmath500 , @xmath501 because @xmath502 , all terms on the right can be made arbitrarily small by by choosing @xmath503 large enough and letting @xmath299 . consider and notice that @xmath504 denote @xmath505 , then @xmath506 and by dominated convergence a similar calculation yields @xmath507 which leads to @xmath356 . for the latter claim , note that only finitely many of @xmath508 are non - zero , so @xmath261 is well - defined , and the result follows by the dominated convergence theorem . the statement is a generalisation of theorem [ thm : general ] , and the proof follows similarly . indeed , condition [ cond : stopping - dependence2 ] and hold by independence , with @xmath372 , so part implies @xmath356 . denoting @xmath509 , a straightforward calculation similar to yields @xmath510 = \frac{{\mathrm{cov}}(\delta_i,\delta_k){\mathbb{e}}(n_i\wedge n_k\mid { \mathcal{f}}_{i , k } ) + { \mathbb{e}}\delta_i{\mathbb{e}}\delta_k { \mathbb{e}}(n_i n_k\mid { \mathcal{f}}_{i , k})}{{\mathbb{e}}(n_i\mid { \mathcal{f}}_{i-1}){\mathbb{e}}(n_k\mid { \mathcal{f}}_{k-1})},\end{aligned}\ ] ] and as in , @xmath511 . let us consider first @xmath194 , which is an average of independent random variables @xmath441 which follow , respectively , the conditional distribution of the single term estimator @xmath90 , given the uniform random variable @xmath442 generating @xmath187 takes value in @xmath443 . the desired variance bound follows from lemma [ lem : strat ] applied with @xmath444 , @xmath445 , @xmath446 and @xmath447 . the sum estimator @xmath195 is similarly stratified version of the average of @xmath89 independent sum estimators . the systematic sampling estimators @xmath201 and @xmath202 , are averages as above , but with the difference that the uniformly distributed random variables on @xmath448 which determine @xmath449 are not independent . we may apply lemma [ lem : strat ] which gives the ( pessimistic ) upper bound on the variance . the residual sampling @xmath212 and @xmath216 may also be seen as stratification of @xmath90 and @xmath137 , but instead of considering @xmath450 , we now let @xmath451 , where @xmath452 it is direct to verify that @xmath453 . let @xmath454 be independent random variables with conditional distribution of @xmath90 ( resp . @xmath137 ) given @xmath455 , respectively , and let @xmath456 be , similarly , independent conditional on @xmath457 . we may apply lemma [ lem : strat ] with @xmath458 , @xmath459 , @xmath460 , @xmath461 and @xmath462 .

starting from the description of the conformal geometry of noncommutative 2-tori in the framework of modular spectral triples , we explicitly compute the local curvature functionals determined by the value at zero of the zeta functions affiliated with these spectral triples . we give a closed formula for the ray - singer analytic torsion in terms of the dirichlet quadratic form and the generating function for bernoulli numbers applied to the modular operator . the gradient of the ray - singer analytic torsion is then expressed in terms of these functionals , and yields the analogue of scalar curvature . computing this gradient in two ways elucidates the meaning of the complicated two variable functions occurring in the formula for the scalar curvature . moreover , the corresponding evolution equation for the metric produces the appropriate analogue of ricci curvature . we prove the analogue of the classical result which asserts that in every conformal class the maximum value of the determinant of the laplacian on metrics of a fixed area is attained only at the constant curvature metric .

1110.3500

in noncommutative geometry the paradigm of a geometric space is given in spectral terms , by a hilbert space @xmath0 in which both the algebra @xmath1 of coordinates and the inverse line element @xmath2 are represented , the latter being an unbounded self - adjoint operator which plays the role of the dirac operator . the local geometric invariants such as the riemannian curvature are obtained in the noncommutative case by considering heat kernel expansions of the form @xmath3 where @xmath4 is the dimension of the geometry . one may equivalently deal with the corresponding zeta functions . thus , it is the high frequency behavior of the spectrum of @xmath2 coupled with the action of the algebra @xmath1 in @xmath0 which detects the local curvature of the geometry . in this paper we shall analyze in depth a specific example , that of the noncommutative two torus @xmath5 whose differential geometry , as well as its pseudo - differential operator calculus , was first developed in @xcite . to obtain a curved geometry from the flat one defined in @xcite , one introduces ( cf . @xcite , @xcite ) a weyl factor , or dilaton , which changes the metric by modifying the volume form while keeping the same conformal structure . both notions of volume form and of conformal structure are well understood ( cf . @xcite ) , and we recall in [ sectprel ] how one obtains the modified dirac operator for the new curved geometry . the starting point is the computation of the value at @xmath6 of the zeta function @xmath7 for the @xmath8-dimensional curved geometry associated to the dilaton @xmath9 . equivalently we are dealing with the term @xmath10 of the heat expansion . this computation was initiated in the late 1980 s ( cf . @xcite ) , and the explicit result which proves the analogue of the gauss bonnet formula was published in @xcite . it was then extended in @xcite to cover the case of arbitrary values of the complex modulus @xmath11 ( which was @xmath12 in @xcite ) . in these two papers only the total integral of the curvature was needed , and this allowed one to make simplifications under the trace which are no longer possible in the case of arbitrary @xmath13 , i.e. to compute the local expression @xmath10 . the technical obstacles for the local computation were overcome by means of * the general rearrangement lemma of [ sectrearang ] ; * the assistance of the computer . while the original computation of @xcite was done by hand , the role of the computer assistance is to minimize the danger of a computational mistake in handling the large number of terms , about one thousand , which arise in the generalized pseudo - differential expressions involved in @xmath10 . the complete calculation of @xmath10 was actually achieved in 2009 and announced ( including by internet posting , although with some typos ) at several conferences ( oberwolfach 2009 and vanderbilt 2011 ) . moreover , the same computation was done independently in @xcite and gave a confirmation of the result . the main additional input of the present paper consists in obtaining an explicit formula for the ray - singer log - determinant of @xmath14 ( which was left open in @xcite ) . by calculating the gradient of the ray - singer log - determinant in two different ways , one obtains new geometric insight as well as a deep internal consistency relation between the different terms of the log - determinant formula . at the same time this elucidates the role of the intricate two operator - variable function occurring in its expression . we now briefly outline the contents of this paper , starting with the description of the local curvature functionals determined by the value at zero of the zeta functions affiliated with the modular spectral triples describing the curved geometry of noncommutative @xmath8-tori . as in the case of the standard torus viewed as a complex curve , the total laplacian associated to such a spectral triple splits into two components , one @xmath15 on functions and the other @xmath16 on @xmath17-forms , the two operators being isospectral outside zero . the corresponding curvature formulas involve second order ( outer ) derivatives of the weyl factor , and as a new and crucial ingredient they involve the modular operator @xmath18 of the non - tracial weight @xmath19 associated to the dilaton @xmath9 . for @xmath15 the result is of the form @xmath20 where @xmath21 is the inner derivation implemented by @xmath22 , @xmath23 @xmath24 is the dirichlet quadratic form @xmath25 and @xmath26 , signifies that @xmath27 is acting on the @xmath28th factor . the operators @xmath29 and @xmath30 are new ingredients , whose occurrence is a vivid manifestation of the genuinely non - unimodular nature of the conformal geometry of the noncommutative @xmath8-torus . the functions @xmath31 and @xmath32 by which the modular derivatives act seem at first of a rather formidable nature , and of course beg for a conceptual understanding . their expressions , arising from the computation , are as follows : @xmath33 @xmath34 one of our new results consists in giving an abstract proof of a functional relation between the functions @xmath35 and @xmath36 . more precisely , denoting @xmath37 we establish by an a priori argument the identity @xmath38 the function @xmath39 is ( up to the factor @xmath40 ) the generating function of the bernoulli numbers , i.e. one has @xmath41 another main result consists in obtaining the following closed formula for the ray - singer determinant : @xmath42 the a priori proof of the functional relation is based on the computation of the gradient of the ray - singer determinant in two different ways . using the left hand side of one obtains a formula involving @xmath43 , while using the right hand side of gives a general expression as shown in theorem [ directgrad ] of [ functrel ] . as a third fundamental result of this paper , we establish the analogue of the classical result which asserts that in every conformal class the maximum value of the determinant of the laplacian for metrics of a fixed area is uniquely attained at the constant curvature metric . this is the content of theorem [ thmmain3 ] , whose proof relies on the positivity of the function @xmath39 . by , @xmath39 is a generating function for bernoulli numbers , known to play a prominent role in the theory of characteristic classes of deformations , where it is used as a formal power series . it is quite striking that in the present context of a conformal ( but not formal ) deformation , @xmath39 appears no longer merely as a formal series but as an actual function , whose _ positivity _ plays a key role . in marked contrast to the ordinary torus , for which @xmath44 and @xmath45 are both constant multiples of the scalar ( or gaussian ) curvature , the local curvature expressions associated to the zeta functions of the two partial laplacians differ substantially . the function @xmath46 of two variables involved in the expression of @xmath45 is related to @xmath47 in a simple fashion , but a new term appears , in the form of an operator @xmath48 applied to the skew quadratic form @xmath49 a fully conceptual understanding of the meaning of this term remains yet to be found . being isospectral outside zero , both partial laplacians have the same ray - singer determinant . this gives rise to a single log - determinant functional , which represents in fact the analytic torsion of the underlying conformal structure . its gradient delivers in turn the appropriate analogue of the notion of scalar curvature . moreover , the corresponding evolution equation for the metric yields the natural analogue of ricci flow . it should be mentioned that a different version of the latter has been proposed , albeit not explicitly computed , in @xcite . we thank farzad fathizadeh and masoud khalkhali for performing independently , cf . @xcite , the central computation of this paper . the preliminary material gathered in this section is essentially borrowed from @xcite in order to provide the necessary background for the present paper . it also serves as a first illustration of the distinctly non - unimodular feature of the conformal geometry of noncommutative @xmath8-tori , which in particular validates the treatment of twisted spectral triples @xcite as basic geometric structures . the modular spectral triples considered below can be understood as special cases of the following general construction . let us start from an ordinary spectral triple @xmath50 which we assume to be even ( and we let @xmath51 be the grading operator ) . using the direct sum decomposition @xmath52 the action of the algebra @xmath1 , the grading operator and the operator @xmath2 take the form @xmath53 were @xmath54 is an unbounded operator from its domain in @xmath55 to @xmath56 and @xmath57 is its adjoint . let now @xmath58 be a positive invertible element . since the commutator @xmath59 $ ] is bounded the multiplication by @xmath60 preserves the domain of @xmath54 and the following operator is self - adjoint @xmath61 we use the notion of modular ( or twisted ) spectral triple in the sense of definition 3.1 of @xcite . let us show that the perturbation @xmath62 of @xmath2 defines a twisted spectral triple on @xmath1 with respect to the inner automorphism @xmath63 . [ tiwst1 ] let @xmath64 be the ( non - unitary ) inner automorphism of @xmath1 associated to @xmath60 . the triple @xmath50 is a @xmath63-twisted spectral triple . let us compute the twisted commutator @xmath65 . one has @xmath66 the upper right element of the matrix gives @xmath67,\ ] ] which is bounded since @xmath68 $ ] is bounded as well as @xmath60 . the lower left element of this matrix gives @xmath69k,\ \ b=\sigma(a),\ ] ] which is also bounded since @xmath70 . [ pertrem]to display the dependence on the grading @xmath51 one can use the following formula for the perturbation @xmath62 of @xmath2 @xmath71 we shall now explain why it is this simple twisting procedure which is appearing naturally when one introduces a weyl factor ( dilaton ) in the geometry of the noncommutative torus . we still need another general notion of transposed spectral triple . given a hilbert space @xmath0 let @xmath72 be the dual vector space . the transposition @xmath73 gives an antiisomorphism @xmath74 where @xmath75 is the opposite algebra of @xmath76 . thus one can associate to any spectral triple @xmath50 the transposed spectral triple as follows . [ tiwst2 ] let @xmath50 be a @xmath63-twisted spectral triple . let @xmath77 be the opposite algebra and @xmath78 the transposed of the unbounded operator @xmath2 . let @xmath79 be the automorphism of @xmath77 given by @xmath80 then the action of @xmath77 in @xmath72 transposed of the action of @xmath1 in @xmath0 defines a @xmath79-twisted spectral triple @xmath81 the boundedness of the twisted commutators @xmath82 implies the boundedness of the twisted commutators @xmath83 note that one can identify the dual vector space @xmath72 with the complex conjugate of @xmath0 by the antilinear isometry @xmath84 @xmath85 one then has the relation @xmath86 [ transposetrip ] given a modular spectral triple @xmath50 the transposed modular spectral triple is given by . let us fix our notations for the noncommutative torus @xmath5 . we let @xmath87 be an irrational real number and consider the ( uniquely determined ) @xmath88-algebra @xmath89 generated by two unitaries @xmath90 which satisfy the multiplicative commutation relation @xmath91 the @xmath8-dimensional torus @xmath92 acts on @xmath93 via the 2-parameter group of automorphisms @xmath94 , @xmath95 , determined by @xmath96 we denote by @xmath97 the subalgebra of smooth elements for this action , i.e. consisting of those @xmath98 such that the mapping @xmath99 is smooth . expressed in terms of the coefficients of the element @xmath100 , @xmath101 the smoothness condition amounts to their rapid decay , i.e. the requirement that the sequences @xmath102 be bounded for any @xmath103 . the basic derivations representing the infinitesimal generators to the above group of automorphisms are given by the defining relations , @xmath104 they are the counterparts of the differential operators @xmath105 , @xmath106 acting on @xmath107 , and behave similarly with respect to the @xmath108-involution : @xmath109 as @xmath87 was chosen irrational , there is a unique trace @xmath110 on @xmath93 , determined by the orthogonality properties @xmath111 and we denote by @xmath112 the hilbert space obtained from @xmath93 by completing with respect to the associated inner product @xmath113 by construction the hilbert space @xmath112 is a bimodule over @xmath93 with @xmath114 and the trace property of @xmath110 ensures that the right action of @xmath1 is unitary . the derivations @xmath115 , viewed as unbounded operators on @xmath112 , have unique self - adjoint extensions , @xmath116 furthermore , they obviously obey the integration - by - parts rule @xmath117 the conformal structures on the classical torus are best parameterized by a complex number @xmath118 , @xmath119 modulo the natural action of @xmath120 by homographic transformations . to @xmath11 one associates the lattice @xmath121 and the quotient complex structure on @xmath122 . the natural isomorphism of the @xmath8-dimensional torus @xmath92 ( with real coordinates @xmath123 as above ) with @xmath124 is given by @xmath125 one thus gets @xmath126 this gives @xmath127 and @xmath128 as linear expressions in @xmath129 and @xmath130 and one finds up to the overall factor @xmath131 that @xmath132 since replacing the modulus @xmath11 by @xmath133 does not affect the complex structure this allows us to transfer the translation invariant complex structures of @xmath134 to @xmath5 . throughout this paper we fix a complex number @xmath135 with @xmath136 and consider the associated translation invariant complex structure , defined by the pair of derivations @xmath137 representing the counterparts of the differential operators @xmath138 , and @xmath139 acting on @xmath107 . our conventions differ slightly from @xcite in which the only case @xmath12 was covered but we prefer to follow the usual convention for the general case . as explained in @xcite , the information on the conformal structure is actually encapsulated in the positive hochschild @xmath8-cocycle on @xmath140 given by the formula @xmath141 thus , as an appropriate analogue of the space of @xmath142-forms on the classical 2-torus , one takes the unitary bi - module @xmath143 over @xmath140 given by the hilbert space completion of the space of finite sums @xmath144 , @xmath145 , with respect to the inner product @xmath146 [ isoform ] the map @xmath147 , @xmath148 is a unitary @xmath140-bimodule isomorphism of @xmath143 with @xmath149 . by definition of the inner product on @xmath143 the operator is unitary and the derivation property of @xmath150 shows that it is an @xmath140-bimodule map . it remains to check that it is surjective . one has @xmath151 and thus @xmath152 which gives the required surjectivity . when viewed as an unbounded operator from @xmath149 to @xmath143 , the operator @xmath150 will be called @xmath153 . in order to implement conformal changes of metric , we consider the family of positive linear functionals parameterized by self - adjoint elements @xmath154 , @xmath155 , defined by @xmath156 _ we shall call a positive linear functional @xmath157 on @xmath93 as in a _ conformal weight with weyl factor @xmath158 and dilaton @xmath9 . _ the normalized functional _ @xmath159 _ is called the _ associated conformal state . each conformal weight @xmath157 determines an inner product @xmath160 on @xmath93 , namely @xmath161 we let @xmath162 denote the hilbert space completion of @xmath93 for the inner product @xmath160 . it is a unitary left module on @xmath93 by construction . note that , whereas for @xmath110 we have the trace relation @xmath163 the functional @xmath164 satisfies instead @xmath165 which is the kms condition at @xmath166 for the 1-parameter group @xmath167 , @xmath168 , of inner automorphisms @xmath169 equivalently , @xmath170 where the modular operator @xmath18 , given by @xmath171 is positive and fulfills @xmath172 the infinitesimal generator of the @xmath173-parameter group @xmath167 is the inner derivation @xmath174 , @xmath175 \ , , \quad x \in a_{\theta}^{\infty } \ , .\ ] ] to correct the lack of unitarity of the action of @xmath93 on @xmath162 by right multiplication , one replaces it by the right action @xmath176 where @xmath177 is the tomita antilinear unitary of the gns representation associated to @xmath157 ; explicitly , with @xmath178 , @xmath179 one thus gets @xmath180 the obtained unitary @xmath140-bimodule is isomorphic to @xmath149 , [ isophi ] the right multiplication by @xmath60 , @xmath181 extends to an isometry @xmath182 and gives a unitary @xmath140-bimodule isomorphism of @xmath149 with @xmath162 . one has for any @xmath183 , @xmath184 this shows that @xmath185 is an isometry . by construction it intertwines the left module structures . moreover one has , using , @xmath186 this shows that @xmath185 intertwines the right module structures . with the complex structure associated to @xmath135 , @xmath136 fixed , the operator associated to the flat metric in the corresponding conformal class on @xmath5 is given by @xmath187 in other words , this is the natural @xmath5 version of the @xmath188-operator , which is isospectral to the usual spin@xmath189 dirac operator on the ordinary torus @xmath134 . the left and right actions for the unitary @xmath190-bimodule structure of @xmath191 both give spectral triples . one can take the transpose in the sense of definition [ transposetrip ] of the spectral triple @xmath192 given by the right action of @xmath190 . this transposed triple is isomorphic to the spectral triple given by the left action of @xmath190 in @xmath193 and the operator @xmath194 if one disregards the grading @xmath51 the spectral triples @xmath195 and @xmath196 are equivalent but this does not hold as graded spectral triples and in fact the equivalence reverses the grading . one can see this distinction even in the commutative case by looking at the equation @xmath197e=0\ , , \ a , b\in \ca , \ \ e=\frac{1+\gamma}{2},\ ] ] which is fulfilled when @xmath198 is antiholomorphic . we now perform a non - trivial conformal change of metric on @xmath5 . let the conformal weight @xmath157 be as above . the varying structure comes from the operator @xmath199 which is given by @xmath153 on @xmath140 but is viewed as an unbounded operator from @xmath162 to @xmath143 , @xmath200 in order to form the corresponding spectral triple we consider the operator @xmath201 where we view @xmath202 both as a left module and a right module over @xmath140 . lemmas and show that an @xmath140-bimodule @xmath203 is isomorphic to @xmath204 by the unitary map @xmath205 let @xmath206 denote the tomita anti - unitary operator on @xmath149 extending the star involution @xmath207 , @xmath100 . we let @xmath208 the direct sum of two copies of @xmath209 acting in @xmath210 . [ unitlem ] let @xmath178 , where @xmath211 is the dilaton of the conformal weight @xmath157 . we let @xmath212 denote the right multiplication by @xmath60 in @xmath149 . * the operator @xmath213 is equal to the self - adjoint unbounded operator @xmath214 * the operator @xmath215 is equal to the self - adjoint unbounded operator @xmath216 let @xmath217 . one has @xmath218 and @xmath219 . thus @xmath220 which gives the first statement . the second statement follows from the compatibility of the star operation with the derivations @xmath221 . [ speclem ] let @xmath178 , with @xmath211 the dilaton of the conformal weight @xmath157 . * the left action of @xmath93 on @xmath203 together with the operator @xmath222 yield a graded spectral triple @xmath223 . * the right action @xmath224 of @xmath93 on @xmath203 together with the operator @xmath222 yield a graded twisted spectral triple @xmath225 , with bounded twisted commutators @xmath226 * the transposed of the modular spectral triple @xmath225 is isomorphic to the perturbed spectral triple @xmath227 \(i ) in order to show that @xmath228 $ ] is bounded , it suffices to check that @xmath229 $ ] is bounded . in turn , the latter easily follows from the derivation property of @xmath230 and the equivalence of the norms @xmath231 and @xmath232 . \(ii ) this follows from lemma [ tiwst1 ] and the third statement which we now prove . \(iii ) this follows from the second statement of lemma [ unitlem ] using . by corollary [ speclem ] , the transposed of the modular spectral triple @xmath225 is simply given by the left action of @xmath190 on @xmath204 and the operator @xmath233 [ define ] the modular spectral triple of weight @xmath157 is @xmath234 where @xmath204 as a left @xmath140-module and @xmath235 is given by . the spectral invariants of the modular spectral triple of weight @xmath157 are obtained by computing zeta functions and heat expansions , i.e. traces of products of an element of @xmath190 ( acting on the left ) by a function of @xmath236 . let @xmath237 be the dolbeault - laplace operator for the flat metric , @xmath238 acting on functions on @xmath5 . [ laplem ] let @xmath178 , where @xmath211 is the dilaton of the conformal weight @xmath157 . * one has @xmath239 * the laplacian on functions is anti - unitarily equivalent to @xmath240 . * the operator @xmath241 is anti - unitarily equivalent to the laplacian @xmath242 on forms of type @xmath142 . this follows from corollary [ speclem ] . [ lem:3zetas ] let @xmath157 be a conformal weight with dilaton @xmath243 . the zeta function of the laplacian on functions is equal to the zeta functions of the operators @xmath15 , @xmath242 and @xmath16 : @xmath244 the operators @xmath245 and @xmath241 have the same spectrum outside @xmath246 , which proves the first equality in . the others follow from lemma [ laplem ] . we digress a little to show that the notion of _ conformal index _ for a manifold , introduced in @xcite , admits a natural extension to the framework of noncommutative geometry . let @xmath247 be a @xmath248-summable spectral triple , which has _ discrete dimension spectrum _ in the sense of @xcite . fix @xmath249 , and let @xmath250 then @xmath251 hence @xmath252 duhamel s formula for the family @xmath253 , @xmath254 allows to write @xmath255 noting that @xmath256 one obtains the identity @xmath257 at this point we make an additional assumption , which stipulates _ the existence of small time asymptotic expansions of the form @xmath258 and more generally , for any @xmath259 , @xmath260 which moreover can be differentiated term - by - term with respect to @xmath261$]_. [ thmconfind ] under the above assumptions the value of the zeta function at the origin @xmath262 is invariant under conformal deformations of the spectral triple @xmath247 . denote by @xmath263 the inverse of @xmath264 restricted to @xmath265 , where @xmath266 stands for the orthogonal projection onto @xmath267 , and consider the zeta function @xmath268 which is related to the theta function by the mellin transform @xmath269 the asymptotic expansion ensures that @xmath270 has meromorphic continuation to @xmath271 , with only simple poles . furthermore , because of the pole of @xmath272 at @xmath273 , @xmath270 is holomorphic at @xmath246 , and its value at @xmath246 is @xmath274 differentiating term - by - term the asymptotic expansion and applying yields the identities @xmath275 in particular , @xmath276 hence @xmath277 an instance where the above hypotheses are satisfied , and hence the result applies , is that of the dilaton field rescaling of the mass in the spectral action formalism for the standard model @xcite . more to the point , the pseudodifferential calculus for @xmath278-dynamical systems @xcite , and especially the elliptic theory on noncommutative tori @xcite , show that the condition is fulfilled in the case of @xmath5 . in particular , all the laplacians in [ spectrip ] admit meromorphic zeta functions , with simple poles and regular at @xmath246 . [ thmconfindlap ] the value at the origin of the zeta function of the laplacian on functions is a conformal invariant , i.e. @xmath279 for any conformal weight @xmath157 on @xmath190 . in view of lemma [ laplem ] , one can replace the laplacian on functions by @xmath280 . consider the family @xmath281 since @xmath282 by using duhamel s formula as in , one sees @xmath283 the variation formulas for the coefficients of the heat operator asymptotic expansion yield in this case the identities @xmath284 in particular , @xmath285 , and the proof is achieved in the same way as that of theorem [ thmconfind ] . this gives a non - computational proof to the gauss - bonnet theorem for the noncommutative @xmath8-torus ( cf . @xcite , @xcite ) . we now focus on the zeta function of the modular spectral triple of weight @xmath157 , as in definition [ define ] , i.e. @xmath286 , in order to compute its local invariants . in order to state our first main result , i.e. theorem [ thmmain ] , we shall first introduce several functions which play a key role in the statement of the basic formula . , even and negative with @xmath287 . ] in the formulation of theorem [ thmmain ] there is an overall factor of @xmath288 where @xmath289 is the imaginary part of @xmath11 but the other functions involved are independent of @xmath11 and we list them below and analyze their elementary properties . one always gets an expression of the form @xmath290 applied to @xmath291 for the first half of laplacian , the function is @xmath292 for the full laplacian , it is @xmath293 for the graded case @xmath294 , even and positive with @xmath295 . ] one always gets an expression of the form @xmath296 applied to @xmath297 the functions @xmath298 of two variables are : for the first half of laplacian @xmath299 for the full laplacian , the formula of theorem [ thmmain ] involves @xmath300 where @xmath301 for the graded case , it involves @xmath302 . the additional skew term is of the form @xmath48 applied to @xmath303 it only appears in the second half of laplacian . the function @xmath304 is @xmath305 which is a symmetric function of @xmath306 and @xmath307 . we now list the elementary properties of the modular curvature functions of two variables . figures [ h0function ] , [ h1function ] , [ sfunction ] ) [ scacurv ] the functions @xmath47 , @xmath308 and @xmath309 fulfill the following properties 1 . they belong to @xmath310 . @xmath311 , @xmath312 and @xmath313 . 3 . @xmath314 , @xmath315 . the smoothness of @xmath36 is clear outside the three lines @xmath316 , @xmath317 and @xmath318 . let @xmath319 be the numerator of the fraction defining @xmath36 . near the first line one gets the expansion @xmath320 @xmath321 which shows that as long as @xmath322 the function @xmath36 is smooth at @xmath323 . the value of @xmath36 on @xmath324 is given by @xmath325 . ] near the second line @xmath326 one gets the expansion @xmath327 which gives the smoothness for @xmath328 on the second line ( and the third similarly ) . one needs to look carefully at what happens at the crossing point @xmath329 . one finds that the taylor expansion of @xmath36 at the point @xmath330 is of the form @xmath331 @xmath332 similarly we let @xmath333 be the numerator of the fraction defining @xmath309 @xmath334 and get the expansion near the line @xmath324 in the form @xmath335 @xmath336 near the second line @xmath326 one gets the expansion @xmath337 ^ 4\ ] ] the denominator of @xmath304 is @xmath338 and this gives the smoothness outside the origin . at @xmath330 one has the taylor expansion @xmath339 . ] we now look at the behavior at @xmath340 . it is enough to show that the function @xmath308 tends to @xmath246 at @xmath340 . we write @xmath46 as the fraction @xmath341 first note the equality @xmath342 which shows that away from the lines @xmath343 the numerator and denominator have the same exponential increase . let @xmath344 . the maximum of @xmath345 on the sphere @xmath346 is reached on the interval @xmath347\}\ ] ] . ] away from the boundary of this interval one can approximate the denominator of @xmath348 by the product of the leading exponentials which gives @xmath349 using this approximation and neglecting terms which are suppressed by an exponential one reduces the function @xmath46 well inside the interval @xmath350 to the fraction @xmath351 on @xmath350 the function @xmath352 reaches its minimum at @xmath353 where @xmath354 fulfills @xmath355 one finds in this way that the maximum of @xmath345 on the sphere @xmath346 is of the order of @xmath356 . one needs to control the size of the function @xmath46 in the neighborhood of the zeros of the denominator . the restriction of @xmath46 to the anti - diagonal @xmath357 is given by the odd function @xmath358 which is equivalent to @xmath359 when @xmath360 . the restriction of @xmath46 to the axis @xmath361 is given by @xmath362 which is equivalent to @xmath363 when @xmath360 . the restriction of @xmath309 to the anti - diagonal @xmath357 is given by @xmath364 which behaves like @xmath365 for @xmath366 and this gives the behavior of the maximum of @xmath367 on the sphere @xmath346 . the minimum of @xmath309 on the sphere @xmath368 is reached on the diagonal @xmath369 where the function reduces to @xmath370 the other properties can be checked in a straightforward manner . with the above notation , we are now ready to express in local terms the value at the origin of the zeta functions of the modular spectral triple of weight @xmath157 @xmath371 for each @xmath372 we consider the zeta function @xmath373 where @xmath374 stands for the orthogonal projection onto @xmath375 . it is related to the theta function by the mellin transform @xmath376 as in the untwisted case , cf . , its value at @xmath246 is related to the constant term in the asymptotic expansion , via @xmath377 the computation of the constant term @xmath378 is quite formidable , as could be expected from the already laborious calculations performed in @xcite and @xcite in the untwisted case , i.e. @xmath379 . for the clarity of the exposition , we postpone giving the technical details until [ tech ] . on the other hand , the additional term is very easy to compute . indeed , @xmath380 is one dimensional and one has ( with @xmath381 the associated state ) @xmath382 one deals in a similar manner with the laplacian @xmath16 and one lets @xmath383 be the orthogonal projection on its one - dimensional kernel , @xmath384 which is independent of @xmath60 , and consists of the constant multiples of the unit @xmath385 , so that @xmath386 [ thmmain ] let @xmath157 be a conformal weight with dilaton @xmath243 and let @xmath178 . the value at the origin of the zeta function associated to the modular spectral triple of weight @xmath157 is given for any @xmath372 by the expression @xmath387 where @xmath388 and for its graded version by @xmath389 where @xmath390 . in order to check the normalization constants we compare this result with the classical formula for the value at @xmath246 of the zeta function of the laplacian on a closed surface @xmath391 @xmath392 where @xmath393 is the scalar curvature ( normalized as being @xmath173 for the unit two sphere ) and @xmath394 the euler - poincar characteristic . one double checks this formula for the unit two sphere , whose laplacian spectrum is the set @xmath395 where the eigenvalue @xmath396 has multiplicity @xmath397 . one has @xmath398 whose constant term @xmath399 agrees with the right hand side @xmath400 of . in a local form one has @xmath401 where @xmath402 is the orthogonal projection on the kernel of the laplacian . to compare this formula with theorem [ thmmain ] we take the half sum of and in the commutative case this reduces to @xmath403 one has @xmath404 and @xmath110 is the state associated to the volume form of the flat two torus with integral spectrum . to check the overall normalization we take @xmath12 so that @xmath405 the torus has coordinates @xmath406 with period @xmath407 , riemannian metric @xmath408 , and the spectrum of the laplacian is the set @xmath409 . thus @xmath410 we now consider the riemannian metric @xmath411 . its volume form is @xmath412 . the scalar curvature is @xmath413 to check this , take the stereographic coordinates on the unit two sphere , so that the metric becomes @xmath411 with @xmath414 . then gives @xmath415 as required . thus gives for arbitrary @xmath60 the local form @xmath416 and this agrees with the right hand side of which is , using , @xmath417 the fact that the local curvature expressions , occurring in on the one hand and those occurring in on the other hand , are sharply different stands in stark contrast with the case of the ordinary torus . for the latter they reduce to @xmath418 and thus are both constant multiples of the gaussian curvature of the conformal metric . in this section we develop the analogue of the osgood - phillips - sarnak functional @xcite , which is a scale invariant version of the log - determinant of the laplacian . we then compute its gradient , whose corresponding flow for riemann surfaces exactly reproduces hamilton s ricci flow @xcite , and therefore yields the appropriate analogue of the scalar curvature . the ray - singer zeta function regularization of the determinant of a laplacian @xcite , as well the related notion of analytic torsion @xcite , make perfect sense in the case of noncommutative @xmath8-tori , due to the existence of the appropriate pseudodifferential calculus @xcite . thus , @xmath419 are well - defined . because @xmath420 and @xmath421 have the same spectrum outside @xmath246 , and so the corresponding zeta functions coincide , the two log - determinants are in fact equal . in order to compute the above determinant , let us consider again the @xmath173-parameter family of laplacians , cf . , @xmath422 differentiating the corresponding family of zeta functions and taking into account one obtains , for @xmath423 , @xmath424 by meromorphic continuation one obtains the identity @xmath425 and taking @xmath426 yields the variation formula @xmath427 applying theorem [ thmmain ] to the conformal weights @xmath428 with dilaton @xmath429 , and retaining only the even part of the expression in the right hand side , yields @xmath430 one has @xmath431 and so by integration along the interval @xmath432 one obtains the following _ conformal variation formula_. [ inside0 ] let @xmath157 be a conformal weight with dilaton @xmath243 . then @xmath433 we now show that this formula simplifies much further . [ inside ] for @xmath434 a function in schwartz space one has @xmath435 the derivation @xmath436 is given by the commutator with @xmath22 and @xmath170 fixes @xmath9 and preserves the trace @xmath110 . thus writing @xmath437 as a fourier transform @xmath438 we thus get that @xmath439 and integrating from @xmath246 to @xmath173 we obtain , under the hypothesis of lemma [ inside0 ] , @xmath440 let us now simplify the last term of . by construction the expression @xmath441 can be expressed as a linear combination of squares of elements of @xmath442 . thus we really need to understand the quadratic form @xmath443 where writing @xmath36 as a fourier transform @xmath444 one has @xmath445 [ doublelem ] let @xmath446 be a schwartz function , such that @xmath447 , then @xmath448 where the function @xmath449 is given by @xmath450 let us define the function of two variables @xmath451 so that , with @xmath452 one has @xmath453 one has , using @xmath454 for all @xmath455 , * @xmath456 for all @xmath455 . * @xmath457 where the last equality follows , using the trace property of @xmath110 , from @xmath458\sigma_{b}(x))\,.\ ] ] thus one gets , using the antisymmetry of @xmath446 and @xmath459 , @xmath460 moreover @xmath461 and one gets @xmath462 which is the required equality . we used the hypothesis that @xmath60 is a schwartz function in order to use freely the fourier transform but since the spectrum of the operator @xmath436 is bounded this hypothesis is not needed as long as we deal with smooth functions . it follows using @xmath463 that @xmath464 where the function @xmath465 is given by @xmath466 we now need to compute the integral in the variable @xmath306 of . [ compute ] @xmath467 where the function @xmath468 is given by @xmath469 one has , by , @xmath470 thus , using , @xmath471 which gives for @xmath472 and one checks that this agrees with by showing that the derivative of @xmath473 is @xmath474 . we can thus simplify and obtain , using @xmath475 , @xmath476 since @xmath477 is isospectral to the dolbeault - laplacian on @xmath134 whose spectrum is the set of @xmath478 for @xmath479 , @xmath480 , @xmath481 remains the same as for the ordinary @xmath8-torus , and is computed by the kronecker limit formula , as in ( * ? ? ? * theorem 4.1 ) ; explicitly , @xmath482 where @xmath483 is the dedekind eta function @xmath484 recalling that the logarithm of analytic torsion for a complex curve @xcite is given by @xmath485 it is perhaps not surprising that the osgood - phillips - sarnak functional @xcite involves the negative of the log - determinant . by analogy , we define the translation invariant functional @xmath486 on the space of selfadjoint elements of @xmath487 by the formula @xmath488 its invariance under rescaling is due to the fact that @xmath489 , cf . theorem [ thmconfindlap ] . indeed , since @xmath490 for any @xmath491 , one has @xmath492 it follows that @xmath493 [ thmmain3 ] the functional @xmath494 has the expression @xmath495 one has @xmath496 for all @xmath9 and equality holds if and only if @xmath9 is a scalar . one obtains using and the classical value for @xmath497 @xmath498 using integration - by - parts with respect to the derivation @xmath499 , one sees that @xmath500 note that the skew term @xmath501)=0\ ] ] vanishes . thus one has @xmath502 this , together with gives . , even and negative . ] in the hilbert space @xmath149 with inner product , the operator @xmath436 which is given by @xmath503 $ ] is bounded and self - adjoint since @xmath504)= \vp_0((-[h , b])^*a)=\langle a , \lmod(b)\rangle\ ] ] since the function @xmath505 is even and strictly negative , it follows that the equality @xmath506 defines a non degenerate positive symmetric inner product on @xmath149 . now by , and the skew adjointness @xmath507 one has @xmath508 and thus @xmath509 this shows that @xmath496 and moreover the equality holds only if @xmath510 which implies that @xmath9 is a constant . _ note that the first term in the right hand side of matches the expression of the functional in the commutative case ( cf . @xcite ) , while the second term is of a purely modular nature and is highly non - linear . _ identifying the tangent space to the dilatons with the selfadjoint elements of @xmath442 , and using the inner product given by @xmath110 , we shall define the gradient of the functional @xmath486 by means of the gteaux differential : @xmath511 [ gradthm ] the gradient of @xmath486 is given by the expression @xmath512 where the functions @xmath39 and @xmath513 are directly related to @xmath35 , @xmath468 and @xmath36 by @xmath514 and @xmath515 consider the @xmath173-parameter family of operators @xmath516 the duhamel ( also the expansional ) formula implies that @xmath517 hence @xmath518 it follows that @xmath519 differentiating at @xmath520 the @xmath173-parameter family of the zeta functions @xmath521 and taking into account the above identity , one obtains for @xmath423 , @xmath522 now taking the derivative @xmath426 gives @xmath523 the calculation of the derivative corresponding to the area term is very easy . indeed , since @xmath110 is a trace , one simply has @xmath524 hence @xmath525 summing up and recalling that by its very definition , cf . , @xmath526 one concludes that @xmath527 to obtain the claimed expression of the gradient we appeal to theorem [ thmmain ] , move one of the exponential factors under the trace @xmath110 , and get @xmath528 one has @xmath529\ , , \ \ \int_{-1}^1 e^{\frac{u \lmod}{2}}du=4\frac{\sinh(\lmod/2)}{\lmod}\ ] ] thus the function @xmath35 gets multiplied by @xmath530 and becomes @xmath531 similarly , one has @xmath532 which gives . [ otherlap ] _ one may wonder if there is a similar manner of using the local formula for @xmath16 of theorem [ thmmain ] to handle @xmath533 . to this end , we start from the equality @xmath421 and the @xmath173-parameter family @xmath534 since @xmath535 one has @xmath536 the duhamel formula @xmath537 implies @xmath538 notice now that @xmath539 reverting us to the expression , which involves the local formula for @xmath15 and not for @xmath16 . thus , one ends up by merely recovering the same expression of the gradient . _ replacing in the definition the fixed inner product by the running one , one could alternatively define the gradient by means of the running inner product , as follows : @xmath540 then theorem [ gradthm ] gives @xmath541 based on the analogy with the standard torus ( comp . * , ( 3.8 ) ) ) , the right hand side @xmath542 can be taken as the appropriate definition of the * scalar curvature * @xmath543 of the conformal metric on the noncommutative torus associated to the given dilaton . the evolution equation becomes @xmath544 unlike the commutative case , the corresponding flow of inner products is not given by the same differential equation . denoting by @xmath545 the hermitian form @xmath546 one has @xmath547 denoting by @xmath548 the resulting hermitian form @xmath549 we conclude that the metric associated to @xmath157 has evolution equation @xmath550 the equation is precisely the analogue of hamilton s ricci flow for the standard torus , which justifies calling the hermitian form @xmath548 the * ricci curvature * of @xmath5 endowed with the metric @xmath545 . [ sectexample ] in order to treat this example it will be convenient to represent the algebra @xmath190 as the crossed product of @xmath551 by the irrational rotation . more specifically we identify the @xmath88-subalgebra generated by @xmath552 with @xmath551 so that one has @xmath553 we then consider a powers - rieffel self - adjoint idempotent of the form @xmath554 one has , taking the @xmath555 real valued , @xmath556 [ propexample ] let @xmath557 be the idempotent given by , @xmath558 where @xmath559 and @xmath560 the associated conformal weight . one has the following closed formula for the ray - singer determinant : @xmath561 where @xmath562 since @xmath563 one gets @xmath564 and with @xmath436 the derivation implemented by @xmath22 one has @xmath565 since @xmath566=-s\left(e^2\delta_j(h)-e\delta_j(h)e\right ) = -s e\delta_j(h).\ ] ] thus the decomposition gives @xmath567 as a sum of eigenvectors for the eigenvalues @xmath568 for the operator @xmath436 . since the function @xmath39 is even we thus get @xmath569 and @xmath570 we now use our formula @xmath571 one has @xmath572 @xmath573 @xmath574 @xmath575 and using we obtain . let us now compute the scalar curvature . it is given by . one has @xmath576 which gives a decomposition in eigenvectors for @xmath436 with eigenvalues @xmath577 . it follows that @xmath578 one has moreover using the decomposition and the vanishing of @xmath579 , @xmath580 we get [ propexamplebis ] let @xmath557 be an idempotent , @xmath558 where @xmath559 and @xmath560 the associated conformal weight . the scalar curvature is given by the formula @xmath581 the above discussion yields the formula @xmath582 but since @xmath563 one has the relation @xmath583 which gives @xmath584 now , as we shall see below , one has the relation @xmath585 which will be understood conceptually as a special case of . we then use @xmath586 and simplify to @xmath587 which is the required equality . note that one checks that each of the two separate terms @xmath588 and @xmath589 have vanishing integral under @xmath110 i.e. the gauss bonnet theorem holds . [ sectnorma ] in this section we explain how to reformulate the above scalar curvature in intrinsic terms involving only the even two - dimensional modular spectral triple @xmath286 . let us first relate the weight @xmath157 to the natural volume form associated to the modular spectral triple . [ volume ] for any @xmath100 one has @xmath590 we shall show that @xmath591 note that since the kernel of @xmath235 is finite dimensional we do not care about the lack of invertibility of @xmath235 and define arbitrarily @xmath592 on the kernel , this does not affect the value of the residue . to prove note that @xmath593 and one gets @xmath594 where the constant @xmath595 comes from the equality @xmath596 which can be checked directly since both sides vanish on @xmath597 for @xmath480 . to compute @xmath595 one just needs the residue at @xmath598 of the zeta function @xmath599 which gives @xmath600 and proves . we shall not need the other part but it follows from the above . using the canonical volume form of the spectral triple instead of @xmath157 in the definition of the scalar curvature thus removes the unpleasant factor @xmath600 in the above formulas . there is one more small adjustment needed to obtain an intrinsic definition ; we have identified above the space of deformed modular spectral triples with the space of self - adjoint elements @xmath601 and the tangent space at a point @xmath9 accordingly by linearity . [ dimtwo ] in @xcite , definition 1.147 , the notion of scalar curvature functional was introduced for spectral triples @xmath602 of dimension @xmath603 by the equality @xmath604 the same formula with @xmath605 instead of @xmath606 works in dimension @xmath607 when @xmath608 with a suitable normalization . for @xmath609 the normalization factor has a pole and the analogue of becomes , @xmath610 where @xmath402 is the orthogonal projection on the kernel of @xmath2 . what the present development shows in particular is that in the even case the above general definition should be refined by using the chiral expression : @xmath611 a number of the above results extend naturally to the general case of dimension two , and relate the variation under inner twisting of the ray - singer torsion of modular spectral triples to their chiral scalar curvature . in this section we shall explain how to perform the direct computation of the gradient of an expression of the form @xmath612 we shall then show how this , together with theorem [ gradthm ] , implies the following relation between @xmath39 and @xmath513 , @xmath613 this relation can then be checked directly and gives the following decomposition @xmath614 @xmath615 the fact that can be proven on a priori ground gives a handle on the complicated two variable functions which appear in theorem [ thmmain ] since by one can deduce the function @xmath36 from @xmath513 . moreover that the function @xmath616 is the generating function for bernoulli numbers since one has @xmath617 [ directgrad ] let @xmath618 be an even schwartz function , then with @xmath619 one has @xmath620 where the function @xmath621 is given by @xmath622 it is enough to prove the statement with @xmath623 replaced by @xmath624 and @xmath625 by @xmath626 where @xmath627 is a derivation equal to @xmath221 or @xmath628 . one has @xmath629 the proof of theorem [ directgrad ] then follows from the equality between the last two terms and the following two general lemmas . [ doubletransfo ] let @xmath618 be an even schwartz function , then for any @xmath630 one has @xmath631 where the function @xmath632 is given by @xmath633 using the fourier transform @xmath634 and the equality @xmath635\qqq z\in a_\theta^\infty,\ ] ] we write @xmath636 since @xmath637 one obtains @xmath638 one has @xmath639 @xmath640 which is of the form @xmath641 and gives for @xmath642 which gives @xmath643 one then performs the integral to obtain . as a simple example we take @xmath644 . in that case one has @xmath645 ^ 2)\ ] ] and the right hand side gives @xmath646 ^ 2)=-2\vp_0([a , x][h , x])= \vp_0(a h(\nabla_1,\nabla_2)(xx))\ ] ] with @xmath647 . let us now consider the term @xmath648 we need to integrate by parts , which is achieved as follows . [ doubletransfobis ] let @xmath618 be a schwartz function , then for any @xmath649 one has @xmath650 where the function @xmath651 is given by @xmath652 one has for any @xmath653 the equality @xmath654 so that @xmath655 and taking @xmath656 we get , @xmath657 where @xmath658 now writing @xmath659 one gets @xmath660 where @xmath661 which is the same as . one has , using integration by parts @xmath662 and using one obtains . the computation follows the same lines as @xcite . the case of the operator @xmath663 is treated at the symbol level as in @xcite . we first give the analogue of lemma 6.1 of @xcite for the operator @xmath664 . this follows from the derivation property of @xmath669 which gives @xmath670 which is the decomposition of the operator as a sum of homogeneous terms . in general we recall the product formula within the algebra of symbols , where with @xmath671 , @xmath672 , one has @xmath673 \ , .\ ] ] one then proceeds as in @xcite for the inductive calculation of the inverse of the symbol of @xmath674 , using @xmath595 as a symbol of order two and @xmath675 and computing to order @xmath676 in @xmath677 the product @xmath678 . by singling out terms of the appropriate degree @xmath679 in @xmath677 , one obtains @xmath680 note that @xmath681 appears on the right in this formula , and we refer to @xcite for detailed explanations . in a similar fashion , collecting terms of degree @xmath682 in @xmath677 one obtains @xmath683 the resulting formula for @xmath684 is quite long and the next step is to perform the integration @xmath685 . note , before starting , that by homogeneity of symbols one has @xmath686 and we thus know that @xmath687 is homogeneous of degree @xmath679 in @xmath595 since @xmath688 moreover the next step in order to obtain the constant term in the heat expansion is to do an integral in the variable @xmath595 of the form @xmath689 where the contour @xmath690 around the positive real axis is chosen in such a way that @xmath691 applying this equality for @xmath6 one gets that @xmath692 thus we can already simplify and fix @xmath693 before we perform the integration in @xmath694 . to start this integration we follow @xcite and perform the change of variables , using @xmath695 and @xmath289 , @xmath696 the jacobian of the change of coordinates is @xmath697 . moreover after changing variables one gets @xmath698 one performs the integration in the angular variable @xmath87 first and the terms one obtains can be organized in the form @xmath699 where one will note the overall multiplication on the right by @xmath681 which for our purpose can not be moved in front as was done in @xcite . t_0,1=i r^2 b_0 k _ 1(k ) b_0 k _ 2(k)+i r^2 b_0 k _ 1(k ) b_0 _ 2(k ) k - i r^4 b_0 k _ 1(k ) k^2 b_0 ^ 2 k _ 2(k ) + -i r^4 b_0 k _ 1(k ) k^2 b_0 ^ 2 _ 2(k ) k - i r^2 b_0 k _ 2(k ) b_0 k _ 1(k)-i r^2 b_0 k _ 2(k ) b_0 _ 1(k ) k + + i r^4 b_0 k _ 2(k ) k^2 b_0 ^ 2 k _ 1(k)+i r^4 b_0 k _ 2(k ) k^2 b_0 ^ 2 _ 1(k ) k+i r^2 b_0 _ 1(k ) k b_0 k _ 2(k ) + + i r^2 b_0 _ 1(k ) k b_0 _ 2(k ) k - i r^4 b_0 _ 1(k ) k k^2 b_0 ^ 2 k _ 2(k)-i r^4 b_0 _ 1(k ) k k^2 b_0 ^ 2 _ 2(k ) k + -i r^2 b_0 _ 2(k ) k b_0 k _ 1(k)-i r^2 b_0 _ 2(k ) k b_0 _ 1(k ) k+i r^4 b_0 _ 2(k ) k k^2 b_0 ^ 2 k _ 1(k ) + + i r^4 b_0 _ 2(k ) k k^2 b_0 ^ 2 _ 1(k ) k. t_0=t_2=-2 r^2 k^2 b_0 ^ 2 k _ 1 ^ 2(k)-4 r^2 k^2 b_0 ^ 2 _ 1(k ) _ 1(k)-2 r^2 k^2 b_0 ^ 2 _ 1 ^ 2(k ) k + + 2 r^4 k^4 b_0 ^ 3 k _ 1 ^ 2(k)+4 r^4 k^4 b_0 ^ 3 _ 1(k ) _ 1(k)+2 r^4 k^4 b_0 ^ 3 _ 1 ^ 2(k ) k - r^2 b_0 k _ 1(k ) b_0 k _ 1(k ) + -r^2 b_0 k _ 1(k ) b_0 _ 1(k ) k+r^4 b_0 k _ 1(k ) k^2 b_0 ^ 2 k _ 1(k)+r^4 b_0 k _ 1(k ) k^2 b_0 ^ 2 _ 1(k ) k + -r^2 b_0 _ 1(k ) k b_0 k _ 1(k)-r^2 b_0 _ 1(k ) k b_0 _ 1(k ) k+r^4 b_0 _ 1(k ) k k^2 b_0 ^ 2 k _ 1(k ) + + r^4 b_0 _ 1(k ) k k^2 b_0 ^ 2 _ 1(k ) k+6 r^4 k^2 b_0 ^ 2 k _ 1(k ) b_0 k _ 1(k)+6 r^4 k^2 b_0 ^ 2 k _ 1(k ) b_0 _ 1(k ) k + -2 r^6 k^2 b_0 ^ 2 k _ 1(k ) k^2 b_0 ^ 2 k _ 1(k)-2 r^6 k^2 b_0 ^ 2 k _ 1(k ) k^2 b_0 ^ 2 _ 1(k ) k+6 r^4 k^2 b_0 ^ 2 _ 1(k ) k b_0 k _ 1(k ) + + 6 r^4 k^2 b_0 ^ 2 _ 1(k ) k b_0 _ 1(k ) k-2 r^6 k^2 b_0 ^ 2 _ 1(k ) k k^2 b_0 ^ 2 k _ 1(k)-2 r^6 k^2 b_0 ^ 2 _ 1(k ) k k^2 b_0 ^ 2 _ 1(k ) k + -4 r^6 k^4 b_0 ^ 3 k _ 1(k ) b_0 k _ 1(k)-4 r^6 k^4 b_0 ^ 3 k _ 1(k ) b_0 _ 1(k ) k-4 r^6 k^4 b_0 ^ 3 _ 1(k ) k b_0 k _ 1(k ) + -4 r^6 k^4 b_0 ^ 3 _ 1(k ) k b_0 _ 1(k ) k. t_1,0^(2 ) = -4 k^2 b_0 ^ 2 k _ 1 _ 2(k)-4 k^2 b_0 ^ 2 _ 1(k ) _ 2(k)-4 k^2 b_0 ^ 2 _ 2(k ) _ 1(k)-4 k^2 b_0 ^ 2 _ 1 _ 2(k ) k + -b_0 k _ 1(k ) b_0 k _ 2(k)-b_0 k _ 1(k ) b_0 _ 2(k ) k - b_0 k _ 2(k ) b_0 k _ 1(k)-b_0 k _ 2(k ) b_0 _ 1(k ) k + -b_0 _ 1(k ) k b_0 k _ 2(k)-b_0 _ 1(k ) k b_0 _ 2(k ) k - b_0 _ 2(k ) k b_0 k _ 1(k)-b_0 _ 2(k ) k b_0 _ 1(k ) k. we check that the coefficient @xmath707 of @xmath708 in the term @xmath54 is non - zero only for even @xmath709 and that the total power of @xmath681 involved in @xmath710 is @xmath711 . thus for @xmath712 as in we get that the general form is a sum @xmath713 in order to put all these terms in a canonical form one moves the powers of @xmath60 to the left using the commutation of @xmath60 with @xmath681 and the rule @xmath714 thus for instance the first term of @xmath715 which is @xmath716 is rewritten as @xmath717 for such terms which are quadratic in the @xmath718 we can use the simple formula @xmath719 which is justified below in [ expansect ] . thus the above term can be written as @xmath720 @xmath721 besides terms which are quadratic in the @xmath718 we also get terms which involve second derivatives of @xmath60 . thus for instance the first term of @xmath722 gives @xmath723 for the terms which involve second derivatives of @xmath60 we need to carefully reexpress them in terms of second derivatives of @xmath724 as we now explain . we write the modular automorphism in the form @xmath725 so that one has the permutation rule @xmath726 the expansional formula can be written as @xmath727 we take @xmath728 and for @xmath729 the term one gets by expanding @xmath730 around @xmath731 , i.e. using purely imaginary arguments @xmath732 , @xmath733 one has a similar expansion for @xmath734 which shows that e.g. @xmath735 is obtained from the coefficient of @xmath736 in @xmath737 . more precisely one writes @xmath737 as @xmath738 @xmath739 @xmath740 where in the second integral one lets @xmath741 which varies from @xmath246 to @xmath173 and @xmath742 which varies from @xmath246 to @xmath743 . in terms of the derivations @xmath744 this gives the formula @xmath745 where one has @xmath746 and @xmath747 note the coefficient @xmath8 in front of @xmath748 since @xmath735 is obtained from the coefficient of @xmath736 in @xmath737 . in order to perform the integration in the radial variable @xmath749 one needs a more general lemma than lemma 6.2 of @xcite . note that only even powers of @xmath749 appear in the expressions for the terms @xmath54 but since one needs to multiply by the jacobian of the change of coordinates , the integration in @xmath749 only involves odd powers of @xmath749 . thus it is natural to let @xmath750 so that @xmath751 . [ permlem ] for every element @xmath752 of @xmath140 and every integers @xmath753 one has , @xmath754 where the function @xmath755 is @xmath756 and @xmath757 signifies that @xmath18 acts on the @xmath28-th factor . let @xmath758 be the inverse fourier transform of the function @xmath759 one has @xmath760 @xmath761 and more generally @xmath762 the role of the polynomials @xmath763 which appear in the numerator is to compensate for the zeros of the denominator in larger and larger strips . thus the imaginary part of the first singularity of @xmath758 is @xmath764 . the inverse fourier transform of the function , defined for @xmath7650,n[$ ] by @xmath766 is @xmath767 so that @xmath768 we now perform in the left hand side of the change of variables @xmath769 , with @xmath770 , and obtain @xmath771 we now choose positive real numbers @xmath772 such that @xmath773 and replace each term @xmath774 by @xmath775 . this is fine for the @xmath306 variable since it accounts for the term @xmath776 , but taking care of the @xmath752 one gets @xmath777 where @xmath778 we set @xmath779 . using we can then write @xmath206 as an integral of terms of the form @xmath780 with respect to the measure given by @xmath781 the term can be written as @xmath782 one has @xmath783 thus one obtains @xmath784 we now replace @xmath779 and replace the term @xmath785 by @xmath786 and we are dealing with the scalar function of @xmath449 variables @xmath787 we can now write @xmath788 with @xmath789 we can perform the integral in @xmath732 one gets that the coefficient of @xmath732 in the exponent is @xmath790 so that the integral in @xmath732 gives the fourier transform of @xmath791 at @xmath792 . this is @xmath793 in the product of these terms from @xmath794 to @xmath795 one gets @xmath796 with the exponent @xmath797 . thus this cancels the term @xmath798 . we thus get @xmath799 which proves the required equality . a. connes , h. moscovici , _ type iii and spectral triples _ , * traces in geometry , number theory and quantum fields * , aspects of mathematics * e 38 * , vieweg verlag 2008 , pp . 5771 ; http://www.arxiv.org/abs/math.oa/0609703 . a. connes , p. b. tretkoff _ the gauss - bonnet theorem for the noncommutative two torus _ , * noncommutative geometry , arithmetic , and related topics * , johns hopkins university press , proceedings of the twenty - first meeting of the japan - u.s . mathematics institute edited by caterina consani and alain connes ( 2011 ) ; arxiv:0910.0188 .

in this paper we investigate the radio - mir correlation at very low flux densities using extremely deep 1.4ghz sub - arcsecond angular resolution merlinvla observations of a 85 field centred upon the hubble deep field north , in conjunction with 24 data . from these results the mir - radio correlation is extended to the very faint (microjy ) radio source population . tentatively we detect a small deviation from the correlation at the faintest ir flux densities . we suggest that this small observed change in the gradient of the correlation is the result of a suppression of the mir emission in faint star - forming galaxies . this deviation potentially has significant implications for using either the mir or non - thermal radio emission as a star - formation tracer of very low luminosity galaxies . [ firstpage ] galaxies : starburst galaxies : high - redshift infrared : galaxies radio continuum : galaxies

0801.1035

since 1970s and 1980s studies of the radio and far - infrared ( fir ) properties of galaxies have shown there to be a tight correlation between their emission in these two observing bands which extends over several orders of magnitude in luminosity @xcite . the advent of the _ infrared astronomical satellite _ ( iras ) all sky survey in 1983 @xcite enabled much larger systematic samples of galaxies to be studied at infrared wavelengths and hence further demonstrated the consistency and tightness of this correlation , albeit for relatively nearby sources @xcite . following _ iras _ , deep observations using the _ infrared space observatory _ ( _ iso _ ) allowed fainter and higher redshift galaxies to be observed at mid - infrared ( mir ) wavelengths . these _ iso _ observations showed that the mir emission from galaxies is loosely correlated with radio emission across a wide range of redshifts , tentatively extending to z@xmath24 @xcite . more recently the launch of the space telescope in august 2003 has greatly increased the sensitivity of mir observations and hence our ability to study the mir - radio correlation . early results , such as from the first look survey ( fls ) , have confirmed that the mir - radio correlation holds for relatively bright star - forming galaxies ( s@xmath3@xmath4jy ) out to at least a redshift of 1 @xcite . radio and infrared emission from galaxies in both the nearby and distant universe is thought to arise from processes related to star - formation , hence resulting in the correlation between these two observing bands . the infrared emission is produced from dust heated by photons from young stars and the radio emission predominately arises from synchrotron radiation produced by the acceleration of charged particles from supernovae explosions . it has however recently been suggested that at low flux density and luminosities there may be some deviation from the tight well - known radio - ir correlation seen for brighter galaxies @xcite . ( 16,5 ) ( -0.5 , 0 ) ( 6.5 , 0 ) -0.5 cm @xcite argue that while the ir emission from luminous galaxies will trace the majority of the star - formation in these sources , in low luminosity galaxies the ir emission will be less luminous than expected considering the rate of star - formation within the source ( i.e. the ir emission will not fully trace the star - formation ) . in this scenario the reduced efficiency of ir production relative to the source star - formation rate ( sfr ) would be the result of inherently lower dust opacities in lower luminosity sources and consequently less efficient reprocessing of uv photons from hot young stars into ir emission . the simple consequence of this is that at lower luminosities the near linear radio - ir correlation l@xmath5l@xmath6 , with @xmath7 ( e.g * ? ? ? * ; * ? ? ? * ) will be deviated from . _ of course such an assertion is dependent upon the radio emission providing a reliable tracer of star - formation at low luminosities which may be equally invalid . _ recently @xcite have presented a statistical analysis of australia telescope compact array ( atca ) 20 cm observations of the 24 sources within two regions ( the _ chandra _ deep field south ( cdfs ) and the european large area _ iso _ survey s1 ( elais ) ) of the wide field survey ( swire ) . in this work @xcite have co - added sensitive ( rms@xmath230@xmath4jy ) radio data at the locations of several thousand 24 sources . using this method they have statistically detected the microjy radio counterparts of faint 24 sources . at low flux densities ( s@xmath8jy ) they confirm the ir - radio correlation but find it to have a lower coefficient ( s@xmath9=0.039s@xmath10 ) than had previously been reported at higher flux densities . this coefficient is significantly different from results previously derived from detections of individual objects ( e.g. @xcite ) and is speculated by @xcite to be the result of a change in the slope of the radio - ir correlation at low flux densities . in this paper , we utilise very deep , high resolution 20 cm observations of the hubble deep field north and surrounding area made using merlin and the vla @xcite in combination with publicly available 24 source catalogues from goods to study the mir - radio correlation for microjansky radio sources . this study extends the flux density limits of the radio - ir correlation by more than an order of magnitude for individual sources and overlaps the flux density regime studied using statistical stacking methods by @xcite . additionally we employ statistical stacking methods , similar to those used by @xcite , to extend the correlation further to still lower flux densities . we adopt h@xmath1175kms@xmath12mpc@xmath12 , @xmath13 and @xmath14=0.7 throughout this paper . extremely deep radio observations of the hdf - n region were made in 1996 - 97 at 1.4ghz using both merlin and the vla . these observations were initially presented in @xcite , @xcite and @xcite . the results from the combined 18 day merlin and 42hr vla observations are described in detail in @xcite . the combined merlin@xmath0vla image has an rms noise level of 3.3@xmath4jy per 02 circular beam making it amongst the most sensitive , high - resolution radio images made to date . using the same methods as described in @xcite these combined merlin and vla observations have been used to image the entire unaberrated field of view , [email protected] arcmin@xmath15 in size , centred upon the original merlin pointing position ( @xmath1636494000 , @xmath171258000 ( j2000 ) ) . this image has an rms noise level of 3.6@xmath4jybeam@xmath12 and has been convolved with a 04 circular beam . these observations have been shown to align with the international coordinate reference frame ( icrf ) to better than 15mas @xcite . as part of the goods enhanced data release ( dr1 + february 2005 ) a catalogue of 24@xmath4 m source positions and flux densities for the goods - n field were released . this source catalogue is limited to flux densities @xmath1880@xmath4jy providing a highly complete and reliable sample . at the time of writing this 24@xmath4 m source catalogue represents the most complete and accurate mid - infrared source list publicly available for the goods - n / hdf - n field . all 24 sources which are detected optically in goods _ hst _ acs images show an accurate astrometric alignment with their optical counterparts implying that the astrometry between these two data - sets and their subsequent catalogues is self - consistent . however , a comparison of the astrometric alignment of the positions of sources catalogued by goods derived from their _ hst _ acs images ( richards _ et al . _ 2007 ; muxlow _ et al . _ _ in prep _ ) shows there to be a systematic offset in declination of @xmath190342 from the radio reference frame . this linear declination correction , although small relative to the resolution at 24 is significant when compared to these high resolution radio data . this linear correction has been applied to the source positions prior to all comparisons between the two data sets . 85 hdf - n field are plotted individually ( small black triangles ) . the median radio flux density logarithmically binned by 24 flux densities are plotted as large filled red circles . the solid red line represents the best - fit line to the binned hdf - n data alone . sources within the cdfs - swire field detected at both 24 and 1.4ghz from norris _ et al . _ ( 2006 ) are plotted as either green open stars ( identified as agn ) or green triangles ( not identified as agn ) . all sources detected at both 1.4ghz and 24 in the first look survey ( fls ) with source position separations of @xmath2015 are plotted in orange @xcite . note the quantisation of the swire and fls points , in this and subsequent plots is a result of the accuracy of the flux densities tabulated in the literature . the blue pluses and fitted lines in the low portion of the plot show the ir - radio correlation derived from stacking radio emission at the positions of 24 sources in the cdfs and elais field by boyle _ ( 2007 ) . a brief summary of the basic characteristics of each of the data sets plotted included in table[tab2].,width=340 ] cccccc survey&radio&instrument&radio senitivity&angular&reference + & survey area&&(@xmath4jybeam@xmath12)&resolution & + hdf - n / goods - n&72.25arcmin@xmath15&vla+merlin&3.3&02@xmath102&muxlow _ et al . _ 2005 , this paper + atlas ( cdfs / elais)&3.7deg@xmath15&atca&20@xmath2160&11@xmath225@xmath23&norris _ et al . _ 2006 + fls&5deg@xmath15&vla&23&5@xmath23&condon _ et al . _ 2003 , fadda _ et al . _ 2006 + the field covered by these goods - n 24 observations contains 1199 24- sources identified with flux densities @xmath1880@xmath4jy . the [email protected] arcmin@xmath15 radio field contains 377 of these sources with 24 flux densities ranging from 80.1 to 1480@xmath4jy . radio emission with the deep merlin+vla data has been measured at the corrected position of each individual 24 source within the radio image . the radio flux density was measured within a series of concentric rings with a maximum radius of 4arcsec centred on each 24 source . statistical analysis of these measurements and a sample of stronger radio sources within the same field shows that the radio flux density increases within progressively larger annuli out to a radius of 15 . for stronger ( @xmath24 ) radio sources the total flux density recovered using this method within a radius of 15 is between 90@xmath25 and 100@xmath25 of the source s total flux density as measured by other means , such as gaussian fitting . additionally the average ( either mean or median ) radial profile of all of the 24 sources shows that almost all of the radio flux density is recovered within radii less than 15 ( @xcite ; beswick _ et al . _ _ in prep _ ) . throughout the rest of this paper the radio flux density recovered within a radius of 15 of each ir source will be assumed to be equal to the total radio source flux density . ( 16,11 ) ( -1.2 , 0 ) ( 5.8 , 0 ) -1 cm as a confirmation of this method several thousand random sky positions within the radio field were selected and cross - referenced to exclude any which were at the positions of known sources at any wavelengths . at each of these blank positions the radio flux density was measured using an identical method as outlined above . the flux densities recorded within each of these annuli showed no positive bias within any ring and followed the same gaussian distribution as derived from the pixel - by - pixel noise statistics of the whole image once all radio sources greater than 40@xmath4jy have been subtracted . as further test of these methods 90 fake , faint ( @xmath26 ) and partially - resolved radio sources were injected into the _ u - v _ data . these data were then fourier inverted and cleaned in an identical manner to the real data and then the flux densities of these fake sources were extracted using the methods outlined above . whilst , as expected , these sources were not detected individually their radio emission , when averaged together , was consistent with the average flux density of the population of fake sources injected into the data . a more detailed description of this method and the radio characteristics of this sample of 24 source , including source catalogues and size information , will be presented in a forthcoming paper . of the 377 24 sources within the radio field 303 were found , by this method , to have total radio flux densities greater than 3 times the local map rms . many of these source have peak flux densities below 3-@xmath27 and hence were not selected using the radio - only detection methods employed by @xcite . the 1.4ghz versus 24 flux densities of all 377 of these source are plotted in left - hand panel of fig[f20vsf24 ] . the control sample , plotted in the right - hand panel of fig[f20vsf24 ] , is derived using an identical radio flux extraction methods and from the same radio data but at randomly assigned locations coincident with no known source position . the flux densities of these sources are also plotted in fig[f20vsf24log ] , along with sources detected in the shallower surveys of the cdfs - swire field by @xcite using atca , and the fls using the vla by @xcite . sources from the southern cdfs - swire field which have been categorised by @xcite as containing agn ( plotted as stars ) are generally situated toward the upper half of this figure and show an expected radio excess when compared to their 24 flux densities . source which were not identified as agn by @xcite are plotted as triangles in this figure . it should be noted that these sources have not been classified as containing an agn but may in fact be either agn , star - forming galaxies or a combination of the two . overlaid in blue in the lower half of fig.[f20vsf24log ] are the flux densities of groups of median stacked infrared sources from the elais and cdfs field from @xcite along with fits to these data . also overlaid , as filled red circles , in this figure are the median values of all of the hdf - n sources logarithmically binned by their 24 flux densities . there is a clear discrepancy between the averaged radio flux densities derived from these hdf - n observations and the flux densities from the elais and cdfs fields @xcite which will be discussed further in the following sections . rrrrrrrr & & & + & & median&mean&&&median&mean + s@xmath28 & number & s@xmath9&s@xmath9&@xmath29&number&s@xmath9&s@xmath9 + ( @xmath4jy ) & & ( @xmath4jy)&(@xmath4jy)&(@xmath4jy)&&(@xmath4jy)&(@xmath4jy ) + 88.8&65&28.9&22.6&0.68&65&26.7@xmath308&27.8@xmath306 + 112.5&72&34.1&32.0&0.62&70&34.0@xmath307&32.9@xmath305 + 138.9&57&49.5&47.7&0.81&54&63.1@xmath309 & 58.1@xmath307 + 174.7&46&58.8&56.1&1.21&44&64.8@xmath3010&60.4@xmath308 + 219.4&49&69.4&90.7&1.68&49&76.1@xmath308 & 90.4@xmath306 + 275.0&22&72.5&79.0&3.01&22&81.9@xmath3013&53.3@xmath304 + using an analysis method similar to @xcite the mean and median images of the radio emission from hdf - n field of the 6 lowest 24 flux density logarithmic bins ( as plotted in figs.[f20vsf24 ] and [ f20vsf24log ] ) have been derived and are shown in fig.[maps ] . each of these images is the statistically combined radio emission from the location of several tens of 24 sources and has been contoured at multiples of three times 3.3@xmath4jybeam@xmath12 divided by @xmath31 where n equals the number of source positions that have been stacked together . as can be seen in these images the off - source noise levels achieved approaches the value expected when co - adding multiple images with near - gaussian noise properties . the co - added image rms achieved in the faintest 24 flux density bin ( 80.0 to 100.25@xmath4jy ) is 0.45 and 0.56@xmath4jybeam@xmath12 in the mean and median co - added images respectively . the flux densities measured directly from these stacked radio images are consistent with the statistical mean and median flux densities within the equivalent bins derived from the individual source flux extraction method described in section 2.3 ( table[tab1 ] ) and plotted in figs.[f20vsf24 ] and [ f20vsf24log ] . in each stacked image the largest angular size of the regions showing elevated radio emission within fig.[maps ] is enclosed within the 15 radius circular area from which the individual source flux densities have been extracted . the sizes of the composite radio sources in these co - added images ( fig.[maps ] ) represent a combination of the true radio source size , any offsets between the ir and radio emitting regions within the sources and any random errors in the source positions , all of which will be convolved with the synthesised beam applied to these radio data ( 04 circular beam in this case ) . no systematic offset is seen between the emission in these composite radio images from their nominal image centres implying that no significant non - random errors between the astrometric alignment of these data - sets remain . further analysis of the radio source sizes and structures will be the subject of a future paper . in fig.[maps ] the mean ( left ) images of the radio emission for the two highest 24 flux density bins are dominated by emission from a few bright sources . several of these brighter radio sources can be seen in fig.[f20vsf24 ] scattered above s@xmath32200@xmath4jy . the presence of these brighter radio sources and the low numbers of images stacked together is clearly evident when the mean and median averaged images are compared . in particular in the second brightest bin ( 197.27 to 247.27@xmath4jy ) two mjy radio sources are present and offset from their equivalent 24 position . it is likely that these bright radio sources , which deviate from the radio - ir correlation , may contain radio agn which would feasibly be spatially offset from any ongoing star - formation and 24 position . an alternative method of projecting the correlation between these 24 and 1.4ghz data is as a function of the commonly defined parameter q@xmath33 , where q@xmath33 = log(s@xmath10/s@xmath9 ) . the mean and median value of 24 for all of the hdf - n sources presented in this study is 0.52 and 0.48 respectively , compared to the value of [email protected] and 0.73 derived from non - agn sources catalogued in observations of the swire field by @xcite . a similar study of brighter sources in the fls @xcite show the mean value to be ( [email protected] ) slightly higher than average value derived here . the distribution of 24 values for all sources in this hdf - n study is shown in fig.[histq24 ] . in figs.[q24_1 ] , [ q24_1b ] , [ q24_2 ] and [ q24_3 ] the parameter 24 is plotted against s@xmath10 , s@xmath9 and redshift . the monochromatic value of 24 represents the slope of the ir - radio correlation and the size of dispersion in 24 is related to the strength of the correlation . in each of these figures no _ k_-correction has been applied to either the radio or ir luminosities since we have incomplete redshifts for the sample , and inadequate spectral information is known about the majority of the individual sources in either the ir or radio bands . a similar study by @xcite has shown that _ k_-corrections made to individual sources result in only a minimal systematic effect on the average value of 24 below redshift @xmath21 . in fig.[q24_1 ] the values of 24 for sources from these hdf - n data , the swire field @xcite and the fls field @xcite are plotted against 24 flux density . in the upper panel both the agn identified and non - agn sources of @xcite are shown whilst in the lower panel the agn sources are excluded for clarity . in the upper panel the agn sources from the swire field , as expected , deviate from the correlation generally showing an excess of 1.4ghz radio emission compared to their 24 emission . the hdf - n field covers a considerably smaller area and was originally chosen to have few bright radio - agn , which is confirmed by the minimal number of hdf - n sources displaying an excess of 1.4ghz radio emission . amongst the hdf - n sources and the non - agn sources from the swire field ( * and fig.[q24_1 ] here ) 24 shows a small trend toward lower values at smaller 24 flux densities . the swire sources plotted here have been selected to have both radio and ir flux density above a certain threshold . the lowest radio flux density of sources within the swire field catalogue @xcite is 100@xmath4jy . a line of constant radio flux density ( 100@xmath4jy , the lowest individual source detections in the swire catelogue ) is overlaid on fig.[q24_1 ] above which sources from the swire study are excluded . the hdf - n radio data plotted here has a sensitivity limit @xmath210 times lower than that of the swire data , thus potentially only excluding sources in the extreme top left - hand part of this plot . furthermore , the sample of sources plotted here has been subject to no radio detection threshold limit hence removing this potential bias completely . the 24 values for the hdf - n data and the sources from @xcite binned as a function of s@xmath10 , along with the ratio of s@xmath10 and s@xmath9 are shown in fig[q24_1b ] . especially within the binned points , these two plots show a tentative trend for low 24 and @xmath34 with declining 24 flux density . the gray areas and dotted line in these two panels show the value of the 24 as derived from the fls @xcite . the hdf - n observations presented here are consistent with those of @xcite at high 24 flux densities . the effect of the 24 sensitivity cut - off ( 80.1@xmath4jy ) in the goods - n data and the equivalent limit in the data of @xcite will , however , positively bias measurements of 24 at low radio flux densities . this is clearly shown in fig.[q24_2 ] . the lower flux density limit imposed by the sensitivity results in an exclusion of sources in the lower portion of fig.[q24_2 ] . this consequently biases the determination of the value of 24 for any complete sample of sources with low radio flux densities . of the 377 source within the 85@xmath185 hdf - n radio field , 259 of these galaxies have published spectroscopic or photometric redshifts . the 24 values for this subset of sources , along with 50 non - agn swire sources , are plotted with respect to their redshift in fig[q24_3 ] . within these two data - sets 24 is seen to slightly reduce at higher redshifts although this effect is small and significantly less than the scatter . these values of 24 have not been _ k_-corrected . the application of a _ k_-correction to these data will result in a small increase in 24 which will increase as a function of z. it is interesting to also note that recent studies of discrete areas within 4 very nearby star - forming galaxies @xcite have shown 24 to vary across the extent of individual sources . however , @xcite derive a mean value of 24@xmath351 when integrating over entire galaxies in their sample , at z@xmath350 , which is consistent with many of the source in both the swire and hdf - n samples but somewhat higher than the average value derived at higher redshifts . using the available redshifts the observed luminosity has been calculated and is plotted in fig.[lumlum ] , including both the sources in the hdf - n region and the non - agn sources in the swire field . as can be seen in fig.[lumlum ] the 24-to-1.4ghz correlation , for individual sources , extends down to l@xmath36whz@xmath12 ( l@xmath37whz@xmath12 ) . 85 field centred upon the hdf - n.,width=340 ] the primary results of this work demonstrate that this sample of individual 24 selected sources follow the mir - radio correlation down to radio flux densities of a few microjy . these results are consistent with shallower independent radio surveys ( such as the atca observations of the swire field ) and seamlessly extend the ir - radio correlation down to lower flux densities ( see fig.[f20vsf24log ] ) . the majority of the ir selected sources within the hdf - n region have radio flux densities less than 100@xmath4jy , with only a few sources which show a significant excess of 1.4ghz flux density , when compared to the ir - radio correlation . this implies that the vast majority of these ir - selected microjy radio sources are primarily driven by star - formation with little significant contamination from agn . this is consistent with the results of @xcite who find that below 100@xmath4jy greater than 70 percent of radio sources are starburst systems . however , the data presented here are based on an ir selected sample rather than a radio selected sample . this will inherently result in the sample being dominated by star - forming systems but will not exclude the presence of a population of weak microjy agn sources . at the positions of the majority of the lowest 24 flux densities sources the peak radio emission is not significantly greater than the image rms . as a consequence most of these sources were not formally identified in the previous radio study @xcite . in many case , especially at low 24 flux densities , the extracted total radio flux density of individual sources is close to the radio sensitivity of these data ( fig[f20vsf24 ] ) . when directly compared with the control sample ( fig[f20vsf24 ] , _ right _ ) , which has been compiled by extracting the radio flux density from blank areas within the same data using identical methods , a significant correlation between the strength of the total radio and 24 emission is evident . however , the radio sensitivity limit of these data precludes the detailed discussion of individual low flux density sources . . all 377 source from the 85@xmath185 hdf - n field are plotted as individual sources ( small black triangle ) . in the upper panel sources detected at both 24 and 1.4ghz in the swire - cdfs field from norris _ et al . _ ( 2006 ) are plotted in green ( small triangles are sources not identified as agn and open stars have been identified as agn ) . sources from the fls are plotted in orange from fadda _ et al . _ ( 2006 ) . in the lower panel all of the identified sources in the hdf - n and the only non - agn sources from norris _ et al . _ ( 2006 ) are plotted . additionally plotted as ` pluses ' and crosses in the lower panel are the values of q@xmath38 as derived from the stacking analysis by boyle _ ( 2007 ) of the swire - cdfs and elais fields respectively . the solid green line overlaid on both panels is a line of constant radio flux density of 100@xmath4jy , the lowest radio flux density of sources in the swire sample plotted here.,width=340 ] ) versus 24 flux density is shown . the individual sources from the hdf - n field are plotted as small black triangles , the median values of these hdf - n source binned as a function of s@xmath39 as filled circles , green stars show the median binned values for sources listed as non - agn within the cdfs - swire sample of norris _ et al . _ ( 2006 ) and the blue ` crosses ' and ` pluses ' show the flux density ratios derived from the stacking analysis of the swire - cdfs and elais fields respectively from boyle _ the overlaid black dotted line is the mean value for ( @xmath40 ) derived by appleton _ this line is equivalent to q@[email protected] with the gray filled box representing the area enclosed by these errors and the flux density range investigated by appleton _ the values of q@xmath38 against 24 flux density are plotted in the lower panel . the symbols within this plot are identical to those in the upper panel , individual hdf - n sources are not included for clarity . the additional diagonal solid green line represents a line of constant radio flux density of 100@xmath4jy , the lowest flux density of sources in the swire sample plotted here ( sources above this line are excluded by this limit from the swire sample ) . this flux density limit will significantly effect the values of the binned swire data points ( green crosses ) negatively biasing the four lowest flux density bins . this bias only effects the swire sample . , width=340 ] . the samples plotted and symbols used are the same as those plotted in fig[q24_1 ] . in both plots the solid line represents a line of constant s@xmath41=80.1@xmath4jy which is equal to the detection threshold of the goods - n sample used.,width=340 ] the radio emission from many of the individual low flux density 24 sources is too faint to be imaged with high signal - to - noise in these radio data . however , it is possible to characterise their average radio emission by either binning the measured radio excess at their locations or by forming composite radio images from many sources . with the relatively limited sample of ir sources which are co - spatial with these radio observations ( 377 in total ) it is only statistically viable to use either binning or image stacking methods at the lowest ir flux densities where the number of sources that can be combined becomes large . overlaid in figs.[f20vsf24 ] and [ f20vsf24log ] are the median flux densities of ir sources in the hdf - n region imaged here . these bins have been logarithmically sampled as a function of 24 flux density and include all sources within these bins regardless of their measured radio flux density . whilst at the lowest emission levels the radio flux densities of individual sources will be subject to moderate or large errors the binned ensemble of these points will robustly represent the flux distribution of sample as a whole . using the radio flux densities extracted at the positions of the individual 24 sources , significant radio emission is detected in every averaged 24 flux density bin . to further test the viability of measuring the statistical radio flux density from sources in this sample below the formal radio detection limit of these merlin+vla data , the images have been stacked at the positions of multiple ir sources , binned as a function of their 24 flux density ( fig.[maps ] ) . these postage - stamp images have been averaged on a pixel - by - pixel basis , where each pixel in the stacked image is either the mean or median value of the pixels within the individual 24 source image stacks . each of the 6 lowest bins , for both the mean and median averaged images , show significant levels of radio emission . the 1.4ghz flux density in these images , extracted via gaussian fitting , is consistent with the binned level of emission derived from measuring individual sources , demonstrating the validity of either method ( table[tab1 ] ) . the sub - arcsecond angular resolution of these merlin+vla observations , in conjunction with their relatively good _ u - v _ coverage , allow the average radio structures of faint ir sources to be investigated . the median and mean images in fig.[maps ] , especially in the two highest flux density bins show some distinct differences . these differences arise due to inclusion of a few bright radio sources which dominate the average emission included in these bins . in particular in the second highest flux density mean image several ` bright ' radio sources are offset from the map centre . the positional offset of radio emission from these sources is real and indicates that , at least in these cases , that the origins of the radio and ir emission are different . the gaussian fitted sizes of the radio emission in the stacked images ( fig[maps ] ) provide an upper limit on the average size of the radio counterparts of these faint ir sources . the largest angular sizes of the radio emission in the median stacked images created from this sample range between 14 and 2 . this is approximately equivalent to a linear size of 10kpc at redshifts beyond 1 . these upper limits on the radio source sizes are consistent with radio emission on galactic and sub - galactic scales and originating within kpc - scale starburst systems . a more detailed analysis of the sizes and structures of the radio emission from both these composite images ( fig.[maps ] ) and the individual sources themselves is beyond the scope of this paper and will be presented in a latter publication . versus redshift . all 259 source from the 85@xmath185 hdf - n field are plotted as individual sources ( black triangle ) with known redshifts . fifty sources from the cdfs - swire field detected at both 24 and 1.4ghz , and redshift information from norris _ et al . _ ( 2006 ) are plotted as green triangle . only sources not identified as agn by norris _ _ 2006 are plotted . no k - correction has been applied to the data points plotted.,width=340 ] two directly comparable studies of the 1.4ghz radio and 24 mir emission from sources in the fls and the swire fields have been made by @xcite and @xcite respectively . @xcite compared observations of the fls made by @xcite using the vla in its b - configuration ( 5restoring beam ) with 24 and 70 observations . these observations had a detection threshold of 90@xmath4jy and 500@xmath4jy at 1.4ghz and 24 respectively . whereas @xcite co - added deep ( @xmath4230@xmath4jy / bm ) 1.4ghz atca observations of the southern swire field at the positions of many thousands of 24 sources , regardless of their detection at radio wavelengths , in order to statistically detect radio emission at the few @xmath4jy level . as a comparison to the hdf - n observations presented here , data points from the swire field @xcite , the fls @xcite and the co - added flux densities derived by @xcite are plotted in figs.[f20vsf24log ] , [ q24_1 ] , [ q24_1b ] and [ q24_2 ] . using the fls @xcite studied mir to radio correlation for 508 sources with redshift information , confirming results from similar previous surveys . following this work @xcite have published a more comprehensive catalogue of @xmath217,000 24 sources from the fls of which 2415 have radio counterparts within 4 . just over 2000 sources with radio counterparts within 15 are plotted in figs.[f20vsf24log ] , [ q24_1 ] and [ q24_2 ] . in general the sources plotted follow the mir to radio correlation but these data contain many outliers with elevated radio emission , characteristic of galaxies containing agn . the study of @xcite included sources with [email protected] and s@xmath4490@xmath4jy . the majority of this sample have redshifts of @xmath451 . the mean ( non-_k_-corrected ) value of q@xmath38 derived by @xcite is [email protected] , which is shown in fig[q24_1b ] . the non - agn sources detected at both 1.4ghz and 24 within the swire field catalogued by @xcite are also plotted in figs.[f20vsf24log ] , [ q24_1 ] and [ q24_2 ] . these sources are limited to @xmath460.1mjy and @xmath47170@xmath4jy at 1.4ghz and 24 respectively . using only the sources within the swire field identified as not being agn by @xcite the mean ( non-_k_-corrected ) values of q@xmath38 is [email protected] . the radio flux density threshold of both of these surveys overlaps with the brighter radio sources in this merlin+vla study . these deep hdf - n merlin+vla observations trace individual sources with considerably fainter flux density than those observed in the swire and fls surveys . these observations ( figs.[f20vsf24log ] , [ q24_1 ] and [ q24_2 ] ) are consistent with the extrapolation to lower flux densities of results found previously ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? however both the swire and fls surveys which sample sources with somewhat higher flux densities sources have mean values of q@xmath38 which are consistent but slightly higher than the value of [email protected] derived from these hdf - n data . averaging together the radio emission from multiple 24 source within the cdfs and elais fields @xcite have traced the radio emission against mir sources with flux densities comparable to those detected in these more sensitive hdf - n observations . @xcite have exploited the large number of mir sources detected in these fields to average together multiple radio images made at the mir source positions and statistically detect radio emission significantly below their radio detection threshold . this method is directly comparable to the methods applied to these radio observations of the hdf - n and is identical to the methods used to produce the stacked images presented in fig.[maps ] and table[tab1 ] . the median stacked radio flux densities of mir sources derived by @xcite show a significantly reduced s@xmath9 relative to s@xmath48 ( i.e. higher q@xmath38 value ) compared with the distribution of individual sources detected in the fls @xcite , the catalogued swire sources @xcite or the hdf - n results presented here . whilst there are several differences between the studies of @xcite and those presented here , such as the sensitivity ( rms@xmath230@xmath4jy versus @xmath23@xmath4jy ) and angular resolution ( 6versus 04 ) of the two surveys , it remains hard to simply reconcile these results . @xcite recognised the inconsistency of their calculated values of q@xmath38(=1.39 ) when compared with 0.84 derived by @xcite and have extensively simulated and tested the correctness of their analysis methods . since the results of @xcite are based on an infrared - selected sample with no radio detection limit , this discrepancy could possibly be explained by the presence of a previously unknown population of faint radio sources with a low radio to infrared flux density ratio . however the values derived by @xcite are not confirmed in the study presented here . a possible alternative explanation may be that the atca observations are underestimating the radio flux density of sources at the lowest flux densities when compared to these other surveys which have used radio data obtained using the vla and/or merlin . it is interesting to note that an earlier deep survey using the atca at 1400mhz ( atesp , australia telescope eso slice project ) @xcite found , upon comparison with nvss flux densities for the same field , that whilst for bright sources the atca and nvss flux density values were consistent , for faint sources the atca results underestimated the radio flux densities by up to a factor of 2 compared with the vla nvss . this underestimation of radio flux density was seen by @xcite to increase with diminishing radio brightness . whilst it remains unclear as to the cause , or reality , of this effect , if applicable to the elais and cdfs fields , it could result in the elevated values of 24 seen by @xcite at very low radio flux densities . it has been suggested , both theoretically @xcite and observationally @xcite , that the ir - radio correlation may evolve at lower flux densities . one of the aims of this study is to investigate this correlation for very low flux density radio sources . derived from the hdf - n data alone the average value of q@xmath38 shows a small trend to reduce as a function of lower values of s@xmath49 ( fig.[q24_1 ] & [ q24_1b ] ) . this can be further demonstrated by splitting these data into two approximately equal samples , above and below s@xmath50=150@xmath4jy . the mean and median values of q@xmath38 for the 187 sources below 150@xmath4jy are 0.45 and 0.36 respectively , compared with 0.56 and 0.54 derived from the 190 sources with s@xmath51150@xmath4jy . these hdf - n results along with results from samples of brighter galaxies @xcite are broadly consistent with average q@xmath38 reducing as a function of decreasing s@xmath49 ( see fig.[q24_1 ] & [ q24_1b ] ) . there are several contributing factors which will affect these results and those from previous studies . these factors include biases introduced by flux density thresholds of observations , the application or non - application of _ k_-corrections and possible agn contamination of sample galaxies . previous studies , such as @xcite , have used samples of sources detected at both radio and infrared wavelengths . the implications of a radio detection limit on these results will be to exclude sources with high q@xmath38 at low s@xmath10 . the line overlaid on fig.[q24_1 ] represents the upper limit imposed on the atca observations of the swire survey by the radio detection threshold of these data @xcite . in a manner similar to that used by @xcite no radio detection limit has been applied during analysis of these merlin+vla data with the radio flux density being recorded for all mir sources regardless of significance of their radio detection . this method will result in errors in the measured radio flux densities for very weak individual radio sources but will not directly bias against larger values of q@xmath38 . however when many sources are statistically averaged together the mean and median value of detected radio emission and hence the derived value of q@xmath38 will be robust even for source considerably fainter than the radio detection threshold of these data . no direct _ k_-correction has been applied to these hdf - n flux density results because in many cases the spectral shapes and redshifts of objects are not known adequately enough to apply any correction with confidence . additionally previous studies , such as by @xcite , have shown that the application of k - corrections to individual sources below z=1 only introduces a small change to the average value of q@xmath38 as a function of redshift . of the sources with known redshifts in our sample 78 percent have redshifts less than 1 , and thus will be minimally affected . within figs.[q24_1 ] and [ q24_2 ] the values of q@xmath38 for individual sources are plotted against their flux densities at each wavelength . these sources sample a wide range of luminosities and redshifts and hence the systematic effects resulting from not applying a _ k_-correction will effect individual sources but will cause only minimal changes in any overall trend of the sample as a function of flux density . however , the effects of the application of a _ k_-correction to these data do warrant further discussion . assuming that the 24 sources identified in this sample are predominately powered by star - formation , at an observed radio frequency of 1.4ghz the emission will be dominated by non - thermal synchrotron emission from supernovae and supernova remnants @xcite as such the non - thermal radio signal will be boosted by a factor of ( 1+@xmath52)@xmath53 , assuming a synchrotron power law ( s@xmath54 ) . this _ k_-correction will become incorrect for the highest redshift sources as the rest - frame radiation will be shifted to higher radio frequencies where the emission from other mechanisms such as thermal free - free emission will become increasingly important . for example a source at redshift of @xmath472.5 the radio emission observed at 1.4ghz will have been emitted at a rest - frame frequency of @xmath475ghz at which point an increasing proportion of the radio emission from a star - forming galaxy will arise from thermal processes resulting in a reduction in the radio spectral slope and hence _ k_-correction that should be required . accurately _ k_-correcting the 24 flux densities requires detailed knowledge of the mid - ir spectral template of the individual sources . in particular , for sources at redshifts @xmath55 significant spectral features from pahs and silicates observed in local starburst galaxies ( e.g. * ? ? ? * ) are shifted into the 24 band which can result in large additional _ k_-correction factors . the inclusion of these line features can result in an increase of a factor of 2 or more in s@xmath10 compared with commonly used starburst spectral templates @xcite . this is equivalent to a change in q@xmath38 of approximately 0.3 . the general effect of the mir _ k_-correction , resulting from a spectral index of a starburst galaxy excluding line emission , will be to boost s@xmath10 as a function of redshift by an amount that will depend upon spectral slope of the galaxy template used . for a luminous starburst this will result in a boosting with a mir spectral slope @xmath56 @xcite . as such the effect of this k - correction applied to the 24 data will be larger than that applied to the 1.4ghz data , hence the overall effect will be to increase the value of q@xmath38 as a function of redshift . uncorrected q@xmath38 is plotted against redshift for the sources with known spectroscopic or photometric redshifts in fig[q24_3 ] . applying _ k_-corrections to these data points will result in a small increase in q@xmath38 for the highest redshift sources plotted . within the scatter of the small sample of sources plotted in fig[q24_3 ] no significant trend is observed as a function of redshift to z@xmath353 . this is consistent with previous studies ( e.g. * ? ? ? * ; * ? ? ? in addition to instrumental and analysis effects outlined above it is possible that the flux densities of sources recorded in this study may include some proportion of contamination by embedded agn @xcite . the effect of the inclusion of sources with significant agn emission is demonstrated by the agn sources from the swire survey @xcite which are overlaid as stars in figs.[f20vsf24log ] , [ q24_1 ] and [ q24_2 ] . in general , sources with a significant agn contribution show an excess of radio emission compared to their infrared luminosity , hence will deviate from the ir - radio correlation . consequently the result of any agn contamination within the hdf - n sample analysed here will be to reduce q@xmath38 . however , the flux densities of the hdf - n sample ( see fig.[f20vsf24log ] ) show only a very small number of sources with excessive radio flux density which deviates significantly from the ir - radio correlation extrapolated from high flux density samples . this is to be expected since it is now generally accepted that below 100@xmath4jy the vast majority of radio sources are predominantly powered by star - formation @xcite . additionally this sample of sources is selected at 24 rather than at radio wavelengths and hence will preferentially select star - forming sources rather than radio agn . thus , throughout this paper no direct distinction has been made between possible agn and non - agn star - forming galaxies in these faint selected hdf - n radio sources . however , a small agn contamination within the sources sampled can not be eliminated . the consequences of any agn contamination may , to some degree , offer an explanation for the lower values of q@xmath38 derived here ( see for example * ? ? ? * ) . in the previous section a small trend in q@xmath38 as a function of 24 flux density has been tentatively identified . this trend is small and only visible in statistically averaged deep radio observations . the much deeper radio surveys that will be made using the newly enhanced radio instruments such as _ e-_merlin and the evla will constrain the faint end of the ir - radio correlation , confirming or refuting any possible deviations . whilst there are several observational and analysis effects which may contribute to the observed trend in the ir - radio correlation ( outlined in the previous sections ) there are also physical mechanisms which may cause an effect similar to that tentatively observed . both the ir and radio emission from star - forming galaxies are indirectly linked with the rate of ongoing star - formation ; simplistically the non - thermal radio emission originates in accelerated charged particles in supernovae and supernova remnants and the ir emission results from the reprocessing by dust of light from hot young stars . this inferred connection is generally considered to result in the ir - correlation . however since both the radio and ir emission are linked with star - formation via complex physical processes it is remarkable that the luminosity in these two bands correlates so accurately over such a large range of luminosities . a recent study by @xcite has provided a detailed investigation of the relationship of the radio and total infrared emission of galaxies and their star - formation rates . in this work several competing physical factors which will affect the ir and radio emission from star - forming galaxies are considered which at low luminosities potentially cause a deviation in the ir - radio correlation similar to that which is very tentatively reported here . if we assume that the non - thermal radio emission from star - forming galaxies is directly proportional to the star - formation rate ( i.e. radio @xmath57 sfr ) , then the systematic reduction in q@xmath38 at low 24 flux density implies that the 24 emission is no longer efficiently tracing the sfr of the galaxy @xcite . observational evidence ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) has shown that galaxy luminosity and dust opacity to uv and h@xmath58 emission are correlated , demonstrating that lower luminosity galaxies have substantially less dust absorption and reddening than high luminosity sources . these studies have also shown that the ratio of far - uv to ir light is significantly lower in less luminous sources , implying that whilst for luminous galaxies the ir provides a good tracer of star - formation at low luminosities much of the far - uv emission from star - formation is not reprocessed as ir emission ( see * ? ? ? * ; * ? ? ? * ; * ? ? ? * for further discussion ) . the consequences of the reduced dust opacity in lower luminosity , presumable smaller , sources will be a reduction in the efficiency of reprocessing of starlight into ir emission and hence will result in an observable decrease in the ratio of ir - to - radio emission from sources at low luminosities . of course it should be reiterated that this makes the assumption that the radio emission is a ` perfect ' tracer of the star - formation which may equally not be true . it has been suggested that the synchrotron radio emission from low - luminosity galaxies can also be suppressed , implying that it is not a true tracer of the sfr @xcite . whilst the connection between non - thermal radio emission and star - formation is complex , it has been discussed , both on the basis of models of the physical mechanisms and observational evidence , by many authors ( see for example * ? ? ? * ; * ? ? ? * ; * ? ? ? a physical explanation for the suppression of the non - thermal radio emission in low - luminosity galaxies is that the accelerating electrons can escape these sources because they have less efficient cosmic - ray confinement than in larger , more luminous sources @xcite . radio surveys of ir sources prior to the very deep results presented here have not detected any significant deviations in the linearity of the radio - ir correlation ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? @xcite argue that this observed linearity at relatively low luminosities implies that both the ir and non - thermal radio emission do not adequately trace the star - formation in low luminosity sources . as such the observed linearity of the ir - radio correlation arises because the two independent physical mechanisms by which the emission in these bands originates both result in an underestimate of the sfr at low luminosities by a similar amount , thus preserving the observed linear nature of the correlation in a conspiratorial way @xcite . however , these two competing physical trends whilst similar in magnitude will diverge for very low luminosity sources resulting in a reduction in the ir - to - radio ratio . this effect can be seen in ( figures 3 and 8 from * ? ? ? * ) and is tentatively hinted at in the observation presented here ( fig.[q24_1 ] ) . not withstanding the factors , discussed in the previous section which may effect this tentative detection of a small deviation in the gradient of the ir - radio correlation , these observational results are broadly consistent with the models of @xcite . if this observed evolution of the ir - radio correlation at low flux densities is real it potentially has significant implications on the use of either ir or radio luminosities for the derivation of star formation rates of faint galaxies in future deep surveys . future observations will be required to confirm this tentative deviation of the ir - to - radio correlation at these extremely low flux densities . these results may be tested further by _ i ) _ deeper radio and infrared observations which detect many faint individual sources at high significance , _ ii ) _ significantly larger area deep surveys that allow robust statistical detections of faint sources to be made , or _ iii ) _ the separation and analysis of the various competing factors which contribute to this effect . the stacked radio images of extremely faint ir sources presented in fig.[maps ] provide a tantalising glimpse of the images which will be routinely available in the near future . the next - generation of radio interferometers , such as _ e-_merlin @xcite and the evla @xcite , will provide images many tens of times more sensitive than can currently be made . future deep radio observations using _ e-_merlin and the evla with integration times comparable to those presented here will produce images with sub - microjy noise levels . surveys using these instruments will easily be able to detect all of the ir sources so far found in the deepest observations with very high confidence , and the high sensitivity of these instruments will allow much larger areas of microjy radio sky to be surveyed . such high - sensitivity and wide - field observations will constrain any low flux density deviations in the radio - ir correlation . the radio observations presented here primarily trace the non - thermal synchrotron radio emission of galaxies with flux densities of a few tens of @xmath4jy . at 1.4ghz @xmath290 percent of the emission of a typical @xmath2l@xmath59 spiral galaxy is non - thermal emission with the remainder coming from thermal bremsstrahlung processes @xcite . whereas the radio synchrotron , as stated previously is indirectly related to star - formation , the bremsstrahlung emission is a more direct tracer of the ongoing star - formation since it arises from thermal radio emission from gas ionised by hot , young stars . at higher radio frequencies ( @xmath4710ghz ) thermal radio emission will begin to dominate over synchrotron emission @xcite tending to flatten the radio spectrum of star - forming galaxies . determining the relative fraction of thermal and non - thermal radio emission in different sources and at different frequencies is non - trivial requiring multi - frequency measurements of the radio flux densities in order to determine the slope of non - thermal emission . such observations will be more feasible using future extremely deep radio observations planned with the new generation of wide - bandwidth radio interferometers , such as _ e-_merlin and the evla , which will allow sensitive multi - frequency radio observations of the distant universe to be made within reasonable observation times . using observations made with these enhanced centimetre instruments along with measurements made using longer wavelength radio interferometers like the giant metrewave radio telescope ( gmrt ) and low frequency array ( lofar ) it will be possible to constrain the radio spectral energy distribution of many star - forming galaxies hence allowing the thermal and non - thermal radio emission to be separated . by disentangling the thermal and non - thermal radio luminosities of many distant star - forming galaxies it will be possible to verify the expected suppression of ir emission from low luminosity sources relative to the thermal radio emission which is a more direct , extinction free , tracer of the star - formation using one of the deepest high - resolution 1.4ghz observations made to date , in conjunction with deep 24 source catalogues from goods , we have investigated the microjy radio counterparts of faint mir sources . these observations confirm that the microjy radio source population follow the mir - radio correlation and extend this correlation by several orders of magnitude to very low flux densities and luminosities , and out to moderate redshifts . this extension of the mir - radio correlation confirms that the majority of these extremely faint radio and 24 sources are predominantly powered by star - formation with little agn contamination . statistically stacking the radio emission from many tens of faint 24 sources has been used to characterise the size and nature of the radio emission from very faint ir galaxies well below the nominal radio sensitivity of these data . using these methods the mir - radio correlation has been further extended and a tentative deviation in this correlation at very low 24 flux densities has been identified . the authors wish to thank ray norris , brian boyle , phil appleton and nick seymour for extremely useful discussions . rjb acknowledges financial support by the european commission s i3 programme `` radionet '' under contract no . this work is based on observations made with merlin , a national facility operated by the university of manchester at jodrell bank observatory on behalf of stfc , and the vla of the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc . this work is based in part on archival data obtained with the spitzer space telescope , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa . support for this work was provided by an award issued by jpl / caltech . appleton , p. n. , fadda , d. t. , marleau , f. r. , frayer , d. t. , helou , g. , condon , j. j. , choi , p. i. , yan , l. , lacy , m. , wilson , g. , armus , l. , chapman , s. c. , fang , f. , heinrichson , i. , i m , m. , jannuzi , b. t. , storrie - lombardi , l. j. , shupe , d. , soifer , b. t. , squires , g. & teplitz , h. i. , 2004 apjs , 154 , 147 beswick , r. j. , muxlow , t. w. b. , thrall , h. , richards , a. m. s. , 2006 , to appear in the proceedings of at the edge of the universe ( 9 - 13 october 2006 , sintra , portugal ) eds . alfonso _ ( astro - ph/0612077 ) _ brandl , b. r. , bernard - salas , j. , spoon , h. w. w. , devost , d. , sloan , g. c. , guilles , s. , wu , y. , houck , j. r. , weedman , d. w. , armus , l. , appleton , p. n. , soifer , b. t. , charmandaris , v. , hao , l. , higdon , j. a. marshall , s. j. , herter , t. l. , 2006 , apj , 653 , 1129 daddi , e. , alexander , d. m. , dickinson , m. , gilli , r. , renzini , a. , elbaz , d. , cimatti , a. , chary , r. , frayer , d. , bauer , f. e. , brandt , w. n. , giavalisco , m. , grogin , n. a. , huynh , m. , kurk , j. , mignoli , m. , morrison , g. , pope , a. , ravindranath , s. , 2007 , apj , 670 , 173 fadda , dario , marleau , francine r. , storrie - lombardi , lisa j. , makovoz , david , frayer , david t. , appleton , p. n. , armus , l. , chapman , s. c. , choi , p. i. , fang , f. , heinrichsen , i. , helou , g. , i m , m. , lacy , m. , shupe , d. l. , soifer , b. t. , squires , g. k. , surace , j. , teplitz , h. i. , wilson , g. , yan , l. , 2006 , aj , 131 , 2859 garrington , simon t. , anderson , b. , baines , c. , battilana , j. a. , bentley , mike n. , brown , d. , burgess , p. , diamond , philip j. , kitching , g. j. , mccool , roshene , muxlow , t. w. , noble , r. g. , roddis , neil , spencer , r. e. , thomasson , p. , 2004 , ground - based telescopes . edited by oschmann , jacobus m. , jr . proceedings of the spie , volume 5489 , pp . 332 - 343 murphy , e. j. , braun , r. , helou , g. , armus , l. , kenney , j. d. p. , gordon , k. d. , bendo , g. j. , dale , d. a. , walter , f. , oosterloo , t. a. , kennicutt , r. c. , jr . , calzetti , d. , cannon , j. m. , draine , b. t. , engelbracht , c. w. , hollenbach , d. j. , jarrett , t. h. , kewley , l. j. , leitherer , c. , li , a. , meyer , m. j. , regan , m. w. , rieke , g. h. , rieke , m. j. , roussel , h. , sheth , k. , smith , j. d. t. , thornley , m. d. , 2006 , apj , 638 , 157 murphy , e. j. , helou , g. , braun , r. , kenney , j. d. p. , armus , l. , calzetti , d. , draine , b. t. , kennicutt , r. c. , jr . , roussel , h. , walter , f. , bendo , g. j. , buckalew , b. , dale , d. a. , engelbracht , c. w. , smith , j. d. t. , thornley , m. d. , 2006 , apj , 651 , l111 muxlow , t. w. b. , richards , a. m. s. , garrington , s. t. , wilkinson , p. n. , anderson , b. , richards , e. a. , axon , d. j. , fomalont , e. b. , kellermann , k. i. , partridge , r. b. , windhorst , r. a. 2005 mnras , 358 , 1159 muxlow , t. w. b. , beswick , r. j. , richards , a. m. s. , thrall , h. j. , 2007 , in proc . conf . _ the 8th european vlbi network symposium on new developments in vlbi science and technology _ , edd . a. marecki et al . ( _ astro - ph/0611951 _ ) , g. , habing , h. j. , van duinen , r. , aumann , h. h. , baud , b. , beichman , c. a. , beintema , d. a. , boggess , n. , clegg , p. e. , de jong , t. , emerson , j. p. , gautier , t. n. gillett , f. c. , harris , s. , hauser , m. g. , houck , j. r. , jennings , r. e. , low , f. j. , marsden , p. l. , miley , g. , olnon , f. m. , pottasch , s. r. , raimond , e. , rowan - robinson , m. , soifer , b. t. , walker , r. g. , wesselius , p. r. & young , e. , 1984 , apj , 278 , l1 norris , ray p. , afonso , jos , appleton , phil n. , boyle , brian j. , ciliegi , paolo , croom , scott m. , huynh , minh t. , jackson , carole a. , koekemoer , anton m. , lonsdale , carol j. , middelberg , enno , mobasher , bahram , oliver , seb j. , polletta , mari , siana , brian d. , smail , ian , voronkov , maxim a. , 2006 , aj , 132 , 2409 richards , a. m. s. , muxlow , t. w. b. , beswick , r. j. , allen , m. g. , benson , k. , garrett , m. a. , garrington , s. t , gonzalez - solarez , e. , harrison , p. a. , kettenis , m. , laing , r. a. , richards , e. a. , thrall , h. , van langevelde , h. j. , walton , n. a. , wilkinson , p. n. , winstanley , n. , 2007 a&a 472 , 805 ( _ astro - ph/0706.3777 _ )

roles of turbulence in the context of magnetic reconnection are investigated with special emphasis on the mutual interaction between flow ( large - scale inhomogeneous structure ) and turbulence . in order to evaluate the effective transport due to turbulence , in addition to the _ intensity _ information of turbulence represented by the turbulent energy , the _ structure _ information represented by pseudoscalar statistical quantities ( helicities ) is important . on the basis of the evolution equation , mechanisms that provide turbulence with cross helicity are presented . magnetic - flux freezing in highly turbulent media is considered with special emphasis on the spatial distribution of the turbulent cross helicity . the cross - helicity effects in the context of magnetic reconnection are also investigated . it is shown that the large - scale flow and magnetic - field configurations favorable for the cross - helicity generation is compatible with the fast reconnection . in this sense , turbulence and large - scale structures promote magnetic reconnection mediated by the turbulent cross helicity .

1105.6343

magnetic reconnection is one of the most important processes in plasma physics . it appears in a variety of phenomena in geo / astrophysics and fusion physics , which include dynamo , solar and stellar flares , magnetospheric substorms , disruptions in fusion devices , etc . in the framework of magnetohydrodynamics ( mhd ) , the diffusion of magnetic field in magnetic reconnection can be described with a basic model proposed by sweet@xcite and parker,@xcite where conservations of the magnetic flux , mass , and momentum are considered . however , sweet parker mechanism of reconnection is too slow to explain typical time scales of eruptive processes observed in laboratory and in space such as solar flares . in order to alleviate this discrepancy due to the slowness of reconnection , several theoretical attempts for the fast reconnection , including the petschek s model,@xcite have been proposed . turbulence is one of the candidate mechanisms that enhance the reconnection rate very much . turbulence in magnetic reconnection has been investigated by several researchers . analysis of magnetic - field observations in the earth s magnetotail suggested the reconnection process in the current sheet has properties of turbulence excited by tearing vortices.@xcite the effects of turbulence on magnetic reconnection were first investigated by matthaeus and lamkin using numerical simulations of pinch sheet . they showed that the electric fluctuations combined with the flow and current filament enhance the dissipation in the reconnection zone , and leads to the rapid reconnection.@xcite in order to dissolve the discrepancy between the typical length scale of phenomena and the dimension required for the current sheet , a fractal nature of the current sheet was suggested.@xcite lazarian and vishniac examined the effects of stochastic field in reconnection and concluded that the presence of the stochastic components of the magnetic field enhances the reconnection rate drastically . they stressed that turbulence alters the relevant scales for reconnection almost independently of the microscopic dissipation mechanisms.@xcite one of the primary effects of turbulence is enhancing the mixing or transport of the system , which is represented by the notion of turbulent transport such as the eddy viscosity , eddy diffusivity , anomalous resistivity , etc . for example , in the mean momentum equation in a turbulent flow , the eddy viscosity @xmath0 enhances the effect of molecular viscosity @xmath1 as @xmath2 with spatiotemporal variation of @xmath0 . the enhanced transport due to turbulence leads to destruction of structure or reduction of spatiotemporal inhomogeneity . in the real - world phenomena , however , roles of turbulence are not so obvious or straightforward . under some circumstances , a large - scale inhomogeneous structure is formed or sustained by turbulence itself . representative situations are seen in turbulent dynamos , where large - scale magnetic fields are generated by some statistical properties of turbulent motion.@xcite this can be regarded as an example of general situations where breakage of symmetry leads to the formation or sustainment of large - scale inhomogeneous structure in turbulence . in the presence of symmetry breakage in turbulence , we may expect some effects that balance or counteract the primary turbulence effects . here , enhanced transports due to turbulence are suppressed , and large - scale inhomogeneous structures are sustained or formed . important question is how to formulate such breakage of symmetry in turbulence which leads to the structure formation . most straightforward method for describing turbulent motions is to directly solve governing equations . however , for realistic turbulence of scientific interests with a huge reynolds number , it is impossible in the foreseeable future to numerically solve all scales of motions ranging from the largest scales of system size to the smallest scales where the energy is dissipated into heat . in such a situation , modeling of turbulent motions provides a useful tool for investigating realistic turbulence . in practice of studying turbulence phenomena that appear in the geo / astrophysical and fusion sciences , one of the important tasks is how to choose quantities that appropriately describe the statistical properties of turbulence . in the studies of turbulence modeling , among several statistical quantities , the turbulent energy density ( per unit mass ) @xmath3 and its dissipation rate @xmath4 $ ] have often been adopted as descriptors of turbulence properties ( @xmath5 : velocity fluctuation , @xmath1 : kinematic viscosity).@xcite the turbulent energy @xmath6 expresses how much fluctuation a turbulent flow has , and the energy dissipation rate @xmath7 expresses how much energy is transferred from large to small scales . in this sense , both @xmath6 and @xmath7 represent the _ intensity _ property of turbulence . the eddy - viscosity representation is supplemented by a model of @xmath0 whose generic form is expressed by _ intensity _ measures of turbulence as @xmath8 . the most popular expression is @xmath9 ( @xmath10 : model constant , @xmath11 : time scale of turbulence ) . however , it is well known in the engineering field that such a simple eddy - viscosity model completely fails if applied to some turbulent flows . the turbulent swirling flow is the representative case . a simple eddy - viscosity model is too dissipative to sustain the centerline dent profile of axial mean velocity that is experimentally observed in the turbulent swirling flow.@xcite in order to alleviate this discrepancy , anomalously small value of @xmath10 has often been adopted in the numerical simulations.@xcite this fact implies that the description based on @xmath6 and @xmath7 is not enough and we have to take into account another turbulent statistical quantity that properly describes the _ structure _ property or breakage of mirrorsymmetry in the turbulent swirling pipe flow.@xcite large - scale rotation and magnetic field are often essential ingredients in the geo / astrophysical and fusion plasma phenomena . if we have large - scale rotational motion or magnetic field , symmetry of turbulence between the directions parallel and antiparallel to the rotation axis or magnetic field is broken . such breakage of mirrorsymmetry is expected to be measured by pseudoscalar turbulent quantities , which change their sign under the inversion of the coordinate system . since pseudoscalar vanishes in a mirrorsymmetric system , a finite value of pseudoscalar indicates the breakage of mirrorsymmetry . the turbulent cross helicity is a pseudoscalar which represents breakage of mirrorsymmetry , more precisely , asymmetry between the directions parallel and antiparallel to the large - scale magnetic field . in the presence of breakage of mirrorsymmetry in turbulence , the primary transports due to turbulence may be suppressed by the cross - helicity effects , and large - scale structures are formed and sustained even in the presence of a strong turbulence . these two effects : ( i ) generation of large - scale structures ; and ( ii ) suppression of turbulent transport ; are two sides of the same coin , the cross - helicity effect . unlike the kinetic and magnetic helicities , the cross helicity can exist even in two dimensional configuration of magnetohydrodynamic plasmas . for physical interpretation of the cross - helicity effect and its application to the dynamo problems , the reader is referred to refs . behavior of turbulent statistical quantities relevant to the transport coefficients provides essential information on the evolution of the large - scale structures . at the same time , how turbulent statistical quantities are spatially distributed and temporally evolve is determined by large - scale flow and magnetic - field structures or mean - field inhomogeneities . as is well known , the mean - velocity shear coupled with the reynolds stress,@xmath12 , is the main cause for the turbulence production in a hydrodynamic ( hd ) flow . in the mhd turbulent flow , as will be shown in sec . [ sec : ii ] , we have several mean - field inhomogeneities other than the mean velocity shear , which may contribute to the energy and cross - helicity generations . large - scale structures , which can be denoted in the broad sense as _ flow _ including the mean flow and global magnetic field , etc . , are determined by _ turbulence _ through the turbulent transport coefficients . on the other hand , properties of turbulence are determined by the large - scale flow structures such as the mean velocity shear , vorticity , magnetic - field shear , electric current , etc . this mutual relationship between the flow and turbulence is schematically depicted in fig . [ fig : flow_turbulence ] . flow turbulence relationship . mean - field structures are determined by turbulence through the turbulent transport coefficients . turbulence properties are determined by mean - field structures through the production rates of turbulent statistical quantities . ] if we see the magnetic reconnection phenomena from this perspective , important points are as follows . magnetic reconnection is considered to be promoted by the enhanced magnetic diffusivity due to turbulence . if the spatiotemporal distributions of the turbulent statistical quantities ( turbulent energy , cross helicity , etc . ) relevant to the turbulent transports are favorable for the magnetic reconnection , the rate of reconnection would be increased . the spatiotemporal distributions of these turbulent quantities obey the balance of the production , dissipation , and transport rates of these statistical quantities . the production mechanism , the most important ingredient , is determined by the large - scale flow structures such as the inflow profiles , global magnetic - field configurations , etc . if the large - scale structures sustained and formed during the magnetic reconnection process are consistent with the favorable distributions of the transport coefficients for the magnetic reconnection , the reconnection would be promoted . otherwise , the reconnection rate should be reduced . therefore , it is the production mechanism of the turbulent statistical quantities due to large - scale structures that should be examined intensively . in this paper , we focus our attention on the turbulent cross - helicity effects , and address the magnetic reconnection from the viewpoint of the mutual relationship between the flow structure and turbulence mediated by the cross helicity . this paper is organized as follows . in sec . [ sec : ii ] , the evolution equation of the turbulent cross helicity is presented . on the basis of this equation , main mechanisms that provide turbulence with the cross helicity are examined . in sec . [ sec : iii ] , the cross - helicity effects in the mean magnetic induction equation are treated with special emphasis on the conditions for the magnetic - flux freezing and the magnetic reconnections . in sec . [ sec : iv ] , the cross - helicity effects in the mean momentum equation are considered . the mean flow configuration affects the reconnection rates very much . by considering the flow induced by the cross - helicity effects , we discuss the influence of the cross helicity in turbulence on the magnetic reconnections . summary and concluding remarks are presented in sec . [ sec : v ] . effects of turbulence on the mean fields in magnetohydrodynamics are represented by the reynolds ( and turbulent maxwell ) stress @xmath13 in the mean velocity ( @xmath14 ) equation and by the turbulent electromotive force @xmath15 in the mean magnetic - field ( @xmath16 ) equation ( @xmath17 : magnetic - field fluctuation ) . note that , in this paper , we adopt the alfvn speed unit for the magnetic field etc . ( @xmath18 , @xmath19 : magnetic field measured in the original physical unit , @xmath20 : magnetic permeability , @xmath21 : density ) . in the case of compressible flows , we have several turbulent transport mechanisms that are intrinsically connected to the density variation . for example , density variation along the mean magnetic field gives an important contribution to the turbulent cross - helicity generation . effects of compressiblity or variable density can be incorporated into the cross - helicity arguments in the context of turbulent dynamos and turbulent transport suppression . however , such arguments will make the description much more complicated . at low plasma beta , incompressibility is considered to be a good approximation although the plasma beta becomes about one at the current - sheet region . considering that the essential physics of magnetic reconnection is not so different in incompressible and compressible cases , here in this paper , as most work following the original sweet parker model , we confine our arguments to the incompressible framework . this treatment never denies the possible importance of compressibility in the magnetic reconnection study especially in the astrophysical context . for some examples of the theoretical extensions to the compressible or variable - density case , the reader is referred to refs . and . as we see in appendix a , from a statistical analytical theory for inhomogeneous turbulence applied to magnetohydrodynamic ( mhd ) turbulence , @xmath22 and @xmath23 are expressed as@xcite [ eq : re_str]@xmath24 [ eq : e_m]@xmath25 where @xmath26 is the turbulent mhd residual energy , @xmath27 and @xmath28 are the strain rates of the mean velocity and magnetic field , respectively , @xmath29 the mean electric - current density , @xmath30 the mean vorticity . here , @xmath31 , @xmath32 , @xmath33 , @xmath34 and @xmath35 are the transport coefficients whose spatial distributions are determined by the properties of turbulence . in eqs . ( [ eq : re_str_exp ] ) and ( [ eq : e_m_exp ] ) , the eddy viscosity and the turbulent magnetic diffusivity , @xmath31 and @xmath34 , are expressed in terms of the turbulent energy @xmath36 . the @xmath28-related transport coefficient @xmath32 in eq . ( [ eq : re_str_exp ] ) and the @xmath37-related coefficient @xmath35 in eq . ( [ eq : e_m_exp ] ) are expressed in terms of the turbulent cross helicity ( velocity magnetic - field correlation in turbulence ) @xmath38 . then , it is important to examine the evolution equations of the turbulent energy and cross helicity , @xmath39 and @xmath40 . in the context of turbulent transport suppression and structure formation , in particular , the production mechanisms of @xmath40 should be intensively investigated . turbulent transports such as the eddy viscosity , turbulent resistivity , dynamo coefficients , etc . are determined by statistical properties of turbulence . a simplest way to treat such statistical properties is adopting one - point turbulent statistical quantities and consider the evolution of these statistical quantities . in magnetohydrodynamic ( mhd ) turbulence , the turbulent mhd energy @xmath39 and the turbulent cross helicity @xmath40 defined by @xmath41 @xmath42 are most important ones among several statistical quantities since they are directly connected to the transport coefficients appearing in @xmath43 , @xmath31 and @xmath32 , and the ones in @xmath44 , @xmath34 and @xmath35 . for detailed description of the transport equations of several statistical quantities , the reader is referred to refs . . in order to see how and how much energy and cross helicity exist in turbulence , we have to consider the transport equations for @xmath39 and @xmath40 . from equations of fluctuation velocity and magnetic fields , @xmath5 and @xmath17 , equations for the turbulent mhd energy ( density ) and turbulent cross helicity ( density ) are written as @xmath45 with @xmath46 . here @xmath47 , @xmath48 , and @xmath49 are the production , dissipation , and transport rates of @xmath50 . they are defined by [ eq : k_evol]@xmath51 @xmath52 @xmath53 [ eq : w_evol]@xmath54 @xmath55 @xmath56 ( @xmath57 : magnetic diffusivity ) . in eqs . ( [ eq : p_k ] ) and ( [ eq : p_w ] ) , @xmath22 and @xmath23 are the the reynolds stress and the turbulent electromotive force defined by eqs . ( [ eq : re_str_def ] ) and ( [ eq : e_m_def ] ) , respectively . in eqs . ( [ eq : t_k ] ) and ( [ eq : t_w ] ) , @xmath58 and @xmath59 are the higher order correlations of fluctuations whose exact expressions are suppressed here . we see from eq . ( [ eq : g_eq ] ) with eq . ( [ eq : w_evol ] ) that there are three main mechanisms that provide turbulence with the cross helicity . they are the effects related to ( i ) the reynold stress , @xmath60 ; ( ii ) the turbulent electromotive force , @xmath61 ; and ( iii ) inhomogeneity along the mean magnetic field , @xmath62 . among these three , the reynolds - stress- and turbulent - electromotive - force - related productions are associated with the drain of the mean - field cross helicity @xmath63 . in this sense , they are related to the cascading property of the cross helicity . on the other hand , the turbulent cross - helicity generation due to the inhomogeneity along the mean magnetic field is not directly related to such cascade processes but arises from a kind of transport with asymmetry of the boundary conditions . in this sense , the inhomogeneity along the mean magnetic field may play a special role in the turbulent cross - helicity generation in the context of externally injecting cross helicity . in the following , we refer to some typical situations in which each of these generation mechanisms may work . the reynolds stress coupled with the mean magnetic shear gives a cross - helicity generation . substituting eq . ( [ eq : re_str_def ] ) into the first term of eq . ( [ eq : p_w ] ) , we have a production due to the reynolds stress as @xmath64 since the sign of @xmath32 is the same as that of @xmath40 as we see from eqs . ( [ eq : nu_k_beta_rel ] ) and ( [ eq : gamma_model ] ) , the second or @xmath32-related term always destroys the cross helicity regardless of the sign of @xmath40 . so , as the production mechanism of the turbulent cross helicity , we write only the @xmath31-related term as @xmath65 this shows that a positive cross helicity is generated if the mean velocity and magnetic shears have the same signs . a typical situation may be caused by a neutral beam injection ( nbi ) in the toroidal direction to tokamak devices in fusion plasmas . in toroidal confinement plasmas , the toroidal magnetic field decreases as the major radius of plasma increases ( @xmath66 , @xmath67 : distance from the major axis ) . if a neutral beam is injected in the center ( minor axis ) region , the toroidal velocity is maximum at the center of the minor radius ( @xmath68 : distance from the minor axis ) and decreases as the minor radius of plasma increases . in this situation , we have negative velocity and magnetic shears ( @xmath69 ) in the half region far from the major axis ( far side ) , and negative velocity and positive magnetic shears ( @xmath70 ) in the half region near the major axis ( near side ) as shown in fig . [ fig : pro_w_re_str ] . this suggests that we have positive and negative cross - helicity generations in the outer- and inner - half regions of plasma , respectively . cross - helicity production due to the neutral beam injection ( nbi ) . the reynolds stress coupled with the mean magnetic - field shear leads to the positive or negative cross - helicity generation . ] the turbulent electromotive force coupled with the mean vorticity @xmath37 gives a cross - helicity generation . substituting eq . ( [ eq : e_m_exp ] ) into the second term of eq . ( [ eq : p_w ] ) , we have a production due to the turbulent electromotive force as @xmath71 since the sign of @xmath35 is the same as that of @xmath40 as seen in eq . ( [ eq : gamma_model ] ) , the second or @xmath35-related term in eq . ( [ eq : pw_em ] ) works for the destruction of the turbulent cross helicity regardless of the sign of @xmath40 . in tokamak plasmas , we have a very strong plasma current @xmath72 . in such a case , the turbulent cross - helicity generation due to the electromotive force may be expressed as @xmath73 since @xmath74 . this shows that the positive or negative turbulent cross helicity is generated if the mean vorticity is aligned with the mean electric current : @xmath75 in tokamak plasmas , we have a strong plasma current in the toroidal direction , @xmath76 ( @xmath77 : toroidal direction in the cylindrical approximation for a torus geometry ) . in the presence of poloidal rotation @xmath78 , we have a mean toroidal vorticity associated with this poloidal rotation , @xmath79 $ ] . then , it is highly probable that we have a generation of positive or negative turbulent cross helicity depending on the direction of poloidal velocity ( fig . [ fig : pro_w_e_m ] ) . cross - helicity production due to the coupling of the large - scale electric current and vorticity . if the large - scale vorticity associated with the poloidal rotation is parallel ( or antiparallel ) to the plasma current , a positive ( or negative ) cross helicity is generated . ] as for the cause of poloidal rotation , we may consider several spontaneous plasma rotation associated with the improved confinement mode , such as local poloidal rotation near the plasma edge in h mode , global poloidal rotation near the internal transport barrier in the negative magnetic shear mode , etc . if we use a nbi with a tilted angle , we have a poloidal rotation in addition to the toroidal rotation . however , direct observation of the turbulent cross helicity in high temperature fusion devices is still very difficult . to validate this type of cross - helicity generation , a numerical simulation of cylindrical geometry mimicking the tokamak plasmas was performed.@xcite in the work , it was shown that a poloidal plasma rotation , which corresponds to the mean vorticity in the toroidal direction , coupled with the plasma current contributes to the cross - helicity generation in turbulence . it was also shown that if the direction of poloidal rotation is changed , the sign of the generated cross helicity changes . if the turbulent mhd energy @xmath80 is inhomogeneously distributed along the mean magnetic field @xmath16 , the cross helicity may be generated in turbulence . a positive ( or negative ) cross helicity is produced for @xmath81 parallel ( or antiparallel ) to @xmath16 : @xmath82 it is ubiquitous in astrophysical phenomena that an ambient magnetic field threads through a plasma gas disk . if intensity of turbulence is not homogeneously distributed in the direction perpendicular to the disk , we have inhomogeneity of turbulence along the mean magnetic field . this mechanism may play a very important role in cross - helicity generation in astrophysical turbulence.@xcite we consider a situation where a large - scale magnetic field threads through a disk from above to below , and the intensity of turbulence is highest at the midplane and decreases with the distance from the midplane ( fig . [ fig : pro_w_inhomo ] ) . in such a case , we have positive and negative cross - helicity generations in the upper- and lower - half regions of the disk . consequently , we have positive and negative distributions of the cross helicity in the upper and lower halves of the disk as @xmath83 cross - helicity production due to the energy inhomogeneity along the large - scale magnetic field . ] we should stress that , in this situation , the intensity of turbulence is maximum at the midplane so that the gradient of turbulent energy , @xmath81 , is zero and changes its sign there . as this result , there is no cross - helicity generation at the midplane , and the magnitude of the cross helicity should vanish there . this is consistent with a basic feature of pseudoscalars , representing the breakage of mirrorsymmetry in turbulence . the primary effect of turbulence in the magnetic induction equation is destruction of magnetic field , which is expressed by the turbulent magnetic diffusivity or anomalous resistivity @xmath34 . if we have no field - generation effects such as the helicity or cross - helicity effect , the turbulent electromotive force is expressed as @xmath84 substituting this into the mean induction equation , we have @xmath85 . \label{eq : mean_b_eq_only_beta}%(16)\ ] ] this shows that the effective magnetic diffusivity is enhanced by turbulence as @xmath86 we should note that at large magnetic reynolds number ( @xmath87 ) , the turbulent magnetic diffusivity is in general much larger than the molecular counterpart as @xmath88 in the absence of any field - generation mechanisms , the mean induction equation is written as @xmath89 . \label{eq : mean_b_eq_beta}%(19)\ ] ] in this case , the magnetic - flux freezing to the plasma motion does not occur since @xmath34 is so large that the effective magnetic diffusivity can not be neglected . on the other hand , if we have some field - generation mechanisms such as the helicity and the cross - helicity effects , the turbulent electromotive force is expressed as @xmath90 then , the mean induction equation reads @xmath91\nonumber\\ & & + \nabla \times \left ( { \alpha { \bf{b } } + \gamma \mbox{\boldmath$\omega$ } } \right ) . \label{eq : mean_b_eq_e_m_full}%(21)\end{aligned}\ ] ] the third term on the right - hand side of eq . ( [ eq : mean_b_eq_e_m_full ] ) represents the field - generation mechanisms due to the turbulent residual helicity coupled with the mean magnetic field , @xmath92 , and the turbulent cross helicity coupled with the mean vorticity , @xmath93 . if such field - generation mechanisms are large enough , they may balance the turbulent magnetic diffusivity effect , @xmath94 , as @xmath95 in this case , the contribution from the turbulent electromotive force is very small , and we have the mean induction equation as @xmath96 this equation has exactly the same form as the induction equation for the instantaneous magnetic field @xmath97 . then , for the flow with very high magnetic reynolds number ( @xmath87 ) , the mean magnetic field @xmath16 is subject to the first term of the right - hand side of eq . ( [ eq : mean_b_eq_supp ] ) . again , we have a magnetic - flux freezing to the plasma motion . although this situation looks very similar to the magnetic - flux freezing in the laminar case , it is completely different from the laminar one . first , in eq . ( [ eq : mean_b_eq_supp ] ) , we have very strong turbulence or fluctuating components of the velocity and magnetic fields . this magnetic - flux freezing occurs only at very high reynolds and magnetic reynolds numbers flow . in this sense , we may call this the turbulent magnetic - flux freezing . second point is that since this state is based on an equilibrium condition ( [ eq : balance_in_e_m ] ) , the `` turbulent magnetic - flux freezing '' highly depends on the mean - field configurations and spatial and temporal distribution of turbulent quantities . if the balance relation between the field destruction and generation breaks , the picture of magnetic - flux freezing can not hold any more . if the field - generation mechanisms are absent , the field - destruction due to turbulence prevails . a typical situation may occur on the midplane of a plasma gas disk or equatorial plane in a spherical convection zone . recalling the fact that pseudoscalar turbulent quantities are descriptors of breakage of mirrorsymmetry , it is quite natural these pseudoscalar quantities are distributed antisymmetricaly with respect to the midplane . for such spatial distributions , the magnitudes of pseudoscalar quantities vanish on the midplane , and very small in the vicinity of it . as a consequence of this disappearance of psedoscalars , a strong effective magnetic diffusivity may cause the magnetic reconnection in the close vicinity of the midplane . in the study of dynamo , the so - called @xmath98 effect has been considered as a basic ingredient of the dynamo process . the @xmath98 effect contains several different physical processes in it . in the context of the toroidal - field generation from the poloidal magnetic field , at least three processes : ( i ) magnetic - flux freezing ; ( ii ) appropriate differential rotation ; ( iii ) magnetic - field reconnection ; should be included in the @xmath98 effect . as has been already mentioned in sec . [ sec : iii ] , in order for a large - scale magnetic - flux freezing to occur in a highly turbulent medium , we need mechanisms that balance or cancel the strong turbulent magnetic diffusivity effect . without such balancing or transport suppression mechanisms , we only have eq . ( [ eq : mean_b_eq_beta ] ) , then no magnetic - flux freezing occurs . the cross - helicity effect , as well as the helicity or @xmath33 effect , is one of such balancing mechanisms . in addition to the magnetic - flux freezing , in order to get a toroidal component of the magnetic field from the poloidal magnetic field , we need a differential rotation of the toroidal velocity . the toroidal velocity should be faster or slower in the lower - latitude region than in the higher - latitude region . without such differential rotation profiles of the toroidal velocity , the toroidal magnetic field can not be generated from the poloidal magnetic field in the framework of the @xmath98 effect . the toroidal rotation observed at the surface of the sun , the prograde rotation in the equatorial or lower - latitude region , is considered to support the scenario of a dynamo that utilizes the @xmath98 effect . during the process generating the toroidal magnetic field from the poloidal magnetic field by differential rotation , we in general need a magnetic - field reconnection , which makes a new configuration possible from the old one . original poloidal field is distorted by the differential rotation . since such distortion is most strong at the midplane or equatorial plane , it is most likely that the reconnection occurs there . then we have a toroidal magnetic - field configuration whose direction is opposite with respect to the midplane . these considerations immediately suggest that the @xmath98 effect utilized for the toroidal magnetic - field generation from the poloidal one is based on a subtle combination of several requirements . it requires a particular type of differential rotation profile that is suitable for the toroidal field generation . it requires a particular spatial distribution of the effective magnetic diffusivity that is convenient for both the magnetic - flux freezing in some region and the magnetic reconnection in other region . in this sense , it is quite natural that model simulations that utilize the @xmath98 effect are very sensitive to the variation of velocity profile including the incorporation of the meridional rotation and the value and spatial distribution of the effective magnetic diffusivity . if the cross - helicity effect is the main balancer for the turbulent magnetic diffusivity effect , the mean induction equation in a stationary state has a special solution as follows . we divide the mean magnetic field and electric - current density as @xmath99 where @xmath100 and @xmath101 are the mean magnetic field and electric - current density without the cross - helicity effect , while @xmath102 and @xmath103 are the mean fields representing the first - order contribution through the cross - helicity effect . substituting eq . ( [ eq : mag_expand ] ) into eq . ( [ eq : mean_b_eq_e_m_full ] ) with the @xmath33-related term dropped , we have equations @xmath104 for the zeroth - order field , and @xmath105 for the first - order field . equation ( [ eq : delta_b_eq ] ) represents the magnetic induction @xmath106 arising from the turbulent cross helicity . equation ( [ eq : delta_b_eq ] ) has a particular solution @xmath107 for a stationary state . here , @xmath108 is a model constant whose value is estimated as @xmath109 [ see eqs . ( [ eq : beta_model ] ) and ( [ eq : gamma_model ] ) ] . here , the proportional coefficient @xmath110 is the turbulent cross helicity scaled by the turbulent mhd energy . this scaled cross helicity is the most important quantity in the cross - helicity dynamo formulation . the value of @xmath110 is not positive - definite , but its absolute value is bounded as @xmath111 the cross - helicity dynamo solution [ eq . ( [ eq : delta_b_sol ] ) ] requires a finite value of turbulent cross helicity for it to work . in the absence of cross helicity in turbulence , eq . ( [ eq : delta_b_sol ] ) gives no contribution at all for the magnetic - field generation . in this sense , the spatial and temporal distributions of the cross helicity should be intensively examined . in order to fully investigate such distributions , we have to solve the evolution equation of @xmath40 [ eq . ( [ eq : g_eq ] ) with eq . ( [ eq : w_evol ] ) ] . however , without considering the full complications of the cross - helicity evolution , we can derive several consequences of eq . ( [ eq : delta_b_sol ] ) as follows . the turbulent cross helicity is a pseudoscalar , which describes breakage of mirrorsymmetry . as has been referred to in sec . i , cross helicity is considered to be a measure of the asymmetry with respect to the magnetic field . if a magnetic field threads through plasmas with the turbulent energy distributed symmetrically with respect to the midplane , we have asymmetry between the directions parallel and antiparallel to the magnetic field . then , we see from eq . ( [ eq : spatio_dis_w ] ) that the spatial distribution of the turbulent cross helicity is antisymmetric . if such an antisymmetric distribution of @xmath40 is coupled with the symmetric toroidal velocity , from eq . ( [ eq : delta_b_sol ] ) we have antisymmetric toroidal magnetic field ( fig . [ fig : toroidal_cross_hel ] ) . toroidal magnetic - field generation due to the cross - helicity dynamo . an antisymmetric toroidal - field configuration is induced from a symmetric velocity profile coupled with an antisymmetric cross - helicity distribution . no differential rotation is required . ] a poloidal magnetic field coupled with the inhomogeneous turbulence leads to antisymmetric distribution of the cross helicity in turbulence [ eqs . ( [ eq : b_grad_k ] ) and ( [ eq : spatio_dis_w ] ) ] . this turbulent cross helicity coupled with the toroidal velocity gives the toroidal magnetic field [ eq . ( [ eq : delta_b_sol ] ) ] . a prominent feature of this scenario is that it does not require a particular differential rotation of the toroidal velocity for the sake of toroidal magnetic - field generation . in the framework of cross - helicity dynamo , differential rotation is not directly related to the toroidal field generation . differential rotation is important only to sustain the energy and cross helicity in turbulence . as we see in eq . ( [ eq : w / k_constraint ] ) , the magnitude of the turbulent cross helicity is bounded by that of the turbulent mhd energy . unless we have enough amount of turbulent mhd energy , we will not have enough cross helicity in turbulence . only in this sense , the differential rotation of plasmas is required for the dynamo process to work . very recent numerical simulations showed that a dynamo model constituted of the evolution equations of the toroidal and poloidal magnetic fields as well as the turbulent cross helicity can reproduce an oscillatory butterfly diagram of the solar magnetic activity without resort to the @xmath98 effect.@xcite these results will be reported in the forthcoming paper . magnetic reconnections are associated with large - scale plasma flows with some typical configurations . we focus our attention on such large - scale flow structures associating with the magnetic reconnection , and examine how turbulence influences such flow structures as well as how such flow structures determine the properties of turbulence . so far we treated only the mean induction equation . here we examine the cross - helicity effects in the equation of motions . for this purpose , we consider the mean vorticity equation @xmath112 \nonumber\\ & + & \nabla \times \left [ { { \bf{f } } + \frac{1}{\beta } \left ( { { \bf{u } } \times { \bf{b } } } \right ) \times { \bf{b } } - \frac{1}{\beta } \frac{\partial { \bf{a}}}{\partial t } \times { \bf{b } } } \right ] , \label{eq : omega_eq}%(29)\end{aligned}\ ] ] which is a direct consequence of the expressions for the reynolds stress @xmath113 [ eq . ( [ eq : re_str_exp ] ) ] and the turbulent electromotive force @xmath23 [ eq . ( [ eq : e_m_exp ] ) ] substituted into the mean velocity equation ( @xmath114 : mean external force , @xmath115 : mean vector potential ) . for the derivation of eq . ( [ eq : omega_eq ] ) , see appendix [ app : b ] . as we have seen in eq . ( [ eq : g_eq ] ) with eq . ( [ eq : w_evol ] ) , it is difficult to sustain a finite cross helicity in turbulence in the absence of the mean magnetic field . in order to clearly understand the role of the turbulent cross helicity in momentum transport , we divide the mean velocity and vorticity as @xmath116 here , @xmath117 and @xmath118 are the mean velocity and vorticity without the mean magnetic - field effects , while @xmath119 and @xmath120 are the mean fields representing the first - order effects of the mean magnetic field through the turbulent cross helicity . substituting eq . ( [ eq : u_expand ] ) into eq . ( [ eq : omega_eq ] ) , we have equations @xmath121 for the zeroth - order field , and @xmath122 \label{eq : delta_omega_eq}%(32)\ ] ] for the first - order field . equation ( [ eq : omega_0_eq ] ) expresses the evolution of the large - scale flow structure @xmath118 , which is subject to the turbulent viscosity @xmath31 and the external force @xmath114 . on the other hand , eq . ( [ eq : delta_omega_eq ] ) represents the variation of the large - scale flow structure @xmath120 arising from the turbulent cross helicity or @xmath35 effect . we see from eq . ( [ eq : delta_omega_eq ] ) that , for a given large - scale flow structure @xmath118 , @xmath123 is a particular solution of the stationary state . equation ( [ eq : delta_u_sol ] ) suggests that we have a variation of momentum transfer that is proportional to the large - scale magnetic field multiplied by the normalized cross helicity . here we should make some remarks on the validity of eq . ( [ eq : delta_u_sol ] ) in the compressible case . in compressible flows , we have several other terms arising from the density variation in eq . ( [ eq : omega_eq ] ) or ( [ eq : mean_omega_eq_exp ] ) . if we neglect the density fluctuation ( @xmath124 ) and consider only the mean density stratification @xmath125 ( @xmath126 : mean density , @xmath127 : fluctuation density ) , one of the main contributions is the baroclinic effect in the mean vorticity equation : @xmath128 if we further assume that the cross helicity does not affect the mean density and pressure ( @xmath129 and @xmath130 or @xmath131 ) , eq . ( [ eq : baroclinic ] ) is reduced to @xmath132 and no contribution to the first - order equation [ eq . ( [ eq : delta_omega_eq ] ) ] . in such a case , the special solution [ eq . ( [ eq : delta_u_sol ] ) ] is obtainable even in the compressible case . it is pointed out that properties of magnetic reconnection such as the scaling law and maximum reconnection rate depend heavily on the form , nature , and value of the imposed boundary conditions.@xcite here we examine the reconnection environment from the viewpoint of large - scale structure . we consider the circumstances or global surroundings of the reconnection region are highly turbulent . such turbulence is a consequence of the large - scale inhomogeneity of the mean fields such as the velocity strain , electric current , etc . in this sense , turbulence in magnetic reconnections here is driven by large - scale structures . at the same time , we also consider the self - generation mechanisms of turbulence in the magnetic - reconnection process . since the argument leading to the result [ eq . ( [ eq : delta_u_sol ] ) ] is quite general , we can consider any general situations including the magnetic reconnection in three dimensions . for the sake of simplicity of description , here we consider a simple situation where a current sheet is associated with a large - scale inflow as in fig . [ fig : flow_in_mr](a ) . magnetic reconnection and turbulent cross helicity . ( a ) magnetic - field lines and streamlines for a large - scale inflow configuration . solid lines : magnetic field , dashed lines : streamline . shaded region represents the reconnection or diffusion region . ( b ) directions of the mean fields and induced velocity due to the turbulent cross helicity . directions of the mean electric - current density @xmath29 and the mean vorticity @xmath118 are shown with the sign of the turbulent cross helicity @xmath40 . directions of the induced velocity due to the cross - helicity effect [ eq . ( [ eq : delta_u_sol ] ) ] are also indicated by the dashed arrows . ] recalling the cross - helicity generation mechanism discussed in sec . [ sec : ii - c ] , we see that the generation mechanism related to the turbulent electromotive force [ eq . ( [ eq : pw_em ] ) ] may be most important in this flow configuration . as we see in fig . [ fig : flow_in_mr](b ) , the large - scale vorticity @xmath118 is spatially distributed antisymmetrically with respect to the midplane of current sheet ( @xmath133 axis , plane with @xmath134 ) , and also with respect to the center plane ( @xmath135 axis , plane with @xmath136 ) . since the large - scale electric - current density @xmath29 is aligned with the negative @xmath77 direction everywhere in this case , the sign of the product of @xmath118 and @xmath29 , @xmath137 , is distributed as @xmath138 consequently , the production of the turbulent cross helicity is positive and negative in the regions with @xmath139 and @xmath140 , respectively . then it is highly probable that the turbulent cross helicity is spatially distributed as @xmath141 as depicted in fig . [ fig : flow_in_mr](b ) . outside of the current sheet region , it follows from eq . ( [ eq : delta_u_sol ] ) that the velocity variation due to the cross - helicity effect is expressed as @xmath142 from the symmetry of the plasma flow , the large - scale vorticity vanishes on the midplane ( @xmath134 ) and center plane ( @xmath136 ) , so we have no cross - helicity - generation mechanism there . the turbulent cross helicity changes its sign across the current sheet ( midplane ) and the center plane . for instance , in the positive @xmath133 region in fig . [ fig : flow_in_mr ] , the sign of @xmath40 changes from positive ( @xmath143 ) to negative ( @xmath144 ) across the current sheet . as this consequence , the magnitude of the cross helicity is small in the vicinity of the current sheet . equation ( [ eq : delta_u_field ] ) thus makes a very small contribution to the velocity in the current - sheet region ( and also in the center - plane region ) . in the absence of the turbulent cross helicity ( @xmath145 ) , we have no induced flows , then the flow configuration remains as it was originally set ( @xmath117 and @xmath118 ) . if the flow is a potential one , we have no mean vorticity . in this situation , we have no cross - helicity generation since @xmath146 everywhere . a stagnation - point - like flow is such an example . on the other hand , in the presence of the turbulent cross helicity , the induced flow due to the cross - helicity effect should alter the flow profile . how much flow configurations are modulated depends on @xmath147 or @xmath148 , the turbulent cross helicity scaled by the turbulent mhd energy . for example , in the case of maximum cross helicity ( @xmath149 ) , the induced flow speed is close to the alfvn speed . as this result , outside the current - sheet region , the inflow is converging toward the central plane ( @xmath136 ) , and in the current - sheet region the outflow is concentrated very narrow region in the vicinity of the midplane ( @xmath134 ) , where the magnitude of @xmath40 is very small . in the mean magnetic equation , the turbulent cross helicity coupled with the large - scale vortical motion gives rise to the induced magnetic field that is aligned with the mean velocity [ eq . ( [ eq : delta_b_sol ] ) ] . the spatial distribution of the turbulent cross helicity , eq . ( [ eq : w_distr_mr ] ) , leads to the induced field : @xmath150 consequently , the magnetic field @xmath151 turns to be a curved field configuration as schematically depicted in fig . [ fig : flow_in_fast_mr ] . combination of the inflow converging toward the @xmath135 axis with the curved magnetic field line gives a x - point - like configuration , which is known to be suitable for the fast reconnection . in the absence of the turbulent cross helicity , such modulations of the large - scale velocity and magnetic fields never occur . this suggests that the presence of cross helicity with a favorable mean - field configuration leads to the enhancement of magnetic reconnection rate as compared with the case without cross helicity . flow and magnetic - field configurations in fast magnetic reconnection . solid lines : magnetic field , dashed lines : plasma flow . shaded region represents the reconnection or diffusion region . signs of the turbulent cross helicity @xmath40 associated with the slow shocks are also indicated . sign of @xmath40 is negative ( or positive ) for the region where the alfvn wave is propagating in the directions parallel ( or antiparallel ) to the large - scale magnetic field . ] as we have already mentioned , the width of the diffusion region in magnetic reconnection is reduced because of the converging flow due to eq . ( [ eq : delta_u_field ] ) and the curvature of the magnetic field due to eq . ( [ eq : delta_b_in_mr ] ) , which would lead to the enhancement of the reconnection rate . here , we estimate the reconnection rate with the aid of the expressions for the cross - helicity effects in the magnetic induction [ eq . ( [ eq : delta_b_sol ] ) ] and in the momentum balance [ eq . ( [ eq : delta_u_sol ] ) ] . as the dimensionless reconnection rate , we adopt the inflow alfvn mach number@xcite @xmath152 where @xmath153 is the magnetic field measured in the alfvn speed unit ( @xmath154 , @xmath155 : alfvn speed , @xmath156 : magnetic field measured in the physical unit ) . let us denote the inflow velocity and magnetic field in the absence of the cross - helicity effects as @xmath157 and @xmath158 , respectively . from eqs . ( [ eq : delta_u_sol ] ) and ( [ eq : delta_b_sol ] ) , the inflow velocity and magnetic field with the cross - helicity effects , @xmath159 and @xmath160 , are expressed as @xmath161 @xmath162 respectively ( fig . [ fig : field_modulation ] ) . this system of equations are written in a matrix form as @xmath163 where @xmath164 , @xmath165 , etc . solving eq . ( [ eq : u_in0_b_in0 ] ) with respect to @xmath166 and @xmath167 , we have @xmath168^{-1 } \begin{pmatrix } 1 & \frac{\gamma}{\beta}\\ \frac{\gamma}{\beta } & 1 \end{pmatrix } \begin{pmatrix } { } ^t{\bf{u}}_{\rm{in}}^{(0)}\\ { } ^t{\bf{b}}_{\rm{in}}^{(0 ) } \end{pmatrix}. \label{eq : u_in_b_in}%(45)\ ] ] then , for given @xmath157 and @xmath158 , we can calculate @xmath159 and @xmath160 . modulation of the velocity ( left ) and magnetic field ( right ) due to the cross - helicity effects . in this figure , the case with the negative cross helicity ( @xmath169 , @xmath170 ) is drawn . ] from eq . ( [ eq : u_in_b_in ] ) , the magnitude of the inflow velocity and magnetic field are given as @xmath171^{-1/2 } \times } \nonumber\\ & & \hspace{20pt } \left [ { { { \bf{u}}_{\rm{in}}^{(0)}}^2 + \left ( { \frac{\gamma}{\beta } } \right)^2 { { \bf{b}}_{\rm{in}}^{(0)}}^2 + 2 \frac{\gamma}{\beta } { \bf{u}}_{\rm{in}}^{(0 ) } \cdot { \bf{b}}_{\rm{in}}^{(0 ) } } \right]^{1/2}\hspace{-12pt } , \label{eq : u_in_magnitude}%(46 ) \end{aligned}\ ] ] @xmath172^{-1/2 } \times } \nonumber\\ & & \hspace{20pt } \left [ { \left ( { \frac{\gamma}{\beta } } \right)^2 { { \bf{u}}_{\rm{in}}^{(0)}}^2 + { { \bf{b}}_{\rm{in}}^{(0)}}^2 + 2 \frac{\gamma}{\beta } { \bf{u}}_{\rm{in}}^{(0 ) } \cdot { \bf{b}}_{\rm{in}}^{(0 ) } } \right]^{1/2}\hspace{-12pt}. \label{eq : b_in_magnitude}%(47 ) \end{aligned}\ ] ] then , the reconnection rate is expressed as @xmath173^{1/2}\hspace{-12pt } , \label{eq : min_exp}%(48)\ ] ] where @xmath174 is the reconnection rate in the absence of the cross - helicity effect , and @xmath175 is the angle between @xmath157 and @xmath158 . in the case of converging inflow and x - point like curved magnetic field such as in fig . [ fig : flow_in_fast_mr ] , the sign of @xmath176 or @xmath177 obeys the quadrupole distribution similar to that of @xmath40 [ eq . ( [ eq : w_distr_mr ] ) ] : @xmath178 then , the sign of @xmath179 in eq . ( [ eq : min_exp ] ) , as well as the @xmath180-related contributions , is always positive . in this sense , in the converging inflow cases it is not the @xmath181 and @xmath176 themselves but the magnitudes of them which determine the modulation of @xmath182 . if @xmath157 is perpendicular to @xmath158 ( @xmath183 ) , eq ( [ eq : min_exp ] ) is reduced to @xmath184^{1/2}\hspace{-12pt}. \label{eq : min_exp_pi/2}%(50)\ ] ] the behaviors of @xmath182 with respect to @xmath185 with several values of @xmath175 in the case of @xmath186 are plotted in fig . [ fig : reconnection_rates ] . here , as an example , we consider the domain with @xmath140 , where @xmath187 . as expected , we have no enhancement of @xmath182 by the cross - helicity effect if @xmath188 . with increase in @xmath189 , @xmath182 increases up to @xmath190 ( at @xmath191 ) . here , we should note that @xmath147 is expressed by the turbulent cross helicity scaled by the turbulent mhd energy as @xmath192 with @xmath193 , which corresponds to the fact that @xmath194 [ eq . ( [ eq : gamma_model ] ) ] is smaller than @xmath195 [ eq . ( [ eq : beta_model ] ) ] but the orders of them are the same . variation of the reconnection rate @xmath182 against the scaled turbulent cross helicity @xmath147 . the angle between the inflow velocity and magnetic field in the absence of the cross - helicity effects , @xmath196 and @xmath158 , is denoted by @xmath175 . : @xmath183 ; @xmath197 : @xmath198 ; - - - : @xmath199 ; @xmath200 , @xmath201 ; @xmath202 , @xmath203 . in all plots , the magnetic reconnection rate without the cross - helicity effects , @xmath204 , is set equal to @xmath205 . ] these results suggest that the magnetic reconnection rate may be substantially enhanced by the cross - helicity effects if the magnitude of the scaled turbulent cross helicity is large enough such as @xmath206 on the other hand , the cross - helicity effect may be negligible for the magnetic reconnection rate in the case of the turbulent cross helicity of @xmath207 in the case of non - converging inflow geometry , the signs of @xmath40 and @xmath176 may be opposite to each other , then the sign of @xmath179 is negative . in such a case , the cross - helicity effects work for reducing the reconnection rate for small @xmath185-value domains since the contribution of @xmath179 is more dominant than that of @xmath208 . in fig . [ fig : reconnection_rates ] , the @xmath182 behavior in such situations ( @xmath209 ) are also plotted . another important point to remark is that the turbulent cross helicity distributions preferable to the reconnection are expected to be sustained in the down - stream region or at the after - reconnection stage . if there is an asymmetry between the directions parallel and antiparallel to the magnetic field , we have a certain production mechanism of the non - vanishing turbulent cross helicity . as is well known , x - point reconnection and slow shock regions are source of the alfvn waves . for instance , the anisotropic ion beams in the plasma sheet boundary layer , which is the so - called psbl ions in the magnetotail , @xcite is known to excite the alfvn waves by the ion cyclotron beam instability . @xcite the reflected ion beams from the slow shock front may excite the alfvn wave in the shock upstream along the magnetic - field line as well.@xcite since the alfvn waves generated by slow shocks propagate in the outward direction , the sign of the cross helicity associated with the alfvn waves is negative ( or positive ) if the large - scale magnetic field is parallel ( or antiparallel ) to the direction of the wave propagation . in terms of turbulent statistical quantities , this mechanism is related to the cross - helicity generation due to the inhomogeneity along the mean magnetic field , @xmath62 [ eq . ( [ eq : b_grad_k ] ) ] . the resultant cross - helicity distribution is consistent with the one expressed by eq . ( [ eq : w_distr_mr ] ) as shown in fig . [ fig : flow_in_fast_mr ] . in relation to the spatial distribution of the turbulent cross helicity , we should note that the quadrupole symmetric distribution of the turbulent cross helicity [ eq . ( [ eq : w_distr_mr ] ) ] may be commonly or ubiquitously observed in the magnetic - reconnection configurations . for example , in the tearing - mode configuration , the large - scale vorticity @xmath118 associated with the plasma flow is distributed as depicted in fig . [ fig : tearing_mode ] . then the production of the turbulent cross helicity due to the coupling of @xmath210 and @xmath118 is expressed as in eq . ( [ eq : j_omega_mr ] ) , leading to the turbulent cross helicity spatially distributed as in eq . ( [ eq : w_distr_mr ] ) . these points should be carefully scrutinized using numerical simulations . tearing - mode configuration and distribution of the turbulent cross helicity . ] magnetic reconnection was viewed from the flow turbulence interaction mediated by the cross - helicity effects . in addition to the enhancement of transport , the primary effect of turbulence , turbulent transport may be effectively suppressed by some other effects of turbulence . the _ structure _ information of turbulence represented by some pseudoscalar statistical quantities such as the turbulent cross , kinetic , current , magnetic helicities is the key for analyzing the transport suppression or structure formation due to turbulence , whereas the _ intensity _ information of turbulence is essential for describing the transport enhancement or structure destruction . in order to investigate the balance between these two effects , we first considered the cross - helicity production mechanisms in turbulence . the cross - helicity effects , which appear in the reynolds stress [ eq . ( [ eq : re_str_exp ] ) ] and the turbulent electromotive force [ eq . ( [ eq : e_m_exp ] ) ] , were examined in the context of the magnetic - flux freezing in turbulence , turbulent dynamo process , and magnetic reconnection . in the arguments on magnetic reconnections , we stressed the importance of the large - scale flow structure and global magnetic - field configuration since these inhomogeneous structures constitute the conditions for the cross - helicity generation . the spatiotemporal distribution of the cross helicity , in turn , determines the transport properties of turbulence , then affects the magnetic reconnection rates . we showed that the large - scale flow and magnetic - field structures associated with the magnetic reconnection are favorable for the cross - helicity generation in some cases , and that the flow and magnetic - field structures induced by the cross - helicity effects are consistently favorable for the enhanced magnetic reconnections . in the absence of the turbulent cross helicity , these mutual interactions between the large - scale structure and turbulence will never appear at all in the magnetic reconnection process . this theoretical scenario of magnetic reconnection based on the mutual interaction between the large - scale structures and the turbulent cross helicity should be validated using numerical simulations . a few points should be remarked on such simulations . first , the theoretical formulation based on the cross - helicity effects through eqs . ( [ eq : delta_b_sol ] ) and ( [ eq : delta_u_sol ] ) extended in this paper is very general , and not limited to the two - dimensional cases . estimate of reconnection - rate modulation and numerical validation of it in three dimensions are desired much especially in the context of astrophysical magnetic reconnection . however , we can start with two - dimensional cases since , unlike the kinetic and magnetic helicities , the cross helicity can exist even in the two dimensional cases . secondly , boundary conditions for the numerical simulations may be important but in a limited sense . actually , it was reported that the conditions on flows are important in determining several features of magnetic reconnection , including the nature of the inflow , the length of the diffusion region , etc.@xcite however , in the scenario presented in the present paper , such boundary conditions are important only in the sense that they affect the spatial distribution of the turbulent cross helicity ( more generally , distributions of relevant turbulent statistical quantities ) through which properties of turbulent transport are determined . numerical simulations validating the present theoretical idea are being prepared , and results will be reported in the forthcoming papers . one of the authors ( ny ) would like to cordially thank alexander g. kosovichev for his encouragement , without which ny might have not even started considering the turbulent effects in magnetic reconnections . part of this work was performed during ny s stay at the center for turbulence research ( ctr ) , stanford university / nasa ames for the ctr summer program 2010 ( june - july , 2010 ) and at the nordic institute for theoretical physics ( nordita ) as a guest researcher ( february and july - august , 2011 ) . from a closure theory for inhomogeneous turbulence applied to magnetohydrodynamic ( mhd ) turbulence , the reynolds ( and turbulent maxwell ) stress @xmath43 and the turbulent electromotive force @xmath23 are known to be expressed as@xcite @xmath211 @xmath212 where @xmath213 is the residual energy , and @xmath214 and @xmath215 are the strain rates of the mean velocity and magnetic fields defined by @xmath216 @xmath217 in eqs . ( [ eq : re_strs_tsdia ] ) and ( [ eq : e_m_tsdia ] ) , @xmath31 , @xmath32 , @xmath34 , @xmath35 , and @xmath33 are the transport coefficients , which are known to be expressed as @xmath218 , \label{eq : alpha_spectral}%(a5)\end{aligned}\ ] ] @xmath219 , \label{eq : beta_spectral}%(a6)\end{aligned}\ ] ] @xmath220 , \label{eq : gamma_spectral}%(a7)\end{aligned}\ ] ] @xmath221 here , @xmath222 is the green s function of mhd turbulence , and @xmath223 , @xmath224 , @xmath225 , @xmath226 , @xmath226 are the spectral functions of the turbulent kinetic helicity , current helicity , kinetic energy , magnetic energy , and cross helicity . equations ( [ eq : alpha_spectral])-([eq : gamma_spectral ] ) show that @xmath33 , @xmath34 , and @xmath35 are determined by the spectral distributions of the residual helicity , energy , and cross helicity , respectively . these coefficients have spatiotemporally non - local nature through the green s function . the simplest possible models for the transport coefficients @xmath33 , @xmath34 , and @xmath35 are given by using turbulent one - point quantities : the turbulent residual helicity @xmath227 , the turbulent mhd energy @xmath228 , and the turbulent cross helicity @xmath229 respectively : @xmath230 @xmath231 @xmath232 ( @xmath233 , @xmath195 , and @xmath194 : model constants ) . here @xmath234 is the time scale of turbulence , which is often expressed by using the turbulent mhd energy scaled by its dissipation rate : @xmath235 if we drop the @xmath33- or helicity - related term in the turbulent electromotive force @xmath23 [ eq . ( [ eq : e_m_tsdia ] ) ] as @xmath236 the mean induction equation ( [ eq : mean_b_eq_e_m_full ] ) yields @xmath237 where the molecular magnetic diffusivity @xmath57 have been dropped as compared with the turbulent counterpart @xmath34 . in a stationary state , the mean magnetic field @xmath238 is a special solution of eq . ( [ eq : mean_b_eq_cross_helicity ] ) provided that the spatial variation of @xmath239 is not so large.@xcite the mean momentum equation in the rotational form is written as @xmath240 where @xmath113 is the reynolds stress defined by eq . ( [ eq : re_str_def ] ) , @xmath114 the mean external force , @xmath241 the mean pressure function . if we substitute expression ( [ eq : re_str_exp ] ) for @xmath113 into eq . ( [ eq : mean_u_eq_rot ] ) , we have @xmath242 then , the mean vorticity equation is given by @xmath243 + \nabla \times { \bf{f}}. \label{eq : omega_eq_exp}%(b3)\end{aligned}\ ] ] in addition to the reynolds stress , which contains fluctuating velocity correlation and the fluctuating lorentz force @xmath244 , the turbulence effects enter the mean velocity equation also through the mean lorentz force @xmath245 . the mean ohm s law is written as @xmath246 if we substitute expression ( [ eq : e_m_exp ] ) for the turbulent electromotive force into eq . ( [ eq : mean_ohms_law ] ) and solve it with respect to @xmath29 , we have the mean electric - current density as @xmath247 ( @xmath115 : vector potential ) . here , we have dropped the molecular magnetic diffusivity @xmath57 as compared with the turbulent one @xmath34 ( @xmath248 ) . then , the mean lorentz force is expressed as @xmath249 here , we should note that the @xmath35- or cross - helicity - related term remains in the mean lorentz force expression . this gives a strong contrast to the @xmath33- or helicity - related term , which disappears due to the alignment of @xmath92 with @xmath29 in eq . ( [ eq : mean_j_exp ] ) . in other words , unlike the turbulent helicity , the turbulent cross helicity contributes to the mean velocity or vorticity equation through the mean lorentz force . substituting eq . ( [ eq : mean_lorentz_force_exp ] ) into the mean lorentz force in eq . ( [ eq : omega_eq_exp ] ) , we obtain the equation for the mean vorticity as @xmath250 \nonumber\\ & + & \nabla \times \left [ { { \bf{f } } + \frac{1}{\beta } \left ( { { \bf{u } } \times { \bf{b } } } \right ) \times { \bf{b } } - \frac{1}{\beta } \frac{\partial { \bf{a}}}{\partial t } \times { \bf{b } } } \right ] . \label{eq : mean_omega_eq_exp}%(b7)\end{aligned}\ ] ] this mean vorticity equation was originally derived in the context of system rotation effect.@xcite as long as the expansion ( [ eq : u_expand ] ) is adopted and the induced vorticity is considered , this equation can be also utilized for cases without a system rotation . e. e. grigorenko , l. zelenyi , m. s. dolgonosov , a. v. artemiev , c. owen , j. -a . sauvard , m. hoshino , and m. hirai , `` non - adiabatic ion acceleration in the earth magnetotail and its various manifestations in the plasma sheet boundary layer , '' to be published in space science review ( 2011 ) . m. hoshino , t. mukai , t. yamamoto , and s. kokubun , `` ion dynamics in magnetic reconnection : comparison between numerical simulation and geotail observations , '' j. geophys . 103 * , 4509 - 4530 ( 1998 ) .

we present a unified model for molecular core formation and evolution , based on numerical simulations of converging , supersonic flows . our model applies to star formation in gmcs dominated by large - scale turbulence , and contains four main stages : core building , core collapse , envelope infall , and late accretion . during the building stage , cores form out of dense , post - shock gas , and become increasingly centrally stratified as the mass grows over time . even for highly - supersonic converging flows , the dense gas is subsonic , consistent with observations showing quiescent cores . when the shock radius defining the core boundary exceeds , where is the isothermal sound speed , a wave of collapse propagates from the edge to the center . during the building and collapse stages , density profiles can be fit by bonnor - ebert profiles with temperature 1.2 - 2.9 times the true value , similar to many observed cores . as found previously for initially static equilibria , outside - in collapse leads to a larson - penston density profile . the third stage , consisting of an inside - out wave of gravitational rarefaction leading to , , is also similar to that for initially - static spheres , as originally described by shu . we find that the collapse and infall stages have comparable duration , , consistent with estimates for observed prestellar and protostellar ( class 0/i ) cores . core building takes longer , but does not produce high - contrast objects until shortly before collapse . the time to reach core collapse , and the core mass at collapse , decrease with increasing inflow mach number . for all cases the accretion rate is early on but sharply drops off ; the final system mass depends on the duration of late - stage accretion , set by large - scale conditions in a cloud .

0904.3568

dense molecular cores are the immediate precursors of new stars on small scales , and understanding how they grow and evolve is fundamental to the theory of star formation @xcite . because many elements are involved in core formation , complete theoretical models have not yet been developed , and it is not yet clear whether a single dynamical effect dominates the overall process , or whether several contributing effects have comparable importance . in one limit that has been studied in some detail , ambient velocities are negligible , and self - gravitating cores form by the slow diffusion of partially - ionized gas through strongly - supporting magnetic fields until a supercritical configuration is reached ( e.g. @xcite ) . in another limit , which has been considered more recently but in much less detail at core scales , magnetic support is negligible , and supersonic turbulence creates and destroys condensations , with some fraction of this gas sufficiently dense and long - lived that it can undergo collapse ( e.g. @xcite ) . as observed clouds are both magnetized and strongly turbulent , the eventual theory for core formation that is developed must account for both processes ; pioneering work towards this goal has begun ( e.g. @xcite ) . because of the technical challenges involved in building comprehensive models and the need to elucidate the contributing physics , it is important to develop simplified models in greater detail . in this contribution , we consider aspects of core growth and evolution in the turbulence - dominated , unmagnetized limit . increasingly detailed observations in recent years provide constraints on theoretical models ( see e.g. the reviews of @xcite ) . one class of observations focuses on the density distribution within cores . one - dimensional ( angle - averaged ) density profiles ( e.g. ) generally show a uniform - density center surrounded by a power - law envelope extending to an outer radius @xmath7 , which is consistent with the density profile of a static , isothermal , unmagnetized bonnor - ebert ( hereafter be ) sphere @xcite . the interpretation in terms of static equilibrium is problematic , however , insofar as many cases show center - to - edge density contrasts exceeding the maximum ratio ( @xmath8 ; here @xmath9 is the central density ) that would be stable against collapse , and would also require central temperatures greater than observed values in order to provide support for the total masses inferred from the integrated continuum emission . in addition , cores are generally not isolated ; rather than being surrounded by a high - temperature , low - density medium with pressure matching the core s outer edge , they are surrounded by moderate - density cold molecular gas representing clumps and filaments within larger clouds . the interpretation of observed density profiles as static solutions is also not unique , in that dynamically - collapsing cores initiated from near - critical equilibrium show the same density profiles as ( supercritical ) static solutions ( see e.g. @xcite , and below ) . concentrations formed within turbulent flows can also have density profiles resembling be spheres @xcite . velocity information can distinguish between static , oscillating @xcite , and collapsing cores , and can potentially also help discriminate how these cores formed out of more diffuse gas . dense , low - mass cores generally have subsonic internal velocity dispersions , whether for isolated cores or for cores found in clusters ( e.g. ) . in cores containing protostars , signatures of infall on small scales ( @xmath10 ) , believed to be indicative of gravitational collapse , have been observed via the asymmetry of molecular lines that trace high - density gas ( e.g. @xcite ) . for starless cores , inward motions are often evident over both small ( @xmath11 ) and larger scales ( @xmath12 ) , sometimes encompassing a whole star - forming complex ( @xcite ; see also ) . small - scale inward motions within cores are subsonic @xcite , while larger - scale motions can be transonic or supersonic ( @xcite infer higher velocities in lower - density gas ) , and may be indicative of converging larger - scale flows in which dense gas builds up in a shock - bounded stagnation region . the relative durations of starless and protostellar ( i.e. containing an accreting embedded class 0 or i object ) stages of core evolution are determined by comparing the relative numbers of the two classes of sources in a given cloud . absolute core lifetimes are further obtained by comparison to the number of t tauri stars with measured ages . several studies using this statistical approach in different clouds have reached similar conclusions : the durations of the starless and accreting stages of cores are comparable . typical starless core lifetimes are estimated at @xmath13 , amounting to a few times the free - fall time @xmath14 measured at the mean core density @xmath15 . with lifetimes considerably below the ambipolar diffusion time for strong magnetic fields @xmath16 ( e.g. @xcite ) , this suggests that observed cores are gravitationally supercritical with respect to the magnetic field . this conclusion is also supported by magnetic field zeeman observations , indicating that cores have mean mass - to - magnetic - flux ratios two times the critical value @xcite . since cores are only identified in millimeter and submillimeter continuum when the @xmath17 exceeds a few @xmath18 , in principle it is possible that an extended period of slow diffusion at lower density precedes the observed core stage . turbulence accelerates ambipolar diffusion , however ( e.g. @xcite ) , so it is also possible that the flux loss needed to reach a magnetically supercritical state may occur more dynamically , at densities below @xmath19 . theoretical modeling of core evolution has a long history . much work has focused on the evolution of unstable thermally - supported equilibria into collapse ( formally resulting in infinite density at the origin ) , followed by accretion of the envelope . self - similar solutions for collapse and/or accretion stages of isothermal spheres were found by @xcite , @xcite , @xcite , and @xcite ; these were later generalized by @xcite . @xcite and @xcite ( hereafter lp ) independently found self - similar solutions which describe the density and velocity prior to the instant of protostar formation ( defined as the instant at which the central density become infinite ) . in the lp solution , the radial velocity approaches a constant value @xmath20 and the density approaches @xmath21 at the instant of central protostar formation , with mass inflow rate @xmath22 . here , @xmath1 is the isothermal sound speed , and the dimensional factor in the accretion rate is given by @xmath23 the analysis of @xcite showed that for an initial profile that is a static singular isothermal sphere , @xmath24 , evolution yields an `` inside - out '' infall solution in which a wave of rarefaction propagates outward at the sound speed . inside of the expansion wave , the mass inflow rate is @xmath25 independent of @xmath26 , and gas accelerates to free - fall ( @xmath27 , @xmath28 ) . @xcite connected and extended the investigations of lp ( which address evolution prior to protostar formation ) with that of shu ( which focuses on the accretion stage ) . he showed that self - similar solutions before and after the point of singularity formation ( i.e. @xmath29 ) can be smoothly matched . this allowed the lp solution to be extended into the accretion phase with similar free - fall behavior near the origin ; @xcite also found a sequence of self - similar solutions valid for all time that approach the @xcite expansion wave solution . many numerical simulations of isothermal collapse have shown that the density in the core approaches a @xmath30 profile at the point of protostar formation , regardless of how collapse is initiated @xcite . these simulations include initiation from a static configuration that is unstable , and initiation from static , stable configurations that are subjected to transient compression , either from enhanced external pressure or a converging velocity field . another feature common to the results from simulations is that the collapse generally begins on the outside , with the infalling region propagating inward as the central density increases . at the time of singularity formation , the central velocity has been found to be comparable to the value @xmath31 derived by lp , with the density normalization also similar to the lp result @xcite . following the instant of protostar formation , the evolution of the mass accretion rate over time depends strongly on the initial conditions , however . simulations with triggered core collapse @xcite are motivated by the fact that star - forming regions are highly dynamic , such that external compression may significantly affect core internal evolution , and enhance the accretion rate by raising the central density . triggering events may be associated with high mass star formation , but even without these highly energetic events , the large - scale turbulence that pervades giant molecular clouds ( gmcs ) can compress initially - quiescent cores . taking this idea one step further , it is interesting to consider not just the core collapse process , but also the core formation process , in a strongly turbulent medium . @xcite conducted one such study , considering how an impulsive converging velocity field can create gravitationally bound , centrally - concentrated cores . core formation induced by supersonic turbulence has also been studied in a number of numerical simulations that focus on the large scales , with much of the emphasis on determining the distribution of core masses for comparison to observed core mass functions and the stellar imf ( see e.g. the review of @xcite ) . however , these studies have not had sufficient resolution to investigate the internal properties of the cores that form . if the mass of a core is built up over time as the post - shock product of colliding supersonic flows , what is the detailed evolution leading up to collapse , and during the accretion phase ? in this paper , we initiate a study of dynamically - induced core formation and collapse by considering perhaps the simplest possible situation : a supersonic , converging , spherical flow . our initial conditions are a uniform low - density medium with no stratification . a dense core forms inside a spherical shock , and over time becomes stratified as its mass grows and it becomes self - gravitating . when the stratification becomes too great , collapse and subsequent accretion occurs in a similar manner to the case of an initially - unstable static equilibrium . we consider cases of varying inflow mach number , and with the large - scale inflow either steady over all time , or shut off after an interval . the condition of spherical inflow that we adopt for this first study is , of course , likely to be rare in real clouds . as the main purpose of this study is to take the first step towards unified models of core formation and collapse in dynamic environments , however , we consider one - dimensional solutions the natural place to start . we shall show that many features consistent with observed cores are evident even in these idealized models , suggesting that they are generic to dynamic core formation scenarios . the present set of simulations , in addition to enabling identification of characteristic evolutionary stages , also serve as a useful reference point for more realistic but more complex simulations . more typical than a converging spherical flow would be a converging planar flow , which yields a dense post - shock stagnation layer in which self - gravitating cores can form . numerical studies that we have begun for supersonic planar inflows show results for core building and collapse that are qualitative similar to the present results for supersonic spherical inflows . the plan of this paper is as follows : in 2 we present the governing equations and describe our problem specification and numerical method . section 3 briefly describes results of collapse initiated from static configurations , demonstrating that we reproduce prior results , and providing a baseline for comparison to our models of dynamic formation and collapse . section 4 presents numerical results for our converging - flow simulations , covering the stage of core formation and evolution up to the point of singularity formation in 4.1 and the subsequent stages in stages in 4.2 . we introduce a breakdown into new physically - defined stages in 4.3 , and quantify the evolution of accretion rates in 4.4 . section 5 summarizes our new results and discusses our findings in the context of previous theory and observations . the equations of motion for a spherically symmetric flow take the form : @xmath32 @xmath33 where @xmath34 is the mass within radius @xmath26 defined by @xmath35 , @xmath36 is the radial velocity , @xmath37 is the gas pressure , and @xmath38 is the density . for prestellar collapse , an isothermal equation of state @xmath39 is often used , because cooling is so efficient that the gas remains at nearly constant temperature during the gravitational collapse @xcite . we adopt an isothermal equation of state . for ease of comparison with previous work , we introduce the following dimensionless variables : @xmath40 @xmath41 @xmath42 @xmath43 @xmath44 here @xmath45 is a fiducial density representing the volume - averaged ambient density in a cloud on large scales , which we shall later set to the uniform density of the converging flow . the unit of velocity is the isothermal sound speed @xmath46 the unit of time is related to the free - fall time at the fiducial density by @xmath47 the unit of length is related to the jeans length at the fiducial density @xmath48 by @xmath49 and the corresponding basic unit of mass is @xmath50 $ ] . the mass unit adopted in equation ( [ m_def ] ) is larger than this by a factor @xmath51 : @xmath52 with the dimensionless variables , the fluid equations become @xmath53 @xmath54 @xmath55 we solve the 1d hydrodynamic equations ( [ cont_dim_eq])-([mass_dim_eq ] ) with the zeus-2d @xcite code , in spherical symmetry . during the evolution to singularity formation ( collapse phase ) , we adopt an inner reflecting boundary condition . for the post - collapse accretion phase , we implement a sink cell @xcite at the origin when the central density reaches a reference value . subsequently , the inner boundary condition is changed to outflow , and the value of the central point mass is tracked via the integrated flow off the grid , with @xmath56 . the sink cell is only implemented after the inflow in the central region become supersonic , so that information from the inner boundary can not propagate into the remainder of the grid . as discussed in 1 , many previous numerical simulations of core evolution have adopted static initial conditions and a fixed total mass ( e.g. , @xcite ) . the adopted initial density profile shapes are consistent with ( or similar to ) a hydrostatic equilibrium , i.e. a be sphere , with the initial density perturbed above the value that can be supported by the internal pressure in order to initiate collapse . these fixed - mass simulations adopt a prescribed external pressure at a low density with an effectively lagrangian outer boundary @xcite , or else a fixed outer boundary with prescribed external pressure and no mass inflow @xcite . @xcite explored both types of boundary conditions , and found very similar results for a given initial cloud density profile . they concluded that the evolution for the fixed - mass case is insensitive to the outer boundary condition for initially - unstable equilibria . this is consistent with the argument of @xcite that the outer boundary condition does not affect evolution up to collapse as long as the free - fall time ( eq . [ tff_eq ] ) is shorter than the cloud crossing time . the ratio of the free - fall time at the mean density to the sound crossing time @xmath57 over the radius of a be sphere is @xmath58 , which approaches @xmath59 as @xmath60 approaches infinity . here , @xmath61 , the dimensionless gravitational potential . for the critical case , the free - fall time is 0.71 times the sound crossing time over the radius . for comparison to previous work , we consider collapse of an initially - static be sphere . for initial conditions , we adopt a critical be sphere , i.e. the outer boundary of the grid is at radius @xmath62 corresponding to dimensionless outer radius @xmath63 . here , @xmath64 is the total core mass divided by its volume . the mass of the critical be sphere is @xmath65 to initiate collapse , density is perturbed above the equilibrium value by @xmath66 . our outer boundary condition is at a fixed pressure , with no inflow . the temporal evolution of the accretion rate for this model is shown in figure [ fig : mfl_pre ] . features similar to those outlined by @xcite are observed in our simulation . the accretion rate peaks at a value approaching the lp prediction at the instant when the central density ( formally ) becomes infinite , and steeply decreases thereafter . @xcite find that the decline in the accretion rate after singularity formation ( unlike the increase in the accretion rate in hunter s self - similar extension of the lp solution ) can be attributed to the variation of velocity with radius in the sphere as it evolves toward collapse . we also considered cases of much larger initial static spheres , with outer boundaries @xmath67 corresponding to highly - unstable be configurations . for these cases , the accretion rate decreases until it reaches a plateau at @xmath68 , consistent with the value reported in @xcite ( when the density is 10% greater than for hydrostatic equilibrium ) , and then further declines to zero after a rarefaction wave propagates inward from the outer boundary to reach the center . similar to the results of previous simulations , we find that the first collapse is `` outside - in , '' with velocities initially nonzero only in the outer parts where the imbalance between gravity and pressure is largest . , as was previously shown by @xcite . ] this is because the inner portions of the sphere , at @xmath69 , initially are equivalent to _ stable _ be solutions . in all of our models initiated from static spheres , the density profile approaches the lp self - similar solution @xmath70 , and the velocity in the inner region approaches @xmath20 , at the moment of singularity formation . before this time , the density profile in the central region is flat with a magnitude that increases over time . the process can be thought of as a wave of compression propagating from the outside to the inside , creating a density distribution in which the ratio of radius to the jeans length at the local density is everywhere constant : @xmath71 . the singularity represents the instant the compression wave converges at the center . supersonic inflow velocities can be achieved without shock formation in the interior of the core ( except at @xmath72 ) because inward acceleration occurs at all radii where the inward gravitational force exceeds the outward pressure force . by construction , the solutions initiated from static configurations all have gravity ( slightly ) exceeding pressure forces everywhere in the initial state , so that the inward acceleration is nonzero . after a singularity forms at the center , accretion begins , and the flow in the interior transitions from the @xmath73 , @xmath74 lp solution to a free - fall solution , with the accreting region propagating from the interior to the exterior in a manner similar to that described by @xcite . thus , while collapse develops in an `` outside - in '' fashion , accretion develops in an `` inside - out '' fashion . note that accretion in any centrally - concentrated configuration should work its way outward from the center , because gravitational collapse times decline outwards @xmath75 , which is @xmath76 for an inverse - square density profile ( true for either the lp profile or the singular isothermal sphere ) in this section , we present the results of our simulations of core formation and evolution for the case of converging spherical , supersonic flows . for these simulations , the outer boundary condition at @xmath77 is maintained at a constant density and inflow velocity . the inflow velocity is characterized by the mach number relative to the isothermal sound speed , @xmath78 . we vary @xmath79 over the range from 1.05 to 7 . for some models ( see below ) , we suppress inflow at the outer radius after collapse occurs to test how the late - time accretion rate is altered . the initial conditions consist of uniform ( low ) density , and uniform inflow velocity equal to the value at the outer boundary . the size of the grid , in terms of the reference length scale given in equation ( [ r0_def ] ) , is @xmath80 , which amounts to a fraction 0.4 of the jeans length at the initial density . note that the radius of a critical be sphere at the same external pressure ( so that @xmath81 ) would be @xmath82 . that is , a critical be sphere confined by the same ambient pressure would be able to fit within our simulation domain , with plenty of room to spare . the size of zones in the radial direction has a constant logarithmic increment , i.e. , @xmath83 , for some @xmath84 , such that @xmath85\delta r_1 $ ] and @xmath86^{-1}$ ] . for all the converging flow simulations in this paper , @xmath87 is set to @xmath88 ; 605 and 597 grids are used during the collapse phase and accretion phase respectively . we begin with a description of the core formation process , which is similar for all of the converging - flow models . because of the reflecting boundary condition at the center and the initial inflow velocities , immediately after we initiate the simulation , a shock forms at the origin and propagates outward . the inflowing matter is compressed by the shock . the shock front divides the converging inflow into two regions : an inner dense post - shock region and an outer low - density region of supersonic inflow . these two regions evolve quasi - independently but are connected by shock jump conditions . under competition between gas pressure and self - gravity , the inner region contracts slowly to begin forming a dense core . as self - gravity starts to overwhelm gas pressure , the dense core enters the collapse phase . at the beginning , the inner region is quasi - hydrostatic , with the velocity a linear function of radius . with negligible density gradient and self - gravity in the early stages , equation ( [ mom_dim_eq ] ) becomes @xmath89 this equation is satisfied by @xmath90 for @xmath91 , @xmath92 , i.e. the coefficient of the linear profile is constant in time . the leftmost lower panels of figures [ fig : be_shock_1.05 ] - [ fig : be_shock_7.0 ] show this linear - velocity behavior in the shocked region , at the time that equals half of the collapse time @xmath93 . throughout this paper , we define the collapse time @xmath93 as the moment that @xmath94 . shown in the leftmost upper panels of figures [ fig : be_shock_1.05 ] - [ fig : be_shock_7.0 ] are the density profiles ; even when the density profile has nonzero gradients , the velocity in the inner region is dominated by the linear term . notice that the leftmost panels have linear - linear scales . the velocity over the whole post - shock region is subsonic and negative , which means the core is slowly contracting . for the @xmath95 models ( see fig . [ fig : be_shock_4.0 ] , [ fig : be_shock_7.0 ] ) , the inner part of the density and velocity profiles oscillate at the beginning of simulations . as the shock front propagates outward , the mass inside the shock increases , and so does self - gravity . after a period of accumulation lasting about 90% of the time until collapse @xmath93 , the slowly - contracting dense region starts to be gravitationally unstable . note that the density and velocity profiles of the regions outside of the shock go through a transient evolution after simulations begin . the density profile evolves from a uniform profile set in the initial condition to a @xmath96 profile consistent with supersonic radial inflow upstream of the shock combines with the steady state mass inflow condition @xmath97 to yield @xmath96 . ] ; the material is also increasingly accelerated by gravity . in figures [ fig : be_shock_1.05 ] - [ fig : be_shock_7.0 ] , the inflection feature in the velocity profiles exterior to the shock corresponds to a wave propagating inward at a speed equal to the inflow speed plus the sound speed . since post - shock velocities are subsonic , gravitational instability is expected to develop at a point when the radius of the inner dense post - shock region becomes comparable to the critical radius of a be sphere . this expectation is indeed borne out by our simulations , which moreover show that the properties of the collapsing dense inner region are similar as those of collapsing cores initialized from hydrostatic be spheres . the collapse follows an `` outside - in '' pattern , starting from the shock front . the central density increases dramatically and the inflow velocity inside the shock becomes supersonic . the collapse propagates inward and establishes a centrally - concentrated density profile @xcite , which approaches the lp density profile @xmath98 , and the velocity approaches @xmath99 . the three panels on the right of figures [ fig : be_shock_1.05 ] - [ fig : be_shock_7.0 ] show the density profile ( top ) and the velocity profile ( bottom ) at three different instants during the outside - in collapse . the density profiles plotted are normalized by the central density and the interior regions are fit by be sphere density profiles . with a fitted temperature @xmath100 . these fits are further discussed below . ] the ratio between the fitted be sphere temperature @xmath100 and the true temperature @xmath101 is also noted in the figures . the first figure of these three shows profiles at the instant when the radius of the post - shock region reaches the critical radius ( see eq . [ radiusbe ] ) of a be sphere at temperature @xmath101 with the same central density i.e. @xmath102 . we define the period after this as the supercritical regime . the center - to - edge density ratio after this point exceeds @xmath103 . the second figure of these three shows the instant when the radius of the post - shock region reaches twice the critical radius of a be sphere . the third figure is at the instant the core collapse @xmath93 ( defined here as @xmath104 ) . the long dashed diagonal line in the third figure shows the lp density profile , which is very close to the numerical solution . the time interval between @xmath105 and collapse for the @xmath106 case in code units is 0.048 ( see eq . [ t0_def ] ) , which corresponds to @xmath107 if the inflowing ambient medium s density is @xmath108 . we note that prior to collapse , the velocities in the dense gas ( inside the shock ) remain small . in particular , for @xmath109 cases , the inflow velocity inside the shock is subsonic throughout the post - shock region before the time when @xmath102 . when @xmath79 is high , figures [ fig : be_shock_4.0 ] and [ fig : be_shock_7.0 ] show that the post - shock velocities remain subsonic up to the instant of collapse . the outward propagation speed of the shock , which from the simulations is approximately constant ( fig . [ fig : shkfrps ] ) at early time , can be obtained using the shock jump conditions . if the shock position is @xmath110 , then provided equation ( [ u_law ] ) is satisfied , the velocity on the downstream side is @xmath111 notice that in order for the downstream velocity to be inward and the shock to be propagating away from the origin , we must have @xmath112 . the isothermal shock jump conditions , with subscript `` @xmath113 '' denoting upstream and `` @xmath114 '' downstream values , are @xmath115 and @xmath116 = d_u[1 + ( u_u - u_{sh})^2]$ ] ; together these imply @xmath117 and @xmath118 . treating the shock speed as approximately constant so that equation ( [ u_d_eq ] ) holds , we can solve for the shock velocity to obtain @xmath119^{1/2})$ ] . for @xmath120 , and taking the upstream velocity as @xmath121 which corresponds to the limit of strongly supersonic inflow , gives @xmath122 = \frac{2}{m_a+ ( m_a^2 + 4)^{1/2}}.\ ] ] the shock speed decreases as the mach number increases , and therefore from equation ( [ u_d_eq ] ) the post - shock flow speed also decreases as the mach number increases . figure [ fig : shkfrps ] shows the position of shock front versus time for @xmath123 . the intercept and the slope are based on the best fit of the linear part where @xmath124 [ 0.0 , 0.04 ] . the intercept is nearly 0 and the slope 0.3061 is the measured shock speed in units of the isothermal sound speed @xmath1 ; equation ( [ eq : ush ] ) predicts a slightly smaller value 0.24 . the analytical solution ( eq . [ eq : ush ] ) for @xmath125 as a function of @xmath79 is plotted as a dotted line in figure [ fig : u_xi_tau_ori ] ( labeled as @xmath126 ) and the shock speeds measured directly from simulations @xmath125 are plotted as asterisks . the analytical approximation is about 15 - 28% below the measured value from the simulations as @xmath79 ranges from 1.05 - 7.0 . using the constant - shock - speed approximation , the immediate post - shock density can also be obtained in terms of the pre - shock density , in the limit @xmath120 and using @xmath127 as : @xmath128 ^ 2,\ ] ] which for @xmath129 is @xmath130 . because of the radial convergence of the inflow in the simulations , @xmath131 varies ; it is initially equal to @xmath132 , but after an initial transient , in the highly - supersonic limit @xmath131 would approach @xmath133 because of mass conservation . it is interesting to investigate how the state of the core when it collapses depends on mach number . figure [ fig : rho_v_collapse ] shows the density profiles and the velocity profiles of the simulated cores at the instant of collapse @xmath93 , for different mach numbers . firstly , it is clear that all the density profiles approach the lp solution , especially for low mach number cases . the inflow velocity inside the shock is supersonic and does not strongly vary over the dense core region . the smaller the mach number is , the closer the inflow speed is to the lp result , @xmath134 . secondly , the radius of the core at the instant of collapse decreases as mach number increases . we plot this dependence in figure [ fig : u_xi_tau_ori ] with diamonds . we quantitatively compare the basic core properties in figure [ fig : u_xi_tau_ori ] , which shows the collapse time @xmath93 and the shock radius @xmath135 at time @xmath93 , both as function of @xmath79 . as noted above , @xmath135 decreases with increasing @xmath79 ; the same is true for @xmath93 . we also recall that @xmath125 decreases with @xmath79 . interestingly , while @xmath136 , and @xmath93 all decrease with @xmath79 , the ratio @xmath137 is nearly constant with @xmath79 : it ranges only from 0.34 to 0.42 ( see fig . [ fig : u_xi_tau_ori ] ) . this result is potentially useful for empirical estimates of core lifetimes , since the inflow velocity @xmath79 , the isothermal sound speed @xmath1 and the radius of dense core are all in principle measurable . if the ratio @xmath137 is taken as a constant @xmath138 , and the shock speed is estimated via equation ( [ eq : ush ] ) , then the life time of cores up to the point of collapse is given in dimensional form by @xmath139/a$ ] . in practice , it may be difficult to measure @xmath79 outside of a core , because the density is much lower than that of the core , and it is difficult to isolate the immediate environment of the core from foreground and background gas . this result is still useful in a statistical sense , however , using the mean mach number of the turbulent flow in a cloud . another direct observable is the core density , so it is interesting to test how the values of the collapse time and radius depend on the mean density in the core at the time of collapse . diamonds in the top panel of figure [ fig : u_xi_tau_nor ] show the core radius in units of @xmath140 , which is @xmath141 , as a function of @xmath79 . we can see @xmath141 is nearly constant , ranging from 4.58 to 3.42 as mach number increases from 1.05 to 7 . taking this as approximately constant , and taking the measured core mean density , the predicted size of core at the time of collapse is @xmath142 . note that this radius is @xmath143 larger than the critial be radius for the same temperature ( see eq . [ radiusbe ] ) . since the post - shock density increases relative to the upstream density approximately as @xmath144 ( see eq . [ eq : rho_d_u ] ) , it is also interesting to test how @xmath145 depends on mach number . in fact this quantity decreases with @xmath79 , as seen in figure [ fig : u_xi_tau_nor ] . to express the core collapse time in terms of observables , we normalize the collapse time using the mean core density . this quantity @xmath146 , is plotted in figure [ fig : u_xi_tau_nor ] as a function of @xmath79 . for reference , the core collapse time normalized using @xmath79 is also plotted in fig . [ fig : u_xi_tau_nor ] . the free - fall time @xmath147 for a uniform sphere in units of @xmath148 is @xmath149 , so that we have @xmath150 as mach number varies from 1.05 to 7 . this time scale is much longer than the observed values @xmath151 for prestellar cores . the reason for this disparity is that during the early part of its evolution , the central density of the core is low , and it would not be identifiable within its surroundings . this is evident in the low contrast between the center and the edge of the core seen in the first frames of fig . [ fig : be_shock_1.05 ] - [ fig : be_shock_7.0 ] . only when the core approaches collapse does the center - to - edge contrast become large . for example , for the @xmath152 model , the center - to - edge density ratio @xmath153 reaches @xmath154 at @xmath155 and reaches @xmath156 at @xmath157 ; this can be compared to the total time until collapse , @xmath158 for this model . for the whole set of models , the observable fraction of the pre - collapse core life is @xmath159% if we choose @xmath160 , or @xmath161% if we choose @xmath162 . taking the period when @xmath160 or @xmath162 as the period over which a core could be observable , e.g. in submilimeter continuum , our simulations give @xmath163 or @xmath164 , respectively . the latter is consistent with observed estimates . another observable aspect of prestellar cores is their density structure . as discussed above , for each solution at the times shown in fig . [ fig : be_shock_1.05 ] - [ fig : be_shock_7.0 ] , we fit a be sphere profile . the density in code units @xmath165 is @xmath166 and the radius in code units is @xmath167 . when the central density is instead used to normalize , the density and radius variables are : @xmath168 and @xmath169 for a be sphere with sound speed @xmath170 , the density profile normalized by the central density is @xmath171 , which is a function of the scaled radius @xmath172 . to fit the density profiles in our simulations to a be sphere , the only free parameter is the ratio of the isothermal sound speed @xmath170 of the be sphere to the sound speed in the simulations , @xmath1 . thus , for any given density profile @xmath173 we adjust the value of @xmath174 until a good match between @xmath171 and @xmath173 is obtained . this procedure yields the fitted temperature ratio : @xmath175 the values obtained for @xmath176 are marked in figures [ fig : be_shock_1.05 ] - [ fig : be_shock_7.0 ] . the range of values we find is 1.23 to 2.89 . this range is consistent with theoretical expectations . as the radius of the be sphere extends to infinity , the density profile approaches the singular solution @xmath177 @xcite corresponding to @xmath178 . the density profile of the collapsed core approaches the lp solution @xmath179 . to match the lp profile with the singular profile therefore requires @xmath180 , which corresponds to a temperature ratio @xmath181 . fits of core profiles to be spheres that indicate values of @xmath100 up to 4.43 times the measured thermal temperature therefore are expected if collapse has taken place ; this does not by itself indicate that magnetic support is present . after the central density becomes singular , the evolution transitions to the infall and accretion stages . for our simulations , we make this transition by implementing an outflow boundary condition at the center when the central density reaches @xmath182 . the initial mass of the central protostar is calculated by integrating the innermost part of the density profile where density is between @xmath183\,\rho_0 $ ] . the specific choice of this density does not significantly affect @xmath184 , or the subsequent evolution since the central density increases dramatically only at the very end of the collapse stage . at the beginning of the accretion phase , the material inside the shock falls onto the protostar . the material approaching the protostar is in a free - fall state @xcite . the region of unsupported infall starts from the center and propagates outward , similar to the expansion wave " described by @xcite . the density profile inside the rarefaction wave changes from @xmath96 to @xmath185 and the velocity profile changes from @xmath186 to @xmath187 . for our simulations , this infall stage ends as the infall rarefaction wave arrives at the shock front . this generally occurs very rapidly ( in less than 10% of @xmath93 ; see below ) . for an initially static density profile @xmath188 , where @xmath189 and @xmath190 is a constant , if the gas pressure is negligible the average speed of the rarefaction wave is @xmath191 . for @xmath192 , which is the lp profile , this yields @xmath193 . for the real case , the initial velocity is nonzero and the gas pressure is non negligible , so that the rarefaction wave propagates at a modified speed . for example , for the @xmath194 model , which has @xmath195 and infall interval @xmath196 , the average speed is @xmath197 . for @xmath198 and @xmath199 , the measured average infall speed of the infall rarefaction wave is @xmath200 and @xmath201 respectively . after the infall rarefaction wave arrives at the shock front , the final accretion stage begins , with material initially outside the shock falling onto the protostar at supersonic speeds . this process is similar to bondi accretion , except that the central mass is growing and the velocity field for our simulations is uniformly converging at large distance . during this stage , the density and the velocity profiles vary @xmath202 and @xmath187 respectively , corresponding to free - fall . the typical density profiles and velocity profiles during the accretion stage for @xmath152 and 4 models are shown in figure [ fig : rho_v_acc ] . three different points in the evolution are shown : the instant of core collapse , the instant when the infall rarefaction wave arrives at the shock front , and a point during the late accretion stage . the transition from the lp profiles to the free fall profiles in density and velocity are clearly evident in the figure . in a real system , the duration of the accretion stage depends on the environment of the protostar , and how long the inflow that creates the core is maintained at large scales . to explore how the late - time evolution is affected by changes in the accretion , we have conducted additional simulations in which the flow inward from the outer boundary is halted at the end of the infall stage ( i.e. when the rarefaction reaches the shock ) . suppression of inflow will affect the mass flux onto the protostar after the rarefaction wave from the boundary reaches the central protostar . we discuss comparison of these models to our standard models in section 4.4 . based on the results presented in 4.1 and 4.2 , we have identified four main stages of protostellar core formation and evolution in a supersonic turbulent medium ( see figure [ fig : stages ] for a schematic depiction ) : \1 . _ core building _ converging flows in a supersonic turbulent medium collide , with post - shock compressed gas accumulating over time in stagnant , shock - bounded regions . if these dense regions are not destroyed by larger scale turbulence , the high density gas will undergo a long contraction process during which gas pressure competes with self - gravity . the typical character of this stage is that the velocity inside the dense gas is subsonic and increases linearly with distance from the center . since the center - to - edge density contrast is relatively low , these clumps are gravitationally subcritical . towards the end of this stage , when the center - to - edge density contrast becomes appreciable , these objects would become observable as prestellar cores . this core building stage lasts @xmath203% of @xmath93 ; only the last @xmath204 would be observable . 2._core collapse _ the core built up during the first stage accumulates enough mass that it becomes gravitationally supercritical , which we operationally define as @xmath205 . self - gravity overwhelms the gas pressure , and the unstable core starts to collapse . the collapse begins near the shock front , where the imbalance between gravity and pressure gradient forces is greatest , and propagates inward . this collapse is an `` outside - in '' process . during core collapse , the central density increases dramatically and the inflow velocity becomes supersonic . as the collapse propagates inward , a density profile @xmath96 is left behind and the velocity increases toward the center . the end of this stage is defined by the instant of protostar formation @xmath93 , when the wave of collapse has reached the center of the core . the density profile approaches the lp profile , @xmath206 . the velocity in the interior of the collapsed core approaches a uniform value comparable to that in the lp solution @xmath20 . the central density is high enough to be optically thick and a protostar forms . because of their large central volume densities , prestellar cores during this collapse stage would be observed as having high peak submm flux densities . this stage lasts for a time @xmath207 , less than 10% of the prestellar core lifetime @xmath93 . _ envelope infall _ during this stage , the high density material inside the shock front falls to the protostar ( or , if angular momentum were included , a circumstellar disk ) . this stage starts at the instant of the protostar formation and ends at the instant when the gravitational rarefaction wave reaches the shock front , clearing out the remnants of the dense gas that accumulated during core building . during this stage , the density and velocity profiles in the interior change from lp profiles to free - fall profiles . since the core contains an embedded protostar , the system would be observationally classified as class 0/i . this stage lasts for a time @xmath208 less than 10% of the prestellar core lifetime . in a real system , the fraction of the envelope mass that eventually reaches the center would depend on the details of the protostellar wind , which would sweep up at least the polar portion of the envelope , creating an outflow . _ late accretion _ during this stage , material from the ambient environment directly accretes to the protostar ( or , more realistically , a disk ) . accretion during this stage is similar to bondi accretion , with free - falling density and velocity profiles @xmath28 and @xmath27 over the whole core region . for a real system , the duration of the late accretion stage , and hence the final stellar mass , depends on ambient conditions far from the protostar . in a real system , the potential of this stage to contribute significantly to the stellar mass would also depend on protostellar winds , which can reverse the accretion . although we have identified these stages based on idealized spherically symmetric models with constant gaseous inflow rates at large distance , we expect that the same stages would be present , in modified form , under more realistic conditions . based on our simulations , the ratios @xmath209 and @xmath210 decrease with increasing @xmath79 , as shown in figure [ fig : dtau ] . the supercritical stage and the infall stage have similar durations , and range from @xmath211 of the prestellar core lifetime @xmath212 , which itself ranges from @xmath213 . the supercritical collapse period ( stage 2 ) @xmath214 and the infall period ( stage 3 ) @xmath208 thus both range over @xmath215 , as shown in figure [ fig : dtau ] . while the time to reach collapse would differ for nonspherical or nonsteady converging large - scale flows , we expect that the character of the evolution would not . we also expect that the ratios @xmath216 and @xmath217 would remain order - unity . figure [ fig : mfl_mass ] shows the temporal evolution of the mass accretion rate and the total integrated mass of the central protostar for @xmath218 and @xmath219 . these can be compared to the mass accretion rate and the integrated central mass for an initially static critical be sphere , as shown in figure [ fig : mfl_pre ] ( see also @xcite , who show similar accretion histories to figure [ fig : mfl_pre ] ) . for both the initially - static collapse and our models that allow for core building from supersonic flows , there is a sharp early peak in the accretion rate . the rise starting from @xmath220 corresponds to the moment of protostar formation at @xmath93 . the smooth decline that follows ( ending at the points marked `` i '' in figure [ fig : mfl_mass ] ) corresponds to the infall stage , as the gravitational rarefaction wave propagates outward . at late times , however , the accretion differs for the initially - static vs. dynamically - built cores . for an initially - static unstable be core , the late accretion steadily declines over time ( fig . [ fig : mfl_pre ] ) . in contrast , for cores formed in the @xmath152 model ( i.e. barely supersonic inflow ) , the late time accretion is nearly constant , and for cores formed in large @xmath79 models , the late accretion rate increases over time ( thin lines in fig . [ fig : mfl_mass ] ) . the early - time peak accretion rates can be compared to the predictions of self - similar models . for the lp profiles at the instant of core collapse , @xmath70 and @xmath221 give @xmath222 , while the self - similar solution for the accretion phase in @xcite predicts @xmath223 . in fact , we do see a jump in @xmath224 above @xmath225 as the evolution transits from the collapse stage to the infall stage . this phenomenon is most clearly evident for the @xmath152 model , which has the highest resolution of the central region because the shock strength is lower than in the high @xmath79 models , yielding a larger core ( see fig . [ fig : rho_v_collapse ] ) at the instant of collapse . the detailed behavior of @xmath224 during the late accretion stage can be understood in terms of various transitions that occur . for @xmath194 , the accretion rate ( see in fig . [ fig : mfl_mass ] ) starts to increase from point @xmath226 until to point @xmath227 , and then decreases . the increases from @xmath226 to @xmath227 occurs as gas stored between the shock front and the outer boundary collapses into the center . the point @xmath227 represents the instant when the density in the whole outer region reaches a profile @xmath28 . after point @xmath227 , the gravitational rarefaction has reached the boundary , and subsequent accretion is limited by the inflow rate imposed at the outer boundary . for the @xmath228 and @xmath219 models shown in figure [ fig : mfl_mass ] , the mass accretion process between point @xmath226 and @xmath227 is similar to that of @xmath152 . however , there is additional transitory behavior before the accretion rate decreases to the inflow rate imposed at the boundary . at the instant corresponding to point @xmath227 , the rarefaction has produced @xmath28 over the whole region . but the velocity profile has @xmath27 only over the inner region . during the stage between point @xmath227 and @xmath229 , the density profile stays almost unchanged but the velocity profile evolves to reach @xmath27 everywhere ( see also fig . [ fig : rho_v_acc ] ) . after point @xmath229 , the accretion rate decreases to the imposed inflow rate . the stage between @xmath227 and @xmath229 is most obvious for the @xmath230 models . as mentioned above , we have also performed models in which inflow to the grid is halted after the point when the rarefaction reaches the shock . the resulting late - stage accretion ( see fig . [ fig : mfl_mass ] ) is initially the same as in our standard models , but then declines over time , after the rarefaction wave reaches the boundary . the mass of the protostar at the end of the infall stage , and the total core mass @xmath231 inside the shock at instant of core collapse , are shown in figure [ fig : core_mass ] . for comparison , we also show the critical be sphere mass ( see eq . [ massbe ] ) using the post - shock density at the time of collapse for @xmath232 ; these are much lower than the actual core mass . because there is continued mass passing through the shock during the infall stage , the post - infall protostellar mass is slightly higher than the core mass inside the shock at the time when core collapses . as the mach number increases , the post - infall protostellar mass and the core mass at @xmath93 both decrease . the protostellar mass ranges over @xmath233 and the core mass at @xmath93 ranges over @xmath234 , taking @xmath108 for the ambient density . star formation takes place in gmcs pervaded by supersonic turbulence , and theoretical models of prestellar ( and protostellar ) cores must take these large - scale supersonic flows into account . here , we have developed models in an idealized , spherically - symmetric framework that nevertheless captures key aspects of the real situation , enabling us to identify and analyze the main stages of core formation and evolution in a dynamic environment . our models differ from previous studies of core evolution in that the cores are not present as either stable or unstable density concentrations in our initial conditions the initial density is everywhere uniform . instead , during the first evolutionary stage cores are built `` from scratch '' by the collision of converging supersonic flows . the boundaries defining the outer edge of a core where the density drops correspond to a shock front across which the temperature is constant and the mass flux is nonzero . the shock front propagates outward , with the mass of the post - shock dense region growing in time . initially , the core is essentially uniform . over time , the mass grows sufficiently so that the core becomes centrally stratified due to self - gravity . observationally , the latter part of this `` core building '' stage corresponds to prestellar cores that have low to intermediate peak brightness . the period over which @xmath235 , and a core would be clearly identifiable in observations , amounts to 1 6 @xmath147 , with the free - fall time defined using the mean core density . when the center - to - edge density contrast exceeds @xmath103 , the core becomes supercritical and a stage of violent `` outside - in '' collapse ensues . the density profile throughout the core approaches @xmath236 , and a protostar forms at the center . we define the instant that collapse is complete and a protostar forms as @xmath212 , or @xmath93 in our dimensionless variables . although the central density becomes very large , the wave of outside - in collapse still leaves most of the core mass in the outer parts . observationally , this core collapse stage corresponds to prestellar cores that have high peak brightness . the period @xmath237 over which cores are supercritical , undergoing outside - in collapse , amounts to less than 10% of @xmath212 , or 0.8 1 @xmath147 . the third stage of evolution is governed by an `` inside - out '' wave of gravitationally - driven rarefaction propagating from the center of the core to the shock front that defines the core s outer edge . the accretion rate onto the star during this infall stage is initially very high , but declines over time . at the end of this infall stage , the dense envelope built during the first stage has plunged into the star . the velocity and density profiles approach free - fall , @xmath4 and @xmath28 , respectively . observationally , this stage corresponds to class 0/i embedded protostars . the period @xmath208 over which cores undergo this inside - out infall is similar to the duration of the previous stage , @xmath214 , and comparable to @xmath147 . during the final stage of evolution , there is no longer a massive envelope . the protostar can continue to accrete from the more distant , lower - density gas in its surroundings . the late - stage accretion rate and total mass accumulated by the system depend on the large - scale environment , rather than the properties established in the core during the building stage . observationally , this stage corresponds to a non - embedded yso that may still be accreting from a disk . the initiation of star formation via outside - in core collapse , followed by inside - out envelope infall , appears to be very robust . the dynamical behavior during these stages of evolution is very similar whether the core is initiated as an unstable equilibrium ( as in previous models ) or is built up dynamically through a shocked converging flow ( as in the present work ) . the larson - penston singular solution with @xmath238 appears to be an `` attractor , '' in that models initiated from stationary equilibria or with different supersonic converging velocities all arrive at this configuration at the moment of protostar formation . prior to the point at which cores become supercritical and outside - in collapse begins , the velocities interior to cores are subsonic , even if they are created by highly supersonic flows . in fact , _ higher _ inflow velocities produce _ lower _ post - shock velocities within the dense core ( see figs . 2 - 5 and eq . this result is consistent with observations showing that dense cores are quiescent in their interiors ( see e.g. , ) . \3 . throughout both the core - building and core - collapse stage , density profiles for cores formed by shocked converging flows can be fit by bonnor - ebert profiles , but with fitted temperatures @xmath100 larger than the true temperature @xmath239 . the range of temperatures fitted for our models with mach numbers up to 7 is @xmath240 . this result is consistent with observational findings ( @xcite ) that effective temperatures greater than directly - measured values ( from fitting seds ) are usually required in order to fit be spheres to observed prestellar cores . the largest possible ratio that could be obtained for an isothermal spherical flow is @xmath241 , so that any observed ratio larger than this suggests that magnetic fields contribute appreciable support , or else the core is anisotropic . @xcite have also recently pointed out that the temperature fit based on matching a be profile may be significantly higher than the true kinetic temperature , for clouds in collapsing stages . \4 . at the time of collapse , for all mach numbers the core size and mean density are closely related . we find that @xmath242 within 15% for mach numbers this radius is @xmath143 larger than the critical radius of a be sphere with the same mean internal density . \5 . as @xmath79 increases , and assuming a given ambient medium density @xmath165 , the time to reach collapse @xmath212 is shorter , the physical size of the core at @xmath212 is smaller , the mean internal density at @xmath212 is higher , and the mass of the core at @xmath212 is lower . for high mach number , the collapse time and collapse radius are related by @xmath244 . the range of core masses at the time of collapse at different @xmath79 is consistent with observed core masses , although the specific dependence of core mass on @xmath79 found in the present work may be sensitive to the spherical converging - flow geometry we have adopted . the durations of the infall stage and collapse ( supercritical ) stage of evolution are comparable for all @xmath79 , and are close to @xmath147 . this is consistent with observations indicating similar lifetimes for prestellar cores and embedded class 0/i accreting protostars . these stages are preceded by an extended core - building stage , during most of which the core would not be observable because its center - to - edge density contrast is low . the mass accretion rate onto the protostar ( or , more realistically , star - disk system if angular momentum were included ) peaks at the beginning of the infall stage at a value @xmath6 , and then declines steeply afterwards as the material stored in the envelope is exhausted . this result appears to hold regardless of how cores form , as it is consistent with earlier work ( see e.g. , @xcite ) for cores initiated from unstable equilibria or which undergo externally - induced compression . later accretion from the ambient medium depends on how long the large - scale cloud maintains a focused converging flow . as noted above , some of our specific conclusions are likely to change for non - spherical geometry , and for time - dependent rather than steady large - scale inflow . furthermore , other elements that are present in real star formation have been entirely omitted in these models ; these elements include rotation , which would lead to disk formation ; protostellar winds , which would sweep up and remove a portion of the envelope during the infall stage , and could prevent late accretion altogether ; and magnetic fields , which would alter the timescales and details of the evolutionary stages . we expect , however , that many of our basic results will carry over even if the idealizations we have adopted are relaxed . while large - scale supersonic converging flows in real gmcs are not generally spherical , the association of observed cores with high density surroundings suggests that the dense gas in post - shock stagnation regions is still the raw material out of which cores are built . we expect that in general core masses and collapse timescales will decrease with increasing density of the post - shock flow , which itself increases with increasing mach number . preliminary simulations of planar converging flows that we have conducted indeed bear out this expectation , showing @xmath245 . for planar converging flows , many cores simultaneously grow and then collapse in the post - shock gas layer ; unless this sheet were viewed exactly edge - on , the density jump at the shock front would not be apparent , and cores would be seen as surrounded by moderate - density gas . we also expect outside - in collapse followed by inside - out infall to be a generic feature of core evolution . although the duration of this pressure - mediated collapse is @xmath5 , it is _ unlike _ free - fall collapse in a crucial way : the core does not remain nearly - uniform . we speculate that the development of stratification during inside - out collapse will suppress growth of perturbations and sub - fragmentation for the non - spherical case . even though the increase in density implies that the local jeans mass becomes smaller and smaller , this is only true in the very center of the core . instead , we expect that collapse of cores built within shocked converging flows will produce single systems , which may be binary ( or multiple ) if the angular momentum is sufficient . non - spherical converging flows that create sheets and filaments of shocked gas and produce many such cores simultaneously could be the progenitors of stellar clusters . for units ) . _ solid curve , right scale : _ evolution of the central point mass ( see eq . [ m0_def ] for units ) . time is shown in units scaled by the density at the outer edge ( see eq . [ t0_def]).,scaledwidth=80.0% ] , at times @xmath246 as noted in the upper panel for each pair . the radius is normalized by the central density , i.e. the abscissa is @xmath247 . for the three upper panels on the right , the solid line is a fit to a be sphere with temperature @xmath100 , with the fitted temperature noted in each panel . dashed vertical lines denote the critical radius of a be sphere with the same central density and sound speed @xmath1 . dot - dashed vertical lines mark twice this critical radius . the time for the leftmost pair is half of the collapse time @xmath93 . the time for the second and third pairs are when the shock reaches the critical be radius and twice that value . the time for the last pair is the instant of collapse @xmath93 ( defined in the simulations as @xmath248 ) . the top - right panel shows with a dashed diagonal line the lp density profile @xmath249 . , scaledwidth=100.0% ] ( as labeled ) . the thick dashed line in the top panel is the lp density profile @xmath70 . for higher @xmath79 cases , the shock front is at smaller radius ( in units of @xmath255 ) and the post - shock speeds in the dense core are lower in magnitude . units of length and velocity are given by equations ( [ r0_def ] ) and ( [ isosp_def ] ) , respectively . density is in units of the gmc ambient value.,scaledwidth=80.0% ] . diamonds show quantities normalized using the mean core density and asterisks show quantities normalized using the mach number ( see text ) . the dot - dashed line in the lower panel shows the free - fall time at density @xmath64 in units @xmath256 , i.e. @xmath257.,scaledwidth=80.0% ] ( left ) and @xmath258 ( right ) models . density profiles ( top ) and the velocity profiles ( bottom ) are each shown at three different instants : solid lines show the instant of core collapse , dashed lines show the instant when the gravitational rarefaction wave arrives at the shock front , and dot - dashed lines show the profiles at a late accretion stage . numbers in the figure show the corresponding time for each instant . in the upper panels , the lp density profile @xmath70 is plotted for reference with thick dashed lines . the transitions from @xmath259 ( early ) to @xmath260 ( late ) and @xmath261 ( early ) to @xmath262 ( late ) are evident.,scaledwidth=100.0% ] . ( 3 ) envelope infall , propagating from inside to outside , and resulting in free - fall onto the protostar . ( 4 ) late accretion of ambient gas onto the protostar . for real systems , evolution would be modified in several ways : converging flows would be non - spherical , angular momentum would lead to disk formation , and outflows would contribute to clearing the envelope.,scaledwidth=100.0% ] ( asterisks ) and the post - collapse infall stage @xmath263 ( diamonds ) to the prestellar lifetime @xmath93 of cores , as a function of mach number . bottom panel shows ratios @xmath216 ( diamonds ) and @xmath217 ( asterisks ) , which range from [ 0.8 , 1 ] , as a function of mach number ; here @xmath147 is computed using the mean density inside the shock at @xmath93.,scaledwidth=80.0% ] and @xmath219 , as labeled . corresponding dashed lines show the results for models in which the inflow to the grid is suppressed after the end of the infall stage . the point marked @xmath226 represents the end of the infall stage , @xmath227 represents the instant when the density profile reaches @xmath28 everywhere , @xmath229 represents the instant when the velocity profile reaches @xmath4 everywhere . units of time , mass , and accretion rate are given by equations ( [ t0_def ] ) , ( [ m0_def ] ) and ( [ accr_def ] ) , respectively.,scaledwidth=80.0% ] of protostar formation ( asterisks ) , and the critical be sphere mass based on the post - shock density at time @xmath93 , all as a function of mach number @xmath79 . the mass unit is given by equation ( [ m0_def]).,scaledwidth=80.0% ]

suppose subgroups in the mapping class group of a closed orientable surface are given and let be the subgroup they generate . we discuss a question by minsky asking when for handlebody subgroups . [ section ] [ thm]lemma [ thm]corollary [ thm]proposition [ thm]question [ thm]remark [ thm]important remark [ thm]definition [ thm]example [ thm]fact [ thm]convention s l

1412.7847

let @xmath5 be a handlebody and @xmath6 the boundary surface . we have an inclusion of mapping class groups , @xmath7 . this subgroup is called a _ handlebody subgroup _ of @xmath8 . the kernel of the map @xmath9 is denoted by @xmath10 . if @xmath11 is a heegaard splitting of a closed orientable @xmath12-manifold , we have two handlebody subgroups @xmath13 with @xmath14 . minsky ( * ? ? ? * question 5.1 ) asked [ question ] when is @xmath15 equal to the amalgamation @xmath16 ? let @xmath17 be the curve graph of @xmath1 and @xmath18 the set of isotopy classes of simple curves in @xmath1 which bound disks in @xmath19 . the _ hempel / heegaard distance _ of the splitting is defined to be equal to @xmath20 . the group @xmath8 acts on @xmath17 by isometries . the stabilizer subgroups of @xmath21 in @xmath8 are @xmath22 . if the hempel distance is sufficiently large , depending only on @xmath1 ( @xmath23 suffices if the genus of @xmath1 is at least two @xcite ) , then @xmath24 is finite @xcite . the following is the main result . it gives a ( partial ) negative answer to question [ question ] . 0.5 cm * theorem [ main ] . * _ for the closed surface @xmath1 of genus @xmath25 and for any @xmath26 there exists a heegaard splitting @xmath27 so that @xmath24 is trivial , @xmath28 is not equal to @xmath29 , and @xmath30 . _ we will prove theorem [ main ] by constructing an example . to explain the idea we first construct a similar example in a certain group action on a simplicial tree ( theorem [ example.tree ] ) , then imitate it for the action of @xmath8 on @xmath17 . by contrast , for the subgroups @xmath31 , ohshika - sakuma @xcite showed [ thm.os ] if @xmath32 is sufficiently large ( depending on @xmath1 ) , @xmath33 is trivial and @xmath34 . that @xmath35 is trivial follows from the fact that @xmath36 are torsion free ( attributed to ( * ? ? ? * proof of prop 1.7 ) in @xcite ) . here is an alternative proof , suggested by minsky , that @xmath37 is torsion free for a handlebody @xmath5 . let @xmath38 be a torsion element . since @xmath39 has finite order , we have a conformal structure on @xmath1 invariant by @xmath39 . moreover , since @xmath39 extends to @xmath5 , by the classical deformation theory of kleinian groups developed by ahlfors , bers , kra , marden , maskit , and sullivan , we have a unique hyperbolic structure on @xmath5 whose conformal structure at infinity is the prescribed one . since the conformal structure is @xmath39-invariant , so is the hyperbolic structure . moreover , since @xmath39 acts trivially on @xmath40 , each geodesic in @xmath5 is invariant by @xmath39 . this implies that @xmath39 is identity on @xmath5 , hence @xmath39 is trivial . 0.5 cm * acknowledgements . * we would like to thank yair minsky for useful comments . let @xmath1 be a closed orientable surface . the mapping class group @xmath8 of @xmath1 is the group of orientation preserving homeomorphisms modulo isotopy . the curve graph @xmath41 has a vertex for every isotopy class of essential simple closed curves in @xmath1 , and an edge corresponding to pairs of simple closed curves that intersect minimally . it is a fundamental theorem of masur and minsky @xcite that the curve graph is @xmath42-hyperbolic . moreover , they show that an element @xmath43 acts hyperbolically if and only if @xmath43 is pseudo - anosov , and that the translation length @xmath44 of @xmath43 is uniformly bounded below by a positive constant that depends only on @xmath1 . it follows that @xmath43 has an invariant quasi - geodesic , called an _ axis _ denoted by @xmath45 , whose quasi - geodesic constants depend only on @xmath1 . a subset @xmath46 in a geodesic space is @xmath47-_quasi - convex _ if any geodesic in @xmath48 joining two points of @xmath49 is contained in the @xmath47-neighborhood of @xmath49 . if @xmath1 is the boundary of a handlebody @xmath5 , then the set @xmath50 of curves that bound disks in @xmath5 is quasi - convex ( i.e. @xmath47-quasi - convex for some @xmath47 ) , @xcite . the stabilizer of the set @xmath51 in @xmath8 is @xmath52 , the mapping class group of the handlebody @xmath5 , i.e. , the group of isotopy classes of diffeomorphisms of @xmath5 . given a @xmath47-quasi - convex subset @xmath48 in @xmath17 , we define the nearest point projection @xmath53 . the nearest point projection is not exactly a map , but a _ coarse map _ , since for a given point maybe there is more than one nearest point , but the set of such points is bounded in diameter , and the bound depends only on @xmath42 and @xmath47 , but not on @xmath48 . in this paper we often take @xmath45 of a pseudo - anosov element @xmath43 as @xmath48 . we may take any @xmath43-orbit instead of @xmath45 . two pseudo - anosov elements @xmath54 are independent ( i.e. , @xmath55 is not virtually cyclic ) if and only if the nearest point projection of @xmath56 to @xmath45 has a bounded image ( cf . @xcite , @xcite ) . in this section we discuss the acylindricity of a group action . this is a key property to prove theorem [ thm.os ] . acylindricity was introduced by sela for group actions on trees and extended by bowditch @xcite . suppose @xmath57 acts on a metric space @xmath48 . the action is _ acylindrical _ if for given @xmath58 there exist @xmath59 and @xmath60 such that for any points @xmath61 with @xmath62 , there are at most @xmath63 elements @xmath64 with @xmath65 . ( here @xmath66 denotes the distance @xmath67 . ) bowditch @xcite showed that the action of @xmath8 on @xmath17 is acylindrical . the following criterion will be useful . [ acylindrical ] suppose @xmath57 acts on a simplicial tree @xmath48 . if the cardinality of the edge stabilizers is uniformly bounded then the action is acylindrical . assume that every edge stabilizer contains at most @xmath68 elements . suppose an integer @xmath58 is given . take @xmath69 . we will show that if @xmath62 then there are at most @xmath70 elements @xmath71 with @xmath72 . indeed , let @xmath73 $ ] be the geodesic from @xmath74 to @xmath75 , and @xmath73'$ ] and @xmath73''$ ] be its subsegments after removing the @xmath76-neighborhood of @xmath77 , and the @xmath78-neighborhood of @xmath77 , respectively . then by the assumption , @xmath79 '' ) \subset [ v , w]'$ ] . moreover , for an edge @xmath80''$ ] near the midpoint , @xmath81 is contained in @xmath73''$ ] and the distance between @xmath82 and @xmath81 is at most @xmath76 . now fix such @xmath82 . then there are elements @xmath83 with @xmath84 , where @xmath85 , such that for any concerned element @xmath71 , there exists @xmath86 with @xmath87 . but since the stabilizer of @xmath82 contains at most @xmath68 elements , there are at most @xmath88 distinct choices for @xmath71 . to explain the background we quote a main technical result from @xcite ( we will not use this result ) . [ free ] let a group @xmath57 act acylindrically on a @xmath42-hyperbolic space @xmath48 . then for a given @xmath89 , there exists @xmath90 with the following property . let @xmath91 be @xmath47-quasi - convex subsets , and @xmath92 torsion - free subgroups . if @xmath93 then + ( 1 ) @xmath94 is trivial . + ( 2 ) @xmath95 . applying the theorem to the action of @xmath8 on @xmath96 with @xmath97 , and @xmath98 , @xmath99 , we obtain theorem [ thm.os ] : if @xmath32 is sufficiently large , depending only on @xmath1 , then @xmath33 is trivial and @xmath100 . to explain the difference between the torsion - free setting of @xcite and ours , we review the proof of theorem [ free ] . we start with an elementary lemma . [ unique ] let @xmath48 be a @xmath42-hyperbolic space and @xmath101 be @xmath47-quasi - convex subsets . let @xmath102 be a shortest geodesic between @xmath49 and @xmath103 . then , 1 . for any @xmath104 with both @xmath105 , and for any shortest geodesic @xmath106 between @xmath49 and @xmath103 , we have @xmath107 . 2 . suppose @xmath39 is an isometry of @xmath48 with @xmath108 . then for any @xmath104 with both @xmath105 , we have @xmath109 . hence for any @xmath104 , we have @xmath110 . 3 . suppose @xmath39 is an isometry of @xmath48 with @xmath111 . for @xmath112 let @xmath113 be a shortest geodesic from @xmath114 to @xmath49 . for an integer @xmath26 assume @xmath115 and @xmath109 . then for any point @xmath116 with @xmath117 , we have @xmath118 for @xmath119 . \(1 ) draw a geodesic quadrilateral with @xmath120 a pair of opposite sides . by @xmath42-hyperbolicity , @xmath114 must be in the @xmath121-neighborhood of one of the three sides not equal to @xmath102 , which must be @xmath106 , for otherwise , @xmath122 or @xmath123 , impossible . \(2 ) put @xmath124 . then for a point @xmath125 satisfying the assumption , by ( 1 ) there is a point @xmath126 with @xmath127 . but @xmath128 since @xmath129 and @xmath130 . by triangle inequality @xmath109 . it then implies @xmath131 for @xmath132 in general . \(3 ) by triangle inequality , for each @xmath133 , we have @xmath134 . let @xmath135 , and draw a geodesic quadrilateral with the corners @xmath136 . then by @xmath42-hyperbolicity , a concerned point @xmath116 is in the @xmath121-neighborhood of the side @xmath137 , hence , as before @xmath118 . \(1 ) set @xmath138 . we fix a constant @xmath139 . let @xmath102 be a shortest geodesic between @xmath49 and @xmath103 . let @xmath140 denote the length of @xmath102 . since @xmath141 , we have points @xmath142 such that @xmath143 and all four of @xmath144 , @xmath145 , @xmath146 , @xmath147 are @xmath148 . if @xmath149 , then both @xmath150 , @xmath151 by lemma [ unique ] ( 2 ) , hence by the acylindricity there are at most @xmath152 such elements , so the order of @xmath153 is @xmath154 . in particular each element in @xmath153 is torsion . since @xmath57 is torsion free , @xmath153 is trivial . \(2 ) + _ claim 1_. let @xmath155 . if @xmath156 then @xmath157 . + to argue by contradiction assume @xmath109 . let @xmath113 be a shortest geodesic from @xmath114 to @xmath49 . apply lemma [ unique ] ( 3 ) with @xmath158 . then for each point @xmath116 with @xmath159 and each @xmath133 , we have @xmath118 . now the subsegment of @xmath113 , except for the end points , that those @xmath160 can belong to has length at least @xmath161 . taking two points near each end of the subsegment , they are moved at most @xmath162 by @xmath163 . but by acylindricity there are at most @xmath152 such elements . since @xmath158 , the order of @xmath39 must be at most @xmath152 , hence trivial , contradiction . it follows from claim 1 that if @xmath104 with @xmath156 , then @xmath114 is not contained in @xmath164 . similarly , + _ claim 2_. let @xmath165 . if @xmath166 then @xmath157 . notice that claim 1 and claim 2 hold if every non - trivial element in @xmath167 has order at least @xmath168 or @xmath169 . now , we apply @xmath170 to @xmath171 , and obtain @xmath172 . put @xmath173 . the path @xmath174\cup \gamma$ ] is roughly a shortest geodesic from @xmath175 to @xmath103 . this is because @xmath176 , claim 1 , and that the geodesic @xmath177 $ ] is contained in the @xmath47-neighborhood of @xmath49 . so , @xmath178 is at least , say , @xmath179 . similarly , now using claim 2 , for any @xmath180 , @xmath181 is at least @xmath182 . to finish , given a reduced word in @xmath183 , @xmath184 , we let the elements @xmath185 successively act on @xmath49 ( or @xmath103 if @xmath186 is empty ) . see figure [ fig0 ] . then as before the distance between @xmath49 and @xmath187 is at least , say , @xmath188 , where @xmath189 is the length as a reduced word . ( here we are using a standard fact in @xmath42-hyperbolic geometry that a piecewise geodesic with each geodesic part long and the `` backtrack '' at each connecting point short is not only a quasi - geodesic , but also a geodesic with the same endpoints follows the path except for the backtrack parts . ) in particular @xmath190 , so @xmath75 is not trivial in @xmath57 . it implies @xmath95 . there is a more general version of theorem [ free ] in which one does not assume that @xmath191 and @xmath192 are torsion free . to state it we introduce the following definition . for an isometry @xmath193 define the _ coarse fixed set _ as @xmath194 . also , we will need a version for more than two subsets in @xmath48 to discuss another application . for that we introduce one more definition . let @xmath195 be mutually disjoint subsets in a @xmath42-hyperbolic space @xmath48 . for a given constant @xmath196 , we say that @xmath197 is _ k - terminal _ if for any other @xmath198 and any shortest geodesic @xmath102 between @xmath198 , the distance between @xmath197 and @xmath102 is at least @xmath68 . [ remark.torsion ] let a group @xmath57 act acylindrically on a @xmath42-hyperbolic space @xmath48 . then for a given @xmath89 , there exists @xmath196 with the following property . let @xmath199 be @xmath47-quasi - convex subsets , and @xmath200 be subgroups for all @xmath201 . assume that there exists a subset @xmath202 containing @xmath197 for each @xmath201 so that : a. for every finite order element @xmath203 we have @xmath204 for each @xmath201 ; b. @xmath205 for all pairs @xmath206 ; and c. each @xmath202 is @xmath68-terminal . then + ( 1 ) @xmath207 is trivial for all @xmath206 . + ( 2 ) @xmath208 . the proof is a slight variation of the proof of theorem [ free ] and is omitted . if @xmath209 does not contain any non - trivial elements of finite order , then we just put @xmath210 . our counterexamples will have the property that @xmath211 ( namely , the two coarse fixed sets intersect although @xmath212 and @xmath213 are far away . claim 1 in the proof of theorem [ free ] ( 2 ) when @xmath39 has infinite order , where @xmath214 is contained in the @xmath215-neighborhood of @xmath49 ) . if the sets @xmath216 satisfy properties ( b ) and ( c ) above for a constant @xmath68 , we say that they are @xmath68-_separated . _ we will show that theorem [ free ] does not hold if we do not assume that @xmath191 and @xmath192 are torsion - free . we construct a counterexample in the action of @xmath8 on @xmath17 ( theorem [ main ] ) . to explain the idea we start with a counterexample when @xmath48 is a simplicial tree . the key geometric feature is that , if we keep the previous notations , @xmath102 and @xmath217 may stay close along an arbitrarily long segment if @xmath218 has finite order ( each point on that segment does not move very much by @xmath218 ) . [ example.tree ] there exists an acylindrical group action on a simplicial tree @xmath48 by a group @xmath57 such that for any number @xmath26 there exist vertices @xmath219 with @xmath220 such that @xmath221 is trivial and @xmath222 is not equal to the free product @xmath223 . we first construct an example with @xmath224 . start with abelian groups @xmath225 with non - trivial torsion elements @xmath226 and @xmath227 , for example , @xmath228 . define the group @xmath229 and let @xmath230 be the bass - serre tree of this graph of groups decomposition . there are two vertices @xmath77 in @xmath230 at distance two whose stabilizers are @xmath49 and @xmath103 . the intersection @xmath231 is trivial in @xmath57 since @xmath232 is trivial in @xmath233 . on the other hand , @xmath234 is not equal to @xmath235 since @xmath57 is the quotient of the free product @xmath235 by the relation @xmath236 . the geometric reason for why @xmath237 is not equal to @xmath238 is that @xmath239 and @xmath240 intersect non - trivially in @xmath230 . the action on @xmath230 is acylindrical by lemma [ acylindrical ] since the edge stabilizer is a conjugate of @xmath241 or @xmath242 . to produce an acylindrical action that works for all @xmath243 we modify the previous example . take the direct product of @xmath244 and the subgroup @xmath245 in @xmath57 . one can write the new group as @xmath246 further , add a new element @xmath247 to @xmath49 with a relation @xmath248 to get @xmath249 and set @xmath250 and @xmath251 this is a two edge decomposition . in the bass - serre tree of this decomposition , consider the `` fundamental domain '' , i.e. the subtree spanned by two vertices @xmath77 at distance two with stabilizers @xmath252 and @xmath103 respectively . let @xmath114 be the vertex between them with stabilizer @xmath253 . see figure [ fig1 ] . now consider the ray based at @xmath114 that contains the vertices @xmath114 , @xmath254 , @xmath255 , @xmath256 , @xmath257 . the stabilizer of every edge on this ray is @xmath258 since both @xmath259 and @xmath247 commute with @xmath218 . so , the intersection of @xmath103 , the vertex group of @xmath75 , and any of the vertex groups along the ray except for @xmath253 is @xmath260 . but for each @xmath261 the subgroup @xmath262 is not equal to @xmath263 since @xmath264 and @xmath265 generate @xmath266 and not @xmath267 . the action of @xmath268 is acylindrical by lemma [ acylindrical ] since any edge stabilizer is a conjugate of @xmath269 or @xmath270 . we will prove the main theorem . [ main ] for the closed surface @xmath1 of genus @xmath271 with @xmath272 and for any @xmath26 there exists a heegaard splitting @xmath27 so that @xmath24 is trivial , @xmath28 is not equal to @xmath29 , and @xmath30 . we will need two properties of pseudo - anosov elements to prove the theorem ( lemma [ elementary ] , lemma [ hempel ] ) . let @xmath1 be a closed surface and @xmath43 a pseudo - anosov mapping class on @xmath1 . the _ elementary closure _ of @xmath43 is the subgroup @xmath273 of @xmath8 that consists of mapping classes preserving the stable and unstable foliations of @xmath43 . equivalently , @xmath273 is the centralizer of @xmath43 in @xmath8 . the group @xmath273 contains a unique finite normal subgroup @xmath274 such that @xmath275 is infinite cyclic . note that @xmath276 and @xmath277 for every @xmath278 . if @xmath279 is a regular cover with deck group @xmath280 and if @xmath281 is a pseudo - anosov mapping class with @xmath282 , then we certainly have @xmath283 for any lift @xmath284 of any power of @xmath43 , but strict inclusion may hold . it is an interesting question whether one can construct @xmath43 so that equality holds for all regular covers . we call such @xmath43 _ prime _ and we discuss a construction of prime pseudo - anosov mapping classes in section [ prime.section ] . for our purposes we need quite a bit less . [ elementary ] suppose @xmath285 is a pseudo - anosov mapping class whose stable and unstable foliations have two singular points , one of order @xmath286 , the other of order @xmath287 , with both @xmath288 odd and relatively prime . let @xmath279 be a double cover with deck group @xmath258 and @xmath289 a lift of a power of @xmath43 to @xmath290 . then @xmath282 and @xmath291 . we first argue that @xmath282 . . then @xmath71 can be represented by a homeomorphism , also denoted @xmath293 , that preserves both measured foliations . in particular , @xmath71 is an isometry in the associated flat metric on @xmath1 with cone type singularities . the homeomorphism @xmath71 fixes both singular points and satisfies both @xmath294 and @xmath295 , since an isometry that fixes a nonempty open set is necessarily the identity . since @xmath288 are relatively prime it follows that @xmath296 . we now argue that @xmath291 . we have @xmath297 . let @xmath298 , then @xmath299 is a finite order homeomorphism that preserves the lift of stable and unstable foliations of @xmath43 . composing with @xmath218 if necessary we may assume that @xmath71 fixes both @xmath286-prong singularities . arguing as above , we see that @xmath294 . since @xmath300 fixes both @xmath287-prong singularities , similarly we have @xmath301 and since @xmath302 we have @xmath296 . we showed @xmath291 . [ ms.elementary ] pseudo - anosov mapping classes as in lemma [ elementary ] exist in every genus @xmath303 . write @xmath304 where @xmath288 are relatively prime odd numbers . for example , we can take @xmath305 and @xmath306 . by the work of masur - smillie @xcite @xmath43 as above exists . suppose @xmath5 is a handlebody and @xmath1 its boundary . let @xmath307 be the set of curves that bound disks in @xmath5 . denote by @xmath308 the closure of @xmath51 , viewed as a subset of @xmath309 . then @xmath310 is nowhere dense in @xmath309 @xcite , and its complement @xmath311 is called the _ masur domain_. hempel @xcite found that if the stable lamination of a pseudo - anosov element @xmath43 is in @xmath311 then @xmath312 . we say a pseudo - anosov element @xmath281 is _ hempel _ for @xmath51 if the nearest point projection of @xmath51 to @xmath45 is a bounded set . @xmath43 is hempel if and only if the end points of @xmath45 are in @xmath311 , @xcite . on the other hand , both endpoints of @xmath45 are in @xmath310 if and only if @xmath45 is contained in a @xmath68-neighborhood of @xmath51 for some @xmath196 since both @xmath51 and @xmath45 are quasi - convex subsets in the @xmath42-hyperbolic space @xmath17 ( cf . since @xmath310 is nowhere dense in @xmath309 and the set of pairs of endpoints @xmath313 of pseudo - anosov mapping classes is dense in @xmath314 , there is a pseudo - anosov element @xmath43 whose stable and unstable laminations are not in @xmath310 , so that @xmath43 is hempel . masur found a condition in terms of the intersection number for a curve to be in @xmath51 ( * ? ? ? * lemma 1.1 ) and used it to prove @xmath310 is nowhere dense ( * ? ? ? * theorem 1.2 ) . the following lemma is proved using his ideas . [ hempel ] let @xmath315 be a double cover between handlebodies with the deck group @xmath258 , @xmath316 , and @xmath317 the set of curves that bound disks in @xmath318 , respectively . if the genus of @xmath1 is @xmath303 , then @xmath8 contains a pseudo - anosov element @xmath43 such that : a. @xmath43 is hempel for @xmath51 , b. @xmath43 lifts to @xmath284 and @xmath289 is hempel for @xmath319 , c. @xmath320 . let @xmath311 be the masur domain for @xmath5 and @xmath321 for @xmath322 . we first find a lamination @xmath323 on @xmath1 that is in @xmath311 such that its lift @xmath324 on @xmath290 is also in @xmath321 . choose a pants decomposition of @xmath1 using curves in @xmath51 and a lamination @xmath325 whose support intersects each pair of pants in this decomposition in 3 ( non - empty ) families of arcs connecting distinct boundary components ( so there are no arcs connecting a boundary component to itself ) . in the proof of ( * ? ? ? * theorem 1.2 ) masur shows that @xmath326 ( for example , take the curve @xmath327 in his proof as @xmath323 ) . this is done by verifying the conditions in lemma 1.1 for @xmath327 with respect to the pants decomposition in the last two paragraphs of the proof of theorem 1.2 . now the lift @xmath324 of @xmath323 to @xmath290 satisfies the same condition with respect to the lifted pants decomposition ( it lifts since our covering is between handlebodies and the boundary curves bound disks ) , so we have @xmath328 . choose a pseudo - anosov homeomorphism @xmath329 both of whose fixed points in @xmath309 are close to @xmath323 and in particular they are in @xmath311 since @xmath311 is open . the lift @xmath268 of @xmath57 ( or its power ) to @xmath290 similarly has endpoints close to @xmath324 and in particular in @xmath321 . it follows that both @xmath57 and @xmath268 are hempel . to finish the proof we need to arrange that @xmath57 has the extra property ( iii ) . let @xmath330 be an arbitrary pseudo - anosov mapping class that satisfies the assumption of lemma [ elementary ] . such @xmath331 exists by corollary [ ms.elementary ] . then @xmath332 also satisfies the assumptions , and hence also conclusion of lemma [ elementary ] for any @xmath261 and has an axis whose endpoints are close to @xmath323 if @xmath261 is sufficiently large . therefore @xmath43 is hempel , and similarly , the lift @xmath289 has an axis whose endpoints close to @xmath324 , therefore @xmath289 is hempel . we prove theorem [ main ] by constructing an example . let @xmath333 with generators @xmath334 , and let @xmath315 be a normal cover between handlebodies with the deck group @xmath331 . if @xmath335 is the genus of @xmath5 , then the genus of @xmath322 is @xmath336 . let @xmath337 be the set of curves in @xmath338 that bound disks in @xmath322 . we have two double covers @xmath339 . let @xmath340 be the set of curves in @xmath341 that bound disks in @xmath342 . put @xmath343 . the genus of @xmath344 is @xmath345 . let @xmath47 be a common quasi - convex constant for @xmath346 , and @xmath42 the hyperbolicity constant of @xmath347 . using lemma [ hempel ] , take a pseudo - anosov element @xmath348 on @xmath344 that is hempel for @xmath349 such that the lift @xmath350 of @xmath348 to @xmath290 is also hempel for @xmath319 , and that @xmath351 . note that @xmath352 are independent pseudo - anosov elements on @xmath290 since their elementary closures are different . in particular , the projection of @xmath353 to @xmath354 is bounded , and vice versa . note that @xmath355 since @xmath356 . set @xmath357 for @xmath26 . then , @xmath358 since @xmath350 centralizes @xmath359 . form the heegaard splitting @xmath360 such that @xmath361 . the surface @xmath290 is fixed but the splitting depends on @xmath63 . we will argue this is a desired splitting . set @xmath362 . in other words , @xmath363 in the heegaard splitting convention . since @xmath364 and @xmath365 , @xmath366 is not the free product of @xmath367 . to prove the theorem we are left to verify @xmath368 as @xmath369 and @xmath370 for any large @xmath26 . [ apart ] @xmath368 as @xmath369 . we claim that there is a constant @xmath49 such that for any @xmath26 , @xmath371 let @xmath372 denote the projection to @xmath353 , and @xmath373 the projection to @xmath354 . as we said they are coarse maps but we pretend they are maps for simplicity . also , we pretend that both @xmath374 are geodesics . let @xmath310 be a common bound of the diameter of the sets @xmath375 , @xmath376 , @xmath377 and @xmath378 . then @xmath310 is a bound of @xmath379 and @xmath380 for all @xmath26 . choose points @xmath381 , @xmath382 , @xmath383 and @xmath384 . now assume @xmath26 is so large that @xmath376 and @xmath379 are far apart , and also @xmath378 and @xmath380 are far apart ( compared to @xmath310 and @xmath42 ) . it suffices to show the above inequality under this assumption . let @xmath385 and @xmath386 be any points , and put @xmath387 . then , by a standard argument using @xmath42-hyperbolicity , the piecewise geodesic @xmath388\cup[x_1,r_1]\cup[r_1,r_2]\cup[r_2,x_2]\cup[x_2,y_2]$ ] is a quasi - geodesic with uniform quasi - geodesic constants that depends only on @xmath310 and @xmath42 . see figure [ fig2 ] . hence the hausdorff distance between the piecewise geodesic and the geodesic @xmath389 $ ] is bounded ( the bound depends only on @xmath310 and @xmath42 ) . it follows that there is a constant @xmath253 such that @xmath390 . on the other hand since @xmath391 and @xmath392 there is a constant @xmath103 such that for all @xmath26 we have @xmath393 . combining them we get a desired estimate with @xmath394 . we note that any geodesic joining a point in @xmath395 and a point in @xmath396 passes through a bounded neighborhood of each of @xmath397 provided that @xmath26 is large enough . the bound depends only on @xmath310 and @xmath42 . see figure [ fig2 ] . in the argument we did not use that @xmath319 ( as well as @xmath398 ) are quasi - convex . to argue @xmath399 , we will need the following lemma from ( * ? ? ? * proposition 6 ) . this is a consequence of the fact that @xmath43 is a `` wpd element '' . [ wpd ] let @xmath43 be a pseudo - anosov mapping class on a hyperbolic surface @xmath1 . there is a constant @xmath90 such that for any @xmath400 the diameter of the projection of @xmath401 to @xmath45 in @xmath17 is larger than @xmath310 , then @xmath402 . [ trivial ] @xmath399 for any large @xmath26 . by lemma [ apart ] choose @xmath63 large such that @xmath403 is very large compared to @xmath42 and @xmath310 . let @xmath102 be a shortest geodesic from @xmath395 to @xmath396 . then as we noted after the proof of lemma [ apart ] , @xmath102 passes through the bounded neighborhood of each of @xmath404 . now let @xmath405 . then we have @xmath406 for any @xmath104 by lemma [ unique ] . since all of @xmath407 are in bounded distance from @xmath102 we conclude each of those four points is moved by @xmath39 a bounded amount ( the bound depends only on @xmath408 ) . but since @xmath409 are contained in @xmath353 and are far apart from each other for any large @xmath26 , we find @xmath410 has a long ( @xmath411 ) projection to @xmath353 , hence @xmath412 by lemma [ wpd ] . by the same reason @xmath413 . we conclude @xmath414 . but since @xmath415 and @xmath416 are independent , @xmath39 must be a torsion element , so @xmath417 . by lemma [ elementary ] , we showed the lemma . we proved the theorem . in view of lemma [ elementary ] we introduce a property that looks interesting for its own sake . we say a pseudo - anosov mapping class @xmath43 is _ prime _ if its stable / unstable foliations are not lifts of any foliations of a ( possibly orbifold ) quotient of @xmath1 . if @xmath43 is prime then @xmath273 is cyclic and @xmath282 . indeed , if @xmath419 then the two foliations lift from @xmath420 , with @xmath274 realized as a group of isometries of @xmath1 using nielsen realization . moreover , [ elementary2 ] ( cf . lemma [ elementary ] ) suppose @xmath421 is a finite cover with the deck group @xmath280 . let @xmath43 be a prime pseudo - anosov element on @xmath1 and @xmath289 a lift of a power of @xmath43 to @xmath290 . then @xmath422 . it is clear that @xmath423 . if the inclusion is strict , then the stable and unstable foliations of @xmath43 can be obtained by pulling back from @xmath424 . so we have a contradiction . note that if we have a prime pseudo - anosov element on @xmath1 , we can use lemma [ elementary2 ] instead of lemma 5.2 in the proof of lemma [ hempel ] and theorem [ main ] . we will give a construction of prime pseudo - anosov elements when the genus of @xmath1 is 3 , so this will also prove the theorem for the genus 5 case . recall that if @xmath425 are 4 vectors in @xmath426 then the cross ratio is @xmath427=\frac{[a , c][b , d]}{[a , d][b , c]}\ ] ] where @xmath428=x_1y_2-x_2y_1 $ ] for @xmath429 . we do not define it when one of @xmath430 , [ b , d ] , [ a , d ] , [ b , c]$ ] is @xmath431 . the cross ratio is invariant under changing signs and scaling individual vectors and applying matrices in @xmath432 . it follows that for any flat structure on the torus the cross ratio for the vectors in the directions of four distinct closed geodesics is ( well - defined and ) rational . a singular euclidean structure ( or just a _ flat _ structure ) on a surface @xmath1 is _ good _ if the cone angle is a multiple of @xmath433 at each singularity . a geodesic segment connecting two singular points , or a closed geodesic is _ good _ if the angle along the geodesic at each singular point is a multiple of @xmath433 . the developing map @xmath434 defined on the universal cover of the complement of the cone points will take a good geodesic to a straight line , or a line segment . so , for any four good geodesics , the cross ratio for the four directions , if they are distinct , is well - defined . next , if @xmath279 is a branched cover between good flat structures , then the cross ratio of four good geodesics in @xmath290 is equal to the cross ratio of their images in @xmath1 , simply because @xmath435 have the `` same '' developing map . in particular , all cross ratios between good geodesics on a torus or a sphere with 4 cone points are rational , and to prove that a particular good flat surface is not commensurable with a torus it suffices to produce four good geodesics whose cross ratio is irrational . [ prime ] suppose @xmath281 is a pseudo - anosov homeomorphism such that : 1 . the stable foliation of @xmath43 has two singular points @xmath114 and @xmath160 , with @xmath286 and @xmath287 prongs respectively , and with @xmath286 and @xmath287 distinct odd primes , and 2 . a flat structure on @xmath1 determined by the stable and unstable foliations has four good closed geodesics with the cross ratio of their ( distinct ) direction vectors @xmath436 irrational . then @xmath43 is prime . we note that there is a 2-parameter family of flat structures determined by the two foliations ; they depend on the choice of the transverse measure in a projective class on each foliation . however , since scaling and linear transformations do not change the cross ratio , the assumption is independent of these choices . let @xmath437 be as in the statement . now suppose @xmath438 is a branched cover of degree @xmath439 and @xmath440 . the local degree of @xmath433 at @xmath114 is either 1 or @xmath286 . it can not be 1 , since at any other preimage of @xmath441 the singularity would have to have @xmath442 prongs , and there are nt any . thus at @xmath114 the map is modeled on @xmath443 , and similarly at @xmath160 it looks like @xmath444 . there are now two cases . _ @xmath445 . it follows that the other points that map to @xmath441 have 2 prongs and so the map there has local degree 2 . thus @xmath446 is odd and away from the images of singular points the foliation @xmath447 is regular ( since otherwise @xmath446 would have to be even ) . thus there are @xmath448 other preimages of @xmath441 , and deleting these and the same for the @xmath287-prong singularity we get that the euler characteristic of @xmath449 is @xmath450 so the euler characteristic of the quotient @xmath290 minus 2 singular points is @xmath451 , i.e. the quotient is the twice punctured @xmath452 , which does not support any pseudo - anosov homeomorphisms ( e.g. the curve complex is finite , see @xcite ) . on the other hand , let @xmath453 be a finite cover of the punctured @xmath1 so that the induced cover to @xmath454 minus two points is regular . some power of @xmath43 lifts to @xmath455 on @xmath453 , and since the cover is regular , a further power of @xmath456 descends to @xmath452 ( since each element , @xmath218 , of the deck group leaves the stable and unstable foliations invariant , so that @xmath457 for some @xmath26 , so @xmath218 and @xmath458 commute ) , contradiction . _ case 2 . _ @xmath459 . again the other points that map to @xmath459 are regular and the map has local degree 2 , so there are @xmath460 such points . here @xmath446 is even and we may have some number , say @xmath461 , of 1-prong singularities @xmath462 in the quotient , with each singularity having @xmath463 preimages where local degree is 2 . now we have that the euler characteristic of @xmath464 is @xmath465 so @xmath466 is odd and the euler characteristic of the quotient @xmath290 minus the singular points is @xmath467 . so , the euler characteristic of @xmath290 is @xmath468 . the only possibilities are @xmath469 and @xmath470 , and the quotient is twice punctured @xmath452 or 4 times punctured @xmath471 . the first possibility is ruled out as in case 1 . in the second case the good flat structure on @xmath1 descends to a good flat structure on @xmath471 with 4 singular points , then lifts to a flat structure on the branch double cover @xmath472 , with four closed geodesics such that the cross ratio of the four direction vectors is @xmath473 $ ] that is not rational , contradiction . indeed , using the same notation as in case 1 , a power of @xmath43 lifts to @xmath456 on @xmath453 that regularly covers @xmath471 minus 4 points . we lift the flat structure on @xmath1 and the stable and unstable foliations of @xmath43 to @xmath453 . then their regular leaves are straight lines . each deck transformation preserves the foliations , so that it is an isometry of @xmath453 , and that the good flat structure on @xmath453 , with cone angle at each singular point at least @xmath474 , descends to a good flat structure of @xmath471 minus 4 points ( and the angle at each puncture is @xmath433 ) . we obtain a good flat structure on @xmath471 with four good closed geodesics and the cross ratio is @xmath473 $ ] . also , the cross ratio will not change when we take a double cover that is a flat torus , contradiction . [ ms.foliation ] regarding the assumption ( 1 ) , if @xmath71 is the genus of @xmath1 , by an euler characteristic count we must have @xmath475 . conversely , the goldbach conjecture predicts that every even integer @xmath476 can be written as a sum of two primes . when the integer is @xmath477 and divisible by 4 , the two primes are necessarily distinct and odd . for example , @xmath478 satisfies the goldbach conjecture . the work of masur - smillie @xcite shows that if @xmath479 and @xmath304 then the surface @xmath1 admits a pseudo - anosov homeomorphism whose stable and unstable foliations have two singular points , one of order @xmath286 , the other of order @xmath287 . [ flat.example ] we now construct an explicit example in genus 3 satisfying the assumption of lemma [ prime ] . we take @xmath480 , @xmath481 . consider the flat square tiled surface @xmath1 pictured below . edges labeled by the same letter are to be identified . if the edges are on opposite sides of the parallelogram they are identified by a translation , and otherwise by a rotation by @xmath433 . the square tiling of @xmath426 induces one on the surface @xmath1 . there are two cone points , with cone angles @xmath482 and @xmath483 respectively . so , @xmath1 has a good flat structure . [ ht ] ( 0,6 ) ( 6,0 ) ; ( 1,6 ) ( 7,0 ) ; ( 0,6 ) ( 1,6 ) ; ( 6,0 ) ( 7,0 ) ; in 1, ... ,5 ( @xmath484 ) ( @xmath485 ) ; ( @xmath486 ) ( @xmath487 ) ; in 0, ... ,7 ( @xmath488 ) ( @xmath489 ) ; in 0, ... ,6 ( @xmath490 ) ( @xmath491 ) ; at ( 0,6 ) [ circle , draw = black , fill = black ] ; at ( 1,6 ) [ circle , draw = black , fill = black ] ; at ( 1,5 ) [ circle , draw = black , fill = black ] ; at ( 2,4 ) [ circle , draw = black , fill = black ] ; at ( 3,3 ) [ circle , draw = black , fill = black ] ; at ( 4,2 ) [ circle , draw = black , fill = black ] ; at ( 6,0 ) [ circle , draw = black , fill = black ] ; at ( 7,0 ) [ circle , draw = black , fill = black ] ; at ( 4,3 ) [ circle , draw = black , fill = black ] ; at ( 5,1 ) [ rectangle , draw = black , fill = black ] ; at ( 6,1 ) [ rectangle , draw = black , fill = black ] ; at ( 5,2 ) [ rectangle , draw = black , fill = black ] ; at ( 3,4 ) [ rectangle , draw = black , fill = black ] ; at ( 2,5 ) [ rectangle , draw = black , fill = black ] ; ( 0.5,5.5 ) ( 1.5,5.5 ) ; ( 2.5,3.5 ) ( 3.5,3.5 ) ; ( 1.5,4.5 ) ( 2.5,4.5 ) ; ( 3.5,2.5 ) ( 4.5,2.5 ) ; ( 4.5,1.5 ) ( 5.5,1.5 ) ; ( 5.5,0.5 ) ( 6.5,0.5 ) ; ( 0.5,5.5 ) ( 0.5,6 ) ; ( 2.5,3.5 ) ( 2.5,4.5 ) ; ( 5.5,0.5 ) ( 5.5,1.5 ) ; ( 6.5,0 ) ( 6.5,0.5 ) ; ( 1.5,4.5 ) ( 1.5,5.5 ) ; ( 3.5,2.5 ) ( 3.5,3.5 ) ; ( 4.5,1.5 ) ( 4.5,2.5 ) ; at ( 0.5,5.5 ) @xmath218 ; at ( 1.5,4.5 ) @xmath492 ; at ( 2.5,3.5 ) @xmath218 ; at ( 3.5,2.5 ) @xmath492 ; at ( 4.5,1.5 ) @xmath114 ; at ( 5.5,0.5 ) @xmath160 ; at ( 1.5,5.5 ) @xmath493 ; at ( 2.5,4.5 ) @xmath39 ; at ( 3.5,3.5 ) @xmath493 ; at ( 4.5,2.5 ) @xmath114 ; at ( 5.5,1.5 ) @xmath39 ; at ( 6.5,0.5 ) @xmath160 ; at ( 0.5,6 ) @xmath494 ; at ( 6.5,0 ) @xmath494 ; at ( 7.5,5.5 ) @xmath495 ; at ( 7.5,4.5 ) @xmath496 ; at ( 7.5,3.5 ) @xmath495 ; at ( 7.5,2.5 ) @xmath496 ; at ( 7.5,1.5 ) @xmath496 ; at ( 7.5,0.5 ) @xmath497 ; at ( 0.5,-1 ) @xmath498 ; at ( 1.5,-1 ) @xmath499 ; at ( 2.5,-1 ) @xmath498 ; at ( 3.5,-1 ) @xmath499 ; at ( 4.5,-1 ) @xmath500 ; at ( 5.5,-1 ) @xmath498 ; at ( 6.5,-1 ) @xmath498 ; we will use thurston s construction of pseudo - anosov homeomorphisms ( * ? ? ? * theorem 14.1 ) to construct @xmath43 . the lines bisecting the squares form three horizontal and three vertical geodesics . the matrix @xmath63 of intersection numbers , where the @xmath501 entry is the intersection number @xmath502 , is @xmath503 the largest eigenvalue of @xmath504 is @xmath505 satisfying the minimal polynomial @xmath506 . then one can choose the lengths and heights of the squares ( making them into rectangles , which gives a tiling of @xmath1 ) so that the twist in the horizontal multicurve is given by the matrix @xmath507 and the twist in the vertical multicurve by the matrix @xmath508 this means that the product of the first and the inverse of the second is @xmath509 whose trace is @xmath510 . so this product is pseudo - anosov and its dilatation is the larger , @xmath511 , of the eigenvalues of @xmath49 and the eigenvector is @xmath512 with @xmath513 and @xmath514 . the heights and widths of the rectangles are coordinates of the @xmath515-eigenvectors @xmath5 of @xmath504 and @xmath322 of @xmath516 . we compute @xmath517 and @xmath518 to show that @xmath43 is prime it suffices to find 4 closed geodesics whose slopes have irrational cross ratio . we take @xmath519 , @xmath520 , @xmath521 , @xmath522 , where @xmath523 and @xmath446 connect second , respectively third , vertex on the lower left side in the figure with the upper right vertex . they are good closed geodesics on @xmath1 based at the round vertex ( to compute the angle at the round vertex , it helps first to identify the two edges labeled by @xmath494 ) . the cross ratio is @xmath524[b , d]}{[a , d][b , c ] } & = \frac{(v_1+v_2)(v_1'+v_2')}{(2v_1+v_2)v_2 ' } = \frac{(\mu^2 - 5\mu+2)\mu}{(\mu^2 - 5\mu+3)(\mu^2 - 5\mu+2 ) } = \frac{\mu}{\mu^2 - 5\mu+3}\\ & = \frac{1}{3\mu^2 - 17\mu+10},\end{aligned}\ ] ] which is irrational ( for the last equality use @xmath506 ) . in this section we discuss another application of theorem [ free ] . for this we need the version stated in theorem [ remark.torsion ] . following dixon @xcite a group @xmath57 is _ invariably generated _ by a subset @xmath1 of @xmath57 if @xmath525 for any choice of @xmath526 . the group @xmath57 is _ ig _ if it is invariably generated by some subset @xmath1 in @xmath57 , or equivalently , if @xmath57 is invariably generated by @xmath57 . @xmath57 is _ fig _ if is is invariably generated by some finite subset of @xmath57 . kantor - lubotzky - shalev @xcite prove that a linear group is fig if and only if it is finitely generated and virtually solvable . gelander proves that every non - elementary hyperbolic group is not ig @xcite . we generalize this result to acylindrically hyperbolic groups . a group @xmath57 is _ acylindrically hyperbolic _ if it admits an acylindrical action on a hyperbolic space and @xmath57 is not virtually cyclic @xcite . examples are non - elementary hyperbolic groups , @xmath527 except for the genus @xmath528 and the number of punctures @xmath529 , and @xmath530 with @xmath531 ( cf . @xcite ) . for @xmath536 define a subset @xmath537 and put @xmath538 @xmath539 is a @xmath71-invariant non - empty set . if @xmath71 is hyperbolic , then @xmath539 is contained in a hausdorff neighborhood of an axis of @xmath540 . this is an easy exercise and we leave it to the reader . this is well known too . let @xmath542 be the set of centers of the orbit of a point @xmath114 by @xmath71 . @xmath542 is invariant by @xmath71 and its diameter is at most @xmath543 , therefore is contained in @xmath544 . we define a function on @xmath48 as follows : @xmath548 . + _ case 1_. @xmath549 . + fix @xmath550 . suppose @xmath551 is given . we will show that @xmath552 $ ] is contained in the @xmath553-neighborhood of @xmath546 . if @xmath554 , then by triangle inequality , @xmath555 for every @xmath556 $ ] , so that @xmath552 \subset x(g , l)$ ] . so assume @xmath557 . to compute @xmath558 for @xmath556 $ ] , draw a triangle @xmath280 for @xmath559 , and let @xmath560 $ ] be a branch point of this triangle , i.e. , the distance to each side of @xmath280 from @xmath523 is at most @xmath42 . since @xmath561 , @xmath562 for any point @xmath563 $ ] . for @xmath564 $ ] , @xmath558 is roughly equal to @xmath565 , with an additive error at most @xmath553 . also it is roughly maximal at @xmath114 on @xmath552 $ ] . ( imagine the case that @xmath48 is a tree and @xmath566 . ) it follows that @xmath552 $ ] is contained in the @xmath553-neighborhood of @xmath546 . now suppose another point @xmath567 is given . then @xmath428 $ ] is contained in the @xmath42-neighborhood of @xmath552\cup [ p , y]$ ] , so that @xmath428 $ ] is contained in the @xmath568-neighborhood of @xmath546 . + _ case 2_. @xmath534 . + then @xmath71 is hyperbolic . to simplify the argument , let s assume that there is a geodesic axis for @xmath71 . then for any @xmath569 , @xmath570 to see this let @xmath571 be a nearest point from @xmath114 . then the hausdorff distance between @xmath572 $ ] and @xmath573\cup [ x',g(x')]\cup [ g(x'),g(x)]$ ] is at most @xmath574 , and the estimate follows . it follows from the above estimate that if @xmath575 , then @xmath573 $ ] is contained in the @xmath576-neighborhood of @xmath546 . for @xmath577 , let @xmath578 be a nearest point from another point @xmath160 to @xmath579 . then @xmath573\cup [ x',y']\cup [ y',y]$ ] is contained in the @xmath576-neighborhood of @xmath546 . but since @xmath428 $ ] is contained in the @xmath580-neighborhood of @xmath573\cup [ x',y']\cup [ y',y]$ ] , @xmath428 $ ] is contained in the @xmath581-neighborhood of @xmath546 . the argument for the case that @xmath71 has only a quasi - geodesic ( with uniform quasi - geodesic constants depending only on @xmath42 ) as an axis is similar and we only need to modify the constants in the argument . we omit the details . let @xmath102 be a geodesic ray from a point in @xmath539 that tends to @xmath286 . then the ray is contained in the @xmath588-neighborhood of @xmath539 . so every point of the ray is moved by @xmath71 by a bounded amount , therefore @xmath286 is fixed by @xmath71 . let @xmath102 be a half of @xmath592 that tends to the point fixed by @xmath71 . then @xmath593 is bounded for @xmath104 . assume that @xmath102 tends to the direction that @xmath39 translates @xmath592 ( otherwise , we let @xmath594 below ) . then for any @xmath26 and @xmath104 , @xmath595 is bounded . now by acylindricity ( apply it to @xmath596 that are far from each other ) , there are only finitely many possibilities for @xmath597 , so @xmath71 commutes with a nontrivial power of @xmath39 . so @xmath71 moves each point in @xmath592 by a bounded amount , therefore @xmath591 . we will use the following result . it follows from the assumption that @xmath57 contains a `` hyperbolically embedded subgroup '' that is non - degenerate , i.e. , proper and infinite , see theorem 1.2 in @xcite . then @xmath57 contains a unique maximal finite normal subgroup @xmath68 and a schottky subgroup @xmath43 so that for every nontrivial @xmath604 any element @xmath605 is either contained in @xmath68 or has a nontrivial power that commutes with @xmath39 . if @xmath68 is trivial , @xmath590 is cyclic . although we will not use this fact , we observe that if @xmath71 has finite order @xmath63 , then the set @xmath609 is @xmath613-quasi - convex . this is because @xmath71 has an orbit whose diameter is at most @xmath614 ( see the proof of lemma [ center ] ) , and @xmath615 is a union of @xmath616- quasi - convex sets all of which contain the bounded orbit . thus @xmath609 is @xmath613-quasi - convex as desired . ( fix a point @xmath114 from the bounded orbit . then for any points @xmath617 , draw a @xmath42-thin triangle for @xmath618 . @xmath619 $ ] is in the @xmath42-neighborhood of @xmath428\cup[x , z]$ ] . ) * @xmath623 and @xmath624 are conjugate for each @xmath201 . * for any @xmath261 , the sets @xmath625 are @xmath68-separated , and moreover , this property holds if we replace @xmath626 with @xmath627 when @xmath628 have finite order . we describe a geometric property we use about the sequence of elements . fix a point @xmath569 . given a constant @xmath536 , if we choose @xmath631 sufficiently large , then for any @xmath261 , and any @xmath632 , the following points are @xmath310-separated : @xmath633 this is an easy consequence of the property such that the embedding of the cayley graph of @xmath43 in @xmath48 using the orbit of the point @xmath114 is quasi - isometric to the image . note that the subsets in the above are @xmath310-separated if we replace the point @xmath114 by a bounded set , possibly taking a larger constant for @xmath631 . to define @xmath634 , choose a sequence @xmath635 such that for each @xmath201 , @xmath636 . this is clearly possible . then by lemma [ projection.bdd ] , the projection of @xmath637 to @xmath638 is bounded . moreover , if the order of @xmath623 is @xmath606 , then @xmath639 , therefore the projection of @xmath640 to @xmath638 is bounded since each @xmath641 has a bounded projection . now take a sequence of sufficiently large constants @xmath642 , depending on the given constant @xmath68 , and put @xmath643 . then for each @xmath261 , the sets @xmath644 are @xmath68-separated since @xmath645 . also we can arrange so that the sets remain @xmath68-separated if we replace @xmath646 with @xmath647 if the order of @xmath624 are finite , maybe for larger constants @xmath648 . let @xmath196 be a constant from theorem [ remark.torsion ] for @xmath649 . let @xmath650 be a set of all conjugacy classes of @xmath57 except for the class for @xmath651 . apply lemma [ conjugation ] to the set @xmath652 and the constant @xmath68 and obtain a new set @xmath653 . for each @xmath654 , @xmath646 is a non - empty , @xmath624-invariant , @xmath547-quasi - convex subset . now , for each @xmath655 , the assumptions ( b ) and ( c ) of theorem [ remark.torsion ] are satisfied by the subgroups @xmath656 and the sets @xmath657 . if @xmath628 has finite order , then take @xmath658 as the desired neighborhood . then by lemma [ conjugation ] they are @xmath68-separated , which implies ( b ) and ( c ) . to finish , first suppose that @xmath664 is a finite set , @xmath665 . then @xmath663 must be a proper subgroup of @xmath57 ( therefore @xmath57 is not ig ) since otherwise @xmath57 contains infinitely many conjugacy classes , a contradiction ( by our assumption , @xmath57 is not virtually cyclic ) . define @xmath669 from @xmath664 by replacing @xmath670 by @xmath667 . @xmath669 contains all non - trivial conjugacy classes of @xmath57 . we claim that the subgroup , @xmath671 , generated by @xmath669 is a proper subgroup in @xmath57 ( so @xmath57 is not ig ) . to see that , define a quotient homomorphism from @xmath57 to the group @xmath672 by sending all @xmath673 to @xmath651 . then the image of @xmath671 is a proper subgroup in @xmath331 , so @xmath671 is a proper subgroup in @xmath57 . + + _ case 2_. assume that @xmath57 contains a non - trivial finite normal subgroup . let @xmath68 be the maximal finite normal subgroup in @xmath57 . then @xmath674 does not contain any non - trivial finite normal subgroup . moreover @xmath268 is acylindrically hyperbolic group . probably this fact is well known to specialists , and we postpone giving an argument till the end ( proposition [ quotient ] ) . by case 1 , @xmath268 contains a proper subgroup @xmath675 that contains all conjugacy classes of @xmath268 . let @xmath676 be the pull - back of @xmath675 by the quotient map @xmath677 . then @xmath331 is a proper subgroup that contains all conjugacy classes of @xmath57 , therefore @xmath57 is not ig . by assumption @xmath57 acts on a hyperbolic space @xmath48 such that the action is acylindrical and @xmath57 is not virtually cyclic . moreover we may assume that the action is co - compact . in fact , we may assume that @xmath48 is a cayley graph with a certain generating set , which is maybe infinite , ( * ? ? ? * theorem 1.2 ) . we will produce a new @xmath57-graph @xmath585 from @xmath48 such that the kernel of the action contains @xmath63 and that @xmath585 and @xmath48 are quasi - isometric . this is a desired action for @xmath268 . for each @xmath63-orbit of a vertex of @xmath48 , we assign a vertex of @xmath585 . note that @xmath57 is transitive on the set of @xmath63-orbits of vertexes of @xmath48 , so @xmath57 acts transitively on the vertex set of @xmath585 . now join two vertices of @xmath585 if the distance of the corresponding @xmath63-orbits in @xmath48 is @xmath651 . @xmath585 is a connected @xmath57-graph and it is easy to check that @xmath585 and @xmath48 are quasi - isometric ( by the obvious map sending a vertex @xmath114 of @xmath48 to the vertex of @xmath585 corresponding to the orbit of @xmath114 ) , so that @xmath585 is hyperbolic , and that the @xmath57-action on @xmath585 is acylindrical . by construction , @xmath63 acts trivially on @xmath585 .

lmc2 has the highest x - ray surface brightness of all know supergiant shells in the large magellanic cloud ( lmc ) . the x - ray emission peaks within the ionized filaments that define the shell boundary , but also extends beyond the southern border of lmc2 as an x - ray bright spur . _ rosat _ hri images reveal the x - ray emission from lmc2 and the spur to be truly diffuse , indicating a hot plasma origin . we have obtained _ rosat _ pspc and _ asca _ sis spectra to study the physical conditions of the hot ( k ) gas interior to lmc2 and the spur . raymond - smith thermal plasma model fits to the x - ray spectra , constrained by hi 21-cm emission - line measurements of the column density , show the plasma temperature of the hot gas interior of lmc2 to be kev and of the spur to be kev . we have compared the physical conditions of the hot gas interior to lmc2 with those of other supergiant shells , superbubbles , and supernova remnants ( snrs ) in the lmc . we find that our derived electron densities for the hot gas inside lmc2 is higher than the value determined for the supergiant shell lmc4 , comparable to the value determined for the superbubble n11 , and lower than the values determined for the superbubble n44 and a number of snrs .

astro-ph0008440

supergiant shells with sizes approaching 1000 pc form the largest structures seen in the interstellar medium ( ism ) . such shells have been detected in the hi 21-cm and h@xmath3 lines in our galaxy , local group galaxies , and more distant galaxies ( tenorio - tagle & bodenheimer 1988 ) . these supergiant shells contain 10@xmath410@xmath5 ergs of kinetic energy , which is most likely contributed by multiple supernovae and energetic stellar winds over several dozen myr . the size of a supergiant shell often exceeds the scale height of the gaseous disk of its host galaxy , thus forming a `` chimney '' through which energy and mass flow into the galactic halo . therefore , supergiant shells play important roles in the global evolution of the ism in a galaxy . it is difficult to study supergiant shells in detail . their distances , and so their sizes and energies , remain uncertain in our galaxy ; furthermore , interstellar obscuration often prevents observations at optical , uv and x - ray wavelengths . for distant galaxies , angular resolution becomes a severe limiting factor . the large magellanic cloud ( lmc ) offers a happy compromise , as it is close ( 50 kpc ; feast 1999 ) and has little obscuration ( a@xmath6 mag ; bessell 1991 ) nine supergiant shells with sizes ranging from 600 to 1400 pc have been reported in the lmc , lmc19 ( goudis & meaburn 1978 ; meaburn 1980 ) . only two of these shells have been studied in detail lmc2 and lmc4 because lmc2 is the most spectacular and lmc4 is the largest . both supergiant shells are filled with hot plasma : diffuse x - ray emission has been detected in lmc2 by wang & helfand ( 1991 ) using _ einstein _ imaging proportional counter ( ipc ) data , and in lmc4 by bomans , dennerl , & krster ( 1994 ) with a 46 ksec _ position sensitive proportional counter ( pspc ) observation . lmc2 is substantially brighter in x - rays than lmc4 and all other supergiant shells in the lmc . the physical and kinematic structure of lmc2 has been reported by points et al . ( 1999 ; hereafter paper i ) . it is found that two different mechanisms are responsible for generating the x - ray - emitting plasma : ( 1 ) local heating by supernova remnants ( snrs ) , e.g. , the emission region enclosed by the northeast rim ; and ( 2 ) outflows from active star formation regions , e.g. , the bright x - ray arc extending from n158 ( notation from henize 1956 ) to n159 ( or snr0540 - 693 to lmcx-1 ; see figure 1b ) . the south end of lmc2 is confusing . diffuse x - ray emission , while peaking within the shell , extends beyond the southern border of lmc2 . it has been speculated that this x - ray spur represents a blowout of lmc2 , but no optical or radio supporting evidence has been found . lmc2 clearly demonstrates the existence of hot gas interior to a supergiant shell , and illustrates different mechanisms of generating the hot gas . spatially resolved spectral analysis of the x - ray emission within lmc2 and neighboring regions is the most effective way to determine the physical conditions of the hot plasma and establish its role in the multi - phase interstellar medium . the x - ray emission from the interior of lmc2 has not been studied since its initial detection with the _ einstein _ ipc . those ipc data were not used to make a detailed spectral analysis , however , because they had low spectral resolution and were contaminated by hard x - rays scattered from lmc x-1 to large off - axis angles ( wang & helfand 1991 ) . to study the distribution of x - ray emission toward lmc2 , we have obtained _ rosat _ pspc and high resolution imager ( hri ) observations of lmc2 . furthermore , to determine the physical conditions of the x - ray emitting plasma interior to lmc2 , we have also obtained _ asca _ solid - state imaging spectrograph ( sis ) observations . this paper is organized as follows : section 2 describes the observations used in this study and their reduction . the distribution of the hot ( @xmath0 k ) plasma and its relationship to the warm ( @xmath7 k ) and cold ( @xmath8 k ) interstellar component of lmc2 is discussed in 3 . in 4 we describe the physical condition of the hot gas . in 5 we compare the physical properties of lmc2 with other objects that exhibit diffuse x - ray emission . we summarize our results in 6 . the x - ray data reported in this paper were obtained using the _ rosat_pspc and hri and the _ asca _ sis . the x - ray images of lmc2 were obtained with the _ rosat _ x - ray telescope , using both the pspc and the hri . as discussed in paper i , we have mosaicked a large number of independent pspc and hri observations to provide extensive angular coverage , and considerably deeper exposure than provided by the _ rosat _ all - sky survey ( rass ) . the data were retrieved through the _ rosat _ public archive at the high energy astrophysics science archive research center ( heasarc ) at the goddard space flight center . the region surrounding lmc2 has been extensively covered by pspc observations of the many sources in the vicinity ( e.g. , snrs & x - ray binaries ) and by the hri survey program of chu & snowden ( 1998 ) which was designed to cover the 30 doradus and bar regions of the lmc with deep and uniform coverage . tables a1 and a2 list the observation sequences that were used for the mosaics of the pspc and hri , respectively . these tables include the number of data sets for each sequence ( individual observations were occasionally processed in sections ) , pointing direction , exposure time , and target name . the data reduction was accomplished using the extended source analysis software ( esas ) package ( snowden & kuntz 1998 ; kuntz & snowden 1998 ) , which is also available through the heasarc . the final pspc and hri mosaics are cast in the same projection with the field center at @xmath9 ( 5@xmath1042@xmath11,@xmath12 ) and pixel size of 30@xmath13 and 20@xmath13 , respectively . the projection is the zenith equal - area azimuthal ( zea ; @xcite ) , which is similar over this solid angle to the more common tangential projection . reduction of the pspc data followed the procedures outlined by snowden et al . ( 1994 ) and demonstrated by snowden & petre ( 1994 ) , and in fact included a reprocessing of data in the latter paper . individual observations were screened for anomalous background conditions , had their residual non - cosmic background components modeled and subtracted , and had their vignetted and deadtime - corrected exposures calculated . only after all of the individual observations were reduced in six statistically - independent energy bands ( see table 1 ) were they cast into the mosaics . because the analysis software can not correct for zero - level offsets ( a constant component of the long - term enhancement non - cosmic background , see @xcite ) , the mosaicking process must correct for the relative offsets between overlapping fields . a single - value deconvolution algorithm was applied to a system of equations comprising all overlaps between separate observations to determine a best fit for all offsets simultaneously . a final constraint was added so that the sum of the offsets were equal to zero . individual observations were then adjusted by multiplying the fitted offset by the exposure map to produce an offset count image , which was then subtracted from the count image . this procedure was run separately on each energy band . table 2 lists some values of interest for the reduction . the true source counts can be seen to completely dominate the non - cosmic background and the average exposures are larger by more than an order of magnitude than that of the rass coverage . although these source counts include emission from 30 doradus , lmc x-1 , and various snrs , lmc2 is well above non - cosmic background . the total coverage of the field is @xmath14 deg@xmath15 . we have extracted a @xmath16 deg@xmath15 sub - image of the lmc2 region from the larger mosaic and present it in figure 1a . the reduction of hri data is simplified from that of pspc data . there is only the total energy band , and only the particle background is modeled and removed . the particle background of the hri completely dominates an hri observation with an average count rate @xmath17 times greater than the cosmic background . it must be removed as the distribution of the particle background is flat across the field while cosmic counts are vignetted . unfortunately , the particle background is also only calibrated to @xmath18% or so , but this is sufficient for the purposes of this paper . the contributions from other non - cosmic background components are ignored as they are `` in the noise . '' because these components are for the most part distributed as a vignetted flat field , they contribute a zero - level offset which is corrected in the mosaicking , which is done in the same manner as for the pspc . the total coverage of the field is @xmath19 deg@xmath15 . in figure 1b we present a sub - image from this mosaic that has the same field - of - view as figure 1a . the x - ray spectra in this paper were obtained with the _ rosat _ pspc and the _ asca _ sis . here we discuss the reduction of the x - ray spectral data . we discuss their analysis in 4 . we extract x - ray spectra of lmc2 from a _ rosat _ pspc observation of the snr dem l 316 ( _ rosat _ sequence no . rp500259n00 ) . the pspc is sensitive in the energy range of 0.1 2.4 kev , and has an energy resolution of @xmath20% at 1 kev . this observation has been discussed in some detail in williams et al . therefore , we only present a brief summary here . the observation was made within the time interval from 1993 january 10 to 1993 february 10 . for this observation , the snr dem l 316 ( @xmath21 5@xmath1047@xmath1195 ; @xmath22 ) was centered in the pspc . the total exposure time for the observation was 4.0 ksec . we processed these data using xselect and ftools , which are available through the heasarc . the data were spatially binned to produce 75 @xmath23 75 pixels . using the binned pspc image , we defined three regions of interest : a bright region in the northern part of lmc2 ( pspc - north ) , a bright x - ray arc in the west ( pspc - arc ) , and a region in the x - ray spur to the south of lmc2 ( pspc - spur ) . these regions are shown in figure 1a . because lmc2 is an extended source and all the regions are located outside the central pspc `` ring '' ( radius 20@xmath24 ) , we selected background regions of the same size and off - axis angle to correct for the charged particle background . during the reduction process we also performed vignetting corrections . we present the regions coordinates , angular sizes , net counts , and effective exposure times in table 3 . our _ asca _ observation is of a bright x - ray region in the northern section of lmc2 ( see figure 1a ) . this observation was performed on 1995 august 25 26 ( _ asca _ sequence no . 53039000 ) . the _ asca _ satellite consists of two instruments that observe simultaneously a pair of sis and a pair of gas imaging spectrographs ( gis ) . the sis has an energy coverage approximately in the 0.5 8 kev range and the gis in the 0.8 10 kev range ( tanaka , inoue , & holt 1994 ) . in this paper , we have only used the sis data because of the low sensitivity of the gis for energies below @xmath25 kev , where diffuse thermal plasma emission is brightest . the sis detectors have good spectral resolution @xmath26 $ ] . the angular resolution of the sis data is poor compared to the _ rosat _ pspc , with a narrow core of @xmath27 diameter and a half - power radius of 3@xmath24 . both of the sis detectors ( sis 0 and sis 1 ) are composed of four separate ccds . not all of the ccds can be on for a given observation as the presence of `` hot '' and `` flickering '' pixels can fill the telemetry with false signals . because of telemetry limitations , our observations were performed in the 2-ccd mode , with part of the bright northern region of lmc2 in the sis 0 ( chips 0 & 1 ) and sis 1 ( chips 0 & 3 ) . we have extracted spectra from sis 0 chip 1 and sis 1 chip 3 , because they image the same area on the sky and lie in a region of bright emission . we reduced and processed the data with ascascreen and ftools . because lmc2 is more extended than the sis field - of - view no region from any of our _ asca _ observations were suitable to perform a background subtraction . therefore , we used the _ asca _ deep field pointings , screened in a manner identical to the lmc2 data , to obtain regions for background subtraction . in table 3 , we also present relevant information on the two _ asca _ sis chips that were used in this paper . we have also extracted spectra in the regions covered by sis 0 chip 0 and sis 1 chip 0 ; however , these data were not analyzed because the count rates were low and these spectra can not be fitted by models with reasonable certainty . lmc2 consists of interstellar gas in several phases . as reported in paper i , we have obtained ( 1 ) _ rosat _ pspc and hri mosaics to investigate the spatial distribution of the hot ionized gas interior to lmc2 , ( 2 ) ccd images of lmc2 in optical emission lines to examine the shell of warm ionized gas , and ( 3 ) 21-cm hi data to study the structure of the cold , neutral gas . below we describe the distribution of the x - ray emitting plasma and its relation to the other gaseous components . the existence of hot ( @xmath0 k ) gas interior to lmc2 was first detected using the _ einstein _ ipc ( wang & helfand 1991 ) . as seen in our _ rosat _ pspc and hri mosaics ( figures 1a & 1b , respectively ) , lmc2 lies toward the brightest and most complex region of diffuse x - ray emission in the lmc . this region includes 30 doradus , lmc2 , the supergiant shell lmc3 , and a bright x - ray spur to the south of lmc2 , as well as snrs ( e.g. , dem l 299 , dem l 316 , and snr [email protected] ) and an x - ray binary ( lmc x-1 ) . these features are labeled on our hri mosaic in figure 1b . a cursory inspection of our pspc and hri mosaics reveals that the x - ray surface brightness of lmc2 is not uniform . the brightest region of diffuse x - ray emission associated with lmc2 is an arc in the southwestern quadrant that appears to extend from snr 0540@xmath28693 to lmc x-1 . a region of low x - ray surface brightness ( @xmath29 , @xmath30 ) lies between this bright x - ray arc and a region of bright x - ray emission in the northeast . the observed variation in x - ray surface brightness can be attributed to non - uniform emission from lmc2 and/or to non - uniform absorption of x - rays by intervening interstellar gas . as discussed below , we find that both of these explanations must be taken into account for the observed x - ray surface brightness toward lmc2 . it is of considerable importance in our study of the physical conditions interior of lmc2 to determine whether the variation in x - ray surface brightness is the result of non - uniform absorption or emission . to investigate this matter we first compare the x - ray mosaics with h@xmath3 emission line images that trace regions of current massive star formation . as the hot gas toward lmc2 is most likely shock - heated by the stellar winds and supernova explosions of massive stars , we expect to see correspondence between x - ray and h@xmath3 features . we present an h@xmath3 image of lmc2 overlaid with x - ray contours in figure 2 . we find three results of note in this comparison : ( 1 ) the x - ray emission from lmc2 is confined by the optical filaments ; ( 2 ) the bright x - ray arc is centered on the hii region n160 . based on the optical and x - ray morphology and high resolution echelle spectra of the h@xmath3 line , we suggest that the arc is formed by an outflow of hot gas from n160 into lmc2 ( paper i ) . ( 3 ) the hii region n163 ( @xmath31 ) is coincident with a small region of low x - ray surface brightness . comparison of the pspc mosaics ( snowden & petre 1994 ) shows that this depression in x - ray surface brightness is more pronounced in the soft x - ray band ( r4 ) than in the hard x - ray band ( r7 ) . therefore , we conclude that n163 lies in front of the x - ray emitting gas and absorbs the soft x - rays from lmc2 . we further compare the x - ray mosaic with the australia telescope compact array ( atca ) aperture synthesis maps of the 21-cm hi emission line toward lmc2 that reveals the presence of neutral atomic gas ( figure 3 ) . we know that hi exists in front and in back of lmc2 . thus , we can not simply expect variations in the hi column density to anti - correlate with the diffuse x - ray emission . instead we use the shadowing effect to look for correlations between features in the hi channel maps with x - ray surface brightness . as reported in paper i , a region of low x - ray surface brightness ( @xmath32 ) is coincident with an hi cloud detected in the hi channel maps . as in the case of the region with low x - ray surface brightness seen toward n163 , this depression is more apparent in the r4 band mosaic than the r7 band mosaic . therefore , this hi gas component most likely lies on the front side of lmc2 and absorbs the x - rays emitted from within lmc2 ( paper i ) . the _ rosat _ pspc and _ asca _ sis data provide x - ray spectra of lmc2that can be used to derive the physical conditions of the hot gas . we will first describe the procedures used to perform the x - ray spectral fitting and the uncertainties involved . then we will use the plasma temperature , foreground absorption column density , and x - ray emissivity obtained from the spectral fits to estimate the electron density , surface mass density , and surface thermal energy density of the hot gas . the observed x - ray spectra are a convolution of the intrinsic spectra , the foreground absorption , and the detector response function . because the foreground absorption and detector response are both energy dependent , we need to assume models of the intrinsic spectra and absorption cross - sections to reduce the number of free parameters . we simulate the observed spectra by folding the assumed intrinsic emission and foreground absorption models with the detector response . the observed spectra are fitted by the simulated spectra ; the @xmath33 of the fits is used to determine the best - fit . the x - ray emission from lmc2 , as seen in the hri mosaic ( figure 1b ) , is truly diffuse , indicating a hot - plasma origin . therefore , we adopt raymond & smith s ( 1977 ) thin plasma emission models for the intrinsic spectra . we use morrison & mccammon s ( 1983 ) effective absorption cross - sections per hydrogen atom to model the foreground absorption . the pspc and sis data were fitted separately because they were extracted from different regions toward lmc2 . the spectral fitting of the _ asca _ sis data was more difficult than the fitting of the _ rosat _ pspc data . the sis data were contaminated by scattered , high - energy photons from the bright nearby source lmc x-1 , an x - ray binary with an accretion disk . therefore , in the fitting procedure for the sis data we added a powerlaw and a blackbody disk component , but held the values of the power law photon index and the blackbody disk temperature constant at the values determined by schlegel et al . ( 1994 ) for lmcx-1 . we allowed the values of the raymond - smith plasma temperature and absorption column to float to the best - fit values . this is not a perfect method to determine the plasma temperature and absorption column of lmc2 because the scattering of the x - ray photons is both energy- and position - dependent . we may fit the sis 0 chip 1 and sis 1 chip 3 data simultaneously with the same model because these data were extracted from the same region on the sky . we performed multiple model fits for each of the extracted regions using xspec in which we allowed the abundance to be 20% , 30% , and 40% solar . although varying the abundance did not noticeably affect the best - fit values of the plasma temperature and absorption column , it did change the parameter space of the @xmath33 grid plots of the confidence contours . we find that the range of possible values for the plasma temperature and absorption column are better constrained for the models with 40% solar abundance . for the remainder of this work , however , we use the results from the 30% solar abundance models , the canonical lmc value . although our choice to use the best - fit values from the 30% solar abundance models may seem arbitrary , this allows us to compare our results with other studies of the hot , ionized medium in the lmc ( 5 ) . we note that the @xmath33 grid plots for a thermal plasma with 40% solar abundance may cover a smaller area of parameter space because the hot gas has been enriched above the canonical lmc value through the actions of stellar winds and supernova explosions of massive stars . we performed a @xmath33 grid search of possible fits in order to determine the uncertainties in the fits for temperature ( @xmath34 ) and absorption column density ( @xmath35 ) . the best - fit plasma temperature and absorption column for a 30% solar abundance raymond & smith ( 1977 ) thermal plasma model for pspc and sis data , as well as the range of the 90% confidence contours , are given in the top section of table 4 . in figures 4 & 5 , we present the x - ray spectra of the different regions , and their @xmath33 grid plots for a thermal plasma with 30% solar abundance , respectively . the grid plots show the range of @xmath33 fits for @xmath34 and @xmath35 for each extracted region ; the three contour levels represent the areas of 68% , 90% , and 99% confidence . to better constrain the range of @xmath35 , and thus to better confine the range of plasma temperatures , we calculate the hi column density toward lmc2 . the galactic values of @xmath35 toward lmc2 are @xmath36 @xmath37 , determined using the ftool `` nh '' which is derived from the hi map of dickey & lockman ( 1990 ) . the lmc values of @xmath35 toward lmc2 are @xmath38 @xmath37 using the parkes multibeam hi survey . the best - fit values of @xmath34 for the fixed @xmath35 model fits , as well as the 90% confidence ranges for @xmath34 and @xmath35 are given in the bottom section of table 4 . in general , the fixed @xmath35 column density is lower than that determined by the floating @xmath35 x - ray spectral fits ( see table 4 ) , similar to results obtained toward 30 doradus ( wang 1999 ) . this appears to be problematic because the parkes multibeam survey samples neutral hithat is both in front of and behind lmc2 . it seems reasonable that the fixed @xmath35 column density should represent an upper limit of the absorption column between us and the x - ray emitting plasma . it should be remembered , however , that x - rays are also absorbed by molecular and ionized gas that are not measured in the 21-cm hi emission - line survey , but are measured by the x - ray spectral fits . observations of the @xmath39 co emission line at 2.6 mm made with the nanten 4-m radio telescope ( fukui et al . 1999 ) do not show co emission toward the lmc2 regions of interest in this investigation . the lack of co emission does not necessarily imply an absence of molecular gas toward the selected regions . the lower dust abundance of the lmc allows uv photons , which would dissociate co molecules , to penetrate more deeply into molecular clouds and form c@xmath40 regions . observations of the [ cii ] 158@xmath41 m emission line toward lmc2 reveal that the distribution of c@xmath40 is more diffuse than the distribution of co ( mochizuki et al . therefore , it is possible that the molecular gas is in a very diffuse phase in which co is rare because of photo - dissociation . our calibrated ccd h@xmath3 images of lmc2 ( paper 1 ) show the emission measure , @xmath42 , where @xmath43 is the electron density and @xmath44 is the path length through the h@xmath3 emitting gas in pc , range from @xmath45 @xmath46 pc in the arc region to @xmath47 @xmath46 pc in the spur region . if the path length through the h@xmath3 emitting gas is comparable to the width of the filaments defining the eastern boundary of lmc2 ( @xmath48 pc ) , the ionized column densities toward the regions range from @xmath49 @xmath37 in the spur to @xmath50 @xmath37 in the arc . it is conceivable that the derived absorption column from the free @xmath35 x - ray spectral fits may be unrealistically higher than the measured absorption column toward the x - ray spectral regions . therefore , we use the fixed @xmath35 spectral fit parameters for the remainder of this work . from the model fits to the pspc and sis x - ray spectra , we can calculate the unabsorbed x - ray flux , and hence the x - ray luminosities , of the selected regions . given a hot gas filling factor @xmath51 , volume @xmath52 , x - ray luminosity @xmath53 , and emissivity @xmath54 , the electron density of the hot plasma of the selected regions will be @xmath55 if @xmath56 . for the regions toward lmc2 and the spur , the volume of the gas is the product of the projected surface area of the region and the path length through the gas . we adopt a path length comparable to the width of the interstellar structure in question ( i.e. , 140 pc for the bright northern region , 230 pc for the bright x - ray arc , and 400 pc for the spur ) . the x - ray emissivity , @xmath54 , is a function of the plasma temperature . from our fixed @xmath35 x - ray model fits to the data , we find that the emissivity of the regions ranges from @xmath57 erg @xmath58 s@xmath59 in the 0.44 2.04 kev energy range . we derive the electron densities for the extracted regions toward lmc2 using the x - ray luminosities and emissivities determined by our fixed @xmath35 spectral fits and present them in table 5 . as seen in figure 1a , the x - ray surface brightness varies from region to region toward lmc2 . because lmc2 covers a large portion of the pspc field - of - view , parts of it are obscured by the pspc window support structure . thus , we can not extract a single x - ray spectrum of lmc2 to determine the integrated x - ray flux of the hot gas . instead , we determine the total count rate of lmc2 from the pspc mosaic in the 0.44 to 2.04 kev energy band . if we assume that lmc2 is filled by hot gas with a plasma temperature @xmath60 kev and an absorption column of @xmath61 @xmath37 , we can use pimms to calculate the intrinsic x - ray flux of lmc2 to be @xmath62 erg @xmath37 s@xmath59 . the intrinsic x - ray flux of the spur is calculated to be @xmath63 erg @xmath37 s@xmath59 for a thermal plasma with temperature @xmath64 kev and a hydrogen absorption column of @xmath61 @xmath37 . we caution the reader , however , that these derived values of the x - ray flux were made assuming a uniform absorption column between us and lmc2 and the spur . therefore , these values do not accurately reflect the global x - ray fluxes of either of these interstellar structures . as x - ray fluxes have now been calculated for both lmc2 and the spur , we may be tempted to derive their physical conditions , such as electron density ( @xmath65 ) , thermal energy ( @xmath66 ) , and hot gas mass ( @xmath67 ) . this approach , however , would ignore the uncertainties inherent in the assumptions we made in our previous calculation of the intrinsic x - ray flux of lmc2 and the spur , such as the emitting volume , temperature , and the emissivity of the hot , thermal plasma . furthermore , we have shown in 3 that the absorption column across lmc2 is non - uniform . therefore , the physical conditions of the hot gas interior to lmc2 and the spur can not be realistically derived using the global x - ray fluxes calculated in 4.2 . instead , we opt to focus on the physical conditions of the hot gas toward the extracted spectral regions because the plasma temperatures , x - ray fluxes , and hot gas emissivities are explicitly determined by the x - ray spectral fits . to further minimize uncertainties in the derived physical conditions of the hot gas toward our extracted spectral regions , we calculate the surface thermal energy density ( @xmath68 ) and the hot gas surface mass density ( @xmath69 ) instead of @xmath66 and @xmath67 because they depend on the assumed path length through the hot gas , but not the volume . in the following analysis we derive the physical properties of the hot gas in our selected regions toward lmc2 . as an example , we calculate the physical properties of the hot gas in the pspc - arc region . the electron density of this region is @xmath70 @xmath71 . the surface mass density of the hot gas is given by @xmath72 where @xmath73 is the effective volume of the emitting gas and @xmath74 is the surface area of the region in arcmin@xmath75 . assuming that @xmath76 , the surface mass density of the hot gas in the arc ranges from 65 to 90 @xmath77 arcmin@xmath78 . if the x - ray emitting gas behaves as an ideal gas , the surface thermal energy density is given by @xmath79 . for the arc , this gives @xmath80 erg arcmin@xmath78 . furthermore , we may also estimate the number of supernova explosions required to produce the hot gas interior to lmc2 , based on our derived value of @xmath68 . for a supernova explosion with a total energy of @xmath81 erg , the thermal energy of the hot gas of the snr is @xmath82 erg and has a radius of @xmath83 pc after 10@xmath84 yr ( slavin & cox 1992 ) . at the distance to the lmc , the surface thermal energy density of the snr would be @xmath85 erg arcmin@xmath78 . thus , if the hot gas interior to lmc2 was produced entirely by a series of supernova explosions @xmath86 yr ago , then the number of supernova explosions required to energize the arc is 125 175 . we have obtained ubv photometry to investigate the star formation history toward lmc2 . those data will allow us to determine the amount of energy massive stars have deposited in the ism of lmc2 and will be compared with the thermal energy that is derived here . the physical properties of the arc , as well as those for the other regions toward lmc2 , are presented in table 6 . before comparing the physical properties of the hot gas interior to lmc2 with the physical properties of hot gas in other interstellar structures , we re - iterate the assumptions that underlie our calculations . the first factor we consider that contributes to the uncertainty in our calculations is the volume of the x - ray emitting gas in each of the regions . in this work we simply assume that the volume of a region is the product of its projected surface area and the path length through the region . because the depth through these regions is uncertain , an increase of the path length by a factor of 2 would alter our determination of the electron density by factor of @xmath87 and our calculations of the the surface thermal energy density and mass by a factor of @xmath88 . another factor affecting the derivation of the physical properties of the hot gas interior to lmc2 is the volume filling factor , @xmath51 . we have determined the physical conditions of the hot gas using a filling factor of @xmath89 and @xmath90 . a filling factor of @xmath89 increases the derived values of n@xmath91 by a factor of @xmath88 and decreases the values @xmath68 and @xmath69 by a factor of @xmath87 , with respect to the values determined using a filling factor of @xmath92 . inspection of figures 1a & 1b reveals that the distribution of the x - ray emission is relatively smooth and does not show much structure . thus , it is likely that the value of the filling factor of the hot gas interior to lmc2 lies between these two bounds . the last factors that need to be considered as contributing to the uncertainty in our results are the best - fits to the plasma temperature , @xmath34 , the absorption column , @xmath35 , and the normalization constant , @xmath93 , defined as @xmath94 . as seen in the @xmath33 grid plots ( figure 5 ) , the confidence contours for the pspc data show that the thermal plasma spectrum can be fitted by a low plasma temperature and high absorption column or by a high plasma temperature and a low absorption column . for example , if we take the plasma temperature of the pspc - north region to be the free @xmath35 model best - fit ( @xmath95 kev ; @xmath96 @xmath37 ) , the derived values of @xmath65 and @xmath69 are @xmath97 times the values derived using the fixed @xmath35 model fit values ( @xmath60 kev ; @xmath98 @xmath37 ) . in this example , the surface thermal energy density is @xmath99 times the value calculated for the fixed @xmath35 model fits and the intrinsic x - ray flux , @xmath100 is increased by a factor of @xmath101 . the hot gas in the ism is produced by shocks from supernova blasts and fast stellar winds from massive stars . consequently , the amount of hot gas and its physical conditions depend upon the spatial distribution and formation history of massive stars . massive stars will form discrete snrs if they are isolated , superbubbles if there is a single burst of star formation such as an ob association or supergiant shells if there are multiple bursts of star formation that have taken place over a period of time in a concentrated area . snrs , superbubbles , and supergiant shells are all filled by hot gas which contributes to the him ( hot ionized medium ; mckee & ostriker 1977 ) component of the ism in galaxies . thus , it is of interest to compare the physical conditions of the hot gas in these interstellar structures . below we compare the physical properties of the hot gas interior to lmc2 with those of : ( 1 ) another supergiant shell in the lmc , lmc4 ; ( 2 ) two superbubbles in the lmc , n11 and n44 ; and ( 3 ) a number of snrs in the lmc . we present the physical properties of these objects in table 7 . the values of the physical properties for these objects were made using the _ rosat _ bandpass of 0.1 2.4 kev and assuming a hot gas filling factor of @xmath92 unless otherwise noted . the pioneering study of diffuse x - ray emission from within a supergiant shell in the lmc was performed on lmc2 by wang & helfand ( 1991 ) using data obtained with the _ einstein _ ipc which operated in the energy range from 0.4 to 4.0 kev . therefore , as a consistency check , we first compare our results of the physical conditions of the hot gas interior to lmc2 to those which they derive . because a detailed spectral analysis of lmc2 could not be performed using the ipc data , wang & helfand ( 1991 ) used the mean hardness ratio of the excess x - ray emission from lmc2 to estimate the temperature of the hot gas . by assuming an optically thin thermal plasma with 50% solar abundance for the emission and a foreground absorption of @xmath102 @xmath37 , they determine the temperature of the x - ray emitting gas to be @xmath103 kev . wang & helfand ( 1991 ) assume the geometry of lmc2 to be spherical and adopt two different radii to determine the volume and physical properties of the 10@xmath84 kgas : ( 1 ) a radius of @xmath104 pc that covers only the arc region , and ( 2 ) a radius of @xmath105 pc that encompasses the entire supergiant shell . they find the density of the hot gas to be @xmath106 @xmath71 or @xmath107 @xmath71 for an adopted radius of 250 pc or 500 pc . their derived values of the unabsorbed x - ray flux , @xmath100 and the electron density , @xmath65 are given in table 7 . using their derived values , we calculate the surface thermal energy density , @xmath68 , to be @xmath108 erg arcmin@xmath78 and @xmath109 erg arcmin@xmath78 for a spherical bubble of hot gas with a radius of 250 pc and 500 pc , respectively . the results obtained using the _ einstein _ ipc data are in good agreement with those we determined using the _ rosat _ pspc data , but not with the sis data . for example , their values for the plasma temperature ( @xmath110 kev ) and the absorption column ( @xmath111 @xmath37 ) fall within the 90% confidence contour limits of our @xmath33 grid - plots of the pspc data , but not the sis data ( see figure 5 ) . a factor that may produce a discrepancy between the x - ray spectral fits to our dataset and the analysis of the ipc data is that wang & helfand ( 1991 ) assume a uniform hydrogen column density toward lmc2 . we have shown in 3.2 that the variations in x - ray surface brightness toward lmc2 are caused by non - uniform absorption and emission . as discussed previously , a higher adopted absorption column will result in a lower value for the plasma temperature . therefore , it is likely that the differences between the observed spectral properties and the derived physical conditions can be attributed to the differences in the adopted absorption column . in addition to lmc2 , diffuse x - ray emission has also been detected in the northern quadrant of the supergiant shell lmc4 ( bomans , dennerl , & krster 1994 ) using the _ rosat _ pspc in the energy range between 0.1 and 2.4 kev . they find that the best fit raymond & smith ( 1977 ) plasma temperature to the data is @xmath112 kev and the best fit absorption column is @xmath113 @xmath37 . if the diffuse x - ray emission from the northern quadrant of lmc4 is representative of lmc4 as a whole , the x - ray flux is @xmath114 erg @xmath37 s@xmath59 for this supergiant shell . assuming that lmc4 is a cylinder with a radius of 600 pc , a height of 1200 pc , and an x - ray emissivity of @xmath115 erg @xmath58 s@xmath59 , they derive an electron density of @xmath116 @xmath71 . using these values we calculate the surface thermal energy density of lmc4 to be @xmath117 erg arcmin@xmath78 . the derived values of @xmath118 , @xmath65 , and @xmath68 are listed in table 7 . the plasma temperature of the hot gas interior to lmc4 ( @xmath112 kev ) is comparable to that we have determined for lmc2 at the 90% confidence level , but the absorption column is about an order of magnitude lower . the primary distinction between the x - ray emitting plasma in lmc2 and lmc4 is that the electron densities in the regions toward lmc2 are a factor of @xmath119 greater than that derived for lmc4 . as we have used the same filling factor and a similar x - ray emissivity for the @xmath0 k gas as bomans et al . ( 1994 ) , the difference in the derived electron densities can be ascribed to the higher unabsorbed x - ray flux determined for lmc2 and/or through uncertainties in the volume of the emitting thermal plasma . if we adopt a path length through our selected regions toward lmc2 comparable to that of lmc4 ( i.e. , @xmath120 pc ) , our derived electron densities would be lowered only by a factor of @xmath121 . therefore , we conclude that the electron density of the hot gas interior to lmc2 is generally higher than the electron density interior to lmc4 . diffuse x - ray emission has been detected in several lmc superbubbles . here we discuss the physical conditions of the hot gas interior to the superbubbles in n11 and n44 , the two most luminous hii regions in the lmc after 30 doradus ( kennicutt & hodge 1986 ) . n44 has the highest x - ray surface brightness among all superbubbles in the lmc ( chu & mac low 1990 ) , while n11 is only moderately x - ray bright ( mac low et al . the physical properties of the hot gas interior to these two superbubbles are given in table 7 . raymond & smith ( 1977 ) spectral model fits to _ rosat _ pspc observations of n11 give a best - fit plasma temperature of @xmath122 kev ( mac low et al . this plasma temperature is higher than the model spectral fits to our pspc data toward lmc2 , but is comparable to plasma temperatures determined by the model fits to the sis data . using their results we calculate the electron density of n11 to be @xmath123 @xmath71 and the surface thermal energy density to be @xmath124 erg arcmin@xmath78 . both pspc and sis data have been obtained for n44 ( magnier et al . because these observations cover the same region on the sky , the pspc and sis x - ray spectra were fit simultaneously with a thermal plasma emission model . the model fits give a plasma temperature of @xmath125 kev for the hot gas interior to n44 ( magnier et al . assuming a filling factor of @xmath89 , they derive an electron density of @xmath126 @xmath71 from which we calculate the surface thermal energy density to be @xmath127 erg arcmin@xmath78 . the plasma temperatures of the hot gas interior to n11 is similar to those we determine for the hot gas interior to lmc2 using the pspc and sis data , but are higher in n44 . the electron density and surface thermal energy density for n11 are on the same order as for the regions toward lmc2 . the electron density of the hot gas interior to n44 is a factor of @xmath128 times higher than that derived for lmc2 , but the surface thermal energy density is @xmath129 times lower . some of the difference between the derived physical conditions of the hot gas interior to lmc2 and n44 may be attributed to the fact that magnier et al . ( 1996 ) used a hot gas filling factor of @xmath89 . the effects of changing the filling factor from @xmath92 to @xmath89 are discussed in 4.3 . supernova explosions are the main producers of the him in galaxies . therefore , it is imperative to understand the physical properties of the hot gas interior to snrs , in order to understand the physical conditions of hot gas interior to superbubbles and supergiant shells . fortunately , the lmc contains a large number of snrs ( @xmath130 ; williams 1999 ) that have been studied extensively at optical and radio wavelengths ( mathewson et al . 1983 ; 1984 ; 1985 ) and x - ray wavelengths ( williams et al . 1997 ; 1999 ; williams 1999 ; hughes et al . williams ( 1999 ) has compiled the most comprehensive list of the x - ray properties of lmc snrs to date . of the lmc snr sample , twenty - three have _ rosat _ pspc data available with sufficient count rates for spectral modeling . for this dataset , williams ( 1999 ) finds that the plasma temperature ranges from @xmath131 to 0.89 kev and that the electron density ranges from @xmath132 to @xmath101 @xmath71 . the published physical conditions of 4 of these snrs are listed in table 7 . clearly , the hot gas interior to snrs in the lmc shows a wide range of physical conditions . this is not surprising as they have different ages and interstellar environments . we see that most of these lmc snrs have plasma temperatures that are higher than lmc2 , but have electron densities that are comparable to it . the supergiant shell lmc2 has the highest x - ray surface brightness of all lmc supergiant shells . we have obtained _ rosat _ pspc and hri mosaics to study the distribution of x - ray emission from lmc2 . comparison of the x - ray mosaics with h@xmath3 images show that the x - ray emission is detected within the shell boundary of lmc2 , indicating the presence of hot ( @xmath0 k ) gas . in general , regions of low x - ray surface brightness are well - correlated with regions of high hi column density , suggesting that some of the variations in x - ray surface brightness toward lmc2 are caused by non - uniform absorption . in addition to the x - ray mosaics , we have also obtained _ rosat _ pspc and _ asca _ sis x - ray spectra to determine the physical conditions of the hot gas interior to lmc2 . model fits of an optically thin thermal plasma ( raymond & smith 1977 ) to the x - ray spectra of selected regions show that the hot gas has a plasma temperature @xmath133 @xmath134 kev with a best - fit of @xmath135 kev for our pspc and sis data . the electron density of the hot gas varies from @xmath136 to 0.09 @xmath71 . the unabsorbed x - ray flux in the 0.44 2.04 kev band is @xmath137 erg @xmath37 s@xmath59 . we have compared the physical properties of lmc2 with other supergiant shells , superbubbles , and snrs in the lmc . the derived electron densities of the lmc2 regions are generally comparable to the values determined for the supergiant shell lmc4 and the superbubble n11 , but are lower than those determined for the superbubble n44 and a sample of snrs in the lmc . the plasma temperature of lmc2 is comparable to that determined for lmc4 and n11 , but is generally lower than the plasma temperatures determined n44 , and snrs in the lmc . here we present tables a1 and a2 which list the individual _ rosat _ pointed observations that were mosaicked to produce the pspc and hri mosaics , respectively . these tables contain the number of data sets for each observation , the pointing direction , exposure time , and target name . bessell , m. s. 1991 , a&a , 242 , l17 bomans , d. j. , dennerl , k. , & krster , m. 1994 , a&a , 283 , l21 chu , y .- h . , & mac low , m .- 1990 apj , 365 , 510 chu , y .- h . , & snowden , s. l. 1996 , _ rosat _ newsletter # 13 , ed . s. l. snowden ( nasa : gsfc ) , 26 dickey , j. m. , & lockman , f. j. 1990 , ara&a , 28 , 215 feast , m. 1999 , pasp , 111 , 775 fukui , y. , et al . 1999 , pasj , 51 , 745 goudis , c. , & meaburn , j. 1978 , a&a , 68 , 189 greisen , e. w. , & calabretta , m. 1995 , asp conf . 77 , astronomical data analysis software and systems iv , ed . r. a. shaw , h. e. payne , & j. j. e. hays ( san francisco : asp ) , 223 henize , k. g. 1956 , apjs , 2 , 315 hughes , j. p. , hayashi , i. , & koyama , k. 1998 , apj , 505 , 732 kennicutt , r. c. , & hodge , p. w. 1986 , apj , 306 , 130 kuntz , k. d. , & snowden , s. l. 1998 , cookbook for analysis procedures for _ rosat _ xrt observations of extended objects and the diffuse background , part ii : mosaics , us _ rosat _ science data center , nasa / gsfc mac low , m .- chang , t. h. , chu , y .- h . , points , s. d. , smith , r. c. , & wakker , b. p. 1998 , apj , 493 , 260 magnier , e. a. , chu , y .- h . , points , s. d. , hwang , u. , & smith , r. c. 1996 , apj , 464 , 829 mckee , c. f. , & ostriker , j. p. 1977 , apj , 218 , 148 meaburn , j. 1980 , mnras , 192 , 365 mochizuki , k. , nakagawa , t. , doi , y. , yui , y. y. , okuda , h. , shibai , h. , yui , m. , nishimura , t. , & low , f. j. 1994 , apj , 430 , l37 morrison , r. , & mccammon , d. 1983 , apj , 270 , 119 points , s. d. , chu , y .- h . , kim , s. , smith , r. c. , snowden , s. l. , brandner , w. , & gruendl , r. a. 1999 , apj , 518 , 298 ( paper i ) raymond , j. c. , & smith , b. w. 1977 , apjs , 35 , 419 schlegel , e. m. , marshall , f. e. , mushotzky , r. f. , smale , a. p. , weaver , k. a. , serlemitsos , p. j. , petre , r. , & jahoda , k. m. 1994 , apj , 422 , 243 slavin , j. d. , & cox , d. p. 1992 , apj , 392 , 131 snowden , s. l. , & kuntz , k. d. 1998 , cookbook for analysis procedures for _ rosat _ xrt observations of extended objects and the diffuse background , part 1 : individual observations , us _ rosat _ science data center , nasa / gsfc snowden , s. l. , mccammon , d. , burrows , d. n. , & mendenhall , j. a. 1994 , , 424 , 714 snowden , s. l. , & petre , r. 1994 , apjl , 436 , l123 staveley - smith , l. 2000 , private communication tanaka , y. , inoue , h. , & holt , s. s. 1994 , pasj , 46 , l37 tenorio - tagle , g. , & bodenheimer , p. 1988 , ara&a , 26 , 145 wang , q. d. 1999 , apj , 510 , l139 wang , q. d. , & helfand , d. j. 1991 , apj , 379 , 327 williams , r. m. , chu , y .- h . , dickel , j. d. , beyer , r. , petre , r. , smith , r. c. , & milne , d. k. 1997 , apj , 480 , 618 willaims , r. m. 1999 , ph.d . thesis , university of illinois at urbana - champaign williams , r. m. , chu , y .- h . , dickel , j. d. , smith , r. c. , milne , d. k. , & winkler , p. f. 1999 apj , 514 , 798 lcccccc pspc - north & 5 44 19 & -69 10 51 & 185 & 34.7 & 717 & 2.7 pspc - arc & 5 41 55 & -69 34 23 & 166.5 & 27.8 & 1255 & 2.7 pspc - spur & 5 44 02 & -70 12 46 & 123 & 34.2 & 512 & 2.8 sis 0 chip 1 & 5 46 04 & -69 13 03 & 108 & 8.4 & 3737 & 30.7 sis 1 chip 3 & 5 46 04 & -69 13 03 & 104 & 8.1 & 3382 & 30.7 lcccc kt ( kev ) & 0.14 & 0.23 & 0.17 & 0.32 @xmath35 ( 10@xmath150 @xmath37)@xmath151 & 0.87 & 0.67 & 0.62 & 0.32 kt ( kev ) & 0.11 0.28 & 0.13 0.32 & 0.07 0.37 & 0.28 0.37 ; 0.60 0.69 @xmath35 ( 10@xmath150 @xmath37)@xmath152 & 0.58 1.08 & 0.48 0.94 & 0.12 1.43 & 0.24 0.42 ; 0.19 0.24 kt ( kev ) & 0.31 & 0.33 & 0.23 & 0.29 @xmath35 ( 10@xmath150 @xmath37 ) & 0.48 & 0.43 & 0.42 & 0.47 kt ( kev ) & 0.09 0.67 & 0.10 0.47 & 0.08 0.45 & 0.27 0.33 ; 0.60 0.69 @xmath35 ( 10@xmath150 @xmath37)@xmath152 & 0.08 1.25 & 0.28 1.05 & 0.07 1.35 & 0.34 0.54 ; 0.12 0.17 lccccccc pspc - north & 140 & 1.67 & 5.0 & 7.43 & 0.09 & 0.07 pspc - arc & 230 & 1.92 & 5.7 & 7.75 & 0.08 & 0.06 pspc - spur & 400 & 1.65 & 4.9 & 5.76 & 0.08 & 0.05 sis & 140 & 0.68 & 2.0 & 7.13 & 0.08 & 0.06 lcccc pspc - north & 0.07 0.09 & 46 65 & 0.32 0.46 & 80 115 pspc - arc & 0.06 0.08 & 65 92 & 0.48 0.68 & 120 170 pspc - spur & 0.05 0.08 & 107 152 & 0.56 0.79 & 140 198 sis & 0.06 0.08 & 41 58 & 0.27 0.38 & 68 95 lccccl lmc2 & 0.43 & 0.4 & 0.02 & 0.46 & wang & helfand ( 1991 ) lmc2 & 0.43 & 0.6 & 0.009 & 0.41 & wang & helfand ( 1991 ) lmc4 & 0.21 & 0.5 & 0.008 & 0.32 & bomans , dennerl , & krster ( 1994 ) n11 & 0.3 & 0.01 & 0.06 & 0.23 & mac low et al . ( 1998 ) n44 & 0.55 & 0.02 & 0.14 & 0.07 & magnier et al . ( 1996 ) dem 316a & 0.8 & 0.009 & 0.28 & 0.16 & williams et al . ( 1997 ) dem 316b & 0.9 & 0.007 & 0.15 & 0.08 & williams et al . ( 1997 ) n11l & 0.3 & 0.004 & 0.85 & 0.13 & williams et al . ( 1999 ) n86 & 0.15 & 0.008 & 0.16 & 0.03 & williams et al . ( 1999 ) lccccl 500138 & 2 & 5 26 36 & -68 50 23 & 27.08 & n144 400148 & 1 & 5 27 47 & -69 54 00 & 5.90 & rx j0527.8 - 6954 400298 & 2 & 5 27 47 & -69 54 00 & 14.85 & rx j0527.8 - 6954 300172 & 3 & 5 32 28 & -70 21 36 & 12.54 & nova lmc 88 # 1 500100 & 2 & 5 35 28 & -69 16 11 & 24.84 & sn1987a 500303 & 1 & 5 35 28 & -69 16 11 & 9.16 & sn 1987 a 600100 & 1 & 5 35 38 & -69 16 11 & 12.16 & region f 300335 & 1 & 5 36 12 & -70 45 00 & 9.90 & 2 new supersoft srcs 500131 & 1 & 5 38 33 & -69 06 36 & 15.45 & n157 400079 & 1 & 5 39 38 & -69 44 23 & 5.20 & lmc x-1 150044 & 1 & 5 40 12 & -69 19 48 & 5.15 & psr 0540 - 69 400052 & 1 & 5 40 12 & -69 19 48 & 6.55 & psr 0540 - 69 400133 & 1 & 5 40 12 & -69 19 48 & 1.72 & psr 0540 - 69 400012 & 1 & 5 46 45 & -71 09 00 & 14.92 & cal87 400013 & 1 & 5 46 45 & -71 09 00 & 14.73 & cal87 500259 & 1 & 5 47 09 & -69 41 59 & 3.92 & dem 316 lccccl 400124 & 1 & 5 18 40.8 & -68 14 24 & 4.55 & lmc h 204 500171 & 3 & 5 19 33.6 & -69 02 24 & 17.32 & 0519.69.0 400657 & 1 & 5 20 31.2 & -69 31 48 & 3.83 & rxj0520.5 - 6932 400153 & 2 & 5 22 26.4 & -67 58 12 & 8.56 & n44c / star 2 400231 & 1 & 5 22 28.8 & -69 26 24 & 3.89 & lmc field 1 600036 & 1 & 5 22 33.6 & -69 18 36 & 1.13 & lmc h 20 500002 & 1 & 5 25 2.4 & -69 38 24 & 27.25 & n132d 600032 & 1 & 5 25 43.2 & -69 13 12 & 0.71 & cal 37 600031 & 1 & 5 25 52.8 & -70 11 24 & 1.26 & cal38 400353 & 2 & 5 26 9.6 & -67 42 36 & 7.94 & lmc field a 400238 & 1 & 5 26 28.8 & -69 46 48 & 3.99 & h211 600033 & 1 & 5 27 38.4 & -69 10 48 & 1.65 & cal 40 201689 & 1 & 5 27 48 & -69 54 00 & 8.50 & rx j 0527.8 - 6954 600647 & 1 & 5 28 19.2 & -68 30 36 & 23.91 & lmc point 017 600646 & 1 & 5 28 19.2 & -68 51 00 & 24.01 & lmc point 016 600641 & 1 & 5 28 19.2 & -69 10 48 & 26.71 & lmc point 011 600782 & 1 & 5 28 19.2 & -69 51 00 & 12.79 & lmc point ao6 - 10 400356 & 1 & 5 28 33.6 & -68 36 36 & 3.16 & w528.8 - 684 400355 & 1 & 5 28 43.2 & -67 25 48 & 4.60 & w529.8 - 672 600777 & 1 & 5 28 50.4 & -69 33 00 & 19.84 & lmc point ao6 - 05 600645 & 1 & 5 32 4.8 & -68 30 36 & 19.00 & lmc point 015 600640 & 1 & 5 32 4.8 & -68 51 00 & 23.98 & lmc point 010 600635 & 1 & 5 32 4.8 & -69 10 48 & 21.23 & lmc point 005 600655 & 1 & 5 32 4.8 & -70 10 48 & 11.99 & lmc point 026 400125 & 1 & 5 32 7.2 & -69 19 12 & 4.92 & lmc h 215 400660 & 1 & 5 32 16.8 & -71 40 12 & 5.52 & rxj0532.3 - 7140 400234 & 2 & 5 32 52.8 & -67 43 12 & 6.39 & lmc field 2 500234 & 1 & 5 33 60 & -69 55 12 & 7.66 & 0534 - 69.9 400352 & 2 & 5 35 4.8 & -70 39 00 & 5.43 & lmc field b 400056 & 1 & 5 35 28.8 & -69 16 12 & 23.47 & sn 1987a 600634 & 1 & 5 35 52.8 & -68 51 00 & 23.62 & lmc point 004 600650 & 1 & 5 35 52.8 & -69 51 00 & 2.10 & lmc point 020 400230 & 1 & 5 35 55.2 & -69 21 00 & 4.86 & lmc field 3 400233 & 1 & 5 35 55.2 & -69 56 24 & 4.98 & lmc field 4 600035 & 1 & 5 36 2.4 & -67 34 48 & 0.78 & cal 60 600639 & 1 & 5 36 7.2 & -68 31 48 & 25.62 & lmc point 009 400644 & 1 & 5 36 16.8 & -67 18 00 & 4.57 & rxj0536.3 - 6718 500036 & 1 & 5 37 52.8 & -69 10 12 & 10.91 & n 157b 400642 & 1 & 5 38 16.8 & -69 24 00 & 2.43 & cal 69 600228 & 1 & 5 38 43.2 & -69 06 00 & 30.09 & 30 dor 400649 & 1 & 5 38 52.8 & -69 01 48 & 2.72 & cal 74 600638 & 1 & 5 39 38.4 & -68 30 36 & 25.73 & lmc point 008 600633 & 1 & 5 39 38.4 & -68 51 00 & 21.50 & lmc point 003 150013 & 1 & 5 39 38.4 & -69 44 24 & 0.80 & lmc x-1 400650 & 1 & 5 39 43.2 & -69 45 00 & 1.20 & lmc x-1 600774 & 1 & 5 40 4.8 & -69 31 48 & 22.34 & lmc point ao6 - 02 150008 & 1 & 5 40 12 & -69 19 48 & 21.59 & snr0540 - 69.3 400236 & 1 & 5 40 45.6 & -70 12 00 & 1.55 & lmc field 6 400349 & 2 & 5 41 26.4 & -68 49 48 & 6.56 & lmc field c 500235 & 1 & 5 43 7.2 & -68 58 48 & 28.24 & 0543 - 68.9 600643 & 1 & 5 43 26.4 & -68 30 36 & 21.45 & lmc point 013 600632 & 1 & 5 43 26.4 & -68 51 00 & 22.49 & lmc point 007 600637 & 1 & 5 43 26.4 & -68 51 00 & 22.49 & lmc point 007 600773 & 1 & 5 43 26.4 & -69 30 36 & 15.99 & lmc point ao6 - 01 600648 & 1 & 5 43 26.4 & -69 51 00 & 1.50 & lmc point 018 600778 & 1 & 5 43 26.4 & -69 51 00 & 8.79 & lmc point ao6 - 06 400121 & 1 & 5 46 28.8 & -68 34 12 & 1.59 & cal 86 400123 & 1 & 5 46 57.6 & -68 52 12 & 1.39 & lmc h 232 500232 & 2 & 5 47 9.6 & -69 42 00 & 20.35 & 0547 - 69.7 600644 & 1 & 5 47 12 & -68 30 36 & 18.76 & lmc point 014 600642 & 1 & 5 47 12 & -68 51 00 & 21.10 & lmc point 012 600636 & 1 & 5 47 12 & -69 10 48 & 20.63 & lmc point 006 600631 & 1 & 5 47 12 & -69 30 36 & 25.65 & lmc point 001 500233 & 2 & 5 47 50.4 & -70 24 36 & 26.41 & 0548 - 70.4 400659 & 1 & 5 47 52.8 & -68 22 48 & 5.29 & rxj0547.9 - 6823

in protoplanetary disks the aerodynamical friction between particles and gas induces a variety of instabilities that facilitate planet formation . of these we examine the so called ` secular gravitational instability ' ( sgi ) in the two - fluid approximation , deriving analytical expressions for its stability criteria and growth rates . concurrently , we present a physical explanation of the instability that shows how it manifests upon an intermediate range of lengthscales exhibiting geostrophic balance in the gas component . in contrast to a single - fluid treatment , the sgi is quenched within a critical disk radius , as large as 10 au and 30 au for cm and mm sized particles respectively , although establishing robust estimates is hampered by uncertainties in the parameters ( especially the strength of turbulence ) and deficiencies in the razor - thin disk model we employ . it is unlikely , however , that the sgi is relevant for well - coupled dust . we conclude by applying these results to the question of planetesimal formation and the provenance of large - scale dust rings . instabilities protoplanetary discs planets and satellites : formation

1609.07930

the assembly of planets is a complex and multi - faceted phenomenon that spans a gulf of some 12 orders of magnitude in length : from micron - sized dust to @xmath0 km planetary cores . it draws on an equally wide range of physical processes : collisions , dust - gas aerodynamics , gravitational collapse , instabilities , and disk structures ( e.g. vortices , dust traps ) , to name but a few ( papaloizou & terquem 2006 , chiang & youdin 2010 , armitage 2010 ) . while it is relatively straightforward to grow cm sized particles from micron sizes , further growth is potentially halted by a number of ` barriers ' ( bouncing , fragmentation , radial - drift ; johansen et al . statistically a small number of ` lucky ' aggregates may hurdle these , but certain collective instabilities can aid aggregation through this difficult size range . these include classical gravitational instability ( gi ; safronov 1969 , ward & goldreich 1973 ) , streaming instability ( youdin & goodman 2005 ) , and the secular gravitational instability ( sgi ; ward 2000 , youdin 2005 ) . it is to the last instability that this paper is devoted . one of the most attractive features of the single - fluid sgi is that its onset is unconditional ; it should always be present . unlike classical gi , which requires the toomre parameter to be less than one , a single fluid analysis presents no analogous restriction : the sgi works no matter how thin or thick the particle sub - disk ( youdin 2005 ) . the instability attacks longer scales preferentially , which ordinarily would be stabilised by the coriolis force ; but particles can shed ( or gain ) angular momentum via areodynamical drag , and hence are not obliged to undergo stabilising epicycles . as a consequence , rings that are radially drifting towards each other continue to do so unimpeded , and the instability can proceed . on small radial scales the sgi is suppressed by dust pressure or gas turbulence , and in fact , for well - coupled dust , turbulence decreases growth rates to potentially insignificant levels ( shariff & cuzzi 2011 , youdin 2011 ) . marginally coupled particles , however , could still be subject to respectable sgi growth rates at certain radii . the sgi has been thoroughly explored in single fluid models , which are applicable when the dust to gas density ratio is tiny ( e.g. ward 2000 , youdin 2005 , shariff & cuzzi 2011 , youdin 2011 , michikoshi et al . 2012 ) . these models assume that the angular momentum bestowed onto , or removed from , the gas disk is negligible . on sufficiently long scales , however , both sides of this momentum transaction must be included and the gas dynamics explicitly calculated . an instability criterion then appears : in a two - fluid model the onset of sgi is no longer unconditional . recently , takahashi & inutsuka ( 2014 , hereafter ti ) made a start on this problem ( see also shadmehri 2016 and takahashi & inutsuka 2016 ) , but there is still much to be established . putting aside the issue of growth rates , an especially important question is : at what radii and for what particle sizes should we expect sgi to exist at all ? the first aim of this paper is to derive clean stability criteria for the sgi . in the limits of strongly - coupled and weakly - coupled particles these can be formulated analytically and involve a variety of parameters , including the gas s toomre parameter and the dust - to - gas density ratio . because they bypass the sgi s full 6th order dispersion relation , these criteria make it relatively easy to assess its prevalence . the criteria also motivate a straightforward physical picture of instability in a two - fluid system . in order for the instability to work , there must exist an intermediate range of lengthscales gas upon which ( a ) dust pressure or turbulent mass diffusion is subdominant , and ( b ) the gas is prevented from executing epicycles , despite its angular momentum transactions with the dust . going to lengthscales longer than the dust pressure ( or diffusion ) scale takes care of the first restriction . but the second can only be satisfied if geostrophic balance holds in the gas fluid , and so we must simultaneously find shortish scales upon which gas pressure is dominant . the existence or not of this intermediate range furnishes us with the stability criterion . the formalism is applied to realistic disk models , where we find that it is unlikely that well - coupled dust is unstable to the sgi at any radius , unless the background turbulence is especially weak . marginally coupled particles , however , can achieve appreciable growth rates in certain circumstances , emphasising that the sgi could help aggregation of solids of cm size . we conclude , however , that sgi is probably unrelated to the dust rings recently observed by alma ( brogan et al . 2015 ) . the paper will be organised in the following way . first , in section 2 , we present the two - fluid razor - thin disk model that we employ , alongside a critical discussion of its shortcomings . the main parameters of the analysis will also be defined . in section 3 we revisit the single - fluid model to fix some ideas and to provide context for the subsequent analysis , while in section 4 we briefly treat a simple two - fluid system where the gas is regarded as incompressible . the main results of the paper are in section 5 , in which we derive analytic stability criteria in relevant limits that are then applied to realistic disk models in section 6 . we draw our conclusions in section 7 , where we discuss the relevance of the sgi in planet and structure formation in protostellar disks . the classical gi and secular gi have primarily been explored with 1d models of a vertically averaged or razor - thin disk . recent notable exceptions are mamatsashvili & rice ( 2010 ) and lin ( 2014 ) , who also capture vertical convection and the magnetorotational instability respectively . a 1d model certainly eases the analysis and it should be a reasonable approximation for unstable modes whose radial wavelength is much greater than the disk thickness ; because the classical gi has minimal vertical structure ( being essentially an f - mode in this limit ; ogilvie 1998 ) , it is also likely that the sgi depends on @xmath1 only weakly . for wavelengths closer to the scale height , a somewhat ad hoc correction may be included ( e.g. shu 1984 ) , which generally works against instability on these shorter lengthscales . in a two - fluid model , however , the razor - thin assumption is complicated by the fact that the particle fluid and the gas fluid possess different thicknesses , and the former can be significantly shorter than the latter . this is a problem for the sgi because the fastest lengthscales are not far from the particle scale height ( youdin 2005 ) , and hence of order or less than the gas scale height . as a consequence , the approximation of a razor - thin disk is not strictly applicable , at least in the description of the gas . in a single fluid model this issue does not crop up because it is assumed that the gas fluid is unperturbed by whatever the dust is doing ; but in a two fluid model this is not the case . it may be that the gas perturbations associated with the sgi are sufficiently small that the disk s vertical structure plays little role . but only calculations in vertically stratified shearing boxes can decide on this point . a second issue is the correct coupling between the two fluids . in a real system , with different gas and dust scale heights , the drag acceleration will be a function of vertical height @xmath1 . moreover , the entire column of gas will not exert drag on the dust if the dust subdisk is much thinner . for consistency , the gas _ external _ to the dust disk should be excluded from a two - fluid razor - thin treatment , with the weighting of the drag force in the momentum equation adjusted appropriately to account for the smaller surface density of the gas subdisk . because most of the mass in both disks is near the midplane , this problem may not invalidate the main qualitative results . it should , however , be kept in mind . a third issue concerns turbulence in the gas , its effects , and how to mathematically describe it . the gas is likely undergoing disordered motions at some ( perhaps all ) radii , though the underlying physics may differ between different regions ( turner et al . the disk may also support turbulence because of the settling of the dust to the midplane , and ensuing vertical kelvin - helmholtz instability ( cuzzi et al . 1993 ) . the consequences of turbulence on particles are several . agitation of the solids produces enhanced velocity dispersions , over and above that arising from particle collisions ( goldreich & tremaine 1978 ) , and thus an appreciable particle pressure that will oppose vertical settling . the precise ` equation of state ' this pressure obeys , however , is difficult to establish . in addition , the random motions induced in the dust can potentially smooth away inhomogeneities in the dust density , and thus lead to diffusion in the continuity equation directly ( youdin 2011 , shariff & cuzzi 2011 , takahashi & inutsuka 2012 ) . the efficiency of this diffusion is pretty much unconstrained and obviously requires numerical exploration ; for instance , the vertical kelvin - helmholtz instability will mix particles effectively in the vertical direction but not necessarily in the horizontal . finally , turbulence will transport momentum , though this effect ( being a straightforward mild damping ) we do not include in this paper ( see ti for its treatment ) . the effects of turbulence on the dust have been mathematically modelled via mean - field theories , and langevin and fokker - planck equations ( e.g. schrpler & henning 2004 , carballido et al . 2006 , youdin & lithwick 2007 ) . given a kolmogorov spectrum of isotropic homogeneous turbulent motions , youdin & lithwick ( 2007 ) derive convenient expressions for the dust velocity dispersion and the turbulent diffusivity of particle mass , in terms of the stopping time and a turbulent efficiency parameter ( described below ) . such a calculation , of course , must assume that the turbulent flux of particles behaves as a fickian diffusion which it need not , especially on scales close to the largest ` eddy ' . other complications could arise from the flux s finite relaxation time and possible anti - diffusive behaviour ( especially on small scales , e.g. frisch 1995 davidson 2000 , cuzzi et al . 2001 ) . our purpose in this paper is not to improve on any of these issues , but it is important to flag them at this point . we persist with the simple models previously employed ( e.g. ti , shadmehri 2016 ) , mainly because they can fix ideas and establish clear results , and presumably give approximately correct predictions . future work , however , should involve vertically stratified models along the lines of mamatsashvili & rice ( 2010 ) and lin ( 2014 ) . as will become clear by section 5 , the two - fluid secular gi is governed by a large number of parameters ; it is hence convenient to define them all in one place . the first key parameter is the dust - to - gas mass ratio , denoted by @xmath2 and defined via @xmath3 where @xmath4 is the background surface density of the dust with size @xmath5 , and @xmath6 is the surface density of the gas located _ within the dust subdisk _ ( see earlier ) . thus @xmath2 is a function of @xmath5 . because the sgi is size - selective it is necessary to distinguish between particles of different sizes , and thus to use separate surface densities for each subspecies . to ease the analysis in this paper we examine each dust species separately , though in reality different species will weakly couple via the gas phase . see shadmehri ( 2016 ) for an attack on a system of interacting gas and two species of particle . generally , @xmath4 is a small fraction of the total solid surface density . in fact , the surface density of mm and cm sized particles can be significantly less than 1% of the total ( e.g. brauer et al . 2008 , windmark et al . 2012a , 2012b ) , though this figure varies greatly with age and as different physical processes are included or neglected ( sticking , bouncing , fragmentation , and mass tranfer , for instance ; garaud et al . note that @xmath6 will also be less than the total gas density , as it only includes gas situated amongst the dust subdisk . this decrease can be quantified by a factor @xmath7 , where @xmath8 and @xmath9 are the scale thicknesses of the gas and dust respectively . we may then write @xmath10 where @xmath11 is the total dust surface density ( including all species ) , and we have assumed that the ratio of total dust to total gas density takes the standard value 0.01 ( chiang & youdin 2010 ) . given the large uncertainties in the second and third factors in eq . ( small and large respectively ) , we simply set @xmath12 for most calculations . inverse stokes number @xmath13 as a function of radius for different particle sizes and two different minimum mass solar nebulae models , as calculated from youdin ( 2011 ) but only using the epstein drag regime . the @xmath14 case corresponds to a nebula with a gas surface density of 2000 g@xmath15 at 1 au , while @xmath16 corresponds to 200 g@xmath15 . [ mmsn].,width=377 ] the second key parameter is the inverse stokes number , which we denote by @xmath13 and define through @xmath17 where @xmath18 is the stopping time of the particles , and @xmath19 is the local orbital frequency . large values of @xmath13 correspond , thus , to strongly coupled particles , and smaller values to weakly coupled particles . the exact value of @xmath13 depends on particle size , naturally , but also on radial location ( e.g. chiang & youdin 2010 ) . fig . [ epsilon ] gives @xmath13 profiles for two different minimum mass solar nebulae ( mmsn ) models . see section 6 for more details . the third parameter is the ratio of dust and gas velocity dispersions , defined through @xmath20 where @xmath21 and @xmath22 denote the velocity dispersion of the dust and gas respectively . to a first approximation , the dust random velocities are controlled by ` kicks ' delivered by the gas turbulence , rather than inter - particle collisions ( youdin & lithwick 2007 ) . thus it is possible to relate @xmath23 to properties of the turbulence , and we do so below . collisional agitation becomes important for larger particles @xmath24 cm , and thus the @xmath23 we use in this paper may be an underestimate for these sizes . the fourth parameter is the toomre @xmath25 of the dust , which describes the onset of classical gi . it is given by @xmath26 where @xmath27 is the gravitation constant and for notational convenience we have dropped the subscript ` @xmath5 ' from the surface density . an analogous expression can be defined for the gas , which we denote by @xmath28 . the two toomre parameters can be related via the following identity : @xmath29 . finally , the turbulent diffusion of solid particles can be quantified via the mass diffusivity @xmath30 , and the diffusion of gas by the analogous @xmath31 . as in youdin ( 2011 ) , we replace these by the dimensionless parameters @xmath32 and @xmath33 , via @xmath34 and @xmath35 , with these related via @xmath36 ( youdin & lithwick 2007 , youdin 2011 ) . the dust s velocity dispersion ( if excited by turbulence ) can also be expressed in terms of @xmath33 and we find @xmath37 thus @xmath33 governs both @xmath30 and @xmath23 . unfortunately , an estimate of its magnitude is one of the great uncertainties in the theory , though it is likely to be smaller ( and possible much smaller ) than the analogous dimensionless coefficient associated with angular momentum transport in the gas ( @xmath38 in a protoplanetary disk ) . for instance , if transport is controlled by magnetocentrifugal winds or strong zonal magnetic fields then the bulk of the disk could even be laminar ( e.g. lesur et al . 2014 , bai 2014 ) . having said that , vertical settling should always lead to disordered motions and some degree of radial diffusion . putting these considerations aside , it is clear that strongly coupled particles ( @xmath39 ) have @xmath40 and @xmath41 ; whereas weakly coupled particles ( @xmath42 ) are diffused less effectively @xmath43 and are far ` colder ' , @xmath44 . though the single fluid analysis of the sgi is well - trodden territory we include it here for completeness and because it helps fix useful ideas that appear later . one especially important theme that crops up is geostrophic balance and the purely azimuthal ` zonal ' flows that ensue . as we are interested in relatively short radial scales it is convenient to employ the shearing sheet model ( goldreich & lynden - bell 1965 ) , whereby a small portion of disk , centred upon a fixed radius @xmath45 , is represented by a cartesian box . in a corotating frame centred on the box , @xmath46 and @xmath47 denote the local radial and azimuthal coordinates , while @xmath19 is the orbital frequency at @xmath45 . the disk is assumed to be razor thin , so that the dust volumetric density is @xmath48 , where @xmath49 is the dust surface density and @xmath50 the dirac delta function ( not the dust - to - gas mass ratio ) . the equations governing the evolution of the dust fluid are given by the continuity , momentum , and poisson equations : @xmath51 where @xmath49 , @xmath52 , and @xmath53 denote the dust surface density , velocity , and pressure . again we have dropped the subscript ` @xmath5 ' from the surface density ; it is understood from here onwards that @xmath49 refers to the surface density of particles of size @xmath5 . the potentials associated with the dust self - gravity and tide are @xmath54 and @xmath55 respectively . the gas velocity is given by @xmath56 , and it interacts with the dust via a drag term whose strength is quantified by the inverse stokes number @xmath13 . in this section it is assumed that there is no appreciable backreaction of the dust on the gas motion . moreover , we neglect the effect of any radial pressure gradient on the gas s orbital rotation . it is thus keplerian and steady : @xmath57 . in realistic disks there is likely to be a fluctuating component of the gas motion due to turbulence , which acts as a forcing term in the dust momentum equation , giving rise to velocity fluctuations in the dust . we will be interested in the large - scale mean dust velocity , rather than these fluctuations ; the latter s effects may be captured by a turbulent pressure tensor in the momentum equation and a turbulent radial flux in the continuity equation . the @xmath58 and @xmath59 terms are the manifestations of these two effects . finally , given that we have assumed a dust pressure , we must stipulate the dust s equation of state , relating @xmath49 , particle velocity dispersion @xmath21 , and @xmath53 . this is not straightforward , especially if the dust velocity dispersion is dominated by the turbulent gas fluctuations . as our intention is to provide a broad physical explanation of underlying physics , rather than detailed modelling , we assume for simplicity that the dust fluid is isothermal , and so @xmath60 , where @xmath21 is constant . the governing one - fluid equations admit an equilibrium characterised by a constant density @xmath61 and the perfect entrainment of the dust in the gas , both undergoing keplerian motion @xmath62 . to this steady state we add small axisymmetric perturbation , @xmath63 , @xmath64 , proportional to @xmath65 , where @xmath66 is a growth rate and @xmath67 is a radial wavenumber . their linearised equations are @xmath68 where the perturbed gravitational potential is given by @xmath69 ( e.g. , binney & tremaine 1987 ) and where we have set @xmath70 for the time being . eliminating the primed variables produces a relatively neat third order dispersion relation : @xmath71 here @xmath72 is the standard expression for the squared frequency of density waves in a 1d inviscid disk . this agrees with multiple examples in the literature , notably in ward ( 2000 ) and youdin ( 2005 ) , and also in youdin ( 2011 ) , and shareef & cuzzi ( 2011 ) , when turbulent diffusion is omitted . it is worthwhile examining the classical case with no drag , i.e. when @xmath73 . the dispersion simplifies and one obtains @xmath74 and @xmath75 . the first pair of solutions corresponds to 1d density waves , which can grow on a band of intermediate wavenumbers @xmath67 girdling @xmath76 the instability criterion requires the toomre parameter @xmath77 to be less than 1 ( e.g. safronov 1969 , goldreich & ward 1973 ) . radial collapse on long wavelengths is impeded by epicyclic motion induced by the inertial forces , whereas short wavelengths modes are stabilised by pressure . ( note that nonlinear non - axisymmetric instability occurs for larger @xmath78 . ) the dust , however , must be rather thin in order to achieve @xmath79 ( e.g. cuzzi et al . 1993 , chiang & youdin 2010 ) the @xmath75 ` quasi - geostrophic ' mode neither grows nor oscillates , but instead corresponds to a steady ` zonal flow ' . from eq . , the fundamental balance is between the coriolis force , on one hand , and self - gravity and pressure , on the other . the mode corresponds to a radially varying sequence of super and sub - keplerian orbital motions ( or ` jets ' ) : @xmath80 where @xmath81 is the associated pressure perturbation . the pressure gradient negates any type of epicyclic motion . this type of flow plays an important part in the secular gi , as will be made clear in the following sections , and indeed in the streaming instability ( jacquet et al . 2011 ) . growth rates of the secular gi in a single fluid for @xmath82 and two different values of @xmath13 , 0.1 ( weakly coupled ) and 100 ( strongly coupled ) . the solid curves represent the full solution to the dispersion relation , whereas the points correspond to the asymptotic estimates and .,width=340 ] being a cubic , the dispersion relation does not yield simple formulae for the various growth rates . it is straightforward , however , to obtain a general stability criterion and asymptotic expressions in the limit of strong and weak drag . the last term in the cubic reveals that linear instability is assured when @xmath83 , which is satisifed for all modes with sufficiently small wavenumbers @xmath84 . hence instability is unconditional , though in practice ( putting aside the disk s global structure and cylindrical geometric effects ) an unstable dust layer must have radial extent greater than @xmath85 or else there will insufficient space for the modes to manifest themselves . in figure [ fig1 ] we plot numerical growth rates of the secular gi for parameters characteristic of the two limits . superimposed are the leading order estimates taken from eqs and below . when @xmath42 we are in the weak drag regime corresponding to larger particles , at larger disk radii , in less massive disks . eq . then yields two density waves : @xmath86 which are unstable , as expected , when @xmath79 and thus correspond to the classical gi . otherwise the two waves are mildly damped by drag . the third secular mode exhibits a growth rate of @xmath87 instability is assured when @xmath88 , in accordance with the stability criterion derived above . the fastest growing mode possesses @xmath89 and the maximum growth rate is @xmath90 the mode always grows , but for large @xmath25 the growth rate can be small . in the limit of small @xmath13 , the unstable mode is , in fact , a modified zonal flow . in eq . , the dominant balance is _ quasi - geostrophic _ , i.e. between the pressure gradient , self - gravity , and the coriolis force ( because @xmath91 and @xmath92 ) . using eq . , with eqs and , gives then precisely the leading order term for @xmath66 . in the strong drag limit , @xmath93 , the sgi grows at a rate of @xmath94 instability occurs for the exact same range of @xmath67 as in the opposed weak drag limit , but the maximum growth rate is slightly altered @xmath95 and the mode works through a different arrangement of forces . in eq . instead of radial geostrophic balance , it is the last three terms that are dominant : radial contraction by self - gravity is met by the drag force and pressure . essentially , the dust has achieved _ terminal velocity_. solving for @xmath96 and using yields the leading term for @xmath66 . in addition , there exist two density waves with @xmath97 . as in the weak coupling limit , they grow exponentially when @xmath79 but are otherwise weakly damped by drag at a rate @xmath98 . for completeness we next consider the impact of turbulent mass diffusivity , focussing especially on the critical wavenumber at which instability is quenched . letting @xmath99 , and taking the limit of weak drag , @xmath100 , the leading order expression for the growth rate picks up a term @xmath101 . the critical @xmath67 is then easy to compute : @xmath102 the importance of turbulent diffusion versus pressure is quantified by @xmath103 . for weakly coupled particles , this parameter is @xmath104 , using the estimates on turbulent diffusion and velocity dispersion from section 2 . this is saying that turbulence is roughly as important as dust pressure , at least in setting the short scale cut - off for sgi . nonetheless , in the remainder of the paper we often omit diffusion when dealing with weakly coupled particles , primarily in order to derive clean expressions . the neglect of diffusion will not change these results qualitatively , but will introduce order 1 corrections . instability criteria should then be regarded as necessary conditions , not sufficient , and maximum growth rates should be understood as upper bounds . in the opposite limit , @xmath39 , the situation is quite different . the growth rate is modified by the term @xmath105 , and the importance of turbulent diffusion on small scales is quantified instead by @xmath106 which is @xmath107 . as a consequence , the critical cut - off for small particles is completely controlled by turbulence , as shown earlier by youdin ( 2011 ) using a heuristic ( but essentially equivalent ) argument . on account of the subdominance of particle pressure in the face of turbulent diffusion when @xmath39 , we may dispense with it entirely . in this case , instability is assured for sufficiently long wavelengths as earlier , with the asymptotic growth given by @xmath108 and a maximum growth rate of @xmath109 where we have introduced the gas toomre parameter in the last equality , and set @xmath110 ( cf . section 2.2 ) . the wavelength of maximum growth is @xmath111 , where @xmath8 is the scale height of the gas disk . given that in this regime @xmath13 is large but @xmath33 is small , it can be hard to establish this characteristic length a priori . before moving on , it should be highlighted that a turbulent mass - diffusion alters the classical gi in unexpected ways . when @xmath73 and @xmath99 , arbitrarily long wavelengths are rendered unstable , though they grow at negligible rates . on the other hand , mass diffusion can not completely stabilise short scales - a pressureless turbulent fluid will be unstable for all @xmath67 ( in contrast to the sgi ) . introducing turbulent momentum diffusion , however , does kill instability for sufficiently large @xmath67 . we omit details of these calculations . as has been commented upon in the literature , the secular gi exhibits two striking features : growth for all finite @xmath25 , and growth upon arbitrarily small @xmath67 . the latter is especially unexpected given that traditional gi prefers intermediate wavelengths , the longer scales stabilised by the dominant coriolis force . how does secular gi overcome the epicyclic response at large scales ? consider two dust rings located at different radii undergoing circular orbital motion . each ring contains a quantity of angular momentum naturally associated with its home radius . suppose the rings are displaced radially towards one another other . though their mutual self - gravitational attraction will attempt to amplify the displacement , the two rings possess an angular momentum incommensurate with their new radial location and hence undergo epicyclic oscillations that thwart any type of gravitational collapse . this is the classical picture of stabilisation at long wavelengths . suppose however that there exists a drag force on both dust rings due to interactions with the background gas . now when the two rings are radially displaced they will exchange angular momentum with the gas , via the last term in eq . . the rings either gain or lose angular momentum until they achieve the amount commensurate with their new radial location . as a result , they do _ not _ undergo epicyclic motion , and self - gravity continues to amplify their radial drift towards one another . as the two rings collapse they continuously shed or gain angular momentum as needed . in this way , the epicyclic restoring forces are negated by the drag . note that this scenario only works if the gas remains an infinite reservoir of angular momentum , which can be removed or added to with no ramifications . this may be a reasonable approximation in cases where the dust density is far less than the gas density , but on some sufficiently long scale even this must break down . the question then arises : on what range of scales does the secular gi operate upon , and under what circumstances may we take the single fluid approximation ? for the smooth running of the sgi , the gas fluid must resist undergoing epicyclic oscillation when perturbed by the dust drag . when and how can this be arranged ? these questions will be explored in the following two sections . as an intermediate step between the single fluid and fully compressible two - fluid models , we briefly analyse the case of a compressible dust disk embedded in an incompressible gas . this situation mimics the case when the dust scale height is far less than the gas scale height , and the unstable motions very subsonic . on the vertical scale of the dust disk , the gas density is effectively constant and the problem is ` vertically local ' as far as the gas is concerned . consequently , the gas density does not contribute to the perturbed poisson equation . the equations governing the evolution of the incompressible gas are @xmath112 where @xmath6 and @xmath113 are the gas surface density and vertically integrated pressure , _ within the dust layer_. because the gas is incompressible @xmath6 is a constant . as earlier , @xmath114 is the gas velocity , and @xmath52 is the dust velocity . the equations governing the dust fluid are those that appear in section 2 , eqs - . to ease the analysis @xmath70 . once again , we assume the standard equilibrium state @xmath115 , @xmath61 , @xmath116 , where the gas pressure and dust surface density is constant . next , axisymmetric perturbations are assumed , denoted by @xmath117 , @xmath118 , @xmath64 , @xmath63 , and these are taken to be @xmath119 . because of the incompressibility condition we immediately obtain @xmath120 , and @xmath121 is computed from the @xmath47-force balance , yielding @xmath122 where we have introduced now @xmath123 which quantifies the dust - to - gas density ratio for a given particle size . this equation states that the gas is azimuthally accelerated by dust drag . simultaneously , a form of radial geostrophic balance holds for the gas , with the coriolis force balanced by the radial pressure pressure , self - gravity , and drag . importantly , the gas pertubation can not undergo epicyclic motion , which might impede the growth of the secular gi . the gas pressure gradient holds the fluid radially ` in place ' and a sequence of azimuthal jets ensues , each accelerated by the dust drag . ( note that the absence of epicycles is a generic feature of incompressible flow confined to the orbital plane . ) the perturbation equations for the dust are the same as eqs - except for the inclusion of the term @xmath124 on the right side of the dust s @xmath47-force balance . eliminating the dependent variables obtains the following quartic dispersion relation : @xmath125s^2 \notag \\ & \hskip2 cm + \epsilon\omega\left[\delta(\overline{\omega}^2 + 2\epsilon^2\omega^2 ) + \overline{\omega}^2-\omega^2\right]s \notag\\ & \hskip4cm+2\delta(\overline{\omega}^2-\omega^2)\epsilon^2\omega^2=0,\label{2fincdisp}\end{aligned}\ ] ] which we now briefly analyse . though eq . may appear rather formidable , an instability criterion appears immediately . putting aside the clasical gi for now , the sgi mode is marginal when the last term is zero , yielding exactly the same instability criterion as in the single fluid model : @xmath126 , and hence instability occurs on all @xmath127 . though we allow for gas perturbations , gas incompressibility restricts these to a form of zonal flow which absorbs or bestows angular momentum as necessary to faciliate instability in the dust . in the weak coupling limit , @xmath128 , the leading order term in the sgi growth rate is obtained by setting @xmath129 and substituting this into . one obtains the quadratic : @xmath130 the resulting solution for @xmath131 agrees with the single fluid expression to leading order in small dust - to - gas fraction @xmath132 . the reason for the dependence on @xmath2 is because the ability to transfer angular momentum between gas and dust is influenced by the relative azimuthal speeds of the two fluids , which depends on @xmath2 via . in the strong coupling limit , @xmath39 , assuming that @xmath133 , a similar analysis reveals that the sgi growth does not depend on @xmath2 at all . in fact , @xmath134 , precisely the same expression as in the single fluid case . having treated simpler models of the dust - gas system , we turn to a fully compressible two - fluid approach , informed by what we have learned so far . the sound speed of the gas @xmath22 and its scale height @xmath135 are assumed finite , with the dust subdisk embedded in the gas , so that @xmath136 and @xmath137 . as in the previous section , we average over the vertical thickness of the dust disk , and thus neglect complications arising on the smaller scales @xmath138 , such as shear instabilities and turbulence . these are included in an ad hoc way , using a turbulent mass diffusivity and enhanced dust pressure . perhaps more importantly , the gas outside the dust disk is completely neglected as far as the onset of instability is concerned . the external gas is ` inert ' both gravitationally and dynamically decoupled . the resulting model is workable but suffers the shortcomings discussed previously in section 2 . we adopt the equations listed in ti in order to describe our system : @xmath139 both dust and gas are assumed isothermal , so that @xmath140 and @xmath141 . as mentioned , by assuming that the dust and gas enclosed in @xmath142 is razor thin , we omit the gravitational influence of gas external to the dust disk . as a consequence , @xmath6 should be understood as the surface density of the gas located within the vertical extent of the dust subdisk . once again we assume a simple background equilibrium of homogeneous density and keplerian shear : @xmath61 , @xmath143 , @xmath144 . to this we add disturbances , denoted by primes , that are @xmath119 . the resulting linearised equations are : @xmath145 where the equilibrium dust - to - gas density ratio is @xmath146 for a given size @xmath5 . finally , the perturbed gravitational potential is obtained from @xmath147 the system of equations - yields a rather involved 6th order dispersion relation : @xmath148 with @xmath149 \\ & \hskip2 cm + \left[(1+\delta)^2\epsilon\omega^2+\overline{\omega}^2_g + \omega^2\right]dk^2 , \\ a_2 & = c^2k^2\overline{\omega}_g^2+\omega\mu^2 - 2c_g^2k^2\pi g\sigma_0|k| \\ & \hskip1 cm + ( 1+\delta)\epsilon^2\omega^2\left[(1+\delta)\mu^2+\delta c^2k^2-\delta c_g^2 k^2\right ] \\ & \hskip2 cm + \left[2(1+\delta)\overline{\omega}_g^2-c_g^2k^2 \right ] \epsilon\omega d k^2 , \\ a_1 & = \epsilon\omega k^2\left[\delta c_g^2(\mu^2-c_g^2k^2)+c^2(\mu^2+\delta c_g^2 k^2 ) \right ] \\ & \hskip1 cm + \left\{\overline{\omega}_g^2\left[1+\epsilon^2(1+\delta)^2\right ] -\epsilon^2\delta(1+\delta)c_g^2k^2 \right\}dk^2\omega^2,\\ a_0 & = d c_g^2 k^4\delta\epsilon\omega^3.\end{aligned}\ ] ] in order to ease the presentation of the coefficients we have introduced the following frequencies @xmath150 recall that @xmath151 is the squared frequency of density waves in the dust fluid given by eq . . in this subsection we analyse the case when mass diffusion in the continuity equation is negligible , @xmath152 , but we retain the velocity dispersion of the dust particles . this situation may adequately describe a disk of weakly or marginally coupled dust and gas , @xmath153 . for this case the dust disk is expected to be thinner and ` colder ' than the gas disk , and so @xmath154 . however , because of the omission of mass diffusion , the instability conditions derived here should be regarded as necessary , not sufficient , and maximum growth rates understood as upper bounds . when @xmath70 the dispersion relation reduces to a quintic . marginality corresponds then to @xmath155 , which provides us with a condition for the onset of instability . in fact @xmath155 is a quadratic equation for @xmath67 , @xmath156 indicating that if instability occurs it takes place on a range of wavenumbers @xmath157 , where @xmath158 and @xmath159 are solutions to the quadratic . in order for such a range to exist , the critical @xmath67 s must be real . this is the case when the discriminant of is positive , and the instability criterion proceeds easily : @xmath160 where the last scaling arises if @xmath132 . alternatively , the criterion can be reworked in terms of the dust s toomre parameter , noting that @xmath161 . we then obtain instability when @xmath162 where in the last approximation we assume that @xmath163 ( always the case , unless turbulence is absent see section 2.2 ) . however expressed , this simple criterion encapsulates clearly the main physical effects . for instance , if we take the limit of negligible dust , @xmath164 , the instability criterion become simply @xmath165 , and the system is always unstable . in this limit the dust s drag on the gas is unimportant and the the two - fluid system reproduces the same stability behaviour as the single fluid model ( section 3 ) : the toomre parameter does not feature . for general @xmath2 , however , stability in the two - fluid disk does in fact depend on the gas s ( or dust s ) toomre parameter : if it is too large then instability is switched off . if @xmath166 , the instability criterion may be written as @xmath167 which may greatly exceed the classical value of 1 , allowing instability to occur in a finite range of conditions . in this case we recover the secular gi . explicit expressions for the growth rate are possible in the limit of @xmath42 . we set aside the classical gi and isolate the sgi mode by setting @xmath168 . collecting terms of order @xmath13 , we obtain @xmath169 to leading order in small @xmath2 and @xmath23 . the last damping term arises from the particle pressure and kicks in at short scales , of order @xmath170 . instability is thus restricted to scales longer than @xmath9 . conversely , the coriolis force acting on the gas ( represented by the @xmath171 ) stabilises long wavelengths , so that instability only occurs on scales shorter than @xmath172 ( as in classical gi ) . the stabilising effect of _ gas _ pressure , on the other hand , does not make an appearance . in summary , instability occurs on a range of intermediate scales . but for this range of unstable wavelengths to exist , the dust pressure must be sufficiently weak or else the last term in swallows the first term . expression can be further simplified if we restrict our attention to lengthscales much shorter than the gas s jeans length @xmath173 , but not so short that the dust pressure dominates the other processes . in other words , we require @xmath174 . if this intermediate range exists , then on it we may approximate @xmath175 in the denominator of the first term in , and obtain @xmath176 assuming further than @xmath177 the last term may be dropped and we have an expression similar to the single fluid one , . the maximum growth rate is easy to compute : @xmath178 using the toomre parameter for the dust . this expression is consistent with eq . , for sufficiently large @xmath25 . but note that it only holds if there is a sufficient separation of scales between the dust and gas pressure scale heights , which is assured if @xmath179 . the wavelength of maximum growth is also similar to that in a single fluid @xmath180 . growth rate of the sgi as a function of radial wavenumber @xmath67 for weakly coupled particles . parameters are @xmath181 , @xmath182 , @xmath183 , and @xmath184 . here @xmath185 is the gas vertical scale height . the solid red line is the full solution to the dispersion relation and the blue dotted solution is the asymptotic growth rate from , width=340 ] in fig . [ fig3 ] we plot the asymptotic growth rate as a function of @xmath67 alongside the full solution obtained numerically from . for the parameters chosen the agreement is respectable , but naturally worsens once we leave the asymptotic regime of small @xmath2 , @xmath23 , and @xmath13 . note that the lengthscale of fastest growth is less than the gas - disk thickness @xmath186 , though much longer than the particle - disk thickness because @xmath187 . having treated the non - turbulent case , more relevant for weakly coupled particles , we now turn to a pressureless dust suspended in a turbulent gas disk , the regime that best describes small dust . thus in eqs - we set @xmath188 , but @xmath189 , and it is assumed that @xmath93 . in addition , the turbulent diffusion is presumed small , @xmath190 . the simplest way to capture the full effects of diffusion is to let @xmath191 , which ( whatever the value of @xmath30 ) will be true on some radial lengthscale . on longer wavelengths persisting with this scaling means we include harmless subdominant terms , while on shorter wavelengths we expect diffusion to quench instability in any case . the resulting equations correspond exactly to those treated in sections 2 and 3 in ti , which we now analyse in more detail and give explicit expressions for the growth rates . when @xmath99 , the onset of instability is difficult to calculate from the dispersion relation because unstable modes possess ( small ) complex frequencies . some progress can be made , however , in the limit of @xmath192 and assuming that @xmath193 ( thus extracting only sgi modes ) and @xmath194 . recall from section 2.2 that in the tight coupling limit @xmath110 . to leading order the dispersion relation becomes the quadratic @xmath195\omega^2s_1 ^ 2 -\left\{\delta c_g^2(\mu^2-c_g^2k^2 ) \right . \notag \\ & \hskip1 cm \left . + \alpha_1(1+\delta)\left[(1+\delta)\overline{\omega}^2-\delta c_g^2k^2\right ] \right\}k^2c_g^2\omega s_1 \notag \\ & \hskip2 cm + \delta c_g^4k^4\alpha_1\omega^2 = 0 . \label{quad}\end{aligned}\ ] ] in eq . ( 13 ) ti present an equivalent expression . for reasonable values of @xmath2 and @xmath196 the coefficient of @xmath197 is positive and so the sign of the growth rates can be determined from the coefficient of @xmath131 . after some manipulation the instability criterion is @xmath198 \sqrt{\frac{1+\delta}{\epsilon\alpha_g \left[\delta+\epsilon\alpha_g(1+\delta)^2\right]}},\ ] ] which agrees with eq . ( 18 ) in ti . taking small @xmath2 , this simplifies to @xmath199 which is remarkably similar to the diffusionless criterion . the smaller the turbulent diffusion @xmath33 , the greater the range of instability . diffusion on small scales has replaced dust pressure in in stabilising the secular gi , but the two processes work exactly the same otherwise . finally , note that differs from the final criteria in ti because they make additional and unnecessary assumptions regarding the relative sizes of @xmath13 , @xmath2 , and @xmath33 . solutions to the quadratic are ugly , but if we take the limit @xmath177 , we obtain @xmath200 this is almost identical to expression . the main difference is that the stabilising term on short scales arises from turbulent diffusion , rather than dust pressure , once again . the long wavelength stabilisation is the same the coriolis force . instability can only occur if these two stabilising scales are sufficiently well separated . growth rate of the sgi as a function of @xmath67 for tightly coupled particles . parameters are : @xmath201 , @xmath202 , @xmath183 , and @xmath203 . the solid red line is the full solution to the dispersion relation , and the blue dots correspond to the asymptotic growth rate .,width=340 ] in fig . [ fig4 ] we plot the asymptotic growth rate as a function of @xmath67 next to the full solution obtained numerically from . for a representative set of parameters the agreement is good , though weakens the smaller @xmath13 and the larger @xmath2 , as expected . note also that at small @xmath67 the approximation fails to capture the bifurcation into two monotonically growing modes with different growth rates . finally , the wavelength of fastest growth is @xmath204 , and thus the approximation of a razor thin disk is only marginally applicable . the approximate maximum growth rate can be obtained by a similar argument to that given in section 5.3.2 . we find to leading order on the intermediate range @xmath205 and for @xmath177 that @xmath206 this expression is similar to the growth rate in the single - fluid pressureless case , , differing only by a half ( on account of the unstable mode appearing as a complex conjugate pair ) . the maximum growth rate is staightforward to compute , and we find @xmath207 with maximum growth occuring on lengthscales @xmath208(as in the single fluid ) . note that a similar expression to ( and identical to ) can also be achieved by enforcing the terminal velocity approximation for the dust in eq . , and then radial geostrophic balance for the gas in eq . . the latter assumption , combined with the remaining equations , ensures that @xmath209 and so , to leading order , the growth rate is the same as in the single fluid treatment . to complete this section we bring together the insights gained from the preceding analyses and construct a physical model for how the secular gi works in a two - fluid disk . as is clear from the asymptotic estimates and , instability occurs only when there exists a range of intermediate wavelengths in which both the gas s tidal forces and dust pressure / turbulence are sufficiently weak . obviously , dust pressure or turbulence will interfere with gravitational collapse on small scales , stopping the clumping of material into rings . instability hence migrates to longer scales where this effect is weak . on sufficiently long scales , however , the tidal force will send a disturbed gas parcel into epicyclic motion , which also impedes clumping . to counteract this effect , instability must then move to shorter scales where the gas pressure is sufficiently strong to block the gas s epicyclic tendency ( via a zonal , or geostrophic , flow ) . criteria and describe necessary conditions that permit some band of wavelengths to satisfy these two requirements . essentially , the two - fluid mode is attempting to be as close to the single - fluid model ( section 3 ) or the two - fluid incompressible model ( section 4 ) as possible . on this intermediate range of lengthscales two dust blobs radially displaced towards one another exchange their angular momentum with the background gas , and hence can gravitationally collapse rather than be sent into epicycles . the gas , however , being no longer an infinite reservoir of angular momentum undergoes a commensurate perturbation . if the gas pressure is sufficiently strong , however , this perturbation takes the form of a zonal flow , not an epicycle , and so the structure of the instability is retained . if we move to longer and longer scales , the gas pressure weakens and so can only support very mild zonal flows , unable to carry appreciable angular momentum from the drifting dust . as a consequence , the instability is suppressed . on the other hand , whereas the gas is in radial geostrophic balance , the dust can fall into either geostrophic balance ( for @xmath128 ) or terminal velocity balance ( for @xmath93 ) . in this section we apply the stability criteria and growth rate estimates of the two - fluid model to reasonable models of protoplanetary disks . because of uncertainties in the parameters and deficiencies in the model itself , we do not attempt to be comprehensive or to hold fast to the quantitative results we obtain . rather the aim is to give a sense of the main trends and qualitative behaviour , and the general range of numbers one might find in a real system . to that purpose we first concentrate on the two populations of strongly and marginally coupled particles , using the simple estimates derived previously . we then solve the full dispersion relationship numerically at each radius . the disk model we use is a variant of the minimum mass solar nebula ( see youdin 2011 ) . the background gas disk surface density is given as a function of disk radius by @xmath210 here @xmath211 is a free dimensionless parameter , and @xmath212 is disk radius in units of au . the temperature of the nebula is given by @xmath213 as a consequence , the gas s toomre parameter is @xmath214 thus at 1 au , @xmath28 is roughly 40 and this falls to about 10 as we approach 100 au . the inverse stoke s number can be approximated by @xmath215 where @xmath216 is particle size in units of mm . we have assumed that the particles always lie in the epstein regime . only the largest particles at the smallest radii enter the stokes drag regime , so to make life simple we omit it . as a result , stability can be determined once @xmath2 , @xmath211 , @xmath33 , particle size , and the disk radius are specified . first consider larger particles with a largish stokes number @xmath217 . from figure 1 , in a standard mmsn these might correspond to @xmath218 10 cm particles at 30 au or more or @xmath219 cm particles at 100 au . in an older less massive disk , this class may also include mm particles but only at 100 au . thus most particles do not fall into this regime . secular gi arises when criterion is fulfilled . taking the standard value for the dust to gas ratio @xmath220 , we only need to estimate the ratio of velocity dispersions . assuming that the particles random motions are controlled by the background turbulence , as in section 2 , we have @xmath221 and thus instability occurs when @xmath222 this criterion includes both the classical gi of a dust layer and the secular gi , in which we are more interested . as discussed earlier , the properties of the turbulence are poorly constrained . we thus allow @xmath33 to vary between @xmath223 ( a perhaps unrealistically low level ) and @xmath224 . next , to fix ideas , we set @xmath225 and find that the critical toomre parameter for the gas is @xmath226 with the lower value corresponding to a thick disk of relatively ` hot ' particles ( @xmath227 ) , and the higher value to a thin disk of colder particles ( @xmath228 ) . our standard mmsn with @xmath14 yields @xmath28 that fall directly in this range . if @xmath227 instability is not possible , but if @xmath229 then instability can occur on most radii . less massive disks exhibit larger @xmath28 ; for instance with @xmath16 , we have @xmath230 at all radii , and thus the existence of sgi is very much conditional on the efficiency of the turbulence . if @xmath231 then cm sized particles , or even smaller , could be unstable in the outer regions of low mass disks . what of the growth rates of the unstable modes ? equation gives an upper bound on @xmath66 , in the regime of larger particles . this can be reworked into @xmath232 where the last equality comes by setting @xmath233 and @xmath234 . for relatively strong turbulence @xmath235 , the e - folding time of an unstable mode is @xmath236 orbits , too long to be relevant at 100 au , but possibly significant at smaller radii , for example 10 cm size particles at @xmath237 au . smaller @xmath33 , of course , yield faster growth on relevant timescales . we next consider well coupled particles , a class that covers most solids of interest ( see figure 1 ) . the relevant stability criterion for this dust is given by . the dimensionless diffusion coefficient for tightly coupled particles is @xmath110 . setting @xmath183 and @xmath238 yields the criterion @xmath239 which is more difficult to satisfy than in the weakly coupled case . if we assume that @xmath240 , the condition becomes @xmath241 where the larger value corresponds to inefficient turbulent diffusion ( @xmath228 ) and the lower bound corresponds to efficient diffusion ( @xmath227 ) . once again , this suggests that instability occurs when the turbulence is sufficiently weak . in fact , given @xmath14 , the sgi is suppressed if @xmath242 . the situation worsens when @xmath243 . the conclusion is that the instability may not be widespread in smaller dust . let us next turn to growth rates , in particular expression . to fix some numbers , we generously set @xmath238 and @xmath183 and after some manipulation obtain @xmath244 if @xmath234 , then we have @xmath245 while @xmath246 yields appreciable growth at all radii , @xmath247 does so only for @xmath248 au . for larger @xmath13 growth times this further reinforces the conclusion that the sgi is only relevant to the dynamics of tightly coupled particles when turbulence is very low indeed . in this subsection we solve the full dispersion relation at each radius of our disk model . figure 1 indicates that particle sizes of a mm and above couple to the gas differently at different radii , potentially passing from the well - coupled to the weakly coupled regimes as we go further out radially . at certain radii @xmath249 and our analytic results are no longer strictly valid , meaning that the dispersion relation must be solved numerically . some stability curves are plotted in fig . [ fig5 ] for two different disk models , @xmath14 ( left panels ) and @xmath16 ( right panels ) , for several values of the turbulence parameter @xmath33 , and for two values of the dust to gas density ratio , @xmath183 ( top row ) and @xmath250 ( bottom row ) . parameter regions above a given curve are subject to instability , and thus for a given @xmath33 and particle size there exists a critical radius within which the sgi is completely suppressed . in fact , for a relatively turbulent disk with @xmath251 , @xmath14 and @xmath183 , all particles smaller than @xmath252 cm are stable , and all particles smaller than 1 mm , when @xmath253 . weaker levels of turbulence , of course , permit instability upon smaller particle sizes and for larger swathes of the disk . but one must drive @xmath33 to levels @xmath254 to obtain sgi at radii @xmath255 au for particles larger than a mm . the sgi s struggles worsen when the disk is less massive ( @xmath16 ) , with mm sized particles unstable only for very low values of @xmath33 . but moving to the lower row of plots , it is immediately clear that increasing @xmath2 improves its range . for example , if @xmath14 and @xmath251 then all particles smaller than mm sizes are stable . if @xmath253 then all particles smaller than @xmath256 mm are stable . on the other hand , when @xmath257 the prospects for instability become increasingly bleak . maximum e - folding times for the sgi at different radii in a standard disk model with @xmath14 and dust - to - gas ratio @xmath183 . the solid curves correspond to cm sized particles , and the dashed curves to mm sized particles . the blue colour represents a turbulent mass diffusivity of @xmath253 , red represents @xmath258 , and green @xmath228.,width=340 ] we also compute the minimum e - folding times for the sgi , for @xmath14 and @xmath183 . these are plotted in figure [ fig6 ] . solid curves correspond to cm sized particles and dashed curves correspond to mm sized particles . we omit smaller particles because they are always in the well - coupled regime treated in section 6.2 . the different colours represent different values of @xmath33 . as is clear , a turbulence level of @xmath259 yields a growth time too long to be important for both particle classes , while a value of @xmath258 yields a growth time of a few @xmath260 years for cm sized particles , at a large range of radii @xmath261 au . milimetre sized particles exhibit appreciable growth only for low turbulence levels @xmath262 and at larger radii @xmath263 au . less massive disks ( such as with @xmath16 ) yield even weaker growth . increasing @xmath2 to 0.1 , exacerbates growth by up to an order of magnitude , whereas decreasing @xmath211 to 0.1 , reduces the growth rate by roughly an order of magnitude . these growth times are on the whole consistent with youdin ( 2011 ) at radii @xmath261 au , but closer to the disk two - fluid effects lead to noticeable departures . the growth times diverge at certain critical radii . these occur , of course , when the sgi is stabilised , as indicated by fig . [ fig5 ] . for a given @xmath33 , any given particle size has a critical radius within which the sgi is stable . the takeaway message is that mm sized particles require very low levels of turbulence to be sgi unstable , and then this is localised to the outer parts of the disk . centimetre sized particles do better , and the sgi may play some role in their dynamics across a range of disk radii and disk properties . finally , older , less massive disks struggle to host the sgi in any form , though a larger dust to gas ratio can mitigate this to some degree . unfortunately , like @xmath33 , the parameter @xmath2 is difficult to constrain . in this paper we have explored the secular gravitational instability ( sgi ) using a simple two - fluid model . despite the complexity of its associated sixth order dispersion relation , analytic stability criteria and growth rates can be obtained in the two limits of weakly and strongly coupled particles . we find that on sufficiently long and short radial scales the sgi is stabilised ; the existence of an unstable range of intermediate scales leads to an explicit instability condition involving the gas s toomre parameter , a distinctive feature of the two - fluid sgi , as opposed to the single - fluid version . the mathematical analysis suggests a straightforward way to understand the instability mechanism . the sgi favours intermediate scales upon which stabilising dust pressure or turbulence is weak , but upon which the gas pressure is strong . the latter condition permits the gas to fall into geostrophic balance : hence when the gas is azimuthally accelerated by the dust drag , it will form a zonal flow rather than undergo epicycles that would disrupt the radially collapsing dust . an assessment of the prevalence of sgi in real disk models is handicapped by uncertainties in two parameters , the strength of the turbulence @xmath33 , and the mass ratio of a certain species of dust to the gas within the dust subdisk , @xmath2 . starting with a fiducial value of @xmath183 , we find that a moderate level of turbulence @xmath253 prohibits the sgi on most radii , and when it does occur it grows too slowly @xmath264 years the timescale of the large - scale evolution of the disk , and of appreciable radial drift . weaker turbulence @xmath258 permits growth for cm sized particles on radii @xmath265 au , with efolding times of a few @xmath260 years . smaller sized particles may be subject to sgi but grow too slowly . it is only for @xmath228 that mm sized particles sustain growth at reasonable levels , and then only for @xmath266 au . increasing @xmath2 improves the situation , of course , and @xmath267 might be the case for particularly well - settled and populous subclasses of particle , though further work is needed to better constrain this parameter . even so , if @xmath268 it may be prove difficult for the sgi to meaningfully impose itself on the disk dynamics . we also discuss the various shortcomings of the razor - thin model we employ , which is especially a problem when the dust and gas disks exhibit different scale thicknesses . these issues no doubt impact quantitatively on our results , but the main qualitative conclusions and our picture of instability should be robust . they can be checked with a suitable vertically stratified analysis akin to mamatsashvili & rice ( 2010 ) and lin ( 2014 ) , which will also provide more reliable quantitative estimates on the stability curves and growth rates . our results extend previous analyses of the sgi , and for larger radii are in relative agreement with youdin ( 2011 ) and shariff & cuzzi ( 2011 ) . a notable difference is that the two - fluid model prohibits sgi on radii less than a critical radius . as a result , the sgi is certainly unviable on radii @xmath269 au , and possibly absent on radii @xmath270 au , the expected regions of planet formation . the prospects for sgi in the cm class of particles on disk radii @xmath237 au are reasonable as long as gas turbulence is not too efficient @xmath271 . the instability could then be an important route by which large aggregates could form further out , leapfrogging the entire range of difficult cm to km sizes . note that our results are only for axisymmetric instability . it is likely , via analogy with classical gi , that non - axisymmetric sgi occurs for larger @xmath28 , in which case our stability curves may need some revision . it has been hypothesised that the sgi generates observed dust ring structures at larger radii in protoplanetary disks ( ti ) . as discussed in section 6 , however , the sgi has great difficulty on radii @xmath265 au for small particle less than a cm in size . the dust - to - gas ratio @xmath2 needs to be increased , and @xmath28 taken to levels approaching 1 in order to obtain instability . while it may be possible to justify increasing @xmath2 , such a low @xmath28 would mean the gas disk is marginally unstable to classical gi . perhaps a more important point is that , while the linear phase of the sgi evolution is axisymmetric , its nonlinear phase will most likely involve a non - axisymmetric breadown into disordered flow , as in classical gi , not the formation of large - scale quasi - steady rings . dedicated nonlinear simulations are required to test what dynamics the sgi exhibits once it reaches nonlinear amplitudes , and how readily it forms planetesimal clumps . this forms the basis of future work . the authors thank the reviewer , dick durisen , for a set of comments that led to a much improved manuscript . hnl acknowledges partial funding from stfc grant st / l000636/1 , and rr from a bridgewater summer internship and from newnham college .

numerical simulations of the interiors of radiation dominated accretion disks show that significant density inhomogeneities can be generated in the gas . here , we present the first results of our study on x - ray reflection spectra from such heterogeneous density structures . we consider two cases : first , we produce a number of toy models where a sharp increase or decrease in density of variable width is placed at different depths in a uniform slab . comparing the resulting reflection spectra to those from an unaltered slab shows that the inhomogeneity can affect the emission features , in particular the fe k and ly lines . the magnitude of any differences depends on both the parameters of the density change and the ionizing power of the illuminating radiation , but the inhomogeneity is required to be within thomson depths of the surface to cause an effect . however , only relatively small variations in density ( on the order of a few ) are necessary for significant changes in the reflection features to be possible . our second test was to compute reflection spectra from the density structure predicted by a simulation of the non - linear outcome of the photon bubble instability . the resulting spectra also exhibited differences from the constant density models , caused primarily by a strong 6.7 kev iron line . nevertheless , constant density models can provide a good fit to simulated spectra , albeit with a low reflection fraction , between 2 and 10 kev . below 2 kev , differences in the predicted soft x - ray line emission result in very poor fits with a constant density ionized disk model . the results indicate that density inhomogeneities may further complicate the relationship between the fe k equivalent width and the x - ray continuum . calculations are still needed to verify that density variations of sufficient magnitude will occur within a few thomson depths of the disk photosphere .

astro-ph0311390

the reprocessing of hard x - rays in relatively cold , optically thick material is a major component of the current x - ray phenomenology of active galactic nuclei ( agn ) and galactic black hole candidates ( gbhcs ) . the evidence for this is based on the shape of the observed spectral continuum . starting with _ ginga _ observations in the late 1980s and continuing on with data collected from _ asca _ , _ rxte _ and _ _ beppo__sax , it was found that the continuum of almost all type 1 seyfert galaxies exhibit a spectral hardening beyond @xmath210 kev @xcite before rolling over at energies @xmath3 kev @xcite . this fact combined with the near ubiquitous presence of a fe k@xmath0 line @xcite is completely consistent with the predictions of compton reflection of x - rays in dense , optically thick material @xcite . moreover , the equivalent width ( ew ) of the fe k@xmath0lines are usually large enough ( ew @xmath2 50300 ev ) that the reflector must cover about half the sky as seen from the x - ray source @xcite , but remain out of the line of sight ( in order to not produce significant absorption ) . in the complex environment of the central engine of an agn , there are many potential locations for the reprocessing gas , such as broad - line region clouds , and the obscuring material of the unification schemes . indeed , reflection from one or both of these sites may be a common occurrence in seyfert galaxies ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? however , in at least some objects ( most famously , mcg6 - 30 - 15 ; e.g. , @xcite ) it was found that the fe k@xmath0 line exhibits a broad red - wing , consistent with originating from the inner regions of the accretion disk @xcite . thus , in these sources , a detailed comparison of the x - ray spectrum with models of compton reflection may yield information on the structure , metallicity , and ionization state of the accretion disk at small schwarzschild radii ( @xmath4 , where @xmath5 is the black hole mass ) . evidence for broad fe k@xmath0lines in gbhcs is becoming more common ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? ? * ; * ? ? ? * ) , so the possibility of performing similar analyses also exist for the galactic accreting black hole systems . calculations of x - ray reflection from optically - thick material have increased in sophistication over the last decade . the initial monte - carlo calculations by @xcite and @xcite assumed a neutral , constant density slab , and were able to quantify the observed strength of the fe k@xmath0 fluorescence line as a function of viewing angle and abundance . soon after , @xcite and @xcite allowed the gas to be ionized by the incoming x - rays and produced reflection spectra that included recombination lines and edges as well as fluorescence lines . @xcite studied the effects of ionization on the fe k@xmath0 line , while @xcite clearly showed that the entire reflection continuum changed shape as a function of the ionization parameter ( @xmath6 , where @xmath7 is the incident x - ray flux , and @xmath8 is the density of the reprocessor ) . more recently , the assumption of constant density slabs has been replaced with more complicated density distributions , such as hydrostatic balance @xcite or constant pressure @xcite . the major difference between these variable density models and the constant density ones is that under certain conditions ( when the compton temperature of the radiation field is very high ) the illuminated atmosphere can be subject to a thermal instability which can split the gas into an outer , low density , completely ionized zone and a deeper , high density , completely recombined zone ( cf . * ) . when this instability occurs the reflection spectrum exhibits only neutral features formed in the lower layer that are slightly broadened due to compton scattering by the hot surface layer otherwise , the reflection spectra looks much like a diluted version of an ionized constant density model @xcite . the features are weaker due to both compton broadening by the hot surface gas , and the lower emissivity ( proportional to density ) of the emitting material . over the last few years it has been possible to fit models which self - consistently predict both the reflection continuum and line emission to real x - ray data . the constant density models of @xcite have been applied to both narrow - line and broad - line seyfert 1 galaxies @xcite as well as gbhcs @xcite . while these are not as sophisticated as the hydrostatic models , they are relatively quick to compute and provide a good parameterization of the data which allows easy comparison between objects . on the whole , when combined with a relativistic blurring function , the models do provide a reasonable description of the hard x - ray spectrum for many accreting black holes . the exceptions to this , such as the narrow - line seyfert 1 1h 0707 - 495 @xcite and mcg6 - 30 - 15@xcite may indicate that more complicated geometries are required than a simple flat accretion disk . as interesting as these fits are , a lingering question remains : are the constant density or hydrostatic assumptions representative of real accretion disks ? where radiation pressure exceeds gas pressure , as likely occurs in the inner regions of agn disks accreting at more than 0.3% of the eddington rate @xcite , models with stress assumed proportional to total pressure are viscously @xcite , thermally @xcite , and convectively @xcite unstable . these instabilities may lead to space or time variation in disk structure . furthermore , the stresses driving accretion in ionized , differentially - rotating disks are now thought to be due to magnetic forces in turbulence resulting from the magneto - rotational instability or mri @xcite . in mhd simulations , gas motions in the turbulence vary over an orbital period @xcite , which is similar to the timescale needed for establishing vertical hydrostatic balance @xcite , therefore , hydrostatic reflection models may not accurately represent the structure of turbulent accretion disks . simulations of the turbulence which include effects of radiation diffusion indicate that density fluctuations may exceed an order of magnitude @xcite . radiation - supported disks with a magnetic field are subject to another dynamical instability , the photon bubble instability or pbi @xcite . photon bubble modes having wavelengths shorter than the gas pressure scale height can grow faster than the orbital frequency @xcite . the non - linear development of the instability may lead to trains of shocks propagating through the disk surface layers @xcite , leading to a way of producing super - eddington luminosities @xcite . density contrasts between shocked and inter - shock regions can exceed one hundred ( n.j . turner et al . 2004 , in preparation ) . will density fluctuations caused by the mri and pbi affect the reflection spectrum ? @xcite and @xcite suggested that they may strengthen the emission features and thus produce an apparent large reflection fraction . but several unknowns remain : since the reflection features are formed within 10 thomson depths ( @xmath9 ) of the surface , what is the magnitude of any density change and where must it lie in the illuminated gas in order to alter the observed reflection spectrum ? do the accretion disk simulations produce the necessary fluctuations ? the answers to these questions are crucial to the interpretation of x - ray spectra of accreting black holes . in this work , we investigate the properties of x - ray reflection spectra from heterogeneous atmospheres . this paper ( paper 1 ) first considers a toy model where a density cut of variable depth and width has been sliced out of a constant density slab ( [ sect : toy ] ) . this will allow a systematic investigation of the effects of a simple density inhomogeneity . we will then consider a more complicated structure produced by calculations of the photon - bubble instability in a radiation - dominated atmosphere with no turbulence ( [ sect : bubbles ] ) . reflection spectra calculated from all these structures will then be compared with the standard constant density ones . in [ sect : discuss ] we discuss the results from this paper and form our conclusions in [ sect : concl ] . in a companion paper ( paper 2 ) , we apply the knowledge gained here to reflection by simulated radiation dominated accretion disks . the simplest test to see if density inhomogeneities affect reflection spectra is to insert a discontinuous jump or drop in density into an otherwise constant density model . in this section , we experiment with placing such steps with different widths at various places in a uniform slab . the calculations used the reflection code described by @xcite and @xcite , and the reader is referred to those papers for the details on the computational procedures . the models are computed by solving the coupled equations of radiative transfer and thermal and ionisation balance in one spatial dimension , using the two - stream approximation for the incoming radiation and a fokker - planck / diffusion method for the emergent spectrum . the solar abundances of @xcite were employed in all the models , as was an incident power - law radiation field ( defined from 1 ev to 100 kev ) with photon index @xmath10 . the following ions were treated : , , , , and . each model started with a constant density slab ( @xmath11 @xmath12 ) defined to have a total thomson depth of @xmath13 . at this optical depth almost all of the high energy photons will interact at least once with the gas . beginning at a depth @xmath14 the density was then multiplied by a factor @xmath15 to create an artificial drop or jump in the density profile at that position ( @xmath9 remained fixed at 6.1 by adjusting the physical width of the altered zones ) . the increase or decrease in density was done discontinuously : there are no zones with densities intermediate between @xmath8 and @xmath16 . the width of the step was arbitrarily defined as @xmath17 , where @xmath18 is a constant . this results in the steps becoming wider as they are placed deeper which potentially allows for stronger observable effects to the reflection spectra . we chose to calculate reflection spectra for 4 different ionization parameters : @xmath19125 , 250 , 500 , & 1000 erg cm s@xmath20 . these values bracket the transition from a neutral 6.4 kev fe k@xmath0 line to an ionized he - like 6.7 kev line @xcite . the soft x - ray features in the reflected emission , such as ly@xmath0 , also vary greatly over this range of @xmath21 @xcite . for each value of @xmath21 , reflection spectra were calculated for steps placed at @xmath22 0.1 , 0.5 , 1 , 1.5 & 2 , with @xmath15 ranging from 10@xmath23 , 10@xmath24 , 0.1 , 0.5 , 2 & 10 . the width of each step was controlled by the value of @xmath18 , and for each @xmath25 pair , models were computed for @xmath26 0.2 , 0.5 , 1 & 2 . thus , 120 different types of density steps were calculated for each @xmath21 . all but two models converged successfully , and the resulting reflection spectra were compared to an unaltered constant density model to assess the significance of any changes ( [ sub : toyres ] ) . as an illustration of the technique , figure [ fig : tempanddens ] plots the temperature and density profiles as a function of @xmath9 for eight @xmath27 erg cm s@xmath20 models . the results are shown for @xmath28 in fig . [ fig : tempanddens](a ) and @xmath29 in fig . [ fig : tempanddens](b ) . in each plot the dotted lines show the density profiles for four different steps ( occurring at @xmath30 & @xmath31 ) inserted into the slab , while the solid lines denote the final temperature profile . all the steps shown here have a width of @xmath32 ( i.e. , @xmath33 ) . the gas temperature reacts as expected : it increases by a factor of few within a density drop , and decreases within a jump . however , the effect of an increase in @xmath8 seems to have a greater impact on the overall temperature profile than a sudden drop in density . [ fig : tempanddens](a ) shows that outside of the step , the temperature profiles are roughly similar , and indeed closely resembles the one from a constant density model ( not shown ) . on the other hand , fig . [ fig : tempanddens](b ) shows that the temperature profiles are quite different for the four models . the differences between these two cases illustrates the major difference between an increase or a decrease in density . if the density changes by a factor @xmath15 then the ionization parameter in that region is changed by @xmath34 . when @xmath35 , the gas in the altered zones is more easily ionized which raises the temperature , but , most importantly , the recombination and free - free emissivity rates ( which scale as @xmath36 ) are reduced . it is this increase in cooling when the gas becomes denser which has the largest impact on the temperature profiles seen in fig . [ fig : tempanddens](b ) . however , the ionized gas in a low density region may be a strong emitter in some energy bands ( e.g. , the 6.7 kev fe k@xmath0 line ) . furthermore , a decrease in density will also lower the opacity to radiation emitted from below . therefore , we now turn to see how the impact of a sudden density change impacts the reflection spectrum . as a first step in determining the effects of a step in density on the reflection spectrum , we simply overplot the predicted emission with that from a constant density slab under the same illumination conditions . four examples of this comparison , one for each value of @xmath21 , are shown in the upper panels of figure [ fig : comparespect ] . since in some cases any differences between the two spectra are difficult to see by eye , we plot in the lower panels the ratio between the constant density and the inhomogeneous model spectra . prior to the ratio being calculated the @xmath10 power - law was added to each spectrum so that each had a reflection fraction of unity . this was done in order for the plot to more accurately represent the ratio of two observed spectra ( in agn & gbhcs the power - law continuum is observed along with the reflected emission ) . figure [ fig : comparespect ] , while only showing four examples , does demonstrate a number of useful points . first , although there are changes , the reflection spectra from the toy inhomogeneous models are not dramatically different from the constant density results ; that is , they still look qualitatively like the spectra first presented by @xcite . secondly , the three variables regarding the density step ( its placement , depth / height , and width ) , as well as the value of @xmath21 , all affect the deviations observed in the spectra . for example , the step illustrated in the @xmath27 erg cm s@xmath20 panel of fig . [ fig : comparespect ] , which corresponds to one of the models shown in fig . [ fig : tempanddens](a ) , results in a negligible change to the reflection spectra above @xmath37 kev . the only major differences in the reflection spectrum are a decrease in intensity of low - energy , , and lines . in this case , the drop in energy was placed too deep into the layer for it to make much difference to the reprocessed emission . however , at larger ionization parameters , the incoming x - rays penetrate further into the slab , and so a step at @xmath38 can have a large impact on the reflection spectrum . this is seen in the @xmath39 erg cm s@xmath20 panel of fig . [ fig : comparespect ] , but more dramatically in the @xmath40 erg cm s@xmath20 plot . in both these cases , the low density step alters the reflection spectra so that they have a shape that resembles ones with a larger value of @xmath21 ( less absorption , broader spectral features ) . these two plots show that the reflection spectra from a inhomogeneous slab can be a mixture of ionization parameters , similar to ones calculated from hydrostatic atmospheres @xcite . of course , when the step causes an increase in density , as is shown in the @xmath41 erg cm s@xmath20 panel , then a spectrum typical of a lower ionization parameter ( stronger absorption and low - energy emission lines ) is mixed in with the results . although fig . [ fig : comparespect ] does show that density inhomogeneities can cause observable effects in the reflection spectra , the results are largely qualitative and selective ( there are @xmath42 other possible panels ) . to better quantify the results , and to obtain a more general overview of the parameter space , we calculated the fe k@xmath0 and ly@xmath0 ews from the total ( reflected+incident ) spectrum for each two - density model and compared it to the values from the corresponding constant density models . the ews were calculated by integrating the spectrum between 6 and 7 kev for fe k@xmath0 , and between 0.6 and 0.7 kev for ly@xmath0 . figure [ fig : ratiopanel ] shows the results of this exercise , where for each @xmath21 we plot the ew ratio ( defined as two - density model / constant density ) for both the fe k@xmath0 ( black ) and ( blue ) lines as a function of @xmath14 . each plot contains 6 panels for the different values of @xmath15 , and the different style of lines within the panel denotes a particular value of @xmath18 . there is plenty of information in this figure , and it is worth going through it in detail , but before doing so there is one general point that can be made . in almost every instance the magnitude of the ew ratio shown in fig . [ fig : ratiopanel ] is proportional to the value of @xmath18 ; that is , the wider the step , the greater the change to the reflection spectrum for those values of @xmath15 and @xmath14 . starting with the @xmath41 erg cm s@xmath20 plot , we find that because of the relatively weak illumination , the density change generally only affects the ews for @xmath43 . the largest changes to the fe k@xmath0ew are when @xmath44 or @xmath45 . in the first case , the increase in ew is due to the appearance of a 6.7 kev component to the fe k@xmath0 line , while the later two instances are a result of a stronger 6.4 kev component . the reason why the ionized line appears in the @xmath46 models but not in the @xmath47 or @xmath48 cases is that the fe is almost entirely fully stripped in the very low density zones and so does not contribute much emission ( as do the other metals ) . the ew varies only slightly , except for when @xmath49 where it drops by @xmath50% due to a low ionization fraction in the overdense region . a similar pattern is seen in the @xmath27 erg cm s@xmath20 models , with the changes now occurring for @xmath51 . there are significant variations in the fe k@xmath0 ew for most values of @xmath15 , with the largest changes ( @xmath52 factor of 8) occurring in the @xmath46 case . the reflection spectrum for the @xmath27 erg cm s@xmath20 constant density model ( seen as the dashed line in fig . [ fig : comparespect ] ) has a very weak fe k@xmath0 line due to auger destruction @xcite . therefore , changes to the density which result in either a 6.7 kev line ( @xmath46 ) or a strong 6.4 kev line ( @xmath49 ) cause a significant increase in the ew . at this value of @xmath21 it seems to be difficult to increase the ew by density steps , but it can be lowered by nearly a factor of two when @xmath49 . increasing the illumination level so that @xmath39 erg cm s@xmath20 results in some changes in the reflected emission with a step down at @xmath53 , but , from fig . [ fig : ratiopanel ] , we see that the most substantial changes still occur at @xmath54 . the fe k@xmath0 ew differs significantly from the constant density value only when @xmath46 or @xmath55 . in these cases the iron in the low density zones becomes highly ionized , but not fully stripped , increasing the column of the he - like species . the line is also broader due to compton down - scattering . again , we see that the spectrum takes on the properties of one from a larger @xmath21 . this is also reflected in the changes in the ew , which is much smaller than the constant density model for @xmath56@xmath55 . however , the opposite behavior ( a larger ew and a smaller fe k@xmath0 ew ) is seen when the step causes an increase in density and low-@xmath21 features are mixed into the reflection spectrum . in most of those models , the 6.7 kev fe line and the ly @xmath0 line are reduced and enhanced in intensity , respectively , although an interesting counter - example is seen when @xmath49 and @xmath57 . in this model , the overdense region is placed at a depth so that both a 6.4 kev and a 6.7 kev fe k@xmath0 line is emitted , causing an increase in the ew . similarly , the ew falls because of a significant transition to emission . finally , the most highly ionized runs ( @xmath40 erg cm s@xmath20 ) show little variation in the fe k@xmath0 ew when @xmath35 . the only exceptions are when the density deficit is near the surface and the high ionization parameter in these zones decreases the amount of 6.7 kev emitting ions . as a result , the line ew falls by about 25% . on the other hand , when @xmath49 , the fe k@xmath0 ew can drop by over a factor of two because the overdense region stops the ionization front at a much shallower depth in the slab . the ew is always smaller than its constant density value when @xmath35 , but can increase by over a factor of 2 ( up to 110 ev ) when there is a density enhancement just beneath the surface . in summary , we find that toy models which place a top - hat function of variable width and depth into a constant density slab , can produce observable effects on the resulting reflection spectra . generally the largest impact on the spectrum , as measured by the fe k@xmath0 and ews , occurs when the step is placed at @xmath58 into the layer . furthermore , changing the density by only relatively small factors of 210 can produce significant effects . the resulting reflection spectra are reminiscent of the one produced by hydrostatic models in so far as they show features from a mixture of ionization parameters . the ew ratios shown in fig . [ fig : ratiopanel ] indicate that density inhomogeneities in the surfaces of accretion disks may have an observable effect on their reflection spectra . those measurements , however , only use a fraction of the information contained in the model spectra . in this section , we take several example reflection spectra calculated from the two - density models and fit them with ones from the constant density models . this method will allow us to quantify any differences between the two scenarios by using all the information over a given energy range . the exercise may also be indicative of how real inhomogeneities will manifest themselves when fitting data with simple models . we chose models that showed large changes in its fe k@xmath0 or ew from each of the four ionization parameters considered ( the exact model parameters are listed in the note to table [ table : stepfits ] ) . simulated observations were made with these model spectra by using the fakeit command in xspecv.11.2.0bp @xcite . the canned _ xmm - newton _ epic - pn response matrix ` epn_sw20_sdy9.rmf ` was used to generate the data assuming an exposure time of 40 ks . as with the ew calculation in the previous section , the incident power - law was added to the reflection spectra so that each simulated spectrum had a reflection fraction , @xmath59 , of unity . each faked dataset was then fit with a grid of constant density models ( calculated with the same version of the reflection code ) between 0.2 and 12 kev as well as between 2 and 10 kev . the fit parameters were the photon index @xmath60 , the ionization parameter @xmath21 , the reflection fraction @xmath59 , and an arbitrary normalization . the results are listed in table [ table : stepfits ] , and the residuals are shown in figure [ fig : stepfits ] . this particular @xmath41 erg cm s@xmath20 model was chosen because the fe k@xmath0 ew increased by over a factor of 2 due to the appearance of a 6.7 kev line . in the 0.212 kev energy band , the constant density model does a decent job of recovering the ionization parameter , but the value of @xmath59 is too low . in contrast , the best fit in the 210 kevrange has a much larger @xmath21 ( 466 erg cm s@xmath20 ) and @xmath59 . evidently , the spectrum in this energy band mimics one from a more ionized reflector . the residuals to the 0.212 kev fit are shown in fig . [ fig : stepfits ] and shows that the low value of @xmath59 is likely due to the overprediction of the line ( as seen in fig . [ fig : ratiopanel ] ) . moreover , the ionized fe k@xmath0 line can not be accounted for by this low @xmath21 . ignoring the low energy data increases the reflection fraction as the spectrum attempts to account for the enhanced fe emission . in both cases , the continuum is fit adequately , but the differences in line emission produce significant residuals . statistically , the fit to the @xmath27 erg cm s@xmath20 two - density model over the 0.212 kev band is much worse than for the previous case . the best fit ionization parameter is lower than the true one , and so overpredicts the strength of the soft emission lines between 0.3 and 0.6 kev ( fig . [ fig : stepfits ] ) . according to fig . [ fig : ratiopanel ] , the fe k@xmath0 ew in this two - density model was @xmath22 times greater than the corresponding constant density model ( although at this @xmath21 the ew is only 20 ev ) , but that the change in the ew was marginal . thus , in this case we find that the overdense region lowers the effective ionization parameter of the reflection spectrum , but this is only discernible through a fit to the soft x - ray emission . indeed , we find a very good and statistically acceptable fit to this two - density model over the 210 kev band . the best fit @xmath21 and @xmath59 are only slightly larger than their true values to account for the enhanced fe k@xmath0 line . the two - density model chosen from the @xmath39 erg cm s@xmath20 suite of calculations is another overdense model with the step appearing at @xmath57 below the surface . as discussed in [ sub : toyres ] , this model includes both 6.4 and 6.7 kev iron emission lines , as well as a weaker - than - expected line . interestingly , this model gives the worst result when fit by the constant density models , with the reduced @xmath61 . as in the @xmath27 erg cm s@xmath20 model , the poor fit is caused almost entirely by an overprediction of the soft x - ray spectral features . the majority of the 0.212 kev continuum is well fit by an ionized reflector ( the best fit @xmath21 is @xmath2630 erg cm s@xmath20 ) , but the emission lines can not be accounted for by this model , resulting in the low value of @xmath59 . the overdense region sitting just below the surface of the reflector produces features that can not be fit by a single-@xmath21 model . the constant density models do a much better job in the 210 kev band , where the best fit ionization parameter is nearly exactly the true one . however , the reflection fraction must be lowered to fit the weaker 6.7 kev fe line . there are still some residuals in this fit between 7 and 8 kev where the overdense region in the two - density model causes additional absorption than what is predicted by the constant density model . our final example considers the case where higher-@xmath21 features are mixed into the reflection spectrum . the @xmath40 erg cm s@xmath20 model used for fitting had a step with a density drop of @xmath56 at @xmath57 . the lower density material became highly ionized and weakened both the fe k@xmath0 and lines ( fig . [ fig : ratiopanel ] ) . the two constant density fits both show the effects of this dilution , as the best fit @xmath21 is greater than the true value and @xmath62 . the residuals shown in fig . [ fig : stepfits ] show that an overprediction of soft x - ray features is again the cause for the poor fit . there is also a significant difference in the curvature of the two spectra noticeable between 1 and 3 kev . this is caused by the lower absorption in the two - density model which flattens the continuum . the idealised two - density models discussed in the previous section show that inhomogeneities in accretion disk atmospheres may affect the reflection spectrum . we next examine reflection from density structures arising in hydrodynamical disk models . effects of the ionization processes on the density structure are neglected . nevertheless , the density arrangement is completely novel in terms of reflection calculations , as it is the first to be considered which did not arise from an ad hoc _ a priori _ assumption ( such as constant density or hydrostatic balance ) . the two - density results indicate that the reflection spectrum is affected by inhomogeneities within the outermost few thomson depths of the disk atmosphere . by contrast , fluctuations resulting from the mri may be strongest on scales comparable to the disk thickness @xcite . the effects of photon bubbles are therefore considered instead . the pbi grows most rapidly at short wavelengths @xcite , and in shakura - sunyaev models typically grows fastest in the disk surface layers ( n.j . turner et al . 2004 , in preparation ) . exploring x - ray reprocessing from photon bubbles is likely to be a good first step toward understanding the reflection properties of radiation - dominated disks . the photon bubble instability was first discussed by @xcite in the context of accreting magnetized neutron stars , and was shown by @xcite to occur in the inner , radiation - supported regions of magnetized accretion disks . for magnetic pressures exceeding the gas pressure , and perturbation wavelengths shorter than the gas scale height , the instability operates when disturbances in radiative flux displace gas along field lines , leading to propagating density variations which grow over time . linear growth is fastest at wavelengths shorter than the gas pressure scale height , but is absent at wavelengths with optical depths much less than unity , as these support no flux variations . an approximate criterion for instability is that photon diffusion carry a greater energy flux than radiation advected at the gas sound speed @xcite . reflection spectra were calculated using a density profile taken from a two - dimensional radiation - mhd simulation of the growth of photon bubbles in a shakura - sunyaev model disk . the simulation was carried out with the zeus mhd code @xcite and its flux - limited radiation diffusion module @xcite . opacities due to electron scattering and bremsstrahlung were included . the domain was a small patch of the disk , centered 34 @xmath63 from a black hole of @xmath64 . the patch extended 1.08 shakura - sunyaev scale - heights either side of the midplane , and its width was one - quarter of its height . the grid consisted of @xmath65 zones . the initial density and temperature profiles were those of a shakura - sunyaev model with @xmath66 , and accretion rate 12% of the eddington rate for a 10% luminous efficiency . during the simulation , no @xmath0-viscosity was applied . differential rotation was neglected , so that the mri was absent , and the material was allowed to cool by radiative losses . photons initially diffused from midplane to boundary in nine orbits . the calculation was started with a magnetic field having a pressure 10% of the midplane radiation pressure , and inclined at @xmath67 . after 3.4 orbits , the pbi had developed into trains of propagating shocks , with density contrasts of up to two orders of magnitude . the vertical density profile used in the subsequent reflection calculations was taken from an arbitrary vertical ray , and is shown in the insets to figure [ fig : bubblespect ] . as with the two - density models , this new density structure was then irradiated by a power - law ( @xmath10 ) of x - rays . in this case , there is no unique ionization parameter so instead the models are distinguished by the incident x - ray flux : @xmath68 , @xmath69 , @xmath70 , and @xmath71 erg cm@xmath24 s@xmath20 . the shakura - sunyaev flux for the assumed disk parameters is @xmath73 erg cm@xmath24 s@xmath20 . all four models converged , and the resulting reflection spectra are presented in figure [ fig : bubblespect ] . the gas temperature and number density are plotted in the insets to each panel , and illustrates how the temperature structure for each case was affected by the density profile . the latter three spectra all exhibit strong 6.7 kev emission lines from he - like iron , while , unusually , the most weakly illuminated model produces both a 6.4 and a 6.7 kev line of about equal strength . the rapid drop in density below one thomson depth causes the surface layers to be easily ionized and able to produce a 6.7 kev emission line even with a relatively small incident flux . on the other hand , the overdense region at @xmath74 produces a lot of soft x - ray line emission , even for highly illuminated situations . thus , the emission below @xmath31 kev is consistently important , despite the incident flux varying by an order of magnitude . to quantify any effects the photon bubbles may have on the reflection spectra we fit the models with a grid of constant density spectra . the procedure was exactly the same as with the two - density models in [ sub : toycomp ] : the power - law was added to the photon bubble spectra and then the sum was used as the basis for a 40 ks _ xmm - newton _ simulation . as before , constant density fits were performed in both the 0.212 kev and 210 kev energy bands . a constant density grid with @xmath75 @xmath12 was used for fitting the photon bubble reflection spectra in order to offset any changes in the reflection spectra due solely to the lower density . the best fit parameters for each value of @xmath7 are listed in table [ table : bubblefits ] , and the residuals are shown in figure [ fig : bubblefits ] . the constant density models had difficulty fitting the spectra over the wide 0.212 kev energy range , with most of the trouble arising from the soft x - ray lines . the best @xmath76 was found with the @xmath68 erg cm@xmath24 s@xmath20 model , where one ionization parameter seemed to provide a good fit to the spectrum at energies @xmath77 kev . the best fit @xmath21 was indicative of an ionized slab , and , indeed , the model could not account for the small 6.4 kev line predicted by the spectrum . despite this problem , the fit was able to recover the correct values of both @xmath60 and @xmath59 . the fit had difficulty below 0.5 kev , however , where it overpredicted the emission . this is most likely a result of extra absorption ( over that predicted for the fitted @xmath21 ) due to the density enhancement just beneath the surface in the photon bubble model . fits to the remaining three models in this energy band all resulted in reduced @xmath78 . a single ionization parameter could not simultaneously account for the shape of the continuum , and the strong soft x - ray and 6.7 kev fe k@xmath0 lines . the best fit values of @xmath59 were always less than the ` true ' value of unity . in these cases the photon bubble models predicted significant emission from a variety of ionization parameters , so that the resulting spectrum exhibited a mixture of features . as a result , a single ionization parameter fit was not adequate . the fits to the photon bubble models improved markedly in the 210 kev energy band with @xmath21 increasing substantially in order to fit the fe k@xmath0line . as a result , all four fits were statistically acceptable . except for the @xmath68 erg cm@xmath24 s@xmath20 model , the reflection fractions again are still underestimated by the constant density fits with the best fit value of @xmath59 decreasing with the @xmath7 of the model . a greater than expected amount of absorption is the likely explanation for this effect . the density enhancement at @xmath79 in the photon bubble model causes more continuum absorption from oxygen then what the reflection models expect ( given , e.g. , the highly ionized fe k@xmath0 line ) . this enhancement becomes more important as @xmath7 is increased and spectral features are formed both above and below it . the extra absorption decreases the magnitude of the continuum and emission , and thus results in a lower @xmath59 when fit with the constant density models . aside from the low value of @xmath59 , reflection from a uniform slab does a good job describing the photon bubble spectra between 2 and 10 kev . currently , the interpretation of the reflection continuum observed in the x - ray spectra of agn is the primary means to glean information regarding the accretion geometry and its radiative environment . it is therefore vital that the models of reflection , on which such conclusions are based , explore a number of different physical situations which may be relevant to accretion physics . motivated by the inhomogeneous nature of recent numerical simulations of radiation - dominated accretion disks , we have begun a study of examining the consequences of such density changes to the x - ray reflection spectrum . our first results , presented in the previous sections , considered reprocessing from a two - density model , where density steps or jumps were inserted into a uniform slab , and a slice from a pbi simulation . the new reflection spectra were then compared with ones calculated assuming a constant density atmosphere , as these are most frequently used in data analysis . we found that the density inhomogeneities can result in observationally important differences between the two cases . as in models with a hydrostatic atmosphere , the reflection spectra can no longer be described by a single ionization parameter , but exhibit features from a mixture of ionization states . however , while in the hydrostatic case the spectra can be adequately described as a diluted constant density model because of a diffuse hot scattering layer on the surface @xcite , the inhomogeneous models can show emission from both low and high @xmath21 material simultaneously , depending on the nature of the density change . this gives the spectra a complexity and richness that the hydrostatic models lacked . the magnitude and characterization of the change in the reflection spectrum depends on both the nature of the inhomogeneity and the strength of the incident x - rays . for the two - density models , where the calculations could be distinguished by the original ionization parameter @xmath21 of the uniform slab , the greatest impact on the fe k@xmath0and ly@xmath0 ews occurred when @xmath80 for @xmath81 . for larger ionization parameters , steps deeper into the gas could make an impact . this correlation is not expected to continue indefinitely however . for example , if an atmosphere had a drop in density at @xmath82 and it was illuminated to the extent that gas was ionized down to this depth , then the additional line emission or absorption caused by the inhomogeneity would be smeared out by compton scattering while escaping the layer . thus , a general conclusion seems to be that density inhomogeneities must be within 2 thomson depths of the surface to have any impact on the reflection spectrum . however , there must also be a lower - limit to the depth of any density change in the reflecting medium for it to be effective in altering the spectrum . in the extreme case of cold reflection ( e.g. , @xmath83 ) ionization effects are important only at @xmath84 from the surface . a density inhomogeniety at such a depth would change such a small amount of gas that it would have a negligible effect on the resulting spectrum . the results of the two - density models also indicate that relatively small changes in density ( say , @xmath85few ) can alter the reprocessed emission . again , this depends on the effective ionization parameter of the layer , but , as was seen in [ sect : bubbles ] , underdense regions beneath the surface of a moderately irradiated atmosphere can result in a substantial 6.7 keviron line . but , if that region was underdense but a large amount , such as 1001000 times lower , then it will be ionized to such an extent that it has very little effect on the outgoing spectrum . similarly , an overdense region below the surface may enhance the 6.4 kev fe k@xmath0line . analogous arguments also apply to the soft x - ray emission lines such as . how realistic are these requirements on the density inhomogeneities ? it certainly appears plausible , even likely , that accretion flows will naturally generate density contrasts greater than a few , especially in the radiation dominated regime where the photon bubble instability can enhance already existing clumpiness . what remains unknown is if the heterogeneous nature of the flow exists to small enough scales that sufficient inhomogeneity remains within a couple of @xmath9 from the disk photosphere . only very high resolution simulations of this region of accretion disks will be able to definitively answer that question . assuming for the moment that the inhomogeneities discussed in this paper occur in reality , what are the immediate observational consequences ? perhaps the most interesting result is that the density jumps can cause the fe k@xmath0 line to vary in a way that is _ completely disconnected from the x - ray continuum_. one of the most well known puzzling properties of the possible fe k@xmath0 ` disklines ' is their lack of response to the variable continuum ( e.g. * ? ? ? . the line flux typically does vary , but it is often not directly correlated with the continuum @xcite . one very striking example , which was presented by @xcite , is an _ xmm - newton _ observation of mkn 841 that was split into two parts , separated by about 15 hours . these authors found that the ew of the narrow fe k@xmath0dropped by @xmath86 between the two observations while the continuum changed by @xmath87@xmath88% . although , we are not explicitly modeling this source , fig . [ fig : ratiopanel ] shows that density steps in the surface can give rise to exactly such changes in the fe k@xmath0 ew , completely independently of the x - ray continuum . a clear way to test if this is the correct model would be to examine the changes of other emission features in the spectrum , but this would require much more sensitive data . the combination of density inhomogeneities with the non - monotonic evolution of the fe k@xmath0 ew due to ionization effects @xcite may naturally result in a poor correlation between the observed flux and the fe k@xmath0 ew . however , since the observed x - rays must be averaged for many kiloseconds before spectral analysis , which is longer than the timescale for a change in the density inhomogeneities , as well as the the ionization / recombination timescale , a relationship between the line and continuum may still be uncovered , but at a weaker level than what may have been previously expected . if the density inhomogeneities beneath the photosphere are relatively small in scale , which should be true for thin disks ( where @xmath89 , and @xmath90 is the disk scale height and maximum size of any density fluctuations ) , then any resulting enhancement or change to the fe k@xmath0line would only occur to a narrow part of the overall line profile , which is determined by the illuminating emissivity . thus , the appearance of narrow fe k@xmath0 lines , such as those recently inferred from observations of ngc 3516 @xcite , ngc 7314 @xcite , and mrk 766 @xcite could conceivably originate from density inhomogeneities in the disk . variable fe k@xmath0 lines from turbulent accretion disks have also been considered by @xcite . these authors assumed the line emissivity was proportional to the local integrated magnetic stress in their numerical simulations , as opposed to our method of assuming a constant illumination and a variable density structure . in reality , a mixture of the two effects would be expected to be ongoing , which should be investigated in future work . the strength of the soft x - ray emission lines are also affected by density inhomogeneities . these lines are one of the dominant sources of cooling for the x - ray heated gas when it reached @xmath91 k. thus , if overdense or underdense regions are introduced into the layer it can alter the rapidity at which the gas cools , thereby changing the emission features . interestingly , these changes may occur only in the soft band , with little effect at higher energies . soft x - ray lines , especially the ly@xmath0 lines of , and , have received some attention recently with the claims that they have been observed to be relativistically broadened in the gratings spectra of some seyfert 1s @xcite . the exact strength of the lines have been the subject to some debate in the literature @xcite , with the difficulties arising because complex warm absorption features must be taken into account in the spectral modeling . the results from the two - density models show that density inhomogeneities can both enhance and diminish the ews of the soft x - ray lines . the constant density fits to both the two - density models and the spectra computed from the photon bubble structure resulted mostly in reflection fractions much less than the true value of unity . if one interprets @xmath59 as a measure of the solid angle subtended by the disk as seen from the x - ray source , then @xmath92 would indicate some form of truncated accretion flow ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? correctly interpreting a fitted value of @xmath59 is fraught with difficulty since fitting neutral reflection models to ionized accretion disks will also give @xmath92 even if @xmath93 @xcite . here , we have shown that ionized disk models can produce erroneous @xmath59 values when there are density inhomogeneities . it is therefore important that other arguments be used ( e.g. , * ? ? ? * ) before drawing conclusions on the accretion geometry from a fitted value of @xmath59 . our conclusions from this first paper in our study on reflection from heterogeneous accretion disks can be summarized as follows : 1 . the reflection spectrum from accretion disks can be altered by clumps or voids in the gas just beneath the photosphere . the effects range from negligible to large and depend on the structure of the inhomogeneity and the illuminating continuum . the greatest effects on the reflected emission for most ionization parameters occurs when the density change is within @xmath86 thomson depth of the surface . 3 . the change in density does not have to be large , even an increase or a decrease by a factor of a few can significantly alter the spectrum from that of a constant density model . density inhomogeneities beneath the disk surface are a possible explanation for the apparent disconnectedness between the fe k@xmath0 line and the x - ray continuum . 5 . the soft x - ray emission lines ( below 1 kev ) are more sensitive to the presence of an inhomogeneity than the harder emission . fitting the spectra produced from models with inhomogeneities with constant density models shows that a mixture of ionization parameters are present , which often results in reflection fractions smaller than unity . constant density models may still be a good means of parameterizing reflection spectra , but only for energies @xmath94 kev . they will have much more difficulty with broadband spectra that cover the soft x - ray lines emitted from the disk . however , in practice this may only be a problem for sources with weak or non - existent warm absorbers . in paper 2 , we continue our investigation by computing reflection spectra from the predictions of 3-d radiation dominated accretion disk simulations . future work will also include a consideration of the dynamical effects of the incident x - rays and multi - dimensional reflection . drb acknowledges financial support by the natural sciences and engineering research council of canada , and thanks all the staff at kitp for their hospitality during his visit . this research was supported in part by the national science foundation under grant no . phy99 - 07949 . armitage , p.j . , & reynolds , c.s . , 2003 , , 341 , 1041 arnaud k.a . , 1996 , in jacoby g. , barnes j. , eds , astronomical data analysis software and systems v , asp conf . vol . 101 , 17 arons , j. , 1992 , , 388 , 561 balbus , s.a . & hawley , j.f . , 1991 , , 376 , 214 balbus , s.a . & hawley , j.f . , 1998 , rev . phys . , 70 , 1 ballantyne , d.r . & ross , r.r . , 2002 , , 332 , 777 ballantyne , d.r . , iwasawa , k. & fabian , a.c . , 2001a , , 323 , 506 ballantyne , d.r . , ross , r.r . & fabian a.c . , 2001b , , 327 , 10 ballantyne , d.r . , ross , r.r . & fabian a.c . , 2002 , , 336 , 867 ballantyne , d.r . , vaughan , s. & fabian , a.c . , 2003 , , 342 , 239 barrio , f.e . , done , c. & nayakshin , s. , 2003 , , 342 , 557 begelman , m.c . , 2001 , , 551 , 897 begelman , m.c . , 2002 , , 568 , l97 bisnovatyi - kogan , g.s . & blinnikov , s.i . , 1977 , , 59 , 111 blaes , o. & socrates , a. , 2001 , , 553 , 987 blaes , o. & socrates , a. , 2003 , , 596 , 509 boller , th . , fabian , a.c . , sunyaev , r. , trmper , j. , vaughan , s. , ballantyne , d.r . , brandt , w.n . , keil , r. & iwasawa , k. , 2002 , , 329 , l1 branduardi - raymont , g. , sako , m. , kahn , s.m . , brinkman , a.c . , kaastra , j.s . & page , m.j . , 2001 , , 365 , l140 chiang , j. , reynolds , c.s . , blaes , o.m . , nowak , m.a . , murray , n. , madajski , g. , marshall , h.l . & magdziarz , p. , 2000 , , 528 , 292 collin , s. , coup , s. , dumont , a .- , petrucci , p.o . & raska , a. , 2003 , , 400 , 437 de rosa , a. , piro , l. , fiore , f. , grandi , p. , maraschi , l. , matt , g. , nicastro , f. & petrucci , p.o . , 2002 , , 387 , 838 done , c. & ycki , p.t . , 1999 , , 305 , 457 done , c. & nayakshin , s. , 2001 , , 546 , 419 dumont , a .- m . , czerny , b. , collin , s. & ycki , p.t . , 2002 , , 387 , 63 eracleous , m. , sambruna , r. & mushotzky , r.f . , 2000 , , 537 , 654 fabian , a.c . , rees , m.j . , stella , l. & white , n.e . , 1989 , , 238 , 729 fabian , a.c . , iwasawa , k. , reynolds , c.s . & young , a.j . , 2000 , , 112 , 1145 fabian , a.c . , ballantyne , d.r . , merloni , a. , vaughan , s. , iwasawa , k. & boller , th . , 2002a , , 331 , 35 fabian , a.c . , vaughan , s. , nandra , k. , iwasawa , k. , ballantyne , d.r . , lee , j.c . , de rosa , a. , turner , a. & young , a.j . , 2002b , , 335 , l1 frank , j. , king , a. & raine , d. , 2002 , accretion power in astrophysics , 3rd edition , cambridge univ . press gammie , c.f . , 1998 , , 297 , 929 george , i.m . & fabian , a.c . , 1991 , , 249 , 352 gliozzi , m. , sambruna , r.m . & eracleous , m. , 2003 , , 584 , 176 hawley , j.f . , gammie , c.f . & balbus , s.a . , 1995 , , 440 , 742 iwasawa , k. , et al . , 1996 , , 282 , 1038 iwasawa , k. , lee , j.c . , young , a.j . , reynolds , c.s . & fabian , a.c . , 2003 , , in press , ( astro - ph/0309432 ) kaspi , s. , et al . , 2001 , , 554 , 216 krolik , j.h . , mckee , c.f . & tarter c.b . , 1981 , , 249 , 422 lee , j.c . , ogle , p.m. , canizares , c.r . , marshall , h.l . , schulz , n.s . , morales , r. , fabian , a.c . & iwasawa , k. , 2001 , , 2001 , 554 , l13 lightman , a.p . & eardley , d.m . , 1974 , , 187 , l1 longinotti , a.l . , cappi , m. , nandra , k. , dadina , m. & pellegrini , s. , 2003 , , 410 , 471 markowitz , a. , edelson , r. & vaughan , s. , 2003 , , in press ( astro - ph/0308312 ) martocchia , a. , matt , g. , karas , v. , belloni , t. & feroci , m. , 2002 , , 387 , 215 mason , k.o . , et al . , 2003 , , 582 , 95 matt , g. , perola , g.c . & piro , l. , 1991 , , 247 , 25 matt , g. , fabian , a.c . & ross , r.r . , 1993 , , 262 , 179 matt , g. , fabian , a.c . & ross , r.r . , 1996 , , 280 , 823 miller , j.m . , 2002a , , 570 , l69 miller , j.m . , et al . , 2002b , , 577 , l15 miller , j.m . , et al . , 2002c , , 578 , 348 miller , j.m . , et al . , 2003 , , in press ( astro - ph/0307394 ) morrison , r. & mccammon , d. , 1983 , , 270 , 119 nandra , k. & pounds , k.a . , 1994 , , 268 , 405 nayakshin , s. & kallman , t.r . , 2001 , , 546 , 406 nayakshin , s. & kazanas , d. , 2002 , , 567 , 85 nayakshin , s. , kazanas , d. & kallman , t.r , 2000 , , 537 , 833 orr , a. , barr , p. , guainazzi , m. , parmar , a.n . & young , a.j . , 2001 , , 376 , 413 padmanabhan , u. & yaqoob , t. , 2003 , in `` high energy processes and phenomena in astrophysics '' ( iau symp . 214 ) , suzhou , china ( astro - ph/0211386 ) perola , g.c . , matt , g. , cappi , m. , fiore , f. , guainazzi , m. , maraschi , l. , petrucci , p.o . & piro , l. , 2002 , , 389 , 802 petrucci , p.o . , et al . , 2002 , , 388 , l5 pounds , k.a . , nandra , k. , stewart , g.c . , george , i.m . & fabian a.c . , 1990 , , 344 , 132 reynolds , c.s . , 1997 , , 286 , 513 reynolds , c.s & nowak , m.a . , 2003 , , 377 , 389 risaliti , g. , 2002 , , 386 , 379 ross , r.r . & fabian , a.c . , 1993 , , 261 , 74 ross , r.r . , fabian , a.c . & brandt , w.n . , 1996 , , 278 , 1082 ross , r.r . , fabian , a.c . & young , a.j . , 1999 , , 306 , 461 ross , r.r . , fabian , a.c . , & ballantyne , d.r . , 2002 , , 336 , 315 raska , a. , dumont , a .- m . , czerny , b. & collin s. , 2002 , , 332 , 799 sako , m. , et al . , 2003 , , 596 , 114 shakura , n.i . & sunyaev , r.a . , 1973 , , 24 , 337 shakura , n.i . & sunyaev , r.a . , 1976 , , 175 , 613 stone , j.m . & norman , m.l . , 1992a , , 80 , 753 stone , j.m . & norman , m.l . , 1992b , , 80 , 791 stone , j.m . , hawley , j.f . , gammie , c.f . & balbus , s.a . , 1996 , , 463 , 656 turner , a.k . , fabian , a.c . , vaughan , s. & lee , j.c . , 2003 , , in press ( astro - ph/0303418 ) turner , n.j . & stone , j.m . , 2001 , , 135 , 95 turner , n.j . , stone , j.m . & sano , t. , 2002 , , 566 , 148 turner , n.j . , stone , j.m . , krolik , j.h . & sano , t. , 2003 , , 593 , 992 turner , t.j . , et al . , 2002 , , 574 , l123 turner , t.j . , kraemer , s.b . & reeves , j.n . , 2003 , , in press ( astro - ph/0310885 ) vaughan , s. & edelson , r. , 2001 , , 548 , 694 wang , j.x . , wang , t.g . & zhou , y.y . , 2001 , , 549 , 891 weaver , k.a . , gelbord , j. & yaqoob , t. , 2001 , , 550 , 261 yaqoob , t. , george , i.m . , nandra , k. , turner , t.j . , serlemitsos , p.j . & mushotzky , r.f . , 2001 , , 546 , 759 yaqoob , t. , george , i.m . , kallman , t.r . , padmanabhan , u. , weaver , k.a . & turner , t.j . , 2003 , , 596 , 85 ycki , p.t . , krolik , j.h . , zdziarski , a.a . & kallman , t.r . , 1994 , , 437 , 597 ycki , p.t . , done , c. & smith , d.a . , 1997 , , 488 , l113 ycki , p.t . , done , c. & smith , d.a . , 1998 , , 496 , l25 ycki , p.t . , don , e c. & smith , d.a . , 1999 , , 305 , 231 cccccccccccccc 125 & 10@xmath95 & & & 2.021@xmath96 & 147@xmath97 & 0.79@xmath98 & 2672/2358 & & & 2.03@xmath98 & 466@xmath99 & 1.2@xmath100 & 1684/1595 + 250 & 10@xmath95 & & & 2.026@xmath101 & 190@xmath97 & 0.92@xmath102 & 4265/2358 & & & 2.03@xmath103 & 327@xmath104 & 1.1@xmath105 & 1590/1595 + 500 & @xmath106 & & & 2.0 & 631 & 0.654 & 7833/2358 & & & 2.02@xmath103 & 507@xmath107 & 0.88@xmath108 & 1638/1595 + 1000 & @xmath109 & & & 2.000@xmath110 & 1244@xmath111 & 0.75@xmath112 & 2994/2358 & & & 2.03@xmath103 & 1282@xmath113 & 0.81@xmath114 & 1639/1595 + cccccccccccccc 10@xmath115 & @xmath116 & & & 1.988@xmath117 & 561@xmath118 & 1.01@xmath103 & 4679/2358 & & & 2.04@xmath103 & 601@xmath119 & 0.96@xmath120 & 1632/1595 + @xmath69 & 10@xmath121 & & & 1.950 & 565 & 0.798 & 5821/2358 & & & 2.03@xmath103 & 1297@xmath122 & 0.73@xmath123 & 1695/1595 + @xmath70 & @xmath124 & & & 1.970 & 593 & 0.926 & 6651/2358 & & & 2.00@xmath103 & 1340@xmath125 & 0.52@xmath126 & 1561/1595 + 10@xmath127 & @xmath128 & & & 2.029 & 491 & 0.611 & 5995/2358 & & & 1.987@xmath129 & 2679@xmath130 & 0.46@xmath126 & 1627/1595 +

we show how to create long range interactions between alkali - atoms in different hyper - fine ground states , with the goal of coherent quantum transport . the scheme uses off resonant dressing with atomic rydberg states . we demonstrate coherent migration of electronic excitation through dressed dipole - dipole interaction by full solutions of models with four essential states per atom and give the structure of the spectrum of dressed states for a dimer . in addition we present an effective ( perturbative ) hamiltonian for the ground - state manifold and show that it correctly describes the full multi - state dynamics . we discuss excitation transport in detail for a chain of five atoms . in the presented scheme , the actual population in the rydberg state is kept small . dressing offers large advantages over the direct use of rydberg levels : it reduces ionisation probabilities and provides an additional tuning parameter for life - times and interaction - strengths .

1011.5483

due to their remarkable properties @xcite , rydberg atoms emerge as a versatile tool in ultra - cold atomic physics . they can be used for diverse topics , such as quantum information , atomic aggregates @xcite , the study of conical intersections @xcite , digital quantum simulations @xcite , electro - magnetically induced transparency @xcite , coherent population trapping @xcite , giant kerr non - linearities @xcite and ultra - long range molecules @xcite , since their long range interactions can be tuned over many orders of magnitude . however excited atoms in rydberg orbitals are more vulnerable to ionisation @xcite or spontaneous decay and more difficult to trap @xcite . recently , several groups proposed to combine the advantages of both , ground - state and rydberg atoms , through `` rydberg - dressing '' schemes @xcite . the core idea is to use off - resonant laser coupling between ground- and rydberg states to create eigenstates of the laser - coupled system in which a small rydberg component is admixed to the ground - state . these resulting _ dressed _ states inherit some of the extreme properties of the rydberg states , while preserving favourable properties of the ground state . moreover , the additional degrees of freedom provided by laser rabi - frequency and detuning increase flexibility and promise potential shaping as well as dynamical control . dressing has been extensively studied for van der waals ( vdw ) interactions , generated when virtual transitions between levels of a two - atom system yield a distance dependent energy shift . several articles have proposed rydberg dressing techniques to induce this kind of interactions for ground - state atoms @xcite . in this context it was shown that dressing ground - state atoms with rydberg levels allows the large rydberg vdw interaction energies to be reduced until comparable to typical energy scales of cold - atom traps , while simultaneously increasing ensemble life - times to those required for thermal equilibration effects @xcite . these are helpful features e.g. , for realising a supersolid phase in rydberg - dressed bose - einstein condensates @xcite . here , we demonstrate that the ideas behind dressing rydberg vdw interactions can also be applied to resonant dipole - dipole interactions of neutral atoms . they occur when transitions between different states of the two - atom system are energetically resonant . in this case , the existence of an interaction potential is linked to electronic state transfer between atoms @xcite . dressing dipole - dipole interactions requires the use of two laser couplings and four electronic states ( two ground- and two rydberg - states ) , in contrast to one coupling and two states that suffice for vdw dressing . hence , the scheme is more involved but also more flexible , offering a larger number of dressed states . it ultimately allows coherent quantum state transport between atoms in different long - lived ground - states over distances of @xmath0-@xmath1 @xmath2 m and for durations of many milli - seconds . similar ideas have previously been applied to trapped rydberg _ ions _ @xcite . however , we are motivated by the possibility to generate potentials in an atomic many - body system that induce different kinds of ( coupled ) electronic and atomic _ motion _ dependent on the overall quantum state of the system . useful applications of such potentials can be found in @xcite . in many systems this goal is easier to achieve with neutral atoms than with ions , since the motion of the latter is typically dominated by their coulomb repulsion or the indispensable trap . furthermore the mapping of state - transfer interactions to normally non - interacting ground states by the laser dressing can be presented more clearly in the atomic case , where complications by trapping , mixing of internal and external dynamics and additional laser induced interactions can be ignored at least for free atoms . finally , see ref . @xcite for dressing techniques targeting the control of atomic motion without linked excitation transport . our results provide an additional handle on the time - scale of excitation transport and life - times , enable time - dependent control over hopping - strengths and can be used to vary the order of magnitude of dipole - dipole forces . within some constraints @xcite , the dressing also enables the use of well established trapping methods . as recently shown , dressed dipole - dipole interactions are an important tool for the realisation of atomic ring trimers @xcite . in particular , the confinement of long - range interacting atoms on a ring , as required in ref . @xcite , is greatly facilitated by the techniques presented here . atomic ring trimers allow the detailed study of wave - function dynamics near and across conical - intersections . dressed dipole - dipole interactions may also prove useful for the study of exciton dynamics in rydberg chains @xcite . to demonstrate dressed dipole - dipole interactions , we employ numerical simulations of a model with four essential states per atom . in this case there are several dressed state manifolds , distinguished by the number of excited rydberg atoms which they contain in the limit of vanishing dressing . we apply van - vleck perturbation theory to obtain analytical expressions for the induced effective interaction and to determine the parameter range within which the scheme can function . the paper is organised as follows : in we present the atomic states considered , describe the resulting hamiltonian and outline the basic principles of the dressing scheme . in we consider an atomic dimer in detail , providing analytical expressions for the effective dressed interactions , diagonalising the full hamiltonian and discussing different available dressing manifolds . in we present two exemplary applications of our results , dressed ring trimers and exciton migration . we conclude in and discuss technical details of our calculations and the underlying atomic physics in the subsequent appendices . we consider a collection of @xmath3 alkali atoms , where the @xmath4th atom is located at the position @xmath5 . the distances between the atoms , @xmath6 , shall be so large that interaction between two atoms can be neglected unless _ both _ of them are in a rydberg state . in this case they experience strong dipole - dipole interactions . as discussed in @xcite the dipole - dipole interaction leads to exciton dynamics similar to that of molecular dye aggregates , molecular crystals or light harvesting systems @xcite . for practical reasons we restrict ourselves to situations where the dynamics can be adequately described using only two `` ground '' states and two rydberg states per atom . with the term `` ground - states '' we refer e.g. , to two different hyperfine levels of the actual alkali - atom ground state . we shall denote the ground states by @xmath7 and @xmath8 and the rydberg states by @xmath9 and @xmath10 . the latter correspond to the @xmath11 and @xmath12 angular momentum components of a rydberg level with large principal quantum number @xmath13 . schematically , the resonant dipole - dipole interaction @xmath14 leads to a coupling between states @xmath15 and @xmath16 , whose matrix elements in the hamiltonian take the form @xmath17 . the transition strength is @xmath18 where @xmath19 is the inter - atomic distance and @xmath2 quantifies the transition dipole moment between the states @xmath9 and @xmath10 . we assume that all atoms are prepared in the @xmath20 azimuthal quantum states and never acquire @xmath21 . with atoms constrained in a two - dimensional ( 2d ) plane , orthogonal to the quantisation axis , this ensures that there is no angular dependence of @xmath14 @xcite . such a geometry covers both scenarios described later in this article . the interaction between atoms @xmath22 and @xmath23 reads @xmath24 to map the strong interactions @xmath25 to the ground states , the atoms are irradiated with far - detuned dressing laser - fields that couple the ground - states and our two selected rydberg states coherently . the relevant level diagram is sketched in fig . [ fig : sketch ] ( a ) . in practice such a coupling is commonly achieved by two ( or multi)-photon transitions @xcite , so that the rabi - frequencies @xmath26 and detunings @xmath27 in the diagram have to be regarded as effective quantities . from the laser transition parameters , we assemble the effective dimensionless dressing parameters @xmath28 they are a measure of how `` far - detuned '' the laser coupling is , and will emerge as crucial quantities controlling the dressing , as it is the case for dressed vdw interactions @xcite . it is important that the state @xmath7 is directly coupled only to @xmath9 and @xmath8 only to @xmath10 . in order for this simple picture with four relevant states per atom to be valid , the detunings of both laser couplings have to be chosen such that all other transitions , coupling to further rydberg states or ground state levels , are so far detuned that they can be safely neglected and @xmath29 with uniquely assigned states @xmath30 . see [ full : levels ] for more details . ] . we show some realistic level diagrams for @xmath31li in fig . [ fig : sketch ] ( b ) to demonstrate how this constraint can be met in practice . throughout the article , we will refer to the states @xmath7 , @xmath9 as the `` s - pair '' and the states @xmath8 , @xmath10 as the `` p - pair '' . using the four states introduced above as a basis for the single atom , an @xmath3-body basis state @xmath32 is written as @xmath33 where @xmath34 describes the electronic state of the atom @xmath35 . for example we write @xmath36 when the first atom is in state @xmath7 , the second in @xmath8 and the third in @xmath9 . after defining operators @xmath37 with @xmath22 , @xmath38 @xmath39 where @xmath4 is the atom - index , the many - body hamiltonian can be written as @xmath40 with @xmath41 and @xmath42 we split @xmath43 into two parts , @xmath44 and @xmath45 , to facilitate our later application of perturbation theory . the atom - light coupling in has been treated in the dipole- and rotating wave approximations , as explained in [ rotwave ] . the _ exciton number _ operator @xmath46 measures the number of atoms in the p - pair of states . it is easy to see that @xmath47 commutes with @xmath43 and hence the exciton number is conserved , see also [ multi_exc ] . of course equation is a simplified description of a complicated multi - state system . for a more complete picture , one could include additional rydberg levels adjacent to @xmath9 , @xmath10 , as shown in fig . [ fig : sketch ] ( b ) , van der waals interactions of all four states , spontaneous decay and loss in the excited states . the latter arises as atoms in rydberg states are lost due to auto - ionization or blackbody radiation @xcite . importantly , parameter regimes can be found where these corrections are minor and suffices , for example for the states and parameters shown in fig . [ fig : sketch ] ( b ) and described in . consider an atomic sample such as a chain , prepared in states @xmath48 , in which the nth atom is in the upper ground state @xmath8 and all others in the absolute ground - state @xmath7 @xmath49 without dressing ( @xmath50 ) this is an eigenstate of the hamiltonian ( [ hamil_dressed ] ) . if we ignore dipole - dipole interactions for the moment , an adiabatic ramp - up of the dressing strengths @xmath26 will change this eigenstate into one with some admixture of rydberg population . allowing then for dipole - dipole interactions finally causes excitation transport via processes of the schematic form : @xmath51 . here @xmath52 symbolizes a transition of both atoms between the ground state manifold and the rydberg state manifold due to the dressing lasers . similarly @xmath53 stands for a transition occurring due to resonant dipole - dipole interactions between rydberg states . in we will confirm this expectation with numerical simulations and later justify and quantify it using perturbation theory . if the lasers are sufficiently off - resonant , the admixed fraction of excited state should remain small throughout . for this reason we can describe the process that we have just sketched as an effective interaction in the space spanned by the @xmath48 . we will refer to it as the _ ground - state manifold_. for later reference we also define spaces spanned by @xmath54 which we call _ gp ( hs ) single excitation manifold_. throughout this paper we focus on a specific realisation of our scheme , sketched in fig . [ fig : sketch ] ( b ) . we consider @xmath31li atoms , whose @xmath55 ground - state has a hyperfine - splitting of @xmath56 mhz @xcite . for the rydberg state @xmath9 we pick a principal quantum number @xmath57 and angular momentum @xmath11 state . for @xmath10 we choose @xmath57 and @xmath12 . the energy difference between the two rydberg states is @xmath58 ghz . we will assume detunings @xmath59 mhz and dressing parameters @xmath60 unless indicated otherwise . we will occasionally refer to @xmath61 ( without index ) , which implies @xmath62 . for the above values of @xmath61 and @xmath63 , our rabi - frequencies are @xmath64 mhz . for the most part , we illustrate the dressing scheme through an atomic dimer with @xmath65 @xmath2 m inter - particle separation . for this distance , the strength of the bare dipole - dipole interaction is @xmath66 mhz , just less than the doubly excited state detuning @xmath67 mhz . the reason for this choice will become clear later . we stress that we have chosen these parameters to be specific , while our findings are more general . the constraints under which the parameters can be varied without invalidating the dressing scheme have been discussed in . all following results are expected to hold for any alkali species if these constraints are met . for the principal demonstration of excitation transport through rydberg dressing , we first treat the simplest possible case , the atomic dimer . in this section , we show dynamics from the initial state @xmath68 . we consider two different ramps of the dressing couplings @xmath26 in the hamiltonian : ( i ) a sudden jump from zero to their final value and , ( ii ) an adiabatic ramp over a small finite time @xmath69 ms , effectively changing the initial state to @xmath70)/{\cal n}$ ] , where @xmath71 is a normalisation factor ( see also ) .. , the ramp ought to be slow on the time - scale of the laser - coupling but fast on that of effective dipole - dipole transport . thus @xmath72 with @xmath73 @xmath2s and @xmath74 ms ( the value is read off and defined as the time it takes for the excitation to oscillate from the first atom to the second and back . ) . ] we expand the time dependent quantum state as @xmath75 , where @xmath32 are the @xmath3-body basis vectors defined in . in this basis , we express the hamiltonian in matrix form and solve the schrdinger equation ( se ) @xmath76 by matrix exponentiation . using the four - state model introduced in we employ the full hamiltonian without further approximations . for two atoms placed a distance @xmath65 @xmath2 m apart , the resulting population - oscillations between the states @xmath77 and @xmath78 can be clearly seen in . on the logarithmic scale of panel ( b ) , we see that also states containing one rydberg excitation or more are populated , but their occupation is roughly suppressed by @xmath79 , indicated by the black solid line . the suppression of excited state population by a factor @xmath80 is well known from vdw dressing techniques @xcite and a basic feature of off - resonant rabi - coupling . as can be seen in , there will be many ground state population oscillations up to the expected spontaneous life - time of the dressed @xmath81-atom state , which we estimate to about @xmath82s as described in . due to the dressing , the time - scale on which the interaction transports the excitation has been massively increased compared to the `` bare '' population oscillation period between rydberg states @xmath83 , which would be @xmath84ns . the comparison of panels ( b ) and ( c ) of shows that the use of the more refined , adiabatically created , initial state has led to a more regular evolution of the excited state populations , whose oscillations now precisely follow those of the two ground states . the population of states with a single rydberg excitation never exceeds @xmath80 . doubly excited states are further suppressed . however , this careful initial state creation is not actually required in order to observe dressed population oscillations , as we have just seen . the results of the preceding section already unambiguously demonstrate excitation transport through full numerical solutions of the problem . to further understand the population - oscillations more intuitively , we now consider the ground - state manifold spanned by @xmath85 , @xmath86 as the _ system _ of interest and its hilbert - space complement as _ environment_. the coupling between the two is treated as a perturbation . we use van vleck perturbation theory @xcite , as outlined by shavitt _ @xcite to derive an effective hamiltonian in terms of the `` system '' only . this scheme conveniently takes care of degeneracies and generates an effective hamiltonian well suited to describe excitation transport . a full analytical diagonalisation of is impractical even for @xmath87 . in contrast the perturbative results for @xmath87 are intuitive and generalise straightforwardly to cases with more atoms . they are also valid over large ranges of parameter space as we shall show below . the aim of van vleck perturbation theory is to find a basis that block - diagonalises the hamiltonian ( [ hamil_dressed ] ) to a given order in the perturbation @xmath45 . for @xmath88 this is achieved by the basis @xmath32 introduced in , since @xmath44 is already fully diagonal . the relative importance of @xmath45 and @xmath44 is governed by the dressing parameters @xmath89 . as the @xmath89 increase , the basis that block - diagonalises @xmath43 becomes more complicated . perturbation theory will thus be valid as long as the dressing parameters @xmath89 are small . we will use the notation @xmath90 with @xmath91 , i.e. @xmath92 is a state whose leading component is @xmath32 as long as @xmath61 remains small and @xmath93 is small ( i.e. r is large ) . in this section we shall only be interested in the block of the effective hamiltonian that governs the dynamics of the effective ground - state manifold @xmath94 and @xmath95 . the details of the calculation are given in [ vanvleck ] , here we merely present the results . in the basis @xmath96 , @xmath97 one obtains @xmath98 where @xmath99 is the @xmath100 identity matrix . we have energy shifts @xmath101 and an explicitly distance dependent part @xmath102 . the light shifts @xmath103 ( @xmath104 ) , corresponding to second ( fourth ) order in @xmath45 , are given by [ eshifts ] @xmath105 @xmath106 where we have used the dressing parameters defined in . these light shifts could allow optical trapping using the same fields that provide the dressing coupling . the position dependent part of the hamiltonian , @xmath102 , can be written as @xmath107 with [ potentials ] @xmath108 @xmath109 where we have defined @xmath110 . as expected the states @xmath111 , @xmath112 acquire small excited state populations of the form : @xmath113 independent of @xmath19 . here @xmath114 is a normalisation factor . let us briefly mention some special cases : for large inter - atomic separations , such that @xmath115 , we can expand the expressions ( [ potentials ] ) and obtain @xmath116 and @xmath117 . the shift @xmath118 , which is independent of @xmath19 , can then be merged with @xmath104 . for shorter distances , the factors @xmath119 become relevant and diverge at @xmath120 , or equivalently at @xmath121 . we will see in the next section that this divergence can be traced to avoided crossings between the perturbed energy eigenvalues . we show in that ( [ potentials ] ) fails only in a fairly narrow region around this avoided crossing . an interesting special case is @xmath122 . here , the equations predict that the dressing effect on the ground - state manifold vanishes , which is confirmed by full simulations . we will however show in that this case has special appeal for dressing within the single excited state manifold . we also derive the contribution to @xmath123 of @xmath124th order in @xmath45 in [ sixth_order ] . we will make use of it in , to illustrate the rate of improvement that can be achieved through higher - order terms of perturbation theory . to get an overview of the consequences of dressing radiation beyond the ground state manifold , we now consider the full energy spectrum of the dimer as a function of inter - atomic separation . as we outline in [ multi_exc ] , the dimer hamiltonian can be brought into block - diagonal form , with three blocks describing exciton numbers of @xmath125 . of these , only the single exciton block is non - trivial , hence we only show the eigenspectrum of its corresponding sub - matrix @xmath126 ( [ hamiltonian_mainblock ] ) here . explicitly we solve @xmath127 besides providing information on the possibilities of dressed excitation transport , the energy eigenvalues as a function of separation can also be viewed as adiabatic ( born - oppenheimer ) potential surfaces for motion of the atoms , as long as this motion is sufficiently slow @xcite . the @xmath19-dependent energy eigenvalues are shown in for @xmath128 mhz and @xmath129 mhz , with other parameters as described in . the slight offset between @xmath130 and @xmath131 makes it easier to grasp the structure of the resulting spectrum . at large distances @xmath19 and for moderate dressing strength @xmath61 , the eigenstates are essentially superpositions of two basis states , and have odd or even symmetry under exchange of atoms @xmath132 and @xmath81 . these superpositions are listed in the caption of fig . [ spectrum_deltaeq ] ( a ) , disregarding normalisation . their energies are then mostly determined by the total detuning of the two superimposed basis states . as the distance @xmath19 is reduced , the dipole - dipole interaction energy of the states consisting mainly of two rydberg atoms ( blue dotted- and red solid lines ) becomes more prominent and eventually leads to avoided crossings between these and the other states of the spectrum . on the large energy scale of fig . [ spectrum_deltaeq ] ( a ) , @xmath133 , the dipole - dipole shift of the @xmath134 states is most prominent . however also the other states with only a small doubly excited state component acquire a space dependent dipole - dipole potential , as can be seen in the close - ups , panels ( b)-(d ) . the energy splitting of the states within the ground - state manifold , panel ( b ) , is of order @xmath135 while that in the singly excited state manifolds , panels ( c - d ) , is of order @xmath136 . in panel ( b ) we additionally display the energy eigenvalues obtained from the effective hamiltonian ( [ h_eff ] ) . the shape of the potential is well reproduced . the dressed potentials shown in ( b)-(d ) have at least two interesting features : ( i ) their overall strength scales like @xmath137 or @xmath138 , it can thus easily be adjusted by choice of laser parameters and even manipulated time - dependently . ( ii ) the shape of the potentials can be modified beyond a simple attractive or repulsive form @xmath139 , due to the appearance of avoided crossings . for example the attractive branch of potentials in the ground - state manifold , shown in fig . [ spectrum_deltaeq ] ( b ) , approaches a finite value for small separations @xmath19 as can be seen from its continuation in fig . [ spectrum_deltaeq ] ( a ) . this could avoid acceleration to high velocities during atomic collisions on the attractive potential , maybe reducing the probability of collisional ionisation @xcite . light induced modifications of potentials that would be strongly attractive in the absence of dressing have already been discussed for the case with a single ground state in ref . @xcite , and proposed as collisional shield for cold polar molecules in ref . however , in our case , once the separation of the atoms is much smaller than the minimal value included in , which is @xmath140 , we have to keep in mind that the simplified four - state model breaks down : the dipole - dipole shift of the dressed states then may become so large that they couple strongly to two - atom states that are not included in our model . let us briefly discuss the case when the detunings have equal magnitudes . as @xmath130 approaches @xmath131 , the asymptotic @xmath141 energy curve ( violet line ) in fig . [ spectrum_deltaeq ] ( a ) is squeezed between the neighbouring ones until the states become degenerate . if we change the sign of one of the detunings and consider the special case @xmath142 , we obtain a particularly symmetric energy spectrum , shown in . as predicted by , the dressing - induced state transfer interaction vanishes between the states @xmath143 and @xmath144 . however symmetrical induced potentials can now be found in the singly - excited state manifold . the strength of these potentials scales as @xmath80 , different from those previously obtained in the ground - state manifold . we will discuss this further in . in this section we revisit dressed dimer dynamics first presented in , to apply the results of . figure [ comparison_pt_fullh ] shows a comparison of the evolution governed by the effective perturbative hamiltonian with that following from the full hamiltonian . already fourth order perturbation theory describes the effective induced hopping period fairly well , as can be seen in panel ( b ) . when we move to the sixth order expressions , which can be found in [ vanvleck ] , the agreement becomes even better : the deviation never exceeds @xmath145 in the time - interval shown . in other simulations we found that perturbation theory agrees better with full calculations if @xmath61 is small , as is expected . even for parameters with poorer agreement between full- and perturbative dynamics ( for example larger @xmath61 or closer to an avoided crossing ) , we typically still obtain the primary features which make these induced potentials interesting : ( i ) we have persistent dressed hopping , ( ii ) the excited state occupancy is suppressed . since we have included the dipole - dipole interaction in the _ unperturbed _ part of the hamiltonian ( [ hunpert ] ) , we do not a priori require it to be small compared to the detuning for the perturbation theory in to be valid . in practice we find that expression ( [ h_eff ] ) works well even for @xmath146 , as long as one stays clear of the avoided crossings at @xmath147 . the large splitting of the eigenstates in the ground state manifold in the vicinity of the avoided crossing , see fig . [ spectrum_deltaeq ] ( b ) , results in comparably fast population - oscillations . the regime where we can find dressed excitation transport on time scales shorter than the life - time of the multi atom state hence is typically entered when @xmath148 approaches @xmath149 , but has not quite reached it yet . the effective population - oscillation - period in is @xmath150ms . here @xmath151 , and the correction factor is @xmath152 due to the vicinity of an avoided crossing . let us reconsider the doubly excited state populations in . for atomic distances where the system is far from all avoided crossings shown in , the doubly excited states would be suppressed by @xmath153 . this gets modified near these avoided crossings . for the distance @xmath65 @xmath2 m as , the system is fairly close to an avoided crossing , thus the doubly excited states are more strongly populated . to conclude this section , we would like to remark on the position degree of freedom of the atoms carrying the interaction . in this article we treat it classically , assuming atoms with a precisely defined separation @xmath19 . in practice , each atom will have a position uncertainty , for example due to the zero - point motion in a harmonic well . the resulting distribution of interatomic distances @xmath19 can lead to a fast de - phasing of population oscillations such as shown in . for the example shown , if we would assume a separation uncertainty of @xmath154 the oscillations de - phase after @xmath155 cycles . we however do not study de - phasing here , since it is not specific to the _ dressed _ interactions . if we considered direct dipole - dipole excitation transport , between the states @xmath15 and @xmath16 only , it would also de - phase after @xmath155 cycles , which however would take place on a timescale of @xmath156 ns due to the by @xmath157 stronger interaction . thus , whether or not de - phasing due to disorder poses a problem depends on the intended application of excitation transport and not on whether or not the interactions arise due to dressing . in particular ref . @xcite discusses an exemplary application where it poses no problem , simulations presented in ref . @xcite _ fully include atomic positioning uncertainties_. also in references @xcite we study systems with interesting combined excitation transport and atomic motion , with interactions that could be interpreted as arising via the techniques considered here . again , de - phasing effects do not pose a problem . the use of chains of rydberg atoms for random walks and other aggregate physics @xcite is limited by spontaneous decay and other incoherent loss processes from the rydberg states . when using a dressing scheme , these decay processes are reduced by construction since the atomic population in rydberg states is kept small . throughout the paper , _ except _ , we thus do not explicitly include loss in the presented dynamical calculations , but instead estimate the life - time of dressed rydberg - aggregates based on their content of actual rydberg population . for example the state @xmath111 will be assigned a total effective decay - rate @xmath158 for two atoms . this scaling of @xmath159 with the dressing parameters follows the same pattern as in vdw dressing @xcite . most of our simulations treat excitation transport via @xmath57 rydberg states of litihum , which have life - times @xmath160 @xmath2s and @xmath161 @xmath2s @xcite . this yields @xmath162ms for @xmath163 and @xmath87 . the corresponding exponential population decay is schematically indicated in fig . [ comparison_pt_fullh ] ( a ) . note , that simple estimates of the life - time of a dressed system using the relations of this section and the results of require excited state populations to be accurately described by perturbation theory . we have seen in that this is the case , even fairly close to the avoided crossings in the spectrum . to ascertain that the simple estimates of loss effects discussed above are valid for our cases , we also calculated results such as shown in with the inclusion of loss and spontaneous decay . consider the density matrix @xmath164 with basis - states as defined in . the prime on the summation symbol indicates that only states within the single exciton manifold ( set @xmath165 in [ multi_exc ] ) are considered . let the time - evolution of @xmath166 be given by the master equation @xmath167 } + \sum_{ak}\bar{\gamma}_{k}{\cal d}[\hat{d}_{k}^{(a)},\hat{\rho } ] + \gamma { \cal l}[\hat{l},\hat{\rho}],\end{aligned}\ ] ] where @xmath168 is the lindblad superoperator @xmath169 = \hat{a}\hat{\rho}\hat{a}^{\dagger } - \hat{a}^{\dagger}\hat{a}\hat{\rho}/2 - \hat{\rho}\hat{a}^{\dagger}\hat{a}/2 $ ] . for @xmath170 define @xmath171 , then @xmath172 acts like @xmath173 in the electronic space of atom @xmath174 and as unity elsewhere . these operators describe spontaneous decay . the superoperator @xmath175 has the form @xmath176 = \hat{a } \hat{\rho}-\hat{\rho}\hat{a}^{\dagger}$ ] , and the operator @xmath177 . here @xmath178 counts the number of excited atoms in the state @xmath32 . the purpose of @xmath175 is to describe loss from the rydberg states to states external to our model @xcite , for example to further rydberg states . for simplicity we assumed an identical loss rate @xmath179 from @xmath9 and @xmath10 . we consider two cases : ( i ) all the population loss from the rydberg levels described by the effective decay rates @xmath180 from @xcite , including black - body effects , ends up in the respective ground - states . here we set @xmath181 and @xmath182 . this is a worst case scenario for our purposes , as all incoherent processes within our four state model de - cohere excitation transport . ( ii ) we split @xmath183 from @xcite into the zero temperature component @xmath184 and the black - body induced component @xmath185 via @xmath186 . this yields @xmath187 @xmath2s , @xmath188 @xmath2s , @xmath189 @xmath2s , @xmath190 @xmath2s . we then assume that population returns to their respective ground - states with a rate @xmath191 , while black - body redistributed population is lost , either due to ionisation or because it leaves the single exciton manifold . ] . we thus set @xmath192 . these calculations , shown in , verify that loss out of the system from the rydberg states @xmath9 and @xmath10 has indeed no other consequences than an overall population decay . if spontaneous decay in the channels @xmath193 and @xmath194 is included , population oscillations additionally show a de - phasing . neither effect is dramatic on the time - scales considered here . in we have pointed out that dressing induced dipole - dipole interactions can be created in the ground - state manifold ( between the states @xmath143 , @xmath144 ) or in two different excited state manifolds ( between the states @xmath195 , @xmath196 or @xmath197 , @xmath198 ) . the latter requires significantly different detunings @xmath130 and @xmath131 ( such like @xmath142 ) to energetically separate the two different singly - excited state manifolds . we have explicitly verified population - oscillations for example between the states @xmath199 and @xmath200 . to know whether the ground or singly - excited state manifold would be more useful for a specific application it is important to consider the life - time of the dressed multi - atom state : the smaller @xmath61 , the longer the life - time , but also the longer the effective hopping period @xmath201 required for a complete population - oscillation that transfers the excitation from one atom in the dimer to its counter - part and back . here , the interaction @xmath202 denotes the off - diagonal entry in the effective hamiltonian of the corresponding manifold , e.g. @xmath203 in for the ground - state manifold . here we estimate the life - time of dressed states more roughly than described in , assuming that ground - state atoms ( @xmath7 , @xmath8 ) do not decay and excited state atoms ( @xmath9 , @xmath10 ) decay with the same rate @xmath204 . for a system of @xmath3 atoms in various states we then again add the decay rates . @xmath202 & @xmath205 + . @xmath206 & @xmath207 & @xmath93 & @xmath208 + . @xmath209 & @xmath210 & @xmath211 & @xmath212 + . @xmath213 & @xmath214 & @xmath215 & @xmath216 + for three different @xmath3-atom states , the life - times determined in this way and the ratio of life - time and hopping period @xmath205 are listed in . as long as @xmath217 , the ratio @xmath218 for the singly - excited state manifold can be larger than for the ground - state manifold . whether dipole - dipole interactions induced through dressing are advantageous over direct dipole - dipole interactions among rydberg states , strongly depends on the problem at hand , as we discuss now for three examples . ( i ) exciton migration : _ for the migration of a single or multiple excitons on a rigid chain , as studied in @xcite , the ratio @xmath218 introduced above gives a direct measure of how many sites the excitation can traverse within the life - time of the whole chain . evidently this measure is never improved by the present dressing technique , whose advantage for this scenario thus only lies in reduced ionization probabilites and simpler atom trapping . in we show an exemplary case where the overall life time in the presence of dressing is long enough to exploit these features . _ ( ii ) adiabatic transport : _ there are dynamical scenarios whose time - scale of interest is not directly given by @xmath219 , such as the adiabatic entanglement and excitation transport that we reported in @xcite . it can be seen that the ratio of the time - scale of transport dynamics to spontaneous life - time of the chain can be improved by factors of about 2 by working in the dressed singly - excited - state manifold . _ ( iii ) ring aggregates : _ in we consider systems of three dipole - dipole interacting atoms trapped in effectively one - dimension on a ring . for such a construction dressing can be highly beneficial . after choosing the dimer as the simplest case to work out the basic details of dressed dipole - dipole interactions , we now briefly present two exemplary applications of our results to larger systems . first , in , we consider a dressed ring - trimer ( @xmath220 atoms confined on a ring ) , a system of interest due to conical intersections of the adiabatic energy surfaces @xcite . then , in , we examine a dressed exciton on an atomic chain with @xmath221 . in ref . @xcite we show that circular rydberg trimers , consisting of three rydberg atoms trapped in one dimension on a ring , exhibit interesting quantum dynamics near conical intersections ( cis ) . for practical realisations , however , dipole - dipole forces are typically _ too large _ for ring confinement . in this case the reduction of interaction strengths , which limits the parameter range where dressing is beneficial for exciton migration , turns into a benefit . as dressing also increases the system s life - time , it greatly facilitates the practical creation of dipole - dipole ring trimers . consider three atoms tightly confined in one dimension on a ring , as sketched in fig . [ trimer : plots ] ( a ) . achieving such confinement is the first point in this section that is simpler for dressed ground state atoms than bare rydberg atoms . on the ring , the bare interaction between atoms @xmath25 is fully determined by the two relative angles @xmath222 and @xmath223 shown in the sketch . adiabatic energies , defined by , then form 2d surfaces @xmath224 . in the absence of laser - couplings , the three surfaces spanned by @xmath225 , @xmath226 , @xmath227 exhibit a conical intersection @xcite , where two of them become exactly degenerate , at @xmath228 . in the presence of laser - couplings , this feature is imprinted also onto surfaces spanned predominantly by ground states @xmath229 , @xmath230 , @xmath231 . conical intersections are of great interest in particular in chemical physics , as they strongly affect the outcome of photo - chemical reactions @xcite . in cold atomic gases of rydberg dressed atoms , they result in interesting non - adiabatic and geometric phase effects @xcite . for a detailed description of this system and the concept of conical intersections we refer to ref . @xcite and references therein . for the purpose of the present article , it is more important to point out that the techniques worked out here are highly beneficial for the practical realisation of this kind of ultra cold conical intersections . this conclusion is reached through an extensive but technical survey of parameter space . we defer full details to a specialised publication , and present here only the core points : in order to realise the scenario sketched in fig . [ trimer : plots ] ( a ) , initially at least three conditions have to be met : + our interest is in dynamics near the conical intersection , where the atoms form an equilateral triangle and are separated by @xmath232 , where @xmath19 is the ring radius . comparing the ( possibly dressed ) interaction strength with the strength of the ring trap , @xmath233 , we have @xmath234 , thus preventing either too small ring radii @xmath19 or to large interaction strength @xmath2 and hence principal quantum numbers @xmath235 . + we can estimate the order of magnitude of the ci transit time @xmath236 from the time - scale set by the classical equations of motion @xmath237 , which holds for a dimer of angular separation @xmath238 . this time scale is @xmath239 . adiabatic dynamics must fulfill @xmath240 , with @xmath219 as in . + thus @xmath241 . + attempting to fulfill the inequalities ( a - c ) for various atomic species leads to two main conclusions : ( i ) lighter atoms are favourable . ( ii ) without dressing ( @xmath242 ) , principal quantum numbers @xmath235 would be too high to avoid strong effects of black - body radiation . if we consider a dressed system however , one more constraint comes into play : + this is to stay clear of avoided crossings such as seen in . a set of parameters where these conditions are met is : @xmath243 , @xmath244 @xmath2 m , @xmath245 mhz and @xmath246 . were given as angular frequency @xmath247 . ] now we consider the dynamical evolution of an eigenstate of , when the interaction parameters @xmath248 are varied in time in a manner representative for near ci dynamics . to this end we extracted the trajectory @xmath249 from full quantum mechanical solutions of the time - dependent schrdinger equation in ref . @xcite , shown in the inset of fig . [ trimer : plots ] ( a ) as dashed line . a second trajectory is taken slightly offset ( dotted line ) . we find that the quantum state dynamics in both cases follows @xmath123 ( see equation ) well , justifying the simplified model used in @xcite . since the energy - spacing @xmath250 ( see ) for @xmath243 is @xmath251 mhz , we could have chosen a much larger detuning , yielding even better agreement between the perturbative and exact solutions . as we have seen in this section , dressing is threefold beneficial for atom trapping in the construction of conical intersections in dressed rydberg ring trimers : ( i ) it greatly widens the available scope for quasi-1d trapping of the atoms , ( ii ) reduces the involved interaction strength such that trapping can be realistically considered in the first place and ( iii ) simultaneously extends the available life - times . we discuss results for a linear chain of @xmath221 equidistant atoms in the following , considering the same physical situation as in , hence the spacing of atoms in the chain is also @xmath252 m . we create the initial state adiabatically as in . for this section only , we will define our excitation to be the state @xmath7 , such that e.g. @xmath253 . the corresponding dressed states have a longer life - time for larger chains . the dressing - induced excitation transport and a comparison of exact and perturbative evolution is shown in . it can be seen that the excitation is transported over many sites of the chain on time - scales shorter than the expected spontaneous life - time of the dressed @xmath0-atom state ( about @xmath254^{-1}=0.14 $ ] s ) . due to the dressing , the time - scale on which the interaction transports the excitation has been increased by a factor of @xmath255 , compared to the direct use of excited rydberg states . in we argued that dressing typically does not favourably alter the possible number of excitation hops , given by the ratio @xmath205 of life - time and hopping period . nonetheless the results of this section show that dressed excitation migration within the available life - time is possible . despite the detrimental effect on @xmath218 , the dressing still enables time - dependent control of the transport , and , as in the previous section , simplifies trapping efforts . we have shown that the dressing of alkali atoms in the ground - state with rydberg - excitations can yield dipole - dipole interactions with excitation - transfer for the ground - state atoms . this generalises existing results for dressing with rydberg van der waals interactions , that do not entail excitation transport . the scheme proposed here is an adaptation of a similar method , introduced in the context of trapped rydberg ions @xcite , to neutral atomic systems . it makes use of two effective laser couplings to rydberg - states , instead of the one required for van der waals dressing . we find that the dressed dipole - dipole transition matrix element between ground state atoms scales like @xmath153 with the dressing parameter @xmath61 , while the life - time of the dressed state increases like @xmath256 . dressed vdw interactions show the same scaling behaviour . we demonstrate in ref . @xcite that this re - arrangement of life - times and interaction time - scales allows the study of conical intersections using rydberg - dressed atoms on a ring . beyond this , the dressing scheme will enable a larger variety of trapping techniqes for the atoms and reduce the likelihood of ionisation . as further example of the flexibility afforded by the dressing , we show that it is possible to induce effective transport of a single rydberg excitation within a chain of ground state atoms . the amplitude of this process scales like @xmath80 . whether the effective interactions discussed in this article are more useful than the direct use of dipole - dipole interactions depends on details of the situation one studies . @xcite we provide an exemplary scenario where they can be highly beneficial . in this section , we derive the rotating - wave hamiltonian from more fundamental expressions for atom - laser interactions , reviewing text - book material , see e.g. @xcite . the hamiltonian for an atom in the presence of an electric field is @xmath257 this expression is valid in the dipole approximation @xmath258 , when the wavelength of the incident light is much larger than typical atomic distances . the operator @xmath259 is the full atomic hamiltonian in the absence of any external field , @xmath260 is the electron charge , @xmath261 the position operator and @xmath262 the electric field at the atomic nucleus , which is treated classically . consider for the moment a single atom only , with the four relevant states @xmath7 , @xmath8 , @xmath9 , @xmath10 and a level scheme as shown in fig . [ fig : sketch ] ( a ) . these states are eigenvectors of the free atomic hamiltonian @xmath259 , thus @xmath263 with @xmath264 . we add two coupling fields @xmath265 , with @xmath266 . here @xmath267 is a unit vector in the @xmath268-direction . for any two states @xmath269 , @xmath270 we define matrix elements @xmath271 . a general quantum state in the chosen subspace is @xmath272 , with equation of motion @xmath273 & = \left [ \begin{array}{ccccc } e_{g } & -\mu_{gs } { \cal e}_{s}(t ) & 0 & 0 \\ -\mu_{gs}^ { * } { \cal e}_{s}(t ) & e_{s } & 0 & 0 \\ 0 & 0 & e_{h } & -\mu_{hp } { \cal e}_{p}(t)\\ 0 & 0 & -\mu_{hp}^ { * } { \cal e}_{p}(t ) & e_{p } \\ \end{array } \right ] \left [ \begin{array}{c } c_{g}\\ c_{s}\\ c_{h}\\ c_{p}\\ \end{array } \right].\end{aligned}\ ] ] here we have used atomic units . now we change the variables @xmath274 of the coefficient vector to @xmath275 , where @xmath276 are arbitrary constants . if the coefficient vector @xmath277 evolved according to @xmath278 for some matrix @xmath279 , @xmath280 evolves according to @xmath281 with @xmath282 , where @xmath283 and @xmath284 are diagonal matrices with elements @xmath285 and @xmath286 . using the specific transformation vector @xmath287 , we obtain @xmath288 & = \left [ \begin{array}{c } 0 \\ -\frac{1}{2}\mu_{gs}^ { * } { \cal e}_{s } f_{-}(\omega_{s},\nu_{s } ) \\ 0 \\ 0 \\ \end{array } \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\ ! \begin{array}{c } -\frac{1}{2}\mu_{gs } { \cal e}_{s } f_{+}(\omega_{s},\nu_{s } ) \\ 0 \\ 0 \\ 0 \\ \end{array } \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\ ! \begin{array}{c } 0 \\ 0 \\ 0 \\ -\frac{1}{2}\mu_{hp}^ { * } { \cal e}_{p } f_{-}(\omega_{p},\nu_{p } ) \\ \end{array } \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\ ! \begin{array}{c } 0 \\ 0 \\ -\frac{1}{2}\mu_{hp } { \cal e}_{p } f_{+}(\omega_{p},\nu_{p } ) \\ 0 \\ \end{array } \right ] \left [ \begin{array}{c } d_{g}\\ d_{s}\\ d_{h}\\ d_{p}\\ \end{array } \right],\end{aligned}\ ] ] where @xmath289 , @xmath290 and @xmath291 stands for @xmath292 . under the _ rotating - wave approximation _ all terms @xmath293 are dropped , as they are oscillating too rapidly to have any effect . we also introduce the detunings @xmath294 and rabi - frequencies @xmath295 and @xmath296 . finally we perform one further variable - change , using @xmath297 to reach @xmath298 & = \left [ \begin{array}{ccccc } 0 & \frac{\omega_{s}}{2 } & 0 & 0 \\ \frac{\omega_{s}^{*}}{2 } & -\delta_{s } & 0 & 0 \\ 0 & 0 & 0 & \frac{\omega_{p}}{2 } \\ 0 & 0 & \frac{\omega_{p}^{*}}{2 } & -\delta_{p } \\ \end{array } \right ] \left [ \begin{array}{c } d'_{g}\\ d'_{s}\\ d'_{h}\\ d'_{p}\\ \end{array } \right].\end{aligned}\ ] ] the hamiltonian in ( [ effective_equation3 ] ) is the basic building block of , as long as only a single atom is concerned . even for more than one atom and with the inclusion of dipole - dipole interactions , the above procedure can be followed . instead of the single atom energies in the first transformation vector @xmath299 , we would employ many - atom energies . since the dipole - dipole interaction only couples energetically degenerate states like @xmath15 to @xmath16 , all complex phase - factors can finally be eliminated to arrive at a many - body version of , which is . in practice the scheme displayed in is complicated by selection rules and transitions are constrained by the availability of laser sources . the coupling between the states @xmath300 can be realised with now broadly established two - photon excitation schemes . these transitions are typically near resonant with some auxiliary middle level , such as @xmath301 in , in order to enhance transition amplitudes . for the right choice of parameters , coherent coupling between @xmath302 and @xmath303 is achieved , see e.g. @xcite . for the coupling between the states @xmath304 two - photon transitions are forbidden due to the selection rule @xmath305 . this necessitates one- or three photon coupling schemes , examples for which are included in . let us briefly discuss both options : + _ ( i ) single photon uv - transition : _ direct transitions from ground- to rydberg levels @xcite suffer from extremely small dipole - transition matrix elements . these scale like @xmath306 with the principal quantum number @xcite . from the matrix element @xmath307 @xcite , from oscillator strengths @xmath308 and transition energies @xmath309 using @xmath310 , @xmath311 is the electron mass . ] we extrapolate @xmath312 , where @xmath313 is the bohr radius . assuming @xmath314 mw of laser - power , focussed to a waist @xmath315 @xmath2 m , one achieves a rabi - frequency @xmath316 mhz as employed in this article . the intensity at the focus is @xmath317 , resulting in a rabi - frequency @xmath318 , where @xmath319 is the electric field , and @xmath260 , @xmath320 as well as @xmath321 fundamental constants . laser light at the uv wavelength is typically created through higher harmonic generation , reducing flexibility and making it more challenging to achieve the required power . + _ ( ii ) three photon transition : _ one would probably employ three different transitions , as indicated in , exploiting near resonant auxiliary levels to enhance the coupling amplitude . in principle it would also be possible to utilise three photons stemming all from the same laser . for li such three - photon , single colour , schemes still can proceed near resonant with the intermediate @xmath322 state @xcite . consider for example the excitation chain sketched in yellow and dark blue in , and denote the rabi - frequencies and detunings of the three transition as @xmath323 , @xmath324 with @xmath325 from lower to higher energies . since both intermediate states decay on a timescale of @xmath156 ns , they have to be far detuned with @xmath326 and @xmath327 of the order of @xmath328 for @xmath329 . here this makes sure that the effective decay rate of the intermediate levels is at least not larger than that due to the rydberg state . if we now consider the effective three - photon rabi frequency : @xmath331 we would require @xmath332 ghz to reach @xmath333 mhz as employed here , rather hard to achieve on the weak rydberg transition . there may though still be applications where much smaller effective rabi - frequencies @xmath334 are sufficient and the three - photon scheme has advantages over the uv transition . the hamiltonian given in is the @xmath3-body generalisation of the results of [ rotwave ] with added dipole - dipole interactions . it conserves the exciton - number , represented by the operator ( [ nexcitons ] ) . consequently the hamiltonian has a block - diagonal structure , with each block describing a given number of excitons . for @xmath87 atoms we explicitly have @xmath335,\:\ : m_{0}= \left [ \begin{array}{cccc } 0 & \frac{\omega_s^*}{2 } & \frac{\omega_s^*}{2 } & 0 \\ \frac{\omega_s}{2 } & -\delta_{s } & 0 & \frac{\omega_{s}^*}{2 } \\ \frac{\omega_s}{2 } & 0 & -\delta_{s } & \frac{\omega_{s}^*}{2 } \\ 0 & \frac{\omega_s}{2 } & \frac{\omega_s}{2 } & -2 \delta_{s } \\ \end{array } \right ] , \:\ : m_{2}=m_{0}\big|_{s\rightarrow p}\end{aligned}\ ] ] @xmath336 . \end{aligned}\ ] ] the bases with respect to which the three blocks are written are @xmath337 , @xmath338 , @xmath339 . in we study the nontrivial part , @xmath126 , of this hamiltonian in more detail . the basic goal of van - vleck perturbation theory was outlined in : to find a basis that block - diagonalises the hamiltonian ( [ hamil_dressed ] ) to a given order in the perturbation @xmath340 . this appendix supplies the details necessary to understand the origin of the results presented in and [ sixth_order ] . we first partition our unperturbed basis @xmath341 into two sets @xmath342 and @xmath343 , the first of which shall span the `` system '' space of interest and the second is its complement in the full hilbertspace ( the `` environment '' ) . the specific basis , @xmath341 , used for the definition of system and environment is given in . this basis is also an eigenbasis of @xmath344 . then we can construct projection operators on the system subspace @xmath345 and its complement @xmath346 . here we consider the example where @xmath347 , thus our system space is the ground - state manifold . we then have @xmath348 . with respect to these partitions the hamiltonian matrix , or similarly any other operator , can be divided into four blocks @xmath349 , @xmath350 , @xmath351 , @xmath352 . out of these , we assemble a block diagonal part @xmath353 and a block - off - diagonal part @xmath354 . since @xmath344 is diagonal in its eigenbasis , we have @xmath355 , @xmath356 . we now aim to find a unitary transformation @xmath357 that yields a block diagonal hamiltonian @xmath358 to a given order in @xmath340 , thus @xmath359 . the effective hamiltonian in the space of interest that we seek , is then given by the block @xmath360 . next , we express the unitary transformation operator @xmath357 as @xmath361 with @xmath362 . we then introduce the condition @xmath363 and @xmath364 for the block diagonal and off - diagonal parts of @xmath365 . other choices are possible , the present one distinguishes the van vleck procedure from other related schemes @xcite . for all operators in the problem , we write a series expansion in orders of the perturbation @xmath340 , most notably @xmath366 , @xmath367 , @xmath368 . as outlined in @xcite , it is then possible to obtain recursive relations for @xmath369}$ ] as series of increasingly deeper nested commutators using the baker - campbell - hausdorff formula . by segregating expansion orders in @xmath340 , the equation is then turned into a recursive set of equations for the @xmath369}$ ] . an elegant method to perform this cumbersome routine is outlined in @xcite , to which we refer for further details . ] . one extracts the full matrix @xmath370 from @xmath369}$ ] as follows . let us introduce the notation @xmath371 with latin indices to denote an eigenvector of @xmath344 within @xmath343 and @xmath372 with greek indices for one within @xmath342 . since we demand @xmath363 we only require matrix elements like @xmath373 or @xmath374 in order to know the whole matrix form of @xmath370 . now we define the resolvent operator . @xmath375 which fulfills : @xmath376 } { |\,{\alpha}\,\rangle}. \label{resolvent_eqn}\end{aligned}\ ] ] for any operator @xmath377 . substituting @xmath378 in and multiplying from the left with @xmath379 we obtain @xmath373 and also @xmath380 . from knowledge of @xmath365 one can infer the expansion orders @xmath381 through commutation relations . a series expansion of @xmath361 to the required order finally yields the perturbed eigenstates via @xmath382 we refer to ref . @xcite for all further technical details . the effective hamiltonian of the dimer in the ground - state manifold spanned by @xmath383 is given in up to fourth order , with odd orders vanishing . here we display the corresponding expression of sixth order , @xmath384 , where we use the definition @xmath385 . we can write @xmath386 with @xmath387 \bigg\ } / ( { { \bar{u } } _ { \mbox{\scriptsize 12}}}^2 - 1)^2 , \label{wpotentials6 } \\ \tilde{u}_{12}^{(6)}(r ) & = \alpha_{s}^2 \alpha_{p}^2 { { u } _ { \mbox{\scriptsize 12}}}\bigg [ \frac { \alpha_{s}^2 + \alpha_{p}^2 } { { { \bar{u } } _ { \mbox{\scriptsize 12}}}^2 - 1 } \cr & + \frac{-2 ( \alpha_{s}^2 + \alpha_{p}^2 ) + 2 { { \bar{u } } _ { \mbox{\scriptsize 12}}}^2 \frac{(\delta_{s}^2 + \delta_{p}^2)}{(\delta_{s } + \delta_{p } ) } \left ( \frac{\alpha_{p}^2}{\delta_{s } } + \frac{\alpha_{s}^2}{\delta_{p } } \right ) } { ( { { \bar{u } } _ { \mbox{\scriptsize 12}}}^2 - 1)^2 } \bigg ] . \label{vpotentials6}\end{aligned}\ ] ] recall that@xmath388 and @xmath389 . finally , for both cases shown in , the effective hamiltonian of has to be adjusted for @xmath390 . in terms of the basis @xmath391 we obtain : @xmath392 , \\ \tilde{u}_{jj}(r ) & = \alpha_{s}^2 \alpha_{p}^2 \left ( \sum_{k\neq j } \frac{1}{1 - { { \bar{u } } _ { \mbox{\scriptsize kj}}}^2 } \right)(\delta_{s } + \delta_{p } ) \delta_{ij } \cr & + \alpha_{s}^2 \alpha_{p}^2 \frac { { { u } _ { \mbox{\scriptsize ij } } } } { 1 - { { \bar{u } } _ { \mbox{\scriptsize ij}}}^2 } ( 1-\delta_{ij}).\end{aligned}\ ] ] 10 [ 1]`#1 ` urlstyle [ 1]doi:#1 m. d. lukin , m. fleischhauer , r.ct , l. m. duan , d. jaksch , j. i. cirac and p. zoller ; _ dipole blockade and quantum information processing in mesoscopic atomic ensembles _ ; phys . lett . * 87 * 037901 ( 2001 ) . h. schempp , g. gnter , c. s. hofmann , c. giese , s. d. saliba , b. d. depaola , t. amthor , m. weidemller , s. sevinli and t. pohl ; _ coherent population trapping with controlled interparticle interactions _ ; phys . lett . * 104 * 173602 ( 2010 ) . v. bendkowsky , b. butscher , j. nipper , j. b. balewski , j. p. shaffer , r. lw , t. pfau , w. li , j. stanojevic , t. pohl and j. m. rost ; _ rydberg trimers and excited dimers bound by internal quantum reflection _ ; phys . * 105 * 163201 ( 2010 ) . i. i. beterov , i. i. ryabtsev , d. b. tretyakov and v. m. entin ; _ quasiclassical calculations of blackbody - radiation - induced depopulation rates and effective lifetimes of rydberg @xmath393 , @xmath394 , and @xmath395 alkali - metal atoms with @xmath396 _ ; phys . a * 79 * 052504 ( 2009 ) . a. v. gorshkov , p. rabl , g. pupillo , a. micheli , p. zoller , m. d. lukin and h. p. bchler ; _ suppression of inelastic collisions between polar molecules with a repulsive shield _ ; phys . * 101 * 073201 ( 2008 ) . m. reetz - lamour , t. amthor , j. deiglmayr and m. weidemller ; _ rabi oscillations and excitation trapping in the coherent excitation of a mesoscopic frozen rydberg gas _ ; phys . lett . * 100 * 253001 ( 2008 ) . l. j. radziemski , r. engleman , jr . and j. w. brault ; _ fourier - transform - spectroscopy measurements in the spectra of neutral lithium , @xmath402li i and @xmath403li i ( li i ) _ ; phys . rev . a * 52 * 4462 ( 1995 ) . g. d. stevens , c. iu , t. bergeman , h. j. metcalf , i. seipp , k. t. taylor and d. delande ; _ precision measurements on lithium atoms in an electric field compared with r - matrix and other stark theories _ ; phys . rev . a * 53 * 1349 ( 1996 ) . y. ralchenko , a. e. kramida , j. reader and nist asd team ( 2010 ) ; nist atomic spectra database ( ver . 4.0.1 ) , [ online ] . available : http://physics.nist.gov/asd [ 2011 , may 27 ] . national institute of standards and technology , gaithersburg , md .

a sample of 27 planetary nebulae ( pns ) in the small magellanic clouds ( smc ) have been observed with the hubble space telescope imaging spectrograph ( _ hst_/stis ) to determine their morphology , size , and the spatial variation of the ratios of bright emission lines . the morphologies of smc pns are similar to those of lmc and galactic pns . however , only a third of the resolved smc pns are asymmetric , compared to half in the lmc . the low metallicity environment of the smc seems to discourage the onset of bipolarity in pns . we measured the line intensity , average surface brightness ( sb ) , and photometric radius of each nebula in h , h , [ oiii ] and 5007 , [ nii ] and 6584 , [ sii ] and 5731 , he i , and [ oi ] and 6363 . we show that the surface brightness to radius relationship is the same as in lmc pns , indicating its possible use as a distance scale indicator for galactic pns . we determine the electron densities and the ionized masses of the nebulae where the [ sii ] lines were measured accurately , and we find that the smc pns are denser than the lmc pns by a factor of 1.5 . the average ionized mass of the smc pns is 0.3 m . we also found that the median [ oiii ] /h intensity ratio in the smc is about half than the corresponding lmc median . we use cloudy to model the dependence of the [ oiii ] /h ratio on the oxygen abundance . our models encompass very well the average observed physical quantities . we suggest that the smc pns are principally cooled by the carbon lines , making it hard to study their excitation based on the optical lines at our disposal .

astro-ph0307351

to understand the final phases of stellar evolution of intermediate mass stars , one has to study the interplay of the stellar and nebular components . during its lifetime , and especially toward the end of its life , a star of mass between 1 and 8 m@xmath3 experiences extreme mass loss , and nuclear burning episodes followed by the occurrence of the dredge - ups that change the chemical mix of its envelope . a planetary nebula is formed by the ejected envelope of its central star at the tip of the agb , and it can be shaped in part by a fast wind from the remnant , during its subsequent evolution . the process of pn formation and evolution has been studied both theoretically and observationally for decades , yet many fundamental questions remain only partially solved . for example , what is the origin of the different morphologies , and what phenomenon , or series of phenomena , decide the final pn morphology ? does progenitor mass alone dictate the dynamic evolution rate and expansion speed ? what is the connection between binary star evolution and the physical properties of pns ( chemistry , morphology , etc . ) ? what is the role of the interstellar medium ( ism ) on pn structure ? how much is the nebular gas mix contaminated by the interaction with the interstellar medium ? to what extent do pns enrich c , n , and o in the ism ? a number of studies based on the comparison of pn and stellar properties with models have shed light on some of these and other questions . even so , observations of planetary nebulae and their central stars are notoriously difficult to compare with adequate theoretical models for a variety of reasons . first , the stellar and nebular components are often observed and analysed separately , yet they are of common origin ; the nebula carries , in its chemical composition , the nuclear burning and dredge up history of the stellar progenitor , and its morphology carries the signature of the wind history of its central star . second , the inferred evolutionary times of the stellar and nebular component within the post - agb phase are unsynchronized . although theoretical models have been developed to explain the nuclear evolution of the agb and post - agb phases , mass loss rates and geometries of the envelope ejection have not yet been successfully modeled from the stellar viewpoint . dynamic theoretical models have explained the observed nebular morphology only if ad hoc assumptions are made about the evolution of the central star . photoionization models are able to predict the strength of optical and uv emission lines that appear during the evolution of pns , but once again without yielding much insight into the numerous details of morphology , the evolution of the progenitor star . we believe that only with large samples of well defined , high quality data sets that allow the identification and significance of the crucial nebular and stellar physical parameters can one make steps forward in this field of study . large amounts of data of high quality will be able to constrain the stellar and nebular models , so that open questions on pn formation and evolution may be addressed . in the quest for acquiring the best observational sets for model testing , pns in the magellanic clouds ( smc , lmc ) deserve special attention . lmc and smc pns have known distances and low field reddening , in contrast with galactic pns . our aim in the last few years has been that of acquiring and analyzing a large sample of lmc and smc pns spectra and images , that , together with the data existing in the literature , permits the construction of an ideal data set for model testing , and for comparison with disk and halo galactic pns . the importance of observing and classifying smc pns morphology , and discerning the evolutionary stage of their central stars , is multifold . first , the ambient metallicity of the smc is closer to that of a primordial galaxy than the lmc or the galaxy are , thus making the smc an unique astrophysical laboratory to study metal depleted yet resolved stellar populations . second , by observing a sizable sample of smc pns with _ hst _ ( i.e. , resolving their morphology and size ) we will be able to study the dependence between morphology and abundances in a low metallicity environment . the space telescope imaging spectrograph ( stis ) has proven to be a very efficient way to acquire broad band and monochromatic images of pns smaller than about 2 . in two previous papers of this series @xcite we have presented the results from the stis slitless spectroscopy of the first set of 29 lmc pns . in this paper , we present the results relative to the observations of 27 smc pns ( _ hst _ program no . in @xmath42 we discuss the observation strategy and the data analysis of this particular observing set ; we include there the rationale for target selection , the data calibration and slitless spectral extraction , the spectral analysis , the photometric radii determinations , and the morphological classification . in @xmath43 we present the scientific results derived from this set of images and spectra , including the discussion of the line intensity , the surface brightness in the different lines , the excitation of the nebulae , and their optical extinction , all discussed across morphological classes . section 4 discusses the results so far and the future endeavors in this study . the analysis of the central stars of smc pns from the images and spectra of the data set of program 8663 will be discussed in a separate paper . our list of 55 targets was compiled with the intent to obtain data of all known smc pns , with accurate coordinates . most targets came from the h@xmath0 survey catalog by @xcite . the majority of these pns are bright enough to be observed in snapshot mode ( less than 1 _ hst _ orbit ) . we included in the target list three objects that have been previously observed with the wide field and planetary camera 1 , and with the faint object camera . this was done to check the consistency of the morphological classification obtained with archived instruments and the stis . our _ hst _ observations were performed in snapshot mode . the targets were selected from the original list by the scheduling specialists at stsci , accordingly to the available observing slots . typically , shorter exposure ( i.e. , brighter targets ) are favored . the program was completed at the 53 @xmath5 level , which is normal for a successful snapshot proposal . among the 29 targets observed ( table 1 ) , 27 are the smc pns analyzed in this paper . two ( mg 2 and ma 1796 ) are misclassified h ii regions , whose images and analysis will be published in another paper . as it turned out , faint targets are slightly underrepresented in our sample . the observations were acquired with stis . all observations were made with the ccd detector , in direct imaging ( 50ccd ) and slitless mode . the spatial scale of the ccd is 0.051 arcsec pixel@xmath6 , corresponding to 0.0144 pc pixel@xmath6 at the distance of the smc . this allows a very good spatial resolution to study pn morphology and size , comparable to what used in ground - based imaging observations of galactic pns . each imaging observation was split in two , to allow easy cosmic rays removal . the slitless spectra were acquired with the g430 m and g750 m gratings for most nebulae . observations with the g430 m grating cover the range 4818 to 5104 at 0.28 pixel@xmath6 , and those with the g750 m grating cover the range 6295 to 6867 at 0.56 pixel@xmath6 . the exposures where planned to have a good signal - to - noise ratio in the [ oiii ] @xmath25007 and h@xmath0 lines . several additional lines have been detected in many pns , including h@xmath1 and [ oiii ] @xmath24959 using g430 m , and [ oi ] @xmath76300 , 6363 , [ siii ] @xmath8 , [ nii ] @xmath76548 , 6584 , he i @xmath26678 , and [ sii ] @xmath76716 , 6731 using g750 m . in some cases the g430 m exposures were skipped , limited by the _ hst _ snapshot duration . the observing log is reported in table 1 , where we list the targets , the observing date , the data set name , the spectral element used in the observations , and the exposure time and number of exposures obtained . aliases are identified in the table notes . the stis data were calibrated using the standard pipeline system , as in the lmc data @xcite . figures 1 through 6 show the observed smc pns in the three observing modes : left panels show the broad band ccd images ; central panels show the h@xmath0 and [ nii ] images ; right panels show , when available , the [ oiii ] 5007 images . spectral analysis of the smc pns have been performed similarly to that of lmc pns @xcite . for most pns observed , the combination of dispersion and spatial scale allows a clear separation of the monochromatic images for all emission lines . exceptions are j 27 , where broad and monochromatic images are at the limit of detectability , and ma 1682 , where the [ nii ] and h@xmath0 images may have partial overlap . no images are severely overlapped , thus the one dimensional spectral extraction was adequate for nebular line flux and ratio analysis . we extract the one - dimensional spectra and applied a photometric calibration using the standard stis calibration pipeline module * x1d * @xcite . we used extraction boxes for the nebulae large enough to encompass all the nebular features , but snug enough as to exclude most of the sky background from the extraction . sky background regions were selected for each object to avoid stray stellar photons from field stars . the background was then averaged and subtracted . we measured emission line intensities with iraf * splot * task , fitting gaussians to individual lines , while estimating the continuum level . in the cases in which the emission lines were notably non gaussian , we estimated the line flux as measured from the area above the continuum . in table 2 we report the measured line intensities for the smc pns in our sample . column ( 1 ) gives the common names ; column ( 2 ) gives the logarithmic h@xmath1 intensities , not corrected for extinction , in erg @xmath9 s@xmath6 ; column ( 3 ) lists the logarithmic optical extinction at h@xmath1 @xcite ; columns ( 4 ) to ( 14 ) give the line intensities for each nebula , relative to h@xmath1 = 100 , not corrected for extinction . line identification are given at the heads of the columns . from the analysis of our spectral line measurements , and given the similarities of this data set to the ones analyzed in @xcite , we confirm that the errors in the line intensities of table 2 ( @xmath10 ) are ( in dex ) : @xmath11 if @xmath12 ; @xmath13 if @xmath14 ; @xmath15 if @xmath16 ; @xmath17 if @xmath18 ; and @xmath19 if @xmath20 . in figure 7 we compare the measured line intensities with those available in the literature . we used the data from selected references @xcite . some authors give the intensity corrected for extinction , so before comparison to the present results we un correct the intensity ratios by using the extinction constant given in the same reference . we have comparisons for several emission lines in mg 8 , mg 13 , smp 1 , smp 6 , smp 11 , smp 13 , smp 14 , smp 17 , smp 25 , and smp 26 . a comparison of our fluxes to with previous values ( fig . 7 ) shows a generally good agreement . the correlation coefficient between the two sets of fluxes is 0.994 , with rms scatter of 0.2 dex . uncertainties in the reference fluxes are typically quoted 20@xmath5 to 50@xmath5 for high and low fluxes respectively , while our errors are generally smaller @xcite . the outlying point in figure 7 corresponds to the [ nii ] @xmath2 6584 flux measurement of target smp 17 . we believe that our measurement is correct , since the reddened [ nii ] @xmath2 6584 to @xmath2 6548 intensity ratio is 2.1 , while the reference does not detect the [ nii ] @xmath26548 line . in table 3 , columns 2 and 3 , we give the positions of the central stars of the nebulae from the continuum images , where observed . alternately , we give the geometric center positions of the pns . columns ( 4 ) and ( 5 ) of table 3 give the photometric radii of the nebulae , measured as described in @xcite , and the nebular dimensions , measured from the 10@xmath5 brightness contour . figures 1 through 6 show the smc pns images , exhibiting the same range of morphological types to those of lmc and galactic pns . the morphological classification used in this paper is the same as that in the other papers of this series @xcite : we classify pns as round , elliptical , bipolar core , bipolar , and pointsymmetric . we distinguish between round and elliptical if the axial ratio is larger than unity by at least 10 percent . bipolarity follows the classic definition @xcite , while bipolar core is defined in @xcite . note that bipolar core pns of this sample are defined as such only if their bc is apparent above the 10 percent intensity contour . morphology is based primarily on the h@xmath0 images , which differs only rarely from that of the [ oiii ] @xmath25007 images @xcite . column ( 6 ) of table 3 gives the morphological classification . in a couple of cases , we use the terminology of round and elliptical pns with inner structures . these structures may be bipolar cores , although we believe that we have the necessary spatial resolution to make the distinction . question marks indicate that the morphological determination is ambiguous or uncertain . in the following paragraphs , we describe in detail the morphological types of our sample pns . * j 4 ( e ) : the field is crowded , making the i d on the clear image a little uncertain . the spectrum is reasonably well exposed , and shows emission in the full set of emission lines . the nebular morphology is elliptical . we detected extremely faint emission in the [ oiii ] frame up to 1.42 from the center of the pn in the spatial direction . * j 18 ( r ) : this pn is round in h@xmath0 and has an elliptical shape in [ nii ] . it does not show a bipolar core . the [ nii ] @xmath26584 emission is brighter than h@xmath0 . * j 23 : this object is apparently spatially unresolved , and its identification on the clear image is difficult . the stellar spectrum shows strong and broad h@xmath0 , and a weaker @xmath26584 [ nii ] line . we also detect he i @xmath26678 . the h@xmath0 velocity width appears to be about 200 - 300 km / s , which is broad for a pn . this object could not be a common symbiotic star , it would have a giant companion with an unusually bright h@xmath0 emission line and a faint underlying continuum @xcite . * j 27 ( b ? ) : this object is roughly box - shaped , and may be the remnant of a bipolar . the nebula is barely detected in the broad - band exposure , at roughly 20 counts / pixel above the background . the spectrum shows faint [ nii ] emission , with some hint of h@xmath0 emission . no central star is detected . * ma 1682 ( b ) : this is an extreme bipolar pn , with a prominent central torus seen edge on . the h@xmath0 and [ n ii ] lines are very strong in the spectrum , and [ sii ] is barely detected ; no other lines are detected . we do nt have a blue spectrum . the central star ( cs ) is easily detected in the continuum image . * ma 1762 ( e(bc ) ) : this is an elliptical pn , with a bi - nebulous inner core . we detected h@xmath0 and marginally [ oi ] @xmath26300 . the cs is easily detected in the continuum image , and in the spectrum . * mg 8 ( e ) : this elliptical pn has a very distinct ring structure , with a hint of ansae ( or arms ) that extend 0.1 along the major axis from the 10@xmath5 contour . the arm - like structure leads to a pointsymmetric classification , but to be conservative we classify this elliptical , possibly bipolar core . h@xmath0 is very strong , and [ nii ] is apparent but weaker , [ sii ] is very weak , and we could also detect a trace of the [ oi ] @xmath26300 emission . h@xmath1 is well exposed , and [ oiii ] is only a bit stronger . the morphology in [ oiii ] differs in detail from the h@xmath0 and other line morphology , in that the emission comes primarily from a ring that is inside the emission from most other lines . the cs is exceptionally bright in the continuum image ; could it be a binary companion to the true cs ? * mg 13 ( e ) : this elliptical pn has a very knotty ring structure , and is classified as e(s ) . h@xmath0 is very strong , but neither [ nii ] nor any other line is detected in the g750 m spectrum . the cs is easily detected in the continuum image . * smp 1 : marginally resolved round or elliptical pn with no detected image of the central star . most emission lines are present , and there may be a trace of he i at 4927 as well . the [ sii ] lines are marginally detected . * smp 6 ( e ) : elliptical pn , the [ nii ] morphology is slightly different from that of the other lines , in that it shows a larger elongation . * smp 8 ( r ) : this round pn has very strong h@xmath0 and [ oiii ] emission , with [ nii ] and he i at the 1@xmath5 level , and [ sii ] weaker still . [ oi ] is marginally detected at 6300 . the cs is detected in the continuum image , but there is a significant contribution from the nebular emission . * smp 9 or j 3 ( r ) : this round pn has a bright hemisphere of emission to the ne , making the whole somewhat asymmetric in appearance . we are not sure that the inner structure clearly discloses a bipolar core , but there is a lot of structure . h@xmath0 and [ oiii ] are very strong , but [ nii ] is fairly weak . the asymmetry is pronounced in the [ nii ] lines . [ sii ] and he i , are also present but very weak , and [ oi ] @xmath26300 may be marginally detected . the cs has been detected in the continuum image . * smp 11 or j 8 ( b ) : this complex bipolar pn has a distinct dark band that divides the bright core into two , unequal lobes . this looks very much like a case of a strong bipolar ring , viewed some distance from the plane . a faint bipolar structure is even evident up to 2 from the center ( particularly to the south ) , but the ribbed , fan - like structure is projected off - center on the sky , and [ nii ] @xmath26548 is blended with the h@xmath0 gaussian wing . h@xmath0 and [ oiii ] are very strong , but [ nii ] @xmath26584 is fairly weak ; these lines all show the same morphological structures . [ sii ] and he i , are also present but very weak , and [ oi ] @xmath26300 is also detected . the cs is detected in the continuum image , and possibly also in the g750 m spectrum . * smp 12 ( e ) : elliptical pn with detected central star . the central emission shows some structure ( es ) . in the [ oiii ] image there is a hint of extended lobe to the east . * smp 13 or j 11 ( r ) : the appearance of this pn is nearly stellar in the continuum , but appears very mildly elliptical in the ionization lines . we classify it as round , since the ellipticity is below the 10@xmath5 level . faint emission in the [ oiii ] image is detected up to 1.42 from the geometrical center . h@xmath0 and [ oiii ] are very strong , [ nii ] , he i , and [ sii ] are very weak but detected . also weak are [ oi ] @xmath76300 , 6363 , and [ s iii ] @xmath26312 . the cs may be detected in the continuum image . * smp 14 ( r ) : this round pn has a distinct structure in the inner core and faint ansae that extend 0.3 from the 10@xmath5 contour . h@xmath0 is very strong , and [ nii ] is detected in the spectrum , but is very weak . [ sii ] and he i , are also present but weak . [ oiii ] is quite strong . the cs is easily detected in the continuum image . * smp 17 ( e ) : this is an elliptical pn , with a very faint outer halo that is detected in the continuum image , in h@xmath0 , and in [ oiii ] at the 0.4@xmath5 contour level , out to about 1.3 . the inner portion is somewhat elliptical in all the observed lines except [ nii ] , where it is ring - like . the [ oiii ] emission shows a marginal bipolar core . h@xmath0 is very strong , and [ nii ] is present but very weak . the cs is detected in the continuum image , but there is a significant contribution from the nebular emission . * smp 18 or j 19 : this small pn with bright h@xmath0 and [ oiii ] emission . the [ nii ] extension is smaller than in the other lines . it is barely resolved when compared to the broad band images of the stars . * smp 19 or j 20 ( r ) : morphologically similar to j 3 , the inner parts of this pn are slightly distorted in the [ oiii ] lines and h@xmath0 . we classify it as r with structures . there is outer emission of 1 radius . * smp 20 : this pn is bright but unresolved . h@xmath0 and [ oiii ] are very strong , with [ nii ] at the 1@xmath5 level , and he i , [ sii ] , and [ oi ] weaker still . the cs is detected in the continuum image , but there is a significant contribution from the nebular emission . the stellar spectrum shows a feature at about 6577 , which is also seen in a few other targets in our sample . * smp 22 ( b ? ) : this pn has a very interesting box / ring shape , with a possible emission `` arm '' toward the east , which is very evident in the [ nii ] images ( could it be part of the ism ? ) . it is classified as bipolar . there are at least three high emission features within the ring . the line profiles of all lines but h@xmath0 and h@xmath1 are split , showing the ring morphology . * smp 23 or j 26 ( e(bc ) ) : elliptical pn with a bipolar core . * smp 24 ( e ) : elliptical pn in the h@xmath0 and [ oiii ] emission lines , it shows some small scale structures in the light of [ nii ] . this nebula may have a faint halo . * smp 25 ( e ) : elliptical planetary nebula with likely broad stellar h@xmath0 emission line . * smp 26 ( p ) : this pn could be pointsymmetric , with two arms that extend from the ne and sw of the center . it is possible that this pn is elliptical with low ionization ansae . h@xmath0 is fairly strong , but [ nii ] is nearly as bright , and shows the arms much more clearly . the line profiles in the 1d spectra are split in all lines except h@xmath0 . no cs is detected in the continuum image . * smp 27 ( r ) : this is a ( very low ellipticity ) elliptical pn . there is a hint of structure in the [ nii ] images only . h@xmath0 and [ oiii ] are very strong , with [ nii ] at the 1@xmath5 level , and he i , [ sii ] , and [ oi ] weaker still . the cs is detected in the continuum image , but there is a significant contribution from the nebular emission . * sp 34 ( r ) : this is a round pn , with a faint outer halo that is detected in the continuum image and in h@xmath0 at the 5 @xmath5 contour level . the inner portion shows significant asymmetry in all the observed lines , in that the western edge is much brighter . h@xmath0 is fairly strong , and [ nii ] is present but much weaker . no cs is detected in the continuum image . we included in our stis target list three objects that were previously observed with the planetary camera 1 and the faint object camera . the rationale to include a few repeat targets from the sample of @xcite was to check the morphological types and dimensions with both methods and to asses the reliability of the pre - costar archival data . all three were observed before the first _ hst _ servicing mission , thus the images suffer for the uncorrected spherical aberration of the telescope mirror . in an _ hst _ archival study @xcite it was found that morphology was well determined with the archived instruments , while dimensions of the nebulae with the photometric method were not always reliable because of the possible presence of fainter extended halos around the images . since program no . 8366 was a snapshot program , we could not guarantee exactly which targets would be observed for this comparison . in the end , targets smp 1 ( n 1 ) , smp 6 ( n 6 ) , and smp 22 ( n 67 ) were re - observed with stis . the first two targets were observed with pc1 earlier , while smp 22 had been observed with foc . smp 1 and smp 6 are small nebulae with very uncomplicated structures . the round ( e ? ) morphology of these two targets was easily seen in the pc1 data set @xcite . the early photometric radius measured for these two pns were 0.12 and 0.152 , respectively for smp 1 and smp 6 was close to our measurements of 0.15 and 0.19 . the situation of the bipolar planetary nebula smp 22 is very different . the complete nebular morphology that we see in most emission lines of the stis spectra , but in particular in the [ nii ] lines , is not as evident in the foc image @xcite , figure 5 . furthermore , we measure a photometric radius of 0.4 , while the radius from the foc image is almost three times larger . we reanalyzed these measurements , and noted that a radius of 0.4 already included about 75@xmath5 of the total flux , but to encircle the 85@xmath5 of the flux required 1.37 radius . from the comparison of the old and new smp 22 images we conclude that the main morphological features of mc pns are reliable from the pre - costar images , even if the detailed morphology was not resolved . on the other hand , the measurements of the photometric radii are not reliable . from our analysis we can see that smc pns are nicely classified using the same morphological scheme as the lmc and galactic pns . in order to increase the sample size for statistical purposes , we include in the present sample three additional smc pns described and classified in @xcite . after eliminating unresolved objects and repeats , we have at our disposal a sample of 30 smc pns whose morphology is well determined . this smc pn sample constitute nearly 50 percent of all known smc pns , thus we consider fairly representative at least of the bright pns . in table 4 we give the statistics of pn morphology for the smc sample as compared to the lmc and galactic samples from @xcite . the lmc and smc samples in columns 2 and 3 of table 4 , have been selected in similar ways , and they have similar observational biases ( but none of the extreme selection biases that affect galactic pn samples toward the galactic plane ) . while the fraction of round pns remains more or less the same in the two samples , the fractions of e and bc in the smc are respectively twice and one - half those of the lmc . the overall frequency of asymmetric pn in the smc is only sixty- percent that of the lmc . this remarkable results strongly suggests that the difference between the magellanic clouds is reflected in their pn populations : something in the environment of the smc may not not be favorable to the formation of bipolar and bipolar core pns . alternately , this may be the result of the low metallicity of the smc , indicating that the production of the higher mass progenitors of pns in the smc has long subsided , and that the low percentage of asymmetric pns in the smc is due to the lack of recent star formation episodes . this issue will be better framed once we have complete abundance analysis for the pns analyzed in this paper . we plot the surface brightness versus radius relation for smc ( fig . 8) and smc and lmc pns together , using the lmc data from @xcite ( fig . 9 ) . in figures 8 and 9 the pns are coded for their morphological types . we exclude from these plots round and elliptical pns ( both smc and lmc ) with inner ( and outer ) structure , to avoid possible misclassifications ( see above ) . these figures confirm for smc pns the trend that we found for lmc pns : surface brightness in the light of [ oiii ] @xmath25007 , h@xmath0 , h@xmath1 , [ oi ] @xmath26300 , [ nii ] @xmath26584 , and [ siii ] @xmath26312 evolves differently for the different morphological types . the surface brightness to radius relation is very tight in all spectral lines , with the exception of the [ nii ] emission line , where a larger spread is present . from the lower left panels of figures 8 and 9 we see that the [ nii ] spread is more extreme for bipolar pns . a possible factor is the higher but more varied nitrogen abundances of bipolar and bc pns . in figure 10 we plot a histogram of the ratio of reddening - corrected fluxes of the [ oiii ] @xmath25007 and h@xmath1 lines ( hereafter `` [ oiii ] /h@xmath1 '' ) for the pns of the smc and the lmc . the median of the smc distribution is a factor of two lower than for the corresponding lmc distribution . specifically , our stis data yield galaxy averages @xmath21[oiii ] /h@xmath1 @xmath22 = 5.7 @xmath23 2.5 and @xmath21[oiii ] /h@xmath1 @xmath24 = 9.4 @xmath23 3.1 . to the best of our knowledge this result is free of object selection biases since both sets of targets were chosen in much the same way . the objects with very low [ oiii ] /h@xmath1 ratios are very low - ionization objects whose central stars are presumably too cool to form much o@xmath25 . however , the [ oiii ] /h@xmath1 ratio tends to reflect that of the pns with the brightest [ oiii ] and h@xmath1 lines . in general we tend to favor targets with hottest central stars : t@xmath26 k. the low median value of [ oiii ] /h@xmath1 in the smc has been noted before , from ground - based measurements @xcite . however , the stis images allow us to distinguish between small , bright hii regions and pns . using the ground - based data from @xcite we derive substantially the same result : @xmath21[oiii ] /h@xmath1 @xmath22 = 4.0 @xmath23 2.8 and @xmath21[oiii ] /h@xmath1 @xmath24 = 9.2 @xmath23 4.2 . for reference , the [ oiii ] /h@xmath1 ratio for galactic pns is of the order of 15 . the [ oiii ] /h@xmath1 emissivity ratio is physically scaled linearly with the o / h abundance and the fractional ionization of o@xmath25 . also it depends exponentially on the local electron excitation temperature , t@xmath27(o@xmath25 ) since electron collisions on the high - energy tail of the free energy distribution excite the transition . of course , t@xmath27(o@xmath28 depends on o / h and o@xmath25/o as well . so interpreting the differences between the [ oiii ] /h@xmath1 ratios of the smc and the lmc is best done using ionization models . we used the ionization model cloudy @xcite to understand the systematic trends in the behavior of [ oiii ] /h@xmath1 for the smc , lmc and the galaxy . we adopted a gas density of 1000 @xmath29 and standard chemical abundances for the three environments . galactic abundances for the pns of the galaxy are those adopted in the paris meeting ( 1985 ; see table 7 in the hazy manual , ferland 1996 ) . we used lmc average abundances of pns quoted in @xcite and smc average abundances from @xcite except for the c / h ratio which comes from @xcite , where we had selected from their table 3 only the low - error data . the stellar ionizing spectrum is assumed to be a blackbody with temperatures and luminosities from the h - burning evolutionary tracks for the appropriate galaxian population by @xcite . our model parameters are summarized in table 5 . the predictions of [ oiii ] /h@xmath1 from cloudy models are shown in figure 11 for the smc , lmc , and galactic pns . the outcomes of the cloudy models are rather insensitive to the adopted stellar properties and the assumed density ( e.g. , fig . but they are very sensitive to the adopted abundance ratios , especially the oxygen abundance . the model values of [ oiii ] /h@xmath1 are shown as a function of stellar temperature over the range of temperatures encountered by a star of mass = 2 m@xmath3 , as it evolves from the agb tip to its maximum post - agb temperature . at stellar temperatures in excess of @xmath30 k the [ oiii ] /h@xmath1 ratio is in rough agreement with the present observations ( see also @xcite ) . it is worth noting that the galactic models presented here are not to be compared directly with the known observed [ oiii ] /h@xmath1 distribution , without accounting for selection effects that hamper galactic pn statistics . we should clarify that the results of figure 11 are valid for the input abundances . if , for example , we chose to use as input the average oxygen abundance for smc ons from @xcite instead of that of @xcite , the [ oiii ] over h@xmath1 emission line would be as high as 7.2 . the cooling processes that determine t@xmath27(o@xmath28 in the smc , lmc and galactic pns are noteworthy . in the galaxy the primary coolants of pns with hot central stars are the optical forbidden lines of [ oiii ] @xmath25007 and other lines of o@xmath31 and o@xmath25 . however , in environments in which o / h is as low as in the smc , the primary coolants may become ultraviolet intercombination lines of c@xmath31 and c@xmath25 . it will be interesting to confirm these predictions with future uv observations . before reaching conclusions and further speculation , let us explore the weight of our assumptions . one is that these models have constant hydrogen density , which is arbitrarily assumed equal to 1000 @xmath29 . by running cloudy for a constant density model , the outer radius is determined by the stroemgen sphere . this makes our models larger as stellar temperature increases , reaching a maximum and then decline with the [ oiii ] @xmath25007/h@xmath1 intensity ratio . but in order to compare the radii of our models with our data we should make a model for each nebula with the correct radius ( i.e. , assuming a density profile that reproduces that pn , for example , as determined in hydrodynamic calculations ) . this will be done in detail in a future paper . for now , let us examine what a difference in the average hydrogen density will make in the [ oiii ] 5007 flux ratio . in figure 12 we plot the [ oiii ] /h@xmath1 ratio against the oxygen abundance for the early and the hottest models in the smc ( triangles ) , lmc ( squares ) , and the galaxy ( circles ) . the early models correspond to models 1 , 6 , and 11 of table 5 , with log n@xmath32=3 @xmath29 . the hot models have been calculated for log n@xmath32=2.5 , 3 , and 3.5 @xmath29 . we see that varying the density does not affect very much the studied intensity ratio , especially for the low oxygen models . the different hot smc models in figure 12 are all within @xmath33 i(5007)/i(h@xmath1 ) = 1 , and their outer radii are 0.11 , 0.25 , and 0.55 parsecs for log n@xmath32=3.5 , 3 , and 2.5 sm@xmath34 respectively . all models described in this paper have filling factor ( @xmath35 ) equal to unity . changing the value of the filling factor will also change the outer radius of the nebular models . for example , the hot galactic model ( with log n@xmath32=3 ) with @xmath35=0.5 has an outer radius about 1.3 times larger than the model with @xmath35=1 . we find that the [ oiii ] @xmath25007/h@xmath1 intensity ratio with the different radius and filling factor is 95@xmath5 the ratio in the unity filling factor model , thus the filling factor assumption is not very important the discussion of the [ oiii ] 5007 line intensity with respect to h@xmath1 . oxygen is usually the major coolant in the oxygen abundant pns . but what other coolants play a role ? in figure 13 we show the intensity of the line relative to h@xmath1 for the major coolants in the smc , lmc , and galactic pns , versus the oxygen abundance . the intensities are from the hottest models for each galaxian mix . cloudy predicts that although the 5007 line is the major coolant for the galactic and lmc pns , in the smc pns the [ civ ] @xmath21548 line is the major coolant . other carbon lines are also important coolant in the smc pns , while the [ oiii ] @xmath24959 and [ oii ] @xmath23727 lines are significant mostly for the galactic models . clearly , more detailed models and observations of [ civ ] @xmath21909 and [ civ ] @xmath21550 are essential to confirm these predictions , but the interpretation is plausible and consistent with the available data . in @xcite we used the i[oiii ] ( 5007 + 4959 ) / h@xmath0 ratio to trace the nebular excitation , and by inference the stellar temperature , of the lmc pns . we can not perform the same analysis here , since it is clear from figure 11 that the [ oiii ] /h@xmath1 ratio versus temperature relation is very non - monotonic . table 6 lists nebular densities determined from strong [ sii ] @xmath76716,6731 lines for eight smc pns ( this paper ) and twelve lmc pns @xcite . we assumed t@xmath36=10@xmath37 k for these estimates . however , the derived densities are extremely insensitive to t@xmath36 @xcite . we find that @xmath38=3.45 , whereas @xmath393.28 , a factor 1.5 different . given the small sample size , a factor of 1.5 is not likely to be very significant . furthermore , there is no discernible trend of n@xmath36 with morphological type in smc or lmc pns . moreover , nebulae with bright [ sii ] lines are strongly biased to those of generally high fluxes or surface brightness , or low ionization . these selection biases render trends in n@xmath36 of limited significance . in addition we estimate masses of pns in the lmc and smc using the method of @xcite . the average ionized mass of the eight smc pns , 0.3 m@xmath3 , is slightly larger than that of lmc pns , 0.2 m@xmath3 . given the uncertainty in the data , and the small data sample , we do not believe that the mass discrepancy between smc and lmc pns is significant . also , we fail to see any obvious trend between nebular mass and morphological type . a sample of 27 smc pns have been observed in imaging a slitless mode with _ hst_/stis to examine their morphology , shape , and fluxes , and to study their evolution . this morphological sample is the first sizable set of smc pns , and represents almost half of the known smc pns . the images and spectra have the same high quality and resolution as our lmc sample . we present the broad images and monochromatic images in the major emission lines , and determine that morphology is easily recognized in most emission lines . we find that the ratio of symmetric to asymmetric pns is remarkably different in the smc and the lmc . specifically , bipolar pns are much rarer in the smc than the lmc ( or the galaxy ) . this new result has significant implications for the relation between stellar population and pn morphology . it is well known from galactic and lmc pn studies that pn morphology correlates with the mass of the progenitor stars . in particular , bipolar pns evolve from relatively massive ( @xmath40 m@xmath3 ) progenitors . thus the low incidence of bipolar pns in the smc probably reflects the low formation rate of these stars in the past @xmath415 gyr . this is in accord with other studies of star formation rates in the lmc and the smc . alternately , or in addition , the low metallicity of the smc may inhibit wind collimation somehow . we also present the measurement of the optical line intensities . we find that the surface brightness declines with radii in most emission lines , adding value to the possible calibration of the surface brightness- nebular radius correlation seen in the lmc . ionized masses and electron densities were calculated where the [ sii ] doublet intensity was available . the resulting ionized masses do not seem to have a relation to the morphology , as already found in the lmc pns . the [ oiii ] /h@xmath1 ratio of magellanic cloud pns has been studied in detail . the factor of two difference between the lmc and smc pns were modeled with cloudy . we find that the cooling models strongly depend on the oxygen content , and that the brightest parts of the [ oiii ] 5007/h@xmath1 luminosity function shifts between smc and lmc pns . we also find a relation between the brightest [ oiii ] 5007/h@xmath1 pns and their morphology , that is different for each magellanic cloud population . we thank george jacoby and orsola de marco for discussing the faint target selection , orsola de marco for her help in completing the phase ii of the _ hst _ program , and massimo stiavelli for his suggestions about cloudy models . this work was supported by nasa through grant go-08663.01-a from space telescope science institute , which is operated by the association of universities for research in astronomy , incorporated , under nasa contract nas -26555 . j 4 & 2000 aug 18 & o65s03030 & g430 m & 840 & 2 + & & o65s03020 & g750 m & 890 & 2 + & & o65s03010 & mirvis & 300 & 2 + j 18 & 2000 oct 3 & o65s08020 & g750 m & 2100 & 2 + & & o65s08010 & mirvis & 300 & 2 + j 23 & 2000 aug 8 & o65s12020 & g750 m & 2100 & 2 + & & o65s12010 & mirvis & 300 & 2 + j 27 & 2000 oct 10 & o65s16020 & g750 m & 2100 & 2 + & & o65s16010 & mirvis & 300 & 2 + ma 1682 & 2000 nov 5 & o65s24020 & g750 m & 2100 & 2 + & & o65s24010 & mirvis & 300 & 2 + ma 1762 & 2000 aug 19 & o65s25020 & g750 m & 2100 & 2 + & & o65s25010 & mirvis & 300 & 2 + ma 1796 & 2000 sep 27 & o65s26040 & g430 m & 200 & 2 + & & o65s26020 & g750 m & 100 & 2 + & & o65s26030 & g750 m & 400 & 2 + & & o65s26010 & mirvis & 120 & 2 + & & o65s26kgq & mirvis & 15 & 2 + mg 2 & 2000 nov 7 & o65s28030 & g430 m & 800 & 2 + & & o65s28020 & g750 m & 1300 & 2 + & & o65s28010 & mirvis & 300 & 2 + mg 8 & 2000 nov 7 & o65s34030 & g430 m & 1200 & 2 + & & o65s34020 & g750 m & 620 & 2 + & & o65s34010 & mirvis & 120 & 2 + mg 13 & 2000 sep 28 & o65s39020 & g750 m & 2100 & 2 + & & o65s39010 & mirvis & 300 & 2 + smp 1 & 2001 feb 6 & o65s40030 & g430 m & 360 & 2 + & & o65s40020 & g750 m & 170 & 2 + & & o65s40010 & mirvis & 120 & 2 + smp 4 & 2001 jan 9 & o65s41030 & g430 m & 460 & 2 + & & o65s41020 & g750 m & 230 & 2 + & & o65s41010 & mirvis & 120 & 2 + smp 6 & 2001 mar 9 & o65s42030 & g430 m & 280 & 2 + & & o65s42020 & g750 m & 190 & 2 + & & o65s42010 & mirvis & 120 & 2 + smp 8 & 2001 jan 23 & o65s43030 & g430 m & 320 & 2 + & & o65s43020 & g750 m & 160 & 2 + & & o65s43010 & mirvis & 120 & 2 + & & o65s43gkq & mirvis & 15 & 1 + smp 9 & 2000 nov 20 & o65s02030 & g430 m & 1100 & 2 + & & o65s02020 & g750 m & 790 & 2 + & & o65s02010 & mirvis & 300 & 2 + smp 11 & 2000 sep 6 & o65s05040 & g430 m & 200 & 2 + & & o65s05020 & g750 m & 60 & 2 + & & o65s05030 & g750 m & 300 & 2 + & & o65s05010 & mirvis & 120 & 2 + & & o65s05lbq & mirvis & 15 & 1 + smp 12 & 2000 oct 14 & o65s44030 & g430 m & 650 & 2 + & & o65s44020 & g750 m & 810 & 2 + & & o65s44010 & mirvis & 300 & 2 + smp 13 & 2000 nov 20 & o65s06040 & g430 m & 52 & 2 + & & o65s06020 & g750 m & 36 & 2 + & & o65s06030 & g750 m & 180 & 2 + & & o65s06010 & mirvis & 120 & 2 + & & o65s06chq & mirvis & 15 & 1 + smp 14 & 2000 sep 17 & o65s45030 & g430 m & 540 & 2 + & & o65s45020 & g750 m & 290 & 2 + & & o65s45010 & mirvis & 120 & 2 + smp 17 & 2001 jan 16 & o65s47040 & g430 m & 160 & 2 + & & o65s47020 & g750 m & 100 & 2 + & & o65s47030 & g750 m & 100 & 2 + & & o65s47010 & mirvis & 120 & 2 + & & o65s47fxq & mirvis & 15 & 1 + smp 18 & 2001 may 16 & o65s09040 & g430 m & 160 & 2 + & & o65s09020 & g750 m & 42 & 2 + & & o65s09030 & g750 m & 210 & 2 + & & o65s09010 & mirvis & 120 & 2 + & & o65s09jiq & mirvis & 15 & 1 + smp 19 & 2000 oct 14 & o65s10030 & g430 m & 330 & 2 + & & o65s10020 & g750 m & 260 & 2 + & & o65s10010 & mirvis & 120 & 2 + smp 20 & 2001 jan 21 & o65s48040 & g430 m & 160 & 2 + & & o65s48020 & g750 m & 80 & 2 + & & o65s48030 & g750 m & 320 & 2 + & & o65s48010 & mirvis & 120 & 2 + & & o65s48yxq & mirvis & 15 & 1 + smp 22 & 2000 oct 14 & o65s49030 & g430 m & 840 & 2 + & & o65s49020 & g750 m & 200 & 2 + & & o65s49010 & mirvis & 120 & 2 + smp 23 & 2000 oct 14 & o65s15030 & g430 m & 230 & 2 + & & o65s15020 & g750 m & 380 & 2 + & & o65s15010 & mirvis & 120 & 2 + smp 24 & 2001 mar 9 & o65s50030 & g430 m & 500 & 2 + & & o65s50020 & g750 m & 150 & 2 + & & o65s50010 & mirvis & 120 & 2 + smp 25 & 2001 mar 5 & o65s51030 & g430 m & 920 & 2 + & & o65s51020 & g750 m & 460 & 2 + & & o65s51010 & mirvis & 120 & 2 + smp 26 & 2000 sep 28 & o65s52030 & g430 m & 1200 & 2 + & & o65s52020 & g750 m & 900 & 2 + & & o65s52010 & mirvis & 300 & 2 + smp 27 & 2001 jan 19 & o65s53040 & g430 m & 160 & 2 + & & o65s53020 & g750 m & 80 & 2 + & & o65s53030 & g750 m & 360 & 2 + & & o65s53010 & mirvis & 120 & 2 + & & o65s53qbq & mirvis & 15 & 1 + sp 34 & 2000 oct 14 & o65s55030 & g430 m & 1000 & 2 + & & o65s55020 & g750 m & 500 & 2 + & & o65s55010 & mirvis & 300 & 2 + j 4 & @xmath42 & 0.168 & 159.1 & 480.4 & 5.1 & 3.8 & 2.5 & 28.9 & 326.3 & 102.5 & 1.9 & 1.6 & 3.0 + j 18 & @xmath43 & & & & 1.736 & & & 107.2 & 100 & 330.4 & & 4.2 & 2.928 + j 23 & @xmath44 & & & & & & & & 100 & 13.9 & & & + j 27 & @xmath45 & & & & & & & & 100 & & & & + ma 1682 & @xmath46 & & & & & & & & 100 & 304.5 & & + ma 1762 & @xmath47 & & & & 0.7829 & & & & 100 & & & & + mg 8 & @xmath48 & 0.133 & 41.4 & 126.5 & 2.0 & 1.6 & & 22.9 & 317.2 & 98.3 & 4.7 & 5.3 & 22.0 + mg 13 & @xmath49 & & & & & & & & 100 & & & & + smp 1 & @xmath50 & 0.287 & 85.0 & 262.1 & 2.0 & 0.9 & 0.7 & 8.9 & 359.3 & 25.7 & 4.3 & 0.3 & 0.7 + smp 6 & @xmath51 & 0.385 & 267.7 & 791.1 & 7.0 & 1.9 & 1.6 & 13.7 & 388.6 & 44.5 & 5.3 & 0.8 & 1.6 + smp 8 & @xmath52 & 0.026 & 198.7 & 592.3 & & & & 1.3 & 291.0 & 2.5 & 3.6 & & + smp 9 & @xmath53 & 0.070 & 318.8 & 959.4 & 1.4 & 1.6 & & 8.1 & 301.4 & 23.5 & & 9.7 & 6.8 + smp 11 & @xmath54 & 0.352 & 110.9 & 351.4 & 5.0 & & & 10.0 & 378.6 & 28.5 & 3.9 & 3.3 & 8.0 + smp 12 & @xmath55 & 0.056 & 203.5 & 613.3 & & & & & 298.0 & & 3.5 & & + smp 13 & @xmath56 & 0.190 & 277.3 & 828.1 & 2.8 & 0.8 & 1.0 & 2.9 & 332.0 & 8.8 & 3.9 & 0.8 & 1.3 + smp 14 & @xmath57 & 0.069 & 311.1 & 928.8 & 2.6 & 1.1 & 1.0 & 4.4 & 301.2 & 10.2 & 1.0 & 1.8 & 3.0 + smp 17 & @xmath58 & 0.064 & 294.0 & 893.2 & 2.8 & 0.3 & 1.2 & 3.7 & 300.0 & 7.6 & 4.0 & 1.1 & 0.5 + smp 18 & @xmath59 & 0.122 & 105.0 & 309.0 & 1.3 & 0.7 & 0.4 & 7.3 & 314.5 & 24.0 & 3.6 & 0.4 & 0.7 + smp 19 & @xmath57 & 0.161 & 282.4 & 847.2 & 4.0 & 1.2 & 0.9 & 3.4 & 324.5 & 7.7 & 2.4 & 1.5 & 2.2 + smp 20 & @xmath60 & -0.019 & 143.8 & 434.9 & & & & 1.0 & 280.8 & 2.8 & 3.9 & & + smp 22 & @xmath61 & 0.165 & 99.1 & 298.2 & 14.8 & & 5.6 & 87.7 & 325.4 & 259.6 & 2.9 & 9.8 & 14.3 + smp 23 & @xmath62 & 0.101 & 271.5 & 828.1 & & & & & 309.2 & 3.6 & 3.6 & & + smp 24 & @xmath59 & 0.047 & 137.9 & 420.1 & 1.9 & 0.9 & & 6.2 & 295.9 & 18.8 & 4.0 & 2.1 & 2.5 + smp 25 & @xmath63 & 0.100 & 108.8 & 328.2 & 3.2 & 1.6 & & 15.8 & 309.0 & 47.6 & 5.6 & 1.3 & 1.9 + smp 26 & @xmath64 & 0.253 & 205.0 & 613.0 & 19.7 & & 5.2 & 111.1 & 349.4 & 327.2 & 2.0 & 14.8 & 17.7 + smp 27 & @xmath65 & 0.040 & 188.8 & 580.1 & 1.4 & 0.5 & 0.4 & 1.8 & 294.2 & 5.2 & 3.8 & 0.2 & 0.6 + sp 34 & @xmath66 & 0.163 & 162.8 & 488.4 & 12.7 & & 22.4 & 19.0 & 325.1 & 55.8 & & 14.3 & 14.6 + j 4 & 0@xmath67 45@xmath68 2730 & @xmath6942153 & 0.32 & 0.27 & e & possible ansae . + j 18 & 0 51 43.39 & @xmath69 00 54.1 & 0.17 & 0.14 & r ? & elongated in [ nii ] . + j 23 & 0 55 30.52 & @xmath70 50 21.3 & 0.38 & & & broad h@xmath0 emission ; unresolved . + j 27 & 0 59 43.48 & @xmath70 57 17.9 & & 2.5 x 1.7 & b ? & very low surface brightness . + ma 1682 & 1 09 03.52 & @xmath70 29 05.2 & 0.83 & 2.86 x 2.17 & b & + ma 1762 & 1 12 40.28 & @xmath70 53 46.4 & 0.71 & 1.45 x 1.26 & e(bc ) & + mg 8 & 0 56 19.59 & @xmath70 06 58.5 & 0.48 & 1.39 x 1.28 & e & inner structure . + mg 13 & 1 43 10.33 & @xmath70 57 03.2 & 1.55 & 1.22 x 1.09 & e & inner structure . + smp 1 & 0 23 58.67 & @xmath69 38 03.8 & 0.15 & & & unresolved . + smp 6 & 0 41 27.52 & @xmath69 47 06.2 & 0.19 & & e & + smp 8 & 0 43 25.17 & @xmath70 38 18.9 & 0.23 & 0.41 x 0.38 & r & + smp 9 & 0 45 20.66 & @xmath69 24 10.5 & 0.55 & 1.20 & r & inner structure . + smp 11 & 0 48 36.61 & @xmath70 58 00.1 & 0.99 & 0.78 x 0.66 & b & possible inner ring . + smp 12 & 0 49 21.00 & @xmath69 52 58.0 & 0.37 & 0.78 x 0.51 & e & + smp 13 & 0 40 51.71 & @xmath69 44 21.3 & 0.19 & 0.20 & r & + smp 14 & 0 50 34.99 & @xmath69 42 57.9 & 0.42 & 0.83 & r & ansae and/or inner structure . + smp 17 & 0 51 56.41 & @xmath71 24 44.6 & 0.25 & 0.50 & e & faint detached 15 halo ; inner ring in [ nii ] . + smp 18 & 0 51 57.97 & @xmath69 20 30.1 & 0.15 & 0.14 & & unresolved . + smp 19 & 0 53 11.14 & @xmath70 45 07.5 & 0.30 & 0.59 & r & outer structure . + smp 20 & 0 56 05.39 & @xmath72 19 24.7 & 0.15 & 0.20 x 0.23 & & unresolved . + smp 22 & 0 58 37.44 & @xmath71 35 49.1 & 0.40 & 0.71 x 0.54 & b ? & possible ansae . + smp 23 & 0 58 42.14 & @xmath70 56 59.6 & 0.30 & 0.66 x 0.60 & e(bc ) & + smp 24 & 0 59 16.09 & @xmath70 01 59.7 & 0.20 & 0.38 & e & + smp 25 & 5 30 33.22 & @xmath72 44 38.4 & 0.19 & & e & + smp 26 & 1 04 17.81 & @xmath69 21 51.2 & 0.28 & 0.61 x 0.57 & p & + smp 27 & 1 21 10.67 & @xmath69 14 35.4 & 0.23 & 0.45 & r & attached outer halo . + sp 34 & 1 12 10.76 & @xmath71 26 50.2 & 0.61 & 0.71 x 0.69 & r & attached outer halo . + lrrrrrl galactic models & & & & & & + 1 & -1.0 & -3.523 & -4.0 & -3.222 & 4.517 & 3.887 + 2 & -1.0 & -3.523 & -4.0 & -3.222 & 4.828 & 3.869 + 3 & -1.0 & -3.523 & -4.0 & -3.222 & 5.032 & 3.823 + 4 & -1.0 & -3.523 & -4.0 & -3.222 & 5.202 & 3.655 + 5 & -1.0 & -3.523 & -4.0 & -3.222 & 5.238 & 3.348 + lmc models & & & & & & + 6 & -1.0 & -3.380 & -3.91 & -3.644 & 4.569 & 3.973 + 7 & -1.0 & -3.380 & -3.91 & -3.644 & 4.797 & 3.964 + 8 & -1.0 & -3.380 & -3.91 & -3.644 & 5.024 & 3.932 + 9 & -1.0 & -3.380 & -3.91 & -3.644 & 5.233 & 3.805 + 10 & -1.0 & -3.380 & -3.91 & -3.644 & 5.298 & 3.481 + smc models & & & & & & + 11 & -1.03 & -3.370 & -4.57 & -4.06 & 4.523 & 3.938 + 12 & -1.03 & -3.370 & -4.57 & -4.06 & 4.838 & 3.936 + 13 & -1.03 & -3.370 & -4.57 & -4.06 & 5.046 & 3.905 + 14 & -1.03 & -3.370 & -4.57 & -4.06 & 5.234 & 3.775 + 15 & -1.03 & -3.370 & -4.57 & -4.06 & 5.292 & 3.474 + lrr smc & & + smp 13 & 3.6 & 0.40 + smp 14 & 3.6 & 0.10 + smp 18 & 3.6 & 0.28 + smp 19 & 3.4 & 0.21 + smp 22 & 3.4 & 0.28 + smp 24 & 3.1 & 0.86 + smp 25 & 3.4 & 0.10 + smp 26 & 3.1 & 0.15 + average & 3.4 & 0.30 + lmc & & + smp 9 & 3.2 & 0.12 + smp 13 & 2.9 & 0.63 + smp 16 & 2.9 & 0.33 + smp 19 & 3.4 & 0.21 + smp 28 & 3.2 & 0.10 + smp 30 & 2.9 & 0.16 + smp 46 & 3.4 & 0.05 + smp 53 & 3.5 & 0.26 + smp 71 & 3.6 & 0.17 + smp-80 & 3.2 & 0.14 + smp 95 & 2.9 & 0.16 + smp 100 & 3.4 & 0.17 + average & 3.3 & 0.21 +

the _ height _ of a trace is the height of the corresponding heap of pieces in viennot s representation , or equivalently the number of factors in its cartier - foata decomposition . let and stand respectively for the height and the length of a trace . we prove that the bivariate commutative series is rational , and we give a finite representation of it . we use the rationality to obtain precise information on the asymptotics of the number of traces of a given height or length . then , we study the average height of a trace for various probability distributions on traces . for the uniform probability distribution on traces of the same length ( resp . of the same height ) , the asymptotic average height ( resp . length ) exists and is an algebraic number . to illustrate our results and methods , we consider a couple of examples : the free commutative monoid and the trace monoid whose independence graph is the ladder graph . * keywords : * automata and formal languages , trace monoids , cartier - foata normal form , height function , generating series , speedup , performance evaluation .

cs0112012

traces are used to model the occurrence of events in concurrent systems @xcite . roughly speaking , a letter corresponds to an event and two letters commute when the corresponding events can occur simultaneously . in this context , the two basic performance measures associated with a trace @xmath2 are its _ length _ @xmath1 ( the ` sequential ' execution time ) and its _ height _ @xmath0 ( the ` parallel ' execution time ) . the ratio @xmath4 captures in some sense the _ amount of parallelism _ ( the _ speedup _ in @xcite ) . let @xmath5 be a trace monoid . define the generating series @xmath6 it is well known that @xmath7 is a rational series @xcite . we prove that @xmath8 and @xmath9 are also rational and we provide finite representations for the series . exploiting the symmetries of the trace monoid enables to obtain representations of reduced dimensions . we use the rationality to obtain precise information on the asymptotics of the number of traces of a given height or length . then , given a trace monoid and a measure on the traces , we study the _ average parallelism _ in the trace monoid . one notion of average parallelism is obtained by considering the measure over traces induced by the uniform distribution over words of the same length in the free monoid . in other terms , the probability of a trace is proportional to the number of its representatives in the free monoid . this quantity was introduced in @xcite and later studied in @xcite . here we define alternative notions of average parallelism by considering successively the uniform distribution over traces of the same length , the uniform distribution over traces of the same height , and the uniform distribution over cartier - foata normal forms . we prove in particular that there exists @xmath10 and @xmath11 in @xmath12 such that @xmath13 furthermore , the numbers @xmath10 and @xmath11 are algebraic . explicit formulas involving the series @xmath7 and @xmath9 are given for @xmath10 and @xmath11 . the present paper is an extended version with proofs of @xcite . we start by introducing all the necessary notions from the theory of trace monoids . the reader may refer to @xcite for further information . in the sequel , a _ graph _ is a couple @xmath14 where @xmath15 is a finite non - empty set and @xmath16 . hence we consider directed graphs , allowing for self - loops but not multi - arcs . such a graph is _ non - directed _ if @xmath17 is symmetric . we use without recalling it the basic terminology of graph theory . given a graph and two nodes @xmath18 and @xmath19 , we write @xmath20 if there is a path from @xmath18 to @xmath19 . fix a finite alphabet @xmath21 . let @xmath22 be a reflexive and symmetric relation on @xmath21 , called the _ dependence _ relation , and let @xmath23 be its complement in @xmath24 , known as the _ independence _ or _ commutation _ relation . the _ trace monoid _ , or _ free partially commutative monoid _ , @xmath25 is defined as the quotient of the free monoid @xmath26 by the least congruence containing the relations @xmath27 for every @xmath28 . the elements of @xmath29 are called _ traces_. two words are representatives of the same trace if they can be obtained one from the other by repeatedly commuting independent adjacent letters . the _ length _ of the trace @xmath2 is the length of any of its representatives and is denoted by @xmath1 . note that we also use the notation @xmath30 for the cardinal of a set @xmath31 . the set of letters appearing in ( any representative of ) the trace @xmath2 is denoted by @xmath32 . the graphs @xmath33 and @xmath34 are called respectively the _ dependence _ and the _ independence graph _ of @xmath5 . let finally @xmath35 denote the canonical projection from @xmath26 into the trace monoid @xmath5 . in the sequel , we most often simplify the notations by denoting a trace by any of its representatives , that is by identifying @xmath36 and @xmath37 . [ ex - k4]let @xmath38 ( the set of subsets of cardinal two of @xmath39 ) . define the independence relation @xmath40 . the dependence graph @xmath41 is the line graph of the complete graph @xmath42 , also called the triangular graph @xmath43 . @xmath44 for notational simplicity , set @xmath45 . the dependence graph is represented on the left of figure [ fi - t4 ] and the independence graph on the right . in the trace monoid @xmath46 , we have @xmath47 . a _ clique _ is a non - empty trace whose letters are mutually independent . cliques are in one - to - one correspondence with the complete subgraphs ( also called cliques in a graph theoretical context ) of @xmath34 . we denote the set of cliques of @xmath29 by @xmath48 . an element @xmath49 is called _ cartier - foata ( cf- ) admissible _ if for every @xmath50 , there exists @xmath51 such that @xmath52 . cartier - foata ( cf ) decomposition _ of a trace @xmath2 is the uniquely defined ( see ( * ? ? ? i ) ) sequence of cliques @xmath53 such that @xmath54 , and the couple @xmath55 is cf - admissible for all @xmath56 in @xmath57 . the positive integer @xmath58 is called the _ height _ of @xmath2 and is denoted by @xmath0 . in the visualization of traces using _ heaps of pieces _ , introduced by viennot in @xcite , the height corresponds precisely to the height of the heap . [ ex - heapk4]consider the trace monoid defined in example [ ex - k4 ] . the set of cliques is @xmath59 . @xmath60 the cf decomposition of @xmath61 is @xmath62 . we have @xmath63 and @xmath64 . we represented the heap of pieces associated with @xmath61 on figure [ fi - heapt4 ] . we define the _ graph of cliques _ @xmath65 as the directed graph with @xmath48 as its set of nodes and the set of all cf - admissible couples as its set of arcs . note that @xmath65 contains as a subgraph the dependence graph @xmath66 . the graph @xmath65 is in general complicated and looks like a maze . [ ex - cliqk4]we continue with the model of examples [ ex - k4 ] and [ ex - heapk4 ] . @xmath67 for simplicity , the graph represented figure [ fi - t4cliq ] is the complement of the corresponding graph of cliques ( the complement of the graph @xmath14 is the graph @xmath68 ) . [ le - stco ] if the dependence graph is connected , then the corresponding graph of cliques is strongly connected . let @xmath41 be the dependence graph , @xmath48 the set of cliques , and @xmath65 the graph of cliques . given @xmath69 , we want to prove that there is a path from @xmath18 to @xmath19 in @xmath65 . we argue by induction on the value of @xmath70 . if @xmath71 , the result follows by the connectivity of the dependence graph @xmath41 . now consider the case @xmath72 . assume first that @xmath73 . let @xmath74 belong to @xmath75 . clearly @xmath76 is cf - admissible . by induction , we have @xmath77 and we deduce that @xmath20 . assume now that @xmath78 . then we have @xmath79 and let @xmath80 . by induction , we have @xmath81 . let us prove that @xmath82 . by connectivity , there exists in @xmath41 a path @xmath83 . for @xmath84 , set @xmath85 if @xmath86 and otherwise set @xmath87 where @xmath36 is the longest trace such that @xmath88 and @xmath89 . by construction , we obtain that @xmath90 is a path in @xmath65 . it completes the proof . the above lemma can be restated as follows : given two cliques @xmath18 and @xmath19 there exists at least one trace in @xmath5 such that the first factor in its cf - decomposition is @xmath18 and the last one is @xmath19 . we now use a standard reduction technique for multi - graphs ( see ( * ? ? ? 4 ) or ( * ? ? ? we partition the nodes of @xmath65 based on their set of direct successors . an _ equitable partition _ of @xmath91 is a partition @xmath92 with the property that for all @xmath93 and @xmath56 the number @xmath94 of direct successors that a node in @xmath95 has in @xmath96 is independent of the choice of the node in @xmath95 . the @xmath97 matrix @xmath98 is called the _ coloration matrix _ corresponding to @xmath99 . in the case of the partition @xmath100 , the coloration matrix is the _ adjacency matrix _ of @xmath65 . [ ex - colk4]we keep studying the model of examples [ ex - k4 ] , [ ex - heapk4 ] and [ ex - cliqk4 ] . consider the partition @xmath99 of @xmath48 defined by @xmath101 it is easily checked that the partition is equitable . the corresponding coloration matrix is @xmath102 a natural family of equitable partitions is the one induced by the non - trivial subgroups of the full automorphism group of @xmath65 . given such a group @xmath103 , the cells of the corresponding partition @xmath104 are the orbits into which @xmath48 is partitioned by @xmath103 . the corresponding coloration matrix is denoted by @xmath105 . an automorphism of @xmath41 induces an automorphism of @xmath65 . indeed , consider an automorphism @xmath106 of @xmath33 . the map @xmath107 can be extended into a map @xmath108 as follows . given @xmath109 with @xmath110 for all @xmath93 , set @xmath111 . note that the definition is unambiguous since the letters @xmath112 commute . it is immediate that @xmath113 is an automorphism of @xmath65 . due to the complex structure of @xmath65 , finding its automorphisms is in general difficult . finding the automorphisms of @xmath41 is often an easier task . this simple observation allows us to focus on the automorphism groups of @xmath33 and to consider their action on the nodes of @xmath114 . when @xmath33 has a great amount of symmetries , the corresponding reduction can be very important ( see section [ sse - hsg ] ) . below we need to consider equitable partitions such that all the cliques in the same cell have a common length . this requirement is always satisfied for the equitable partitions associated with automorphism groups . [ ex - colok4]the model is the one of examples [ ex - k4 ] , [ ex - heapk4 ] , [ ex - cliqk4 ] , and [ ex - colk4 ] . the symmetric group @xmath115 of degree 4 is a non - trivial group of automorphisms of @xmath41 . it is of index @xmath116 in the full automorphism group @xmath103 of @xmath41 . the partition of @xmath117 induced by @xmath115 ( or by @xmath103 ) is @xmath118 and @xmath119 . the coloration matrix is given by @xmath120 let @xmath121 $ ] be the _ height and length generating function _ defined by @xmath122 where @xmath123 and @xmath124 are commuting indeterminate and @xmath125 is the number of traces of height @xmath126 and length @xmath127 . set @xmath128 and @xmath129 . then @xmath130 and @xmath131 are respectively the generating functions of the height and of the length . the _ mbius polynomial _ @xmath132 of the graph @xmath133 is defined by @xmath134 it is well known ( * ? ? ? ii ) that @xmath131 is equal to the inverse of the mbius polynomial , i.e. @xmath135 . in particular , it is a rational series . [ th - a ] let @xmath136 be a trace monoid and let @xmath48 be the set of cliques of @xmath133 . define the matrix @xmath137^{\mathfrak c \times \mathfrak c}$ ] by setting @xmath138 if @xmath139 is cf - admissible and 0 otherwise . define also @xmath140^{1 \times \mathfrak c}$ ] and @xmath141^{\mathfrak c \times 1}$ ] . the height and length generating function is then given by @xmath142 where @xmath143 is the identity of @xmath144 $ ] and @xmath23 is the @xmath145 identity matrix . proposition [ th - a ] states that @xmath146 is a rational series of @xmath144 $ ] and that @xmath147 is a finite representation of it . [ co - a ] the series @xmath7 and @xmath9 are rational , and we have @xmath148 and @xmath149 . proposition [ th - a ] and corollary [ co - a ] , although easy to prove , do not seem to appear in the literature . in the case of the length generating series , the rationality is not new but corollary [ co - a ] provides a new formula for @xmath7 . there exist related results in the context of directed animals . indeed there is a bijection between directed animal of width @xmath126 on a 2d triangular lattice and traces in the monoid @xmath150 with @xmath151 and @xmath152 . the precise asymptotics for such directed animals are derived in @xcite with the same method as in the proof of proposition [ th - a ] . more generally , the method of proof of proposition [ th - a ] can be viewed as an instance of the transfer matrix method ( * ? ? ? 4.7 ) . in the context of trace monoids , the idea of working with the alphabet of cliques @xmath117 to study the height function appeared in @xcite and was later used in @xcite . let @xmath153 be an equitable partition of @xmath48 such that all the cliques in @xmath154 have a common length @xmath155 . let @xmath156 be the coloration matrix . define the matrix @xmath157^{s\times s}$ ] by @xmath158 . define @xmath159^{1 \times s}$ ] and @xmath160^{s \times 1}$ ] . then formula holds when replacing @xmath161 , and @xmath162 by @xmath163 , and @xmath164 . the proof is similar to the one below . as recalled above , with each trace is associated its unique cf decomposition . we associate with a path @xmath165 in @xmath65 the sequence of its nodes @xmath166 . by construction , the cf decomposition of the trace @xmath167 is precisely @xmath168 . in other words , the cf decompositions of traces are in one - to - one correspondence with the paths in @xmath65 . the contribution of the trace @xmath2 to the series @xmath8 is @xmath169 . the weight of the path @xmath165 in the weighted automaton @xmath147 is @xmath170 this completes the proof of the result . it is easily checked that the series @xmath146 is not recognizable in general . we recall that @xmath171 is a _ recognizable _ series of @xmath144 $ ] if there exists @xmath172 , and @xmath173 , such that @xmath174 for all @xmath126 and @xmath127 . [ ex - thk4]we persevere with the model of examples [ ex - k4 ] , [ ex - heapk4 ] , [ ex - cliqk4 ] , [ ex - colk4 ] , and [ ex - colok4 ] . the height and length generating function is given by @xmath175 setting @xmath176 , we check that the length generating function is the inverse of the mbius polynomial , i.e. @xmath177 . setting @xmath178 , we obtain the height generating function @xmath179 . the taylor expansion of the series @xmath8 around 0 is @xmath180 for instance , there are 126 traces of length 5 and height 3 , or 71910 traces of length 8 and height 6 . we now use proposition [ th - a ] to provide some precise results on the asymptotics of the number of traces of a given length or height . given a complex function analytic at the origin , a _ singularity _ is a point where the function ceases to be complex - differentiable . dominant _ singularity is a singularity of minimal modulus . throughout the paper , given a series @xmath181 $ ] , we set @xmath182 . when applicable , we denote the modulus of the dominant singularities of @xmath31 ( viewed as a function ) by @xmath183 . classically , see @xcite , the asymptotic growth rate of @xmath184 is linked to the values of the dominant singularities . [ le - rho=1 ] we have @xmath185 or @xmath186 if and only if @xmath150 is the free commutative monoid over @xmath21 . we have @xmath187 , and @xmath188 ( the ` exponential growth formula ' ) . it implies that @xmath189 and @xmath190 . assume there exists @xmath191 with @xmath192 . then all the traces @xmath193 with @xmath194 are of length @xmath195 and height @xmath195 . it implies that @xmath196 and that @xmath197 . it implies in turn that @xmath198 and @xmath199 . assume now that @xmath150 is the free commutative monoid . by direct computation or using the results from section [ sse - fcm ] , we get @xmath200 and @xmath201 . it implies that @xmath185 and @xmath186 . [ le - clique ] let @xmath41 be a connected dependence graph . then @xmath7 and @xmath9 have a unique dominant singularity which is positive real and of order 1 . it follows ( see @xcite ) that when @xmath41 is connected , we have @xmath202 and @xmath203 , with @xmath204_{|y=\rho_l}$ ] and @xmath205_{|x=\rho_h}$ ] . the proof of proposition [ le - clique ] is based on the representation given in proposition [ th - a ] . for convenience reasons , the proof is included in the proof of proposition [ pr - lambda2 ] and given in appendix . [ le - clique2 ] let @xmath41 be a non - connected dependence graph . let @xmath206 be its partition into maximal connected subgraphs . denote by @xmath207 , the corresponding length and height generating functions . then one has : \1 ) the series @xmath7 has a unique dominant singularity equal to @xmath208 , and whose order is @xmath209 ; \2 ) the series @xmath9 has a unique dominant singularity equal to @xmath210 . its order is @xmath211 if @xmath150 is the free commutative monoid , and @xmath212 otherwise . let @xmath213 and @xmath214 denote the respective orders of @xmath215 in @xmath7 and @xmath216 in @xmath9 . it follows from the above proposition ( see @xcite ) that we have @xmath217 , and @xmath218 with @xmath219_{|y=\rho_l}$ ] and @xmath220_{|x=\rho_h}$ ] . we have @xmath221 where @xmath222 is defined in . it implies directly the result on @xmath215 . consider now the height generating function . we prove the result by induction on @xmath223 . assume first that @xmath224 and set @xmath225 . we have @xmath226 . it implies that @xmath227 applying proposition [ le - clique ] , we obtain @xmath228 , with @xmath229 , and @xmath230 , with @xmath231 . consider first the case @xmath232 and @xmath233 . we have @xmath234 the same type of identity also holds for the second term in . going back to , we then obtain @xmath235 hence we have @xmath236 and the order of @xmath216 in @xmath9 is 1 . we consider now the case @xmath237 and @xmath238 . by lemma [ le - rho=1 ] , we get that @xmath150 is the free commutative monoid over two letters . applying , we get that @xmath239 . hence we have @xmath186 and the order of @xmath216 in @xmath9 is @xmath116 . by symmetry , the last case to consider is @xmath232 and @xmath238 . by proposition [ le - clique ] , we have @xmath240 . we also have @xmath241 . simplifying , we obtain that @xmath242 . it implies that @xmath243 and that the order of @xmath216 in @xmath9 is 2 . consider now the case @xmath244 . let @xmath245 and @xmath246 be a partition of @xmath41 in two subgraphs such that @xmath245 is connected . the induction hypothesis applies to @xmath246 and the proof follows exactly the same steps as above the results on @xmath7 in proposition [ le - clique ] and proposition [ le - clique2 ] can be restated as results on the smallest root of the mbius polynomial of a non - directed graph . they improve on a recent result by goldwurm and santini @xcite stating that the mbius polynomial has a unique and positive real root of smallest modulus . our proof of proposition [ pr - lambda2 ] follows several of the steps of @xcite . one central difference is that we work with cartier - foata representatives instead of minimal lexicographic representatives . proving the strengthened statements while working with the latter does not appear to be easy . a _ matching _ in a ( non - directed ) graph is a subset of arcs with no common nodes . the _ matching polynomial _ of a graph is equal to @xmath247 , where @xmath248 is the number of matchings of @xmath126 arcs . hence , the matching polynomial of a graph @xmath103 is equal to the mbius polynomial of the complement of the line graph of @xmath103 . matching polynomials have been studied quite extensively . it is known for instance that all the roots of a matching polynomial are real @xcite . it implies that the same is true for the mbius polynomial of a graph which is the complement of a line graph . for a general graph , the result is not true and one has to settle for the weaker results in proposition [ le - clique ] and proposition [ le - clique2 ] . consider for instance the graph with nodes @xmath249 and arcs @xmath250 . it is the smallest graph which is not the complement of a line graph . its mbius polynomial is @xmath251 , which has two non - real roots . we want to address questions such as : what is the amount of ` parallelism ' in a trace monoid ? given several dependence graphs over the same alphabet , which one is the ` most parallel ' ? to give a precise meaning to these questions , we define the following performance measures . let @xmath252 denote the set of traces of length @xmath195 of the trace monoid @xmath5 . we equip @xmath252 with a probability distribution @xmath253 and we compute the corresponding average height @xmath254=\sum_{t\in { \m}_n } p_n\{t\ } h(t)\:.\ ] ] assuming the limit exists , we call @xmath255/n$ ] the _ ( asymptotic ) average height_. obviously this quantity belongs to @xmath256 $ ] , where @xmath257 is the maximal length of a clique . clearly the relevance of the average height as a measure of the parallelism in the trace monoid depends on the relevance of the chosen family of probability measures . this may vary depending on the application context . a very common choice is to consider uniform probabilities . it is the natural solution in the absence of precise information on the structure of the traces to be dealt with . let us consider different instances of uniform probabilities over traces . let @xmath258 be the uniform probability distribution over @xmath259 which is defined by setting @xmath260 , for every @xmath261 . we set @xmath262 , i.e. @xmath263 . the limit below exists : @xmath264}{n } = \lim_n \frac{\sum_{w\in \sigma^n } h(\psi(w))}{n|\sigma|^n } \:.\ ] ] this is proved using markovian arguments in @xcite . the existence of @xmath265 can also be proved using sub - additive arguments . more precisely , it is shown in @xcite that @xmath266 is recognized by an automaton with multiplicities over the @xmath267 semiring , which provides a different proof of the existence of @xmath265 . in fact a stronger result holds . consider a probability space @xmath268 . let @xmath269 be a sequence of independent random variables valued in @xmath21 and uniformly distributed : @xmath270 . the probability distribution of @xmath271 is then the uniform distribution over @xmath259 . it is proved in @xcite that @xmath272 except for small trace monoids , @xmath265 is neither rational , nor algebraic . the problem of approximating @xmath265 is np - hard @xcite . non - elementary bounds are proposed in @xcite . exact computations for simple trace monoids are proposed in @xcite . a software package named ers @xcite enables to simulate and compute bounds for @xmath273 . a natural counterpart of the above case consists in considering the uniform probability distribution over @xmath252 , i.e. @xmath274 for every @xmath275 . assuming existence , we define the limit @xmath276}{n } = \lim_n \frac{\sum_{t\in \m_n } h(t)}{n|{\m}_{\,n}|}= \lim_n \frac{\sum_{m \in \n } m f_{m , n } } { \sum_{m \in \n } nf_{m , n}}\:.\ ] ] dually , let @xmath277 be the set of traces of height @xmath58 , and let @xmath278 be the uniform probability measure on @xmath277 , i.e. @xmath279 for every @xmath280 . the average length of a trace in @xmath277 is equal to @xmath281=\sum_{t \in { } _ m \m } \tilde{q}_m\{t\ } |t| $ ] . assuming existence , we define the limit @xmath282}{m}= \lim_m \frac{\sum_{t\in { } _ m \m } |t|}{m|{}_m \m |}= \lim_m \frac{\sum_{n \in \n } n f_{m , n } } { \sum_{n \in \n } mf_{m , n}}\:.\ ] ] the quantity @xmath11 is an _ ( asymptotic ) average length_. the analog of @xmath265 and @xmath10 is then the quantity @xmath283 . [ pr - lambda2 ] the limits @xmath10 in and @xmath11 in exist . furthermore , @xmath10 and @xmath11 are algebraic numbers . the proof , based on proposition [ th - a ] , is rather long and we postponed it to the appendix . in fact , the proof of proposition [ pr - lambda2 ] provides a formula for @xmath10 and @xmath11 . define @xmath284 and @xmath285 . then , with the notations of section [ se - hlgf ] , we have @xmath286_{|y=\rho_l}}{k_l\rho_l [ l(y)(\rho_l - y)^{k_l}]_{|y=\rho_l } } , \ \gamma_{\m } = \frac { [ \tilde{g}(x)(\rho_h - x)^{k_h+1}]_{|x=\rho_h}}{k_h\rho_h [ h(x)(\rho_h - x)^{k_h}]_{|x=\rho_h}}\:.\ ] ] in this section , we use some basic results on markov chains , for details see for instance @xcite . let @xmath287 be the adjacency matrix of @xmath65 . we associate with @xmath288 , the markovian matrix @xmath289 we define the vector @xmath290 by @xmath291 for all @xmath93 . we define the probability measure @xmath292 on @xmath277 as follows : for a trace @xmath293 with cartier - foata decomposition @xmath294 , we set @xmath295 . an interpretation for the family @xmath296 is as follows . consider a markov chain @xmath297 on the state space @xmath117 with transition matrix @xmath298 and with initial distribution @xmath299 . then @xmath300 . equivalently , given a trace @xmath2 of height @xmath58 , we get a trace @xmath301 of height @xmath302 by picking at random and uniformly an admissible clique @xmath303 and by setting @xmath304 . this can be loosely described as a ` uniform probability on cf decompositions ' . the average length of a trace in @xmath277 is equal to @xmath281=\sum_{t \in { } _ m \m } r_m\{t\ } |t|$ ] . assuming existence , the analog of @xmath305 or @xmath283 is then the _ ( asymptotic ) average height _ @xmath306}\:.\ ] ] let @xmath307 be defined by @xmath308 . it can be interpreted as the limit distribution of the markov chain @xmath297 . according to the ergodic theorem for markov chains ( theorem 4.6 in @xcite ) , the limit exists in and we have @xmath309 when @xmath41 is connected , it follows from lemma [ le - stco ] that @xmath298 is irreducible . then @xmath165 is entirely determined by @xmath310 and @xmath311 ( perron - frobenius theorem , see @xcite ) . it implies that @xmath312 is explicitly computable and rational . when @xmath41 is non - connected , @xmath312 is still explicitly computable and rational according to proposition [ le - critcf ] . consider an equitable partition @xmath313 such that all the cliques in @xmath314 have a common length @xmath155 . there exists an analog of corresponding to this partition . let @xmath315 be the markovian matrix associated with the coloration matrix @xmath316 . let @xmath317 be defined by @xmath318 . then , we have @xmath319 . assume that @xmath41 is _ non - connected _ and let @xmath206 be the maximal connected subgraphs of @xmath41 . we now propose formulas to express the average height of @xmath41 as a function of the ones of @xmath320 . first , it is simple to prove using and the strong law of large numbers ( see also theorem 5.7 in @xcite ) that we have @xmath321 [ le - crit ] denote by @xmath322 the length generating function of @xmath320 . define @xmath323 . then , we have @xmath324 where @xmath325 , and @xmath326 . the proof uses proposition [ pr - lambda2 ] and is given in appendix . there seems to be no simple way to write @xmath327 as a function of @xmath328 , as illustrated by the example of section [ sse - fcm ] . [ le - critgamm ] define @xmath329 . then , we have @xmath330 if @xmath331 . if @xmath332 , that is if @xmath150 is the free commutative monoid , we have @xmath333 . the proof is given in appendix . proposition [ le - critgamm ] is the counterpart of proposition [ le - crit ] for @xmath11 , but it is more precise . [ le - critcf ] let @xmath298 be defined as in section [ sse - cf ] . let @xmath334 be the set of cliques of @xmath335 . define the matrix @xmath336 of dimension @xmath337 as follows : @xmath338 if @xmath339 and @xmath340 otherwise . define the vectors @xmath341 of dimension @xmath117 as follows : @xmath342 if @xmath343 and @xmath344 otherwise . set @xmath345 , where @xmath346 . then we have @xmath347 the graph of cliques @xmath65 of @xmath133 can be decomposed in its maximal strongly connected subgraphs ( mscs ) . replacing each mscs by one node , we define the _ condensed _ graph of @xmath65 . final _ mscs are the mscs without any successor in the condensed graph . according to lemma [ le - stco ] , the final mscs are precisely the ones with sets of nodes @xmath334 where @xmath348 . remark that the non - negative matrix @xmath336 is such that @xmath349 for every @xmath350 . in particular , it implies that @xmath351 is invertible . define @xmath345 for every @xmath348 . the quantities @xmath352 can be interpreted in terms of the markov chain @xmath297 defined in section [ sse - cf ] : we have @xmath353 ( theorem 4.4 in @xcite ) . let @xmath354 be the restriction of @xmath17 to the index set @xmath355 and let @xmath356 be the markovian matrix associated with @xmath354 . let @xmath357 be the unique probability distribution on @xmath334 such that @xmath358 ( perron - frobenius theorem , chapter 1 in @xcite ) . according to the ergodic theorem for markov chains ( theorem 4.2 in @xcite ) , we have @xmath359 . define the vector @xmath360 of dimension @xmath117 by @xmath361 if @xmath362 and @xmath363 otherwise . the vector @xmath364 is then the unique limit distribution of @xmath297 . by the ergodic theorem for markov chains ( theorem 4.6 in @xcite ) , we obtain . in terms of computability , the simplest quantity is @xmath312 and the most complicated one is @xmath265 . this is reflected by the fact that @xmath312 is rational , that @xmath10 and @xmath283 are algebraic , and that @xmath265 is in general not algebraic , see for instance . another point of view is to compare the families of probability measures @xmath365 , and @xmath366 associated respectively with @xmath367 and @xmath312 . a family of probability measures @xmath368 defined on @xmath369 ( or @xmath370 ) is said to be _ consistent _ if we have @xmath371 for all @xmath372 . in this case , there exists a unique probability measure on infinite traces whose finite - dimensional marginals are the probabilities @xmath368 . consistency is a natural and desirable property . clearly the families @xmath373 and @xmath366 are consistent . on the other hand , the families @xmath374 and @xmath375 are not . it is also interesting to look at the asymptotics in @xmath195 of the empirical distribution of @xmath376 or @xmath377 . for @xmath378 , let @xmath379 denote the probability measure concentrated in @xmath74 . it follows from that we have @xmath380 with the arrow standing for ` convergence in distribution ' . similarly , it follows from the ergodic theorem for markov chains that we have @xmath381 the notations being the ones of section [ sse - cf ] . there are no such concentration results for @xmath374 and @xmath375 . to check this , consider the case of the free commutative monoid over two letters . we obtain easily that @xmath382 where @xmath383 is the uniform distribution over the interval @xmath384 $ ] and @xmath385 is the uniform distribution over the interval @xmath386 $ ] . consider two dependence graphs @xmath387 and @xmath388 with @xmath389 . the intuition is that @xmath390 should be ` more parallel ' than @xmath391 . in accordance with this intuition , it is elementary to prove that @xmath392 . however , the corresponding inequalities do not hold for @xmath10 and @xmath312 . consider for instance the trace monoids over three or four letters whose average heights are given in section [ se-234 ] . this raises some interesting issues on how to interpret these quantities . on the other hand , we conjecture that the inequality @xmath393 is satisfied . consider the dependence graph @xmath41 with @xmath394 . the corresponding trace monoid @xmath150 is the free commutative monoid over the alphabet @xmath21 , which is isomorphic to @xmath395 . set now @xmath396 . a direct application of yields @xmath397 . consider now @xmath312 . the final maximal strongly connected subgraphs of @xmath65 are precisely the cliques of length 1 . in particular , they are of cardinality 1 . applying the results in section [ sse - cf ] , we get @xmath398 . let us compute @xmath10 and @xmath11 . using the methodology of sections [ se - hlgf ] and [ se - aah ] is feasible , but there are simpler methods . consider @xmath11 first . by a counting argument , we get @xmath399 we obtain @xmath400 and @xmath401 . we deduce that @xmath402 . let us now compute the average height @xmath10 . the length generating function is @xmath403 . applying a result from carlitz @xcite , we have @xmath404 using , we obtain @xmath405 the last equality is a classical identity for harmonic summations ( see chapter 6.4 in @xcite ) . asymptotically in @xmath126 , we have @xmath406 . this is to be compared with @xmath397 and @xmath407 . consider now the trace monoid @xmath150 obtained as the direct product of the free monoids @xmath408 , with @xmath409 and @xmath410 . equivalently , the dependence graph is @xmath41 with @xmath411 , @xmath412 , and @xmath413 for all @xmath93 . clearly , we still have @xmath414 and @xmath398 . the formulas in and still hold when replacing @xmath124 by @xmath415 . we deduce that @xmath10 is still given by . on the other hand , we have @xmath416 , a value which can also be obtained using proposition [ le - critgamm ] . hence the value of @xmath283 does not depend on the value of @xmath303 , @xmath410 , and is different from the value obtained for @xmath417 . in view of proposition [ th - a ] , the simplest sets of cliques are those with the property that the clique partition according to the length is equitable , so that the dimension of the corresponding coloration matrix reduces to the maximal size of a clique . this holds if the full automorphism group of @xmath41 or @xmath65 acts transitively on the sets of cliques of the same length . this is in particular the case when the dependence graph is the _ triangular graph _ , i.e. the line graph of the complete graph @xmath418 , or the _ square lattice graph _ , i.e. the line graph of the complete bipartite graph @xmath419 . a particularly simple class of independence graphs is the class of node and arc - transitive triangle - free graphs . in this case , the coloration matrix associated with the full automorphism group is of dimension @xmath420 . let us consider a family of graphs of this type . let @xmath34 be the _ ladder graph _ , i.e. @xmath421 and @xmath422 . the corresponding dependence graph is known as the _ cocktail party graph _ @xmath423 . the full automorphism group is the wreath product @xmath424 $ ] of the symmetric group of degree @xmath195 with @xmath425 . the corresponding partition of @xmath117 is @xmath426 with @xmath427 and @xmath428 . the coloration matrix is @xmath429 the computation of @xmath430 was worked out by brilman ( see @xcite , proposition 14 ) : @xmath431 we compute @xmath8 using the reduced representation induced by the partition . the dominant singularity of @xmath7 is @xmath432 and we obtain @xmath433 the dominant singularity of @xmath9 is @xmath434 , and we get @xmath435 considering the markovian matrix @xmath436 and using , we obtain @xmath437 we check that @xmath438 and that @xmath439 @xmath440 . remark that @xmath441 . by specializing the above results to @xmath442 , we get the average heights for the triangular graph @xmath43 considered in examples [ ex - k4 ] , [ ex - heapk4 ] , [ ex - cliqk4 ] , [ ex - colk4 ] , [ ex - colok4 ] , and [ ex - thk4 ] . we have @xmath443 this section is devoted to the proof of propositions [ pr - lambda2 ] , [ le - crit ] , and [ le - critgamm ] . @xmath444 + we give the proof for @xmath10 . the one for @xmath11 is similar ( and easier ! ) . recall that @xmath445 is the length generating function . define @xmath446 assuming existence of the limit in , we have @xmath447 . according to pringsheim s theorem ( * ? ? ? 7.21 ) , @xmath7 and @xmath103 have a positive real dominant singularity . they are denoted respectively by @xmath215 and @xmath448 according to the previous conventions . since we have @xmath449 , it implies that @xmath450 . let @xmath213 be the order of @xmath215 in @xmath7 . assume that @xmath215 is the unique dominant singularity in @xmath7 . assume that the order of @xmath215 in @xmath103 is @xmath451 and is strictly larger than the one of the other singularities of modulus @xmath215 ( there might exist several dominant singularities for @xmath103 , see section [ sse - fcm ] ) . then the limit in exists and we have , @xmath452_{|y=\rho_l}}{k_l\rho_l \cdot [ l(y)(\rho_l - y)^k_l]_{|y=\rho_l}}\:.\ ] ] in particular , @xmath10 is an algebraic number . the above assumptions on the dominant singularities of @xmath7 and @xmath103 ensure that the sequences @xmath453 and @xmath454 do not have an oscillating behavior . it remains to prove that these assumptions actually hold . we work with the representation @xmath147 of @xmath8 given in the statement of proposition [ th - a ] . we have @xmath455 where @xmath456 stands for the determinant and adj ( . ) for the adjoint of a matrix . set @xmath457 and @xmath458 . it follows that we have @xmath459 . by differentiating @xmath8 , we get @xmath460 set @xmath461 . the above equations imply that the set of singularities of @xmath7 ( resp . @xmath103 ) is included in the set of singularities of @xmath462 . in particular , a dominant singularity of @xmath7 ( resp . @xmath103 ) has a greater modulus than a dominant singularity of @xmath462 . the next step consists in transforming the triple @xmath463 into another triple @xmath464 of dimension @xmath465 , where @xmath466 , and where we set @xmath467 and @xmath468 . before formally defining it , we illustrate the construction on the figure below . as usual we view a triple as an automaton with multiplicities , i.e. as a weighted graph with input and output arcs . we have represented the portion of the automata @xmath463 and @xmath464 corresponding to the cliques @xmath18 and @xmath19 where @xmath469 , and @xmath470 is cf - admissible . @xmath471 [ fi - transfo ] consider the index set @xmath472 let us define @xmath473 and @xmath474 as follows : @xmath475 in an automaton , an _ input _ ( resp . _ output _ ) _ node _ is a node with an input ( resp . output ) arc . a _ successful path _ is a path from an input node to an output node . there is a one to one mapping between successful paths in the automata @xmath463 and @xmath464 : to the successful path @xmath476 in @xmath463 corresponds the successful path @xmath477 @xmath478 in @xmath479 , and vice versa . note that the lengths of corresponding paths do not coincide . using this correspondence , we get that @xmath480 let us prove that @xmath481 given a matrix @xmath482 of dimension @xmath195 , we have @xmath483 where @xmath484 is the set of permutations of @xmath485 , and where @xmath486 is the sign of a permutation . the permutations having a non zero contribution to the determinant are the ones which correspond to a partition into simple cycles of the nodes of the graph of @xmath482 . we have seen above that there is a one - to - one correspondence between successful paths in the graphs of @xmath487 and @xmath488 . there is also clearly a one - to - one correspondence between simple cycles in ( the graphs of ) @xmath487 and @xmath488 . when comparing the simple cycles of @xmath489 and @xmath490 , one needs to be more careful . let @xmath31 be the set of simple cycles of @xmath489 and let @xmath491 be the one of @xmath490 . to the simple cycle @xmath492 in @xmath31 , there corresponds the simple cycle @xmath493 in @xmath491 . a simple enumeration shows that @xmath494 given @xmath495 ( resp . @xmath491 ) , we denote by @xmath496 the contribution of @xmath303 to @xmath497 ( resp . @xmath498 ) . more precisely , for @xmath499 and setting @xmath500 ( resp . @xmath501 ) , we set @xmath502 consider @xmath503 . we have @xmath504 let @xmath505 be the corresponding cycle of @xmath491 . then we have @xmath506 we check that @xmath507 except in the case @xmath508 . in this last situation , we have @xmath509 and @xmath510 . however , this difference is precisely compensated by the contribution to @xmath498 of the simple cycles in @xmath511 . we conclude that @xmath512 . let @xmath513 be the maximal connected subgraphs of @xmath41 . let @xmath117 be the set of cliques of @xmath133 and let @xmath314 be the one of @xmath514 . for @xmath515 , define @xmath516 . note that we have @xmath517 . the set @xmath117 is partitioned by the sets @xmath518 . let @xmath65 be the graph of cliques of @xmath150 . using lemma [ le - stco ] , we get that the maximal strongly connected subgraphs of @xmath65 are the subgraphs with sets of nodes @xmath518 . clearly , there is a path in @xmath65 from a node in @xmath519 to a node in @xmath520 if and only if @xmath521 . it implies the following . the restriction of the matrix @xmath17 to the index set @xmath519 , denoted by @xmath522 , is irreducible . now range the index set @xmath117 according to the order @xmath523 where @xmath524 , is an ordered list of the subsets of @xmath525 satisfying the property : @xmath526 . then the matrix @xmath17 is block upper - triangular with the blocks @xmath527 on the diagonal . an analog statement holds for @xmath528 , replacing @xmath519 by @xmath529 . we denote by @xmath530 the restriction of @xmath528 to the index set @xmath531 . we have @xmath532 given an index set @xmath31 and @xmath348 , define @xmath533 by @xmath534 and @xmath535 . for @xmath536 , define @xmath537 the coefficient @xmath538 can be interpreted combinatorially as the number of paths from @xmath93 to @xmath56 with weight @xmath539 in the automaton @xmath463 . in particular , we have @xmath540 . with a proof similar to the one of , we get @xmath541 consider @xmath542 . let @xmath543}$ ] denote the matrix obtained from @xmath487 by replacing the line and the column @xmath18 by a line and a column of zeros . then we have @xmath544})}{\det(i - a(1,y ) ) } \:.\ ] ] let @xmath545}$ ] denote the matrix obtained from @xmath528 by replacing the line @xmath546 and the column @xmath547 by a line and a column of zeros . with a proof similar to the one of , we get @xmath548})= y\det(i - y\tilde{a}_{[u]})\:.\ ] ] assume that @xmath18 belongs to @xmath549 and let @xmath550}$ ] denote the restriction of @xmath545}$ ] to the index set @xmath551 . using , , , and , we obtain @xmath552})}{\det ( i -y\tilde{a } ) } = \frac{y\det ( i - y\tilde{a}_{u,[u]})}{\det ( i -y\tilde{a}_u ) } \:.\ ] ] we have @xmath550}\leq \tilde{a}_u$ ] ( for the coordinate - wise ordering ) and @xmath550}\neq \tilde{a}_u$ ] . we have seen above that @xmath553 is irreducible . according to the perron - frobenius theorem for irreducible matrices ( see for instance @xcite , chapter 1.4 ) , it implies that the spectral radius of @xmath550}$ ] is strictly less than the one of @xmath553 . now , the roots of the polynomial @xmath554})$ ] , resp . @xmath555 , are the inverses of the non - zero eigenvalues of @xmath550}$ ] , resp . hence the possible simplifications between the numerator and the denominator in the right - hand side of do not involve any dominant singularity . we conclude that the dominant singularities of @xmath557 are precisely the dominant singularities of @xmath558 . we have @xmath559 for all @xmath195 . it implies that a dominant singularity of @xmath7 has a smaller modulus than a dominant singularity of @xmath557 . we deduce that a dominant singularity of @xmath7 has a smaller modulus than a dominant singularity of @xmath558 for all @xmath383 , hence a smaller modulus than a dominant singularity of @xmath462 . using that @xmath560 , we obtain the same result for @xmath103 . we conclude that the modulus of the dominant singularities of @xmath7 , @xmath103 , and @xmath462 are equal . furthermore , the sets of dominant singularities of @xmath7 and @xmath103 are included in the set of dominant singularities of @xmath462 . since @xmath561 , the set of dominant singularities of @xmath462 is also equal to the set of inverses of maximal eigenvalues of @xmath528 . let @xmath562 denote the spectral radius of @xmath528 . first assume that @xmath563 . according to lemma [ le - rho=1 ] , @xmath150 is the free commutative monoid over @xmath21 . the analysis of section [ sse - fcm ] applies . in particular , the limit @xmath10 in exists and is given in . it is obviously algebraic and even rational . hence proposition [ pr - lambda2 ] is satisfied in this case . from now on , we assume that @xmath564 . let us specialize for a moment to the case where @xmath41 is connected . using the above analysis , the matrix @xmath528 is irreducible . for any @xmath565 , we have @xmath566 . we conclude that @xmath528 is primitive . by perron - frobenius theorem for primitive matrices ( @xcite , chapter 1.1 ) , the matrix @xmath528 has a unique eigenvalue of maximal modulus which is positive real and of multiplicity 1 . we conclude that @xmath215 is the unique dominant singularity of @xmath7 and @xmath103 . we conclude also that the order of @xmath215 is 1 in @xmath7 , and at most 2 in @xmath103 . since @xmath567 , we deduce that the order of @xmath215 in @xmath103 is 2 . we have just proved that the result of proposition [ le - clique ] holds for @xmath7 . the proof of proposition [ le - clique ] for @xmath9 is similar ( and easier ) . let us come back to the general case for @xmath41 . since we have now proved proposition [ le - clique ] , we are allowed to use proposition [ le - clique2 ] ( the proof of the latter requires the former ) . we conclude that in all cases , @xmath7 has a unique dominant singularity . it remains to study the set of dominant singularities of @xmath103 . to do this , we study the set of eigenvalues of @xmath528 of maximal modulus . fix a subset @xmath515 and consider the restricted matrix @xmath530 . let @xmath568 denote the spectral radius of @xmath530 . we distinguish between two cases . _ case ( i)_. assume there exists @xmath569 such that @xmath570 is different from the free monoid @xmath571 , or equivalently such that @xmath572 is not empty . then there exists @xmath573 such that @xmath574 . it implies that the cyclicity of the matrix @xmath530 is 1 . since @xmath530 is irreducible , we deduce that it is primitive . according to the perron - frobenius theorem for primitive matrices ( @xcite , chapter 1.1 ) , the matrix @xmath530 has a unique eigenvalue of maximal modulus which is positive real and of multiplicity 1 . _ case ( ii)_. assume now that @xmath575 for all @xmath569 . it implies that @xmath576 for all @xmath577 . the cyclicity of @xmath530 is @xmath578 and @xmath530 is not primitive as soon as @xmath579 . however in this case , we are able to completely compute the spectrum of @xmath530 . set @xmath580 . it is more convenient to work with @xmath581 . using the same arguments as in the proof of , we get @xmath582 we also have @xmath583 and @xmath584 is the matrix of dimension @xmath585 whose entries are all equal to 1 . the eigenvalues of @xmath584 are 0 with multiplicity @xmath586 and @xmath587 with multiplicity 1 . we have @xmath588 it follows that the non - zero eigenvalues of @xmath530 are @xmath589 all with multiplicity 1 . in particular , we have @xmath590 . according to , the spectral radius of @xmath528 is given by @xmath591 define @xmath592 and @xmath593 . using the above analysis , we can distinguish between two situations . first , assume that @xmath594 . according to case ( i ) , it implies that @xmath595 is the only eigenvalue of maximal modulus of the matrix @xmath528 . we conclude that @xmath596 is the only dominant singularity of @xmath103 . second , assume that there exists @xmath597 . according to case ( ii ) , it implies that @xmath598 let @xmath599 denote the power set of a set @xmath31 . we deduce easily from that @xmath600 we conclude that the set of maximal eigenvalues of @xmath528 is precisely given by @xmath601 set @xmath602 . the multiplicity of the eigenvalue @xmath587 is at least @xmath578 ( and it is exactly @xmath578 if @xmath603 ) . for a complex and non positive real maximal eigenvalue , the multiplicity is exactly the number of appearances of the eigenvalue in . the maximal such multiplicity is equal to @xmath604 and attained for the eigenvalue @xmath605 . it follows that the maximal order of a complex and non positive real dominant singularity in @xmath462 is @xmath604 . using , we conclude that the maximal order of a complex and non positive real dominant singularity in @xmath103 is @xmath604 . now we also have @xmath606 . the mbius function of @xmath607 is @xmath608 . we deduce that @xmath609 the order of the singularity @xmath610 in @xmath7 is consequently at least @xmath578 . since we have @xmath449 , we deduce that one of the dominant singularities of @xmath103 must be of order @xmath611 . since we have @xmath612 , the only possible choice is @xmath610 . we conclude that the positive real dominant singularity of @xmath103 has a strictly larger order than all the other dominant singularities . it completes the proof . @xmath444 + the notations are borrowed from the statement of proposition [ le - crit ] . to avoid trivialities , assume that @xmath613 . let @xmath614 and @xmath615 . let @xmath7 , @xmath616 , and @xmath617 be the respective length generating functions of @xmath41 , @xmath245 , and @xmath246 . by construction , we have @xmath618 . let @xmath619 and @xmath620 be the order of @xmath621 and @xmath622 in their respective series . according to proposition [ le - clique2 ] , we have @xmath623 with @xmath624 @xmath625 , and @xmath626 with @xmath627 . furthermore , we have @xmath628 , hence @xmath629 since @xmath630 , the series @xmath631 is convergent . furthermore , we obtain easily that @xmath632 we deduce that @xmath633 . let @xmath634 be an increasing map from @xmath635 to @xmath635 such that @xmath636 and @xmath637 . define @xmath638 where @xmath639 . we have @xmath640 since @xmath636 , we have @xmath641 and @xmath642 . let @xmath643 and note that @xmath644 . we have @xmath645 using the inequality @xmath646 , we obtain @xmath647 we now consider the terms @xmath648 . given a trace @xmath2 , we can decompose it as @xmath649 with @xmath650 and @xmath651 . consider a trace @xmath2 such that @xmath652 and @xmath653 . we have @xmath654 and @xmath655 , where @xmath257 is the maximal length of a clique . using that @xmath637 , we obtain that , for @xmath195 large enough , @xmath656 . hence , we have , for @xmath195 large enough , @xmath657 given @xmath658 and @xmath659 , we have @xmath660 which depends on @xmath18 only via its length . we deduce that @xmath661 replacing in , we conclude that @xmath662 . @xmath444 + the notations are the ones of the statement of proposition [ le - critgamm ] . assume first that @xmath150 is the free commutative monoid over @xmath21 . according to the results of section [ sse - fcm ] , we have indeed @xmath663 . assume now that @xmath150 is not the free commutative monoid . let @xmath246 be a maximal connected subgraph of @xmath41 and let @xmath664 and @xmath665 . denote respectively by @xmath666 and @xmath667 the height generating functions of @xmath668 , and @xmath669 . we choose @xmath246 so that @xmath670 is different from the free commutative monoid . according to lemma [ le - rho=1 ] , it implies that @xmath671 . we are going to prove the following equalities @xmath672 formula follows easily from the above . assume first that @xmath673 . according to lemma [ le - rho=1 ] , it implies that @xmath233 . applying propositions [ le - clique ] and [ le - clique2 ] , we have @xmath674 and @xmath675 with @xmath676 and @xmath677 . using and performing the same type of computations as in the proof of proposition [ le - clique2 ] , we get @xmath678 we define the maps @xmath634 , @xmath679 and @xmath680 as in the proof of proposition [ le - crit ] . consider the set @xmath681 using the same type of arguments as in the proof of proposition [ le - crit ] , it is easily seen that @xmath682 . set @xmath683 and note that @xmath684 . we have @xmath685 using the inequality @xmath686 , where @xmath257 is the maximal length of a clique , we obtain @xmath687 using the equality @xmath688 , we obtain @xmath689 where the last equality is obtained exactly in the same way as in the proof of proposition [ le - crit ] . assume now that @xmath690 . then we have @xmath671 and @xmath238 . it implies that @xmath691 and @xmath241 . using , we obtain that @xmath692 . now , by a direct computation , we get , for @xmath658 , @xmath693 define the set @xmath694 . we have @xmath695 . it implies that @xmath696 this completes the proof . we give the values of the average heights for all the trace monoids over alphabets of cardinality 2 , 3 , and 4 . on the tables below , a trace monoid is represented by its ( non - directed ) dependence graph . for readability , self - loops have been omitted in the dependence graphs . we have not represented the free monoids for which @xmath697 . let us denote the dependence graphs in table iii , listed from top to bottom , by @xmath698 . the graph @xmath699 is the cocktail party graph @xmath700 , hence the values of the average heights can be retrieved from section [ sse - hsg ] . more generally , most of the values in the table can be computed using the results from the paper . the exceptions are @xmath265 for @xmath701 and @xmath702 . for @xmath703 and @xmath704 , the value of @xmath265 can be computed by applying proposition 12 from @xcite . let us concentrate on @xmath707 . let @xmath269 be a sequence of independent random variables valued in @xmath21 and uniformly distributed : @xmath708 . define @xmath709 , then @xmath297 is a markov chain on the state space @xmath710 . let @xmath74 be the letter such that @xmath711 for all @xmath712 . define @xmath713 . an elementary argument using the strong law of large numbers then shows that @xmath714/e[t]$ ] . it follows that @xmath715 this expression involves non algebraic generalized hypergeometric series . by truncating the infinite sum and upper - bounding the remainder using the inequality @xmath716 , we get the following exact bounds : @xmath717\:.\ ] ] another formula for @xmath707 involving multiple contour integrals and due to alain jean - marie is given in ( * ? ? ? 13 ) . at last , let us comment on the value of @xmath11 for @xmath703 . using the results from section [ sse - upt ] , we get @xmath724 where @xmath725 is the smallest root of the equation @xmath726 . numerically , we have @xmath727 and @xmath728 . in this case , cardan s formulas are of no use ( they provide an expression of the real @xmath725 as a function of the cubic root of a complex number ) . let us conclude by going back to the original motivation of comparing the degree of parallelism in different trace monoids . we claim for instance that there is some strong evidence that @xmath699 is ` more parallel ' than @xmath703 . indeed we have @xmath729 and @xmath730 . the authors would like to thank mireille bousquet - mlou and xavier viennot for pointing out several relevant references . we are also grateful to alain jean - marie for sharing with us his knowledge on the difficult problem of computing @xmath273 . a. bertoni , m. goldwurm , and b. palano . a fast parallel algorithm for the speed - up problem of traces . in _ proceedings of the workshop on trace theory and code parallelization _ , number 263 - 00 in rapporto interno , univ . degli studi di milano , pages 2936 , 2000 . s. gaubert and j. mairesse . performance evaluation of timed petri nets using heaps of pieces . in p. bucholz and m. silva , editors , _ petri nets and performance models ( pnpm99 ) _ , pages 158169 . ieee computer society , 1999 . heaps of pieces , i : basic definitions and combinatorial lemmas . in labelle and leroux , editors , _ combinatoire numrative _ , number 1234 in lect . notes in math . , pages 321350 . springer , 1986 .

for a characteristic variety with controlled-singularities , we state conditions which imply that a divisorial sheaf is cohen - macaulay or at least has depth at certain points . this mirrors results of kollr for varieties in characteristic zero . as an application , we show that relative canonical sheaves are compatible with arbitrary base change for certain families with sharply-pure fibers .

1207.1910

in the paper @xcite , kollr proved that sheaves @xmath4 satisfy strong depth conditions if @xmath5 is locally @xmath6-linearly equivalent to a divisor @xmath7 such that @xmath8 is slc or klt . these results generalized ( * ? ? ? * lemma 3.2 ) , @xcite and ( * ? ? ? * corollary 5.25 ) . because depth conditions can be interpreted as vanishing of local cohomology , these results were described as a local version of the kawamata - viehweg vanishing theorem . in this paper , we obtain characteristic @xmath0 analogs of the main results of @xcite . this is particularly interesting because the ( global ) kawamata - viehweg vanishing theorem is false in positive characteristic @xcite . we replace the klt and slc conditions by strongly @xmath2-regular and sharply @xmath2-pure singularities respectively ( such characteristic @xmath0 singularity classes are known as @xmath2-singularities ) . for the convenience of the reader , we recall that by @xcite * klt pairs correspond philosophically to strongly @xmath2-regular pairs , and * slc pairs correspond philosophically to sharply @xmath2-pure pairs . similar to @xcite , we can apply our results on depth to prove base change for relative canonical sheaves . * ( a special case of ) . * _ ( _ cf._@xcite ) let @xmath9 be a flat morphism of finite type with @xmath10 , @xmath11 equidimensional fibers to a smooth variety and let @xmath12 be a @xmath13-divisor on @xmath1 avoiding all the codimension zero and the singular codimension one points of the fibers . further suppose that @xmath14 is @xmath6-cartier , @xmath15 and @xmath16 is sharply @xmath2-pure for every @xmath17 . then @xmath18 is flat over @xmath19 and compatible with arbitrary base change . _ is hoped to be useful in constructing a moduli space for varieties of general type in positive characteristics . see @xcite for further explanation , and also for examples for which the above compatibility does not hold . we also remark here that @xmath18 behaves surprisingly well with respect to base - change . it obeys base - change for example when the fibers are cohen - macaualay ( * ? ? ? * theorem 3.6.1 ) . in particular , this pertains to families of normal surfaces . in contrast , the higher reflexive powers , @xmath20}$ ] for @xmath21 , are not compatible with base change in the surface case ( * ? ? ? * section 14.a ) . similar differences between canonical and pluricanonical sheaves have been observed earlier @xcite , ( * ? ? ? * remark 4.4 ) . the technical result on depth used to prove is as follows . it is a characteristic @xmath0 version of ( * ? ? ? * theorem 3(1 ) ) , also compare with ( * ? ? ? * lemma 3.2 ) , ( * ? ? ? * theorem 4.21 ) , ( * ? ? ? * theorem 1.5 ) and ( * ? ? ? * theorem 1.2 , 1.5 ) . * . * _ ( _ cf . _ * theorem 3(1 ) ) ) suppose that @xmath22 is local , @xmath10 and @xmath11 and that @xmath23 is an @xmath24-divisor on @xmath25 with no common components with the singular locus of @xmath1 and such that @xmath8 is sharply @xmath2-pure . set @xmath26 to be the closed point and assume that @xmath27 is not an @xmath2-pure center of @xmath8 . suppose that @xmath28 is another @xmath24-divisor and that @xmath29 is integral for some @xmath30 relatively prime to @xmath31 . further assume that @xmath32 is any rank-1 reflexive subsheaf of @xmath33 such that @xmath34 ( here @xmath35 denotes reflexive power ) . then @xmath36 _ another interesting depth statement , again completely analogous to a theorem of kollr is below . in the introduction we phrase it in the language of frobenius splittings @xcite , but in the text it is phrased slightly more generally . * _ ( _ cf . _ * theorem 3(2 ) ) ) suppose that @xmath37 is an @xmath10 local ring with frobenius splitting @xmath38 which is not compatibly split with @xmath39 . additionally suppose that @xmath40 is any union of compatibly @xmath41-split subvarieties of @xmath25 such that no irreducible component of @xmath40 coincides with an irreducible component of @xmath1 . suppose that @xmath42 is the ideal defining @xmath40 , then @xmath43 _ the other main statement on depth we obtain , , asserts that if @xmath8 is strongly @xmath2-regular and @xmath44 is an integral divisor linearly equivalent to @xmath45 , then @xmath4 is cohen - macaulay . compare with ( * ? ? ? * theorem 2 ) . one should also compare the above results on depth , as well as the related characteristic zero results , to ( * ? ? ? * theorem 4.8(vi ) ) where aberbach and enescu showed that the depth of an @xmath2-pure ring @xmath22 is always @xmath46 than the dimension of the minimal @xmath2-pure center ( ie , of the dimension of @xmath22 modulo the splitting prime , which we know is equal to the @xmath47-dimension of @xcite by @xcite ) . _ acknowledgements : _ the authors began working on this project at the workshop _ acc for minimal log discrepancies and termination of flips _ held at the american institute of mathematics and organized by tommaso de fernex and christopher hacon . the authors would also like to thank florian enescu for valuable discussions as well as thank the referee , jnos kollr and sndor kovcs for many useful comments on a previous draft . throughout this paper , all schemes are noetherian , separated , of equal characteristic @xmath0 and @xmath2-finite . note that any such scheme @xmath1 is automatically locally excellent by @xcite and also has a dualizing complex by @xcite . in particular , we are implicitly assuming all schemes are locally excellent and possess dualizing complexes . little will be lost to the reader if he or she considers only schemes that are essentially of finite type over a perfect field . we remind the reader of some special divisors on non - normal schemes . we follow the notation of ( * ? ? ? * section 16 ) . for an @xmath10 reduced local ring @xmath22 , set @xmath25 . we define a _ w - divisor _ ( or _ weil divisor _ ) to be a formal sum of codimension one subsets of @xmath1 _ whose generic points are not singular points of @xmath1_. this has the same data as divisors on the regular locus of @xmath1 or as rank-1 @xmath10 submodules @xmath32 of @xmath48 ( the total ring of fractions of @xmath22 ) such that @xmath49 as a subset of @xmath48 , for every codimension 1 singular point @xmath50 of @xmath1 . later in the paper , we will need to instead work with the more general notion of weil divisorial sheaves @xmath51 , rank one reflexive subsheaves of @xmath33 that are invertible in codimension 1 . in the non - local setting , such divisors are simply formal sums of irreducible subschemes that satisfy this definition locally . we now set @xmath52 and @xmath53 . note we have containments : @xmath54 one can also form @xmath55 and @xmath56 similarly , but the natural maps @xmath57 are not necessarily injective ( * ? ? ? * section 16 ) . given @xmath58 , we use @xmath59 to denote @xmath60 ( such roundings are not necessarily well defined for @xmath61 or @xmath62 ) . finally , given @xmath63 , we use @xmath64 ( or @xmath65 ) to denote the corresponding subsheaf of @xmath66 ( or of @xmath48 ) in the usual way . note that @xmath5 is effective if and only if @xmath67 . now we move away from divisors . suppose that @xmath22 is a ring of characteristic @xmath0 . following @xcite , we say a _ cartier subalgebra _ @xmath68 is a graded subring of the graded ring @xmath69 where multiplication is done by frobenius twisted composition_e$ ] and @xmath70_d$ ] , then @xmath71 . see the aforementioned sources for more details . ] such that the zeroth graded piece @xmath72_0 = \hom_r(r , r ) \cong r$ ] . we note that even though we call @xmath68 a _ cartier subalgebra _ , it is not an @xmath22-algebra because @xmath73_0 $ ] is not necessarily central . [ ex.cartieralgebras ] given an @xmath10 and @xmath11 ring @xmath22 , set @xmath25 and assume that @xmath74 on @xmath1 ( for example , if @xmath22 is normal , @xmath7 is simply an @xmath24-divisor ) . we can form the cartier subalgebra @xmath75 where @xmath76_e : = \hom_r(f^e _ * r ( \lceil ( p^e - 1)\delta \rceil ) , r ) \subseteq \hom_r(f^e _ * r , r).\ ] ] suppose that @xmath77 is as above and @xmath78_e$ ] for some @xmath79 . then we can form the cartier subalgebra @xmath80 generated by @xmath81_0 $ ] and @xmath41 . explicitly , this is the direct sum @xmath82 . now we define sharply @xmath2-pure pairs and @xmath2-pure centers . if @xmath68 is a cartier subalgebra on @xmath22 , then we say that the pair @xmath83 is _ sharply @xmath2-pure _ if there exists some @xmath84_e$ ] for some @xmath85 such that @xmath86 . in particular , if @xmath87 is a pair as in , then we say that @xmath87 is _ sharply @xmath2-pure _ if the associated @xmath88 is sharply @xmath2-pure . if @xmath89 is sharply @xmath2-pure , then we simply say that @xmath22 is _ @xmath2-pure_. if @xmath83 is a pair as above , then an ideal @xmath90 is called _ @xmath68-compatible _ if @xmath91 for all @xmath84_e$ ] and all @xmath92 . in the case that @xmath93 , we will sometimes simply say that @xmath94 is _ @xmath41-compatible_. an irreducible closed set @xmath95 , for some @xmath96 , is called an _ @xmath2-pure center _ if the following two conditions hold : * the localization @xmath97 is sharply @xmath2-pure , and * for every for @xmath92 and @xmath84_e$ ] , we have @xmath98 ( in other words , if @xmath99 is @xmath68-compatible ) . likewise we say that @xmath100 is an @xmath2-pure center of @xmath87 if it is an @xmath2-pure center of @xmath88 where @xmath75 is associated to @xmath7 as in . we also define strongly @xmath2-regular pairs . if @xmath22 is a local ring , a pair @xmath83 is called _ strongly @xmath2-regular _ if the only proper @xmath68-compatible ideals of @xmath22 are @xmath101 and @xmath22 itself . if @xmath22 is not local , then we say @xmath83 is _ strongly @xmath2-regular _ if every localization is . a pair @xmath87 is strongly @xmath2-regular if @xmath88 is strongly @xmath2-regular . given a pair @xmath8 , all of the above definitions generalize to the non - affine setting by requiring them to hold at each stalk . the notion of cartier subalgebras is somewhat more subtle in the non - affine setting however ( but we will not need such generalities ) . we recall some facts about compatible ideals and @xmath2-pure centers . [ lem.factsaboutsplittings ] suppose that @xmath83 is a pair and @xmath102 are ideals . * the set of @xmath68-compatible ideals are closed under sum and intersection . * a prime ideal @xmath99 is @xmath68-compatible if and only if @xmath103 is @xmath104-compatible . * if @xmath105 are cartier subalgebras and @xmath94 is @xmath68-compatible , then @xmath94 is @xmath106-compatible . * given @xmath107 , we have @xmath108 if and only if @xmath109 is @xmath80-compatible . * suppose that @xmath38 is surjective . some @xmath96 is @xmath41-compatible if and only if it is @xmath110-compatible where @xmath111 * if @xmath94 is @xmath41-compatible , then there exists a map @xmath112 such that the following diagram commutes : @xmath113^{{\varphi } } \ar[d ] & r \ar[d ] \\ f^e _ * ( r / i ) \ar[r]_{{\varphi}/i } & ( r / i ) . } \ ] ] furthermore , @xmath114 is @xmath41-compatible if and only if @xmath115 is @xmath116-compatible . ( this statement can also be done with cartier subalgebras , but we will not need it ) . * @xmath83 is strongly @xmath2-regular if and only if for every @xmath117 , there exists a @xmath84_e$ ] for @xmath79 , in fact one may take @xmath118 to be any larger multiple , such that @xmath119 . \(v ) follows from the argument of ( * ? ? ? * proposition 4.1 ) . ( vii ) can be found in this generality in ( * ? ? ? * proposition 3.23 ) . the rest are obvious . our next goal is to give an example of a @xmath41-compatible ideal that will be crucial in later sections . the main idea is that frobenius maps and frobenius splittings induce maps on local cohomology . those induced maps can then be thought of as acting directly and explicitly on classes . for the convenience of the reader not already familiar with this construction , recall that if @xmath25 and @xmath120 , then for any coherent @xmath121-module @xmath32 , we have @xmath122 for @xmath123 and also that @xmath124 . we can then use the cohomology description of sheaf cohomology to define @xmath125 . for a more thorough description of local cohomology by the complex , see for example ( * ? ? ? * section 3.5 ) . we now consider frobenius action on local cohomology . the frobenius map @xmath126 yields @xmath127 . given a class @xmath128 = [ \dots , z_j , \dots ] \in h^i_{\bm}(r)$ ] , we have @xmath129 ) \in h^i_{\bm}(f _ * r)$ ] . but certainly @xmath129 ) = f^e _ * [ z]^{p^e } = f^e _ * [ \dots , z_j^{p^e } , \dots]$ ] is identified with raising the entries of @xmath128 $ ] to the @xmath130th power . now we do the same computation with a frobenius splitting . suppose that @xmath38 is an @xmath22-linear frobenius splitting , and so we have a map @xmath131 induced by @xmath41 . certainly @xmath132 ) = [ \dots , { \varphi}(f^e _ * y_j ) , \dots]$ ] . but now observe that for any @xmath128 \in h^i_{\bm}(r)$ ] and @xmath133 we have that @xmath134^{p^e } ) ) = { \varphi}(f^e _ * [ \dots , r z_j^{p^e } , \dots ] ) = [ \dots , { \varphi}(f^e _ * r ) z_j , \dots ] = { \varphi}(f^e _ * r ) [ z].\ ] ] ( _ cf . _ * theorem 4.1 ) ) [ lem.localcohomologycompatiblewithsplittings ] suppose that @xmath37 is a local ring . then @xmath135 is compatible with every splitting that sends @xmath136 to @xmath137 . splittings are necessarily surjective . ] @xmath38 of frobenius @xmath138 . we have the following composition @xmath139 & f^e _ * r \ar[r]^{{\varphi } } & r\\ 1 \ar@{|->}[r ] & f^e _ * 1 \ar@{|->}[r ] & 1 . } \ ] ] now suppose that @xmath140 . then choose @xmath128 \in h^i_{\bm}(r)$ ] . we want to show that @xmath141 = 0 $ ] . now , it follows from the cohomology description of local cohomology , and @xmath41 s action on it , that @xmath142^{p^e } ) ) = { \varphi}(f^e _ * r).[z]\ ] ] which completes the proof . we also recall the following fact . we include the proof because the method will be generalized later . [ lem.stronglyfregulariscohenmacaulay ] if @xmath83 is strongly @xmath2-regular , then @xmath22 is normal and cohen - macaulay . we first note that by ( vii ) , the strong @xmath2-regularity hypothesis implies that there exists a frobenius splitting @xmath143 such that @xmath144 is an isomorphism for some @xmath79 . it then easily follows that @xmath22 must be reduced since if not , the map @xmath138 is not injective . normality follows since the conductor ideal is compatible with every @xmath145 by the argument of ( * ? ? ? * proposition 1.2.5 ) . for the cohen - macaulay condition , by working locally we assume that @xmath146 is a local domain . by local duality ( * ? ? ? * chapter v , theorem 6.2 ) , each @xmath147 is matlis dual to @xmath148 for some normalized dualizing complex @xmath149 . since an element @xmath150 annihilates a finitely generated @xmath22-module if and only if @xmath151 annihilates the matlis dual of a module , it follows that there exists @xmath152 such that @xmath153 for all @xmath154 . ( vii ) then implies that there exists @xmath84_e$ ] such that the composition @xmath155 is an isomorphism . taking local cohomology for @xmath154 gives us an isomorphism : @xmath156 where the middle map is the zero map . thus @xmath157 which completes the proof . we also state a generalization of ( * ? ? ? * theorem 4.3 ) , similar computations were done in @xcite . [ lem.s2g1impliesfrobeniussplityieldsdivisor ] suppose that @xmath25 is @xmath10 and @xmath11 and sharply @xmath2-pure . then there exists an element @xmath158 such that @xmath159 is @xmath6-cartier and @xmath8 is sharply @xmath2-pure . a surjective map @xmath160 induces an effective weil divisorial sheaf @xmath161 by ( * ? ? ? * proposition 2.9 ) such that @xmath162 . we would like to show that @xmath161 can be identified with an element of @xmath163 . at the singular height one points @xmath50 of @xmath1 , @xmath164 is already gorenstein . thus we can consider the map @xmath165 which generates @xmath166 as an @xmath167-module . set @xmath39 to be the maximal ideal of @xmath168 and notice that @xmath39 is the conductor ideal since @xmath2-pure rings are seminormal @xcite and in particular the conductor is radical . but then @xmath169 by the proof of ( * ? ? * proposition 1.2.5 ) . now , we know @xmath170 is equal to @xmath171 for some @xmath172 . we want to show that @xmath173 is a unit , which would prove that @xmath161 is trivial at @xmath50 . since @xmath174 is surjective , we see that @xmath175 and thus @xmath173 is a unit . this implies that the weil divisorial sheaf @xmath176 coincides with @xmath164 and thus @xmath177 as desired . finally , set @xmath178 . we conclude by recalling a well known lemma on the height of annihilators of local cohomology modules . however , because we lack a reference , we provide a proof . [ lem.dimensionsoflocalcohomology ] suppose that @xmath37 is a local ring and suppose that @xmath32 is a finitely generated @xmath22-module which is @xmath179 in the sense for all @xmath180 . note that here we use @xmath181 not @xmath182 . ] of @xcite . set @xmath183 . suppose that @xmath184 is the minimum over the dimensions of the components of @xmath185 . then @xmath186 for @xmath187 . note @xmath188 may not be the same as @xmath189 since @xmath125 is not finitely generated . set @xmath190 to be a normalized dualizing complex on @xmath1 ( recall that all our rings are excellent and possess dualizing complexes ) . by local duality in the form of ( * chapter v , theorem 6.2 ) , it is equivalent to prove that @xmath191 . suppose this is false , and thus that @xmath192 is an irreducible component of dimension @xmath193 for some @xmath187 . set @xmath194 to be the generic point of @xmath100 ( which we also view as a prime ideal ) . by localizing at @xmath194 , we see that @xmath195 ) ) = \myh^{-i + t } \myr \hom_{r_{\gamma } } ( m_{\gamma } , \omega_{r_{\gamma}}^{\mydot}))\ ] ] is supported at a point . the shift by @xmath196 $ ] is necessary to keep the dualizing complexes normalized . thus @xmath197 by local duality again . now , @xmath198 . also observe that @xmath199 ( this is why the @xmath184 is necessary since we do not know what component of @xmath185 we will be restricting to ) . since @xmath200 is still @xmath179 , we see that @xmath201 for @xmath202 . but then @xmath203 since @xmath204 and @xmath205 . setting @xmath206 we obtain a contradiction . our goal in this section is to prove several results on the depths of sheaves on schemes with controlled @xmath2-singularities . first we prove our result for pairs @xmath87 which are strongly @xmath2-regular , this is the simplest case . * theorem 2 ) ) [ thm.kollartheorem2 ] suppose that @xmath37 is local and that @xmath207 is strongly @xmath2-regular . further suppose that @xmath28 is such that @xmath208 for some integral divisor @xmath5 and some integer @xmath30 realtively prime to @xmath31 . then @xmath4 is cohen - macaulay . by possibly multiplying @xmath173 with an integer , we may assume that @xmath209 . choose , using , @xmath210 such that @xmath211 for all @xmath212 . note that since @xmath213 , @xmath214 and then @xmath215 is strongly @xmath2-regular as well by ( iii ) . therefore , by ( vii ) there exists an @xmath79 and a splitting @xmath41 such that the composition @xmath216 is an isomorphism . by replacing @xmath118 by a multiple if necessary , we may assume that this @xmath79 also satisfies the condition from the hypothesis . twisting by @xmath4 , reflexifying , and applying @xmath217 we obtain the following composition which is also an isomorphism . @xmath218 however , the map labeled @xmath219 is the zero map for @xmath220 since @xmath221 thus @xmath222 as desired . if one assumes that @xmath8 is purely @xmath2-regular ( an analog of purely log terminal @xcite ) , the same result holds by the same proof . the point is that we may take @xmath151 annihilating @xmath147 and which simultaneously does nt vanish along the support of any component of @xmath223 . * corollary 5.25 ) ) if @xmath37 is local and @xmath207 is strongly @xmath2-regular , then for every @xmath6-cartier integral divisor @xmath5 , @xmath4 is cohen - macaulay . if the index of @xmath5 is not divisible by @xmath31 then the statement is a special case of by setting @xmath224 . hence assume that the index @xmath225 of @xmath5 is divisible by @xmath31 . choose then an effective divisor @xmath226 linearly equivalent to @xmath5 and set @xmath227 , @xmath228 for some integer @xmath229 . in this situation @xmath173 is relatively prime to @xmath31 and @xmath230 furthermore , for @xmath229 , @xmath231 is strongly @xmath2-regular . hence , we may apply for @xmath7 replaced by @xmath232 and the above choices of @xmath173 , @xmath5 and @xmath223 . this concludes our proof . before moving on to the sharply @xmath2-pure pairs , we need a lemma on the existence of certain frobenius splittings . [ lem.splittingnotcompatiblewithm ] suppose that @xmath83 is any pair where @xmath68 is a cartier subalgebra on a local ring @xmath37 . suppose @xmath39 is not @xmath68-compatible . then there exists some frobenius splitting such that @xmath233 . there obviously exists a map in @xmath72_e$ ] , @xmath234 , such that @xmath235 . it follows that @xmath236 . we have two cases : suppose that there is a unit @xmath237 such that @xmath238 . thus @xmath239 . consider the map @xmath240 and notice that @xmath241 which shows that @xmath41 is a splitting . also notice that @xmath39 is not @xmath41-compatible since @xmath41 is a unit multiple of @xmath143 . thus we have found our @xmath41 . since we have already handled case 1 , we may assume that @xmath242 for all units @xmath243 . choose @xmath244 such that @xmath245 . now then @xmath246 since @xmath137 is a unit . thus @xmath247 is a unit . but this is a contradiction since @xmath151 is assumed not be a unit and so @xmath248 is a unit . [ rem.thereexistsanicedivisor ] if , in addition to the hypotheses of , @xmath22 is @xmath10 and @xmath11 , then by applying the argument of to the splitting @xmath41 constructed in the proof of , we obtain a @xmath6-divisor @xmath7 on @xmath25 such that * @xmath159 is cartier . * @xmath8 is sharply @xmath2-pure . * @xmath249 is not an @xmath2-pure center of @xmath8 . the second two statements follow since @xmath250_e$ ] . * theorem 3(2 ) ) ) [ thm.kollartheorem3(2 ) ] suppose that @xmath146 is a local @xmath10-ring , and @xmath68 is a cartier - subalgebra on @xmath22 such that @xmath83 is sharply @xmath2-pure and @xmath251 is not an @xmath2-pure center . additionally suppose that @xmath252 is any union of @xmath2-pure centers of @xmath83 . we also assume that @xmath40 and @xmath1 have no common irreducible components . if @xmath253 is the ( radical ) ideal defining @xmath40 , then @xmath43 since shrinking @xmath68 is harmless , we set @xmath93 for some splitting @xmath38 which by is not compatible with the origin @xmath251 . indeed , by ( iii ) we can only increase the number of centers when restricting a cartier subalgebra . we have the long exact sequence : @xmath254 and recall we are trying to show that @xmath255 for @xmath256 . since @xmath22 is @xmath10 , @xmath257 . since @xmath258 and @xmath40 is reduced , @xmath259 and so @xmath260 . thus @xmath261 . it is now sufficient to prove the case when @xmath262 . this implies that @xmath263 ( since @xmath40 and @xmath1 have no common components ) . furthermore , we can assume that every component of @xmath40 has dimension at least 2 . indeed , suppose that @xmath264 is an irreducible component of @xmath40 such that @xmath265 . if @xmath266 is the union of the other components of @xmath40 and @xmath267 , then @xmath268 ( for dimension reasons and since we working in a local ring ) . but this implies that @xmath27 is an @xmath2-pure center since intersections of @xmath2-pure centers are unions of @xmath2-pure centers by ( i ) . thus we can assume that @xmath269 is @xmath137-dimensional . but then @xmath270 , which contradictions our assumption . by , we know that @xmath271 is compatible with @xmath272 however , if @xmath273 , then since @xmath22 is @xmath10 and of dimension @xmath3 , @xmath274 by . but @xmath271 is radical ( since @xmath68 is sharply @xmath2-pure ) so @xmath275 . but @xmath251 is not an @xmath2-pure center , this is a contradiction . we conclude that @xmath276 . now we come to @xmath277 . again , since @xmath278 is reduced , @xmath278 is s1 . furthermore , since @xmath40 has no 1-dimensional components we can apply to conclude that @xmath279 can either be @xmath39-primary or @xmath22 . suppose it is @xmath39-primary . since @xmath280 is still a splitting , it follows that @xmath281 is @xmath280-compatible and also radical and so equal to @xmath282 . but then @xmath39 is @xmath41-compatible by basic facts about frobenius splitting or by ( vi ) . we conclude that @xmath283 . this forces @xmath284 to be zero and completes the proof . [ rem.kolthminfsplitlanguage ] another way to state a special case of using the language of frobenius splittings is as follows : _ suppose that @xmath37 is an @xmath10 local ring with frobenius splitting @xmath38 which is not compatibly split with @xmath39 . additionally suppose that @xmath40 is any union of compatibly split subvarieties of @xmath25 such that no irreducible component of @xmath40 coincides with an irreducible component of @xmath1 . suppose that @xmath42 is the ideal defining @xmath40 , then @xmath43 _ now we come to our main technical result on depth . it is a characteristic @xmath0 version of ( * ? ? ? * theorem 3(1 ) ) but also compare with ( * ? ? ? * lemma 3.2 ) , ( * ? ? ? * theorem 4.21 ) , ( * ? ? ? * theorem 1.5 ) and ( * ? ? ? * theorem 1.2 , 1.5 ) . * theorem 3(1 ) ) ) [ thm.kollartheorem3(1 ) ] suppose that @xmath22 is local , @xmath10 and @xmath11 , @xmath25 and that @xmath74 is such that @xmath87 is sharply @xmath2-pure . set @xmath26 to be the closed point and assume that @xmath27 is not an @xmath2-pure center of @xmath87 . suppose that @xmath28 is another element of @xmath285 and that @xmath286 is integral for some @xmath30 relatively prime to @xmath31 . further assume that @xmath32 is any rank-1 ( along each component of @xmath1 ) reflexive coherent subsheaf of @xmath33 such that @xmath34 ( here @xmath35 denotes reflexive power).}$ ] . we do not use that notation since it might be confused with frobenius power . ] then @xmath287 first observe that it is harmless to assume that @xmath288 since otherwise the statement is trivial since @xmath32 is reflexive and thus @xmath10 by ( * ? ? ? * theorem 1.9 ) . we may also assume that @xmath289 is an ideal sheaf since we are working locally . we thus identify @xmath32 with an ideal of @xmath22 also denoted by @xmath32 . finally , by replacing @xmath173 by a power if necessary , we may assume that @xmath290 for some @xmath79 . using , we can find @xmath41 a splitting , not compatible with @xmath39 , making the following composition an isomorphism : @xmath291 twisting by @xmath32 and reflexifying ( which we denote by @xmath292 ) , we obtain @xmath293 using the fact that @xmath294 , we have a composition @xmath295 that is an isomorphism ( note the first map is not the usual inclusion of ideal sheaves via frobenius ) . certainly @xmath296 since @xmath32 is reflexive and thus @xmath10 by ( * ? ? ? * theorem 1.9 ) . we now study @xmath297 . since @xmath32 is @xmath10 , it follows that either @xmath298 is equal to @xmath22 or it is @xmath39-primary by . since we have an injection @xmath299 , it follows that @xmath298 is at the very least radical ( since if @xmath300 kills @xmath297 , so does @xmath173 ) . in particular , if @xmath301 , then it must be @xmath39 . fix @xmath128 \in h^2_{\bm}(m)$ ] , and recall that we are considering @xmath32 as an ideal . if @xmath302 , and local cohomology classes are treated as classes on @xmath303 as explained earlier . ] , then since @xmath304 we have @xmath305 thus we have a class @xmath306 \in h^2_{\bm}\big(f^e _ * ( { \mathcal{o}}_x((p^e - 1)\delta ' ) \tensor m^{(p^e - 1 ) } \otimes m)^{**}\big).\ ] ] now , when we apply @xmath307 to this class , it is just applied component - wise . thus @xmath308 ) = [ z]$ ] . for any @xmath309 , it follows that @xmath310 ) = { \varphi}(f^e _ * r).[z]$ ] . in particular , if an arbitrary @xmath311 annihilates all classes @xmath312 \in h^2_{\bm}(f^e _ * ( { \mathcal{o}}_x((p^e - 1)\delta ' ) \tensor m^{(p^e - 1 ) } \otimes m)^ { * * } ) \cong h^2_{\bm}(f^e _ * m)$ ] , then @xmath313 also annihilates all such @xmath128 \in h^2_{\bm}(m)$ ] . this proves that @xmath39 is @xmath41-compatible , a contradiction . here we list the most important corollaries of the results of . the characteristic zero analog of many of them are already mentioned in @xcite . we still state them here for the sake of completeness and we give a full proof of our main motivation , the compatibility of the relative canonical sheaf with base change . in section [ sec : applications_lemmas ] some lemmas are gathered while in section [ sec : applications_corollaries ] the promised corollaries are presented . [ fact : s_d_hyperplane ] ( * ? ? ? * theorem 1.2.5 ) let @xmath314 be a coherent sheaf on a noetherian scheme @xmath1 , @xmath315 a cartier divisor on @xmath1 containing a point @xmath316 , such that the local equation of @xmath315 at @xmath316 is not a zero divisor of @xmath317 ( in other words , it is a regular element for @xmath318 ) . then [ lem : s_d ] if @xmath321 is a morphism of noetherian schemes , @xmath322 is a coherent sheaf on @xmath1 flat over @xmath19 , such that @xmath323 is @xmath324 for every @xmath17 ( i.e. , @xmath314 is relatively @xmath324 over @xmath19 ) and @xmath325 is an @xmath324 coherent sheaf on @xmath19 , then @xmath326 is @xmath324 as well . @xmath330{ega_iv_ii } } } \\ \geq & \underbrace{\min\big\{d , \dim_{\so_{f , p } } \left(\sf|_f \right)_p\big\ } + \depth_{\so_{y , q } } \sg_{q}}_{\textrm{$\sf|_f$ is { $ \textnormal{s}_d$ } } } \\ = & \min\big\{d + \depth_{\so_{y , q } } \sg_q , \dim_{\so_{f , p } } \left(\sf|_f \right)_p + \depth_{\so_{y , q } } \sg_q\big\ } \\ \geq & \underbrace{\min\big\{d + \depth_{\so_{y , q } } \sg_q , \dim_{\so_{f , p } } \left(\sf|_f \right)_p + \min\{d , \dim_{\so_{y , q } } \sg_q\}\big\}}_{\textrm{$\sg$ is { $ \textnormal{s}_d$ } } } \\ = & \min\big\{d + \depth_{\so_{y , q } } \sg_q , \min\{\dim_{\so_{f , p } } \left(\sf|_f \right)_p + d,\dim_{\so_{f , p } } \left(\sf|_f \right)_p + \dim_{\so_{y , q } } \sg_q\}\big\ } \\ = & \underbrace{\min\big\{d + \depth_{\so_{y , q } } \sg_q , \min\{\dim_{\so_{f , p } } \left(\sf|_f \right)_p + d,\dim_{\so_{x , p } } ( \sf \otimes f^ * \sg)_p \}\big\}}_{\textrm{$\sf$ is flat over $ y$ and \cite[corollaire 6.1.2]{ega_iv_ii } } } \\ \geq & \min\big\{d + \depth_{\so_{y , q } } \sg_q , \min\{d,\dim_{\so_{x , p } } ( \sf \otimes f^ * \sg)_p \}\big\ } \\ \geq & \min\big\{d , \dim_{\so_{x , p } } ( \sf \otimes f^ * \sg)_p \big\ } . \end{array}\ ] ] [ lem : gorenstein_base ] if @xmath321 is a flat morphism of finite type to a gorenstein scheme , then @xmath331 and @xmath332 $ ] are locally isomorphic . in particular if @xmath1 is relatively gorenstein over @xmath19 , then it is gorenstein , and @xmath18 and @xmath333 are locally isomorphic over @xmath19 . [ [ restricting - divisors - by - f - adjunction - the - f - different ] ] restricting divisors by @xmath2-adjunction : the @xmath2-different ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ suppose that @xmath12 is a @xmath6-divisor on an @xmath10 and @xmath11 variety @xmath1 and that @xmath159 is cartier . in fact , everything we say even holds more generally if @xmath12 is a @xmath335-weil divisorial sheaf , which is intuitively something like a weil divisor having components also in the singular locus-@xmath336-divisors . further suppose that @xmath5 is a reduced cartier divisor on @xmath1 that is itself @xmath10 and @xmath11 and which has no common components with @xmath7 . we now explain how we can construct a canonical @xmath335-weil divisorial sheaf ( not necessarily a @xmath335-weil divisor ) which we call @xmath337 on @xmath5 . here the subindex @xmath2 means that this is the @xmath2-singularity counterpart of the usual different known from minimal model program theory . however , contrary to the usual different , the construction of @xmath338 goes through without any further assumption requiring that @xmath7 is @xmath6-cartier at certain points . without loss of generality we can assume that @xmath25 and that @xmath22 is a local ring and that @xmath339 . the fact that the divisor is cartier implies that is a free @xmath340-module . choose a generator of this module @xmath41 , which we can also view as an element of @xmath341 since @xmath7 is effective . now define a new map @xmath234 by the rule @xmath342 . certainly @xmath343 and thus @xmath143 induces a map @xmath344 . note that for every height one prime @xmath50 containing @xmath345 , _ i.e. _ , a minimal associated prime of @xmath5 , we have that @xmath346 ( since @xmath7 and @xmath5 have no common components ) . furthermore , @xmath347 is regular ( since @xmath348 is reduced and zero dimensional and hence regular ) . it follows from inspection that @xmath349 is non - zero at every such @xmath50 . by ( * theorem 2.4 ) , it follows that @xmath349 induces an effective @xmath335-weil divisorial sheaf on @xmath5 . it is straightforward to verify that @xmath350 is independent of the choice of @xmath118 and @xmath41 and so : [ obs.fadjunction ] notice now additionally that @xmath349 ( corresponding to @xmath337 ) is surjective if and only if @xmath143 ( corresponding to @xmath351 ) is surjective . in other words , @xmath352 is sharply @xmath2-pure near @xmath5 if and only if @xmath353 is sharply @xmath2-pure . it will be useful for us to note that if @xmath353 is sharply @xmath2-pure , then @xmath337 is in fact an honest @xmath335-weil divisor and not just a @xmath335-divisorial sheaf . suppose not , then the weil divisorial sheaf @xmath354 must properly contain @xmath355 , even at some generic point of the non - normal locus . a contradiction can then be obtained from the fact that the conductor is already compatible with every @xmath145 ( this last fact follows from the argument of ( * ? ? ? * proposition 1.2.5 ) ) . the above introduced @xmath2-different @xmath337 is equal to @xmath350 in most cases when the latter is defined . [ lem : restriction_equals_different ] with the notation above , suppose additionally that @xmath7 is @xmath335-cartier at all of the height - two primes primes of @xmath22 containing @xmath356 ( the codimension 2 points of @xmath1 that are contained inside @xmath5 ) . then @xmath337 coincides with the restriction @xmath350 of @xmath7 to @xmath5 . at each of those codimension 2 points @xmath357 , @xmath358 is already gorenstein ( since @xmath359 is @xmath11 and so @xmath360 is gorenstein ) . it is enough to prove the result at each such @xmath361 , so fix one such @xmath361 . further choose @xmath79 as above and also sufficiently divisible such that @xmath44 is cartier at @xmath361 . we can thus write @xmath362 for some @xmath363 . since @xmath358 is gorenstein , we can choose @xmath364 generating the set as an @xmath365-module . consider the map @xmath366 defined by the rule @xmath367 . certainly @xmath368 restricts to a map @xmath369 as above . furthermore , @xmath370 generates the @xmath371-module @xmath372 by the diagrams in ( * ? ? ? * proof of proposition 7.2 ) . it follows that @xmath373 restricts to @xmath374 and hence corresponds to the naive restriction @xmath375 . this proves the lemma . [ lem : base_change ] let @xmath376 be a flat morphism of finite type with @xmath10 , @xmath11 equidimensional fibers to a smooth variety @xmath19 and @xmath377 , such that @xmath14 is @xmath6-cartier and @xmath15 . assume also that @xmath378 is a smooth cartier divisor such that for @xmath379 , @xmath7 does not contain any component of @xmath100 and @xmath380 is sharply @xmath2-pure . is sharply @xmath2-pure implies that @xmath381 is a @xmath335-divisor and not simply a @xmath335-weil divisorial sheaf . ] then @xmath382 . by , both @xmath1 and @xmath100 are @xmath10 . similarly , both are @xmath11 by . by @xmath2-adjunction , _ cf . _ * main theorem , proposition 7.2 , remark 7.3 ) , @xmath383 is sharply @xmath2-pure in a neighborhood of @xmath100 . hence , so is @xmath384 . we now claim : suppose that @xmath385 was an @xmath2-pure center of @xmath8 contained in @xmath100 . let @xmath50 denote the generic point of @xmath40 and now we work in @xmath386 with maximal ideal @xmath39 corresponding to @xmath50 . for any element @xmath387 of @xmath388_e$ ] , we notice that @xmath389 since @xmath40 is an @xmath2-pure center . choose @xmath390 to be the defining equation of the cartier divisor @xmath100 in @xmath164 . it follows from construction that @xmath391_e = [ \sc^{\delta + w}]_e$ ] . in other words , for any @xmath392_e$ ] , we can write @xmath393 for some @xmath250_e$ ] . with this notation , for any @xmath133 we have @xmath394 . this proves that @xmath395 is not sharply @xmath2-pure at @xmath50 , the generic point of @xmath40 . but we assumed that @xmath396 was sharply @xmath2-pure , a contradiction which proves the claim . we return to the proof of . by then , for every @xmath397 , @xmath398 however , by , @xmath333 and @xmath18 are isomorphic locally , and then in the above inequality @xmath333 can be replaced by @xmath18 . then by , @xmath399 is @xmath10 . to be precise , to apply , one needs to prove a priori that the local equation of @xmath100 is not a zero - divisor of @xmath18 . for this it is enough to show that @xmath18 is @xmath400 , which follows using again that locally @xmath18 and @xmath333 are isomorphic and that @xmath333 is @xmath10 by ( * ? ? ? * corollary 5.69 ) . therefore @xmath399 is @xmath10 indeed . however , so is @xmath401 by using ( * ? ? ? * corollary 5.69 ) again . furthermore , @xmath399 and @xmath401 are isomorphic on the relative gorenstein locus , since the relative canonical sheaf is compatible with base - change for gorenstein morphisms ( * ? ? ? * theorem 3.6.1 ) . therefore ( * ? ? ? * theorems 1.9 and 1.12 ) yields the statement of the lemma . [ lem : relative_canonical_reflexive ] suppose that @xmath19 is a scheme of finite type over an algebraically closed field . if @xmath321 is a projective , flat , relatively @xmath10 and @xmath402 , equidimensional morphism , then @xmath18 is reflexive . let @xmath100 be the non - relatively gorenstein locus . fix a finite surjective morphism @xmath407 over @xmath19 , after possibly shrinking @xmath19 ( _ cf . _ ( * ? ? ? * proof of corollary 24 ) ) . set then @xmath408 , @xmath409 . let @xmath410 be the natural inclusion . with the above choices , @xmath404 is satisfied for all @xmath17 . for the other condition , notice that it is enough to prove that the natural homomorphism @xmath411 is an isomorphism . however @xmath412{hassettkovacsreflexivepullbacks } using that $ \omega_{\bp^n_y / y}$ is flat and relatively $ s_2 $ } } \cong \pi _ * \omega_{x / y},\end{gathered}\ ] ] and the composition of the above isomorphisms is the natural homomorphism @xmath411 . * lemma 3.2 ) , ( * ? ? ? * theorem 4.21 ) , @xcite , ( * ? ? ? * theorem 1.5 ) , ( * ? ? ? * theorem 1.2 , 1.5 ) ) suppose that @xmath25 is @xmath10 and @xmath11 . if @xmath1 is @xmath2-pure and @xmath26 is not an @xmath2-pure center of @xmath1 , then @xmath413 we may assume that @xmath25 for a local ring @xmath37 with @xmath249 . since @xmath1 is @xmath10 and @xmath11 , by using , we can assume that there exists some @xmath12 such that @xmath159 is cartier , such that @xmath8 is sharply @xmath2-pure and such that @xmath27 is not an @xmath2-pure center of @xmath8 . now the second statement follows from by setting @xmath414 and setting @xmath415 . the first statement also follows from by setting @xmath416 and @xmath417 . to prove our main corollary , we need to introduce a generalization of @xmath419 to the case when @xmath5 has higher codimension . we focus only on our case of interest , that is , when @xmath5 is the fiber over a smooth base . let @xmath9 be a flat morphism of finite type with @xmath10 , @xmath11 equidimensional fibers . further suppose that @xmath19 is a smooth variety , and @xmath377 is such that @xmath7 does not contain any component of any fiber . fix a point @xmath17 . working locally , we may assume that @xmath420 and @xmath25 for local rings @xmath421 and @xmath37 . further assume that @xmath422 is regular system of generators with @xmath423 regular . then , let @xmath424 and define then @xmath425 . the only question is whether this construction of @xmath426 is independent of the choice of @xmath427 . following the method of @xmath2-adjunction , multiplying by each @xmath427 successively , we take a map corresponding to @xmath7 and @xmath38 and obtain another map @xmath428 defined by the rule @xmath429 . we then restrict this map to @xmath430 by modding out by @xmath431 and so obtain @xmath432 . choosing different @xmath188 s is simply choosing a different set of generators @xmath433 for @xmath431 which yields @xmath434 . to complete the proof of the claim , it is sufficient to show that these maps differ only by multiplication by a unit . we use @xmath435}$ ] to denote the ideal generated by the @xmath130th powers of the generators of @xmath431 . since @xmath435 } : \bn = \langle ( f_1 \cdots f_n)^{p^e - 1 } \rangle + \bn^{[p^e ] } = \langle ( g_1 \cdots g_n)^{p^e - 1 } \rangle + \bn^{[p^e]}$ ] , see for example ( * ? ? ? * proposition 2.1 ) , it follows that @xmath436 for some unit @xmath437 , elements @xmath438 and @xmath439 . but now we see that @xmath440 since any multiple of @xmath441 will be sent into @xmath442 . this completes the proof . [ cor : relative_canonical_sheaf_base_change ] ( _ cf._@xcite ) let @xmath9 be a flat morphism of finite type with @xmath10 , @xmath11 , equidimensional fibers to a smooth variety and let @xmath377 be such that it does not contain any component of any fiber . additionally assume that @xmath14 is @xmath6-cartier , @xmath15 and @xmath443 is sharply @xmath2-pure for every @xmath17 . then @xmath18 is flat over @xmath19 and compatible with arbitrary base change . we claim that @xmath18 is flat over @xmath19 and relatively @xmath10 . by ( * lemma 2.13 ) , flatness follows as soon as we prove that the restriction of @xmath18 to every fiber is @xmath400 . on the other hand relatively @xmath10 means the stronger condition that the above restrictions are @xmath10 . therefore to show the claim , it is enough to prove that @xmath444 is @xmath10 for every @xmath17 . by ( * corollary 5.69 ) , @xmath445 is @xmath10 and hence , it is enough to show that @xmath446 locally around every point @xmath447 . we thus replace both @xmath1 and @xmath19 by @xmath448 and @xmath449 , respectively . therefore , we may assume that there is a sequence of smooth subvarieties : @xmath450 , such that @xmath451 is a cartier divisor in @xmath188 . set @xmath452 and @xmath453 with @xmath454 . note that then @xmath455 . by applying ( backwards ) inductively and possibly further restricting @xmath1 around @xmath27 , one obtains that @xmath456 and hence @xmath457 is sharply @xmath2-pure for all @xmath458 ( in fact , this also implies all the @xmath335-weil divisorial sheaves @xmath459 are honest divisors ) . finally , applying inductively again yields that @xmath460 for all @xmath458 , and consequently @xmath446 . this finishes the proof of our claim . by our claim and ( * ? ? ? * corollary 3.8 ) @xmath18 and all its pullbacks are reflexive . hence , by restricting to the relatively cohen - macaulay locus ( whose complement has codimension @xmath461 ) and using ( * ? ? ? * proposition 3.6 ) , for any morphism @xmath462 , @xmath463 . by , the appearance of @xmath426 can be replaced by an actual `` geometric '' restriction , if we assume the following : @xmath464 in particular , this is satisfied if @xmath465 does not contain the singular codimension one points of the fibers . indeed , let @xmath466 be a codimesnion 1 point of a fiber @xmath430 . if @xmath466 is in the singular locus of @xmath430 , then @xmath467 and hence @xmath7 is cartier at @xmath466 . otherwise , @xmath1 is smooth around @xmath466 , and hence @xmath468 is cartier at @xmath466 . in particular then by @xmath15 , we obtain that @xmath7 is @xmath335 cartier at @xmath466 . in either cases @xmath7 satisfies , and therefore in the special case of stated in , the use of ordinary restriction of @xmath7 was legitimate . [ cor : relative_canonical_sheaf_base_change_projective ] let @xmath9 be a flat projective morphism with @xmath10 , @xmath11 equidimensional fibers . further suppose that @xmath19 is a reduced , separated scheme of finite type over an algebraically closed field , and @xmath7 a @xmath6-weil - divisor that avoids all the codimension zero and the singular codimension one points of the fibers . additionally assume that there is a @xmath469 , such that @xmath470 is cartier in relative codimension one and @xmath471}(n \delta)$ ] be the intersection of the relative gorenstein locus and the locus where @xmath470 is cartier . set @xmath472 for the natural inclusion . the sheaf @xmath471}(n \delta)$ ] is the reflexive hull of @xmath473 , i.e. the unique reflexive sheaf that restricts on @xmath303 to the above sheaf . it can be obtained as @xmath474 . indeed , @xmath474 is reflexive by ( * ? ? ? * corollary 3.7 ) and it is unique by ( * ? ? ? * proposition 3.6 ) . ] is a line bundle and that @xmath16 is sharply @xmath2-pure for every @xmath17 . then @xmath18 is flat and compatible with arbitrary base change . first , we need some preparation about pulling back @xmath7 . suppose @xmath475 is a morphism and set @xmath476 , @xmath477 , and @xmath478 the induced morphisms . then a natural pullback @xmath223 of @xmath7 can be defined as follows . let @xmath479 be the open set where @xmath356 is gorenstein and @xmath7 is @xmath6-cartier . then , pull @xmath480 back to @xmath481 , and finally extend it uniquely over @xmath482 . this extension is unique , since @xmath483 . we claim that @xmath484}(n \delta ) \cong \omega_{x'/y'}^{[n]}(n \delta').\ ] ] indeed , notice that by construction @xmath485}(n \delta)$ ] and @xmath486}(n \delta')$ ] agree over @xmath481 , that is , in relative codimension one . notice also that since @xmath471}(n \delta)$ ] is assumed to be a line bundle , so is @xmath485}(n \delta)$ ] , and therefore @xmath485}(n \delta)$ ] is reflexive . on the other hand , since @xmath486}(n \delta')$ ] is defined as a pushforward of a line bundle from relative codimension one , it is reflexive by ( * ? ? ? * corollary 3.7 ) . therefore by ( * ? ? ? * proposition 3.6 ) , holds . in particular , @xmath486}(n \delta')$ ] is a line bundle . the main consequence of the previous paragraph , is that the conditions of the corollary are invariant under pullback to another reduced , separated scheme @xmath487 of finite type over @xmath488 . that is , @xmath482 , @xmath489 and @xmath223 defined above satisfy all the assumptions of the corollary . let us turn now to the actual proof of the corollary . first , we may assume that @xmath19 is connected . second , according to ( * ? ? ? * corollary 24 ) and , there is a locally closed decomposition @xmath490 , such that if @xmath491 , then @xmath492 is flat and commutes with base - change if and only if @xmath491 factors through some @xmath493 . now , for every irreducible component @xmath487 of @xmath19 , there is a regular alteration @xmath494 ( * ? ? ? * thorem 4.1 ) . by the above discussion @xmath495 satisfies the assumptions of the corollary , and hence also of . therefore , @xmath496 is flat and compatible with arbitrary base - change . hence @xmath497 factors through one of the @xmath188 . in particular , since the image of @xmath497 is the component @xmath487 , @xmath498 . that is , every irreducible component of @xmath19 is contained in one @xmath188 . however , @xmath19 is connected , therefore all irreducible components of @xmath19 are contained in the same @xmath188 , and hence by the reducedness of @xmath19 , @xmath499 . in the case of @xmath500 , if instead of assuming that @xmath430 is sharply @xmath2-pure , one assumes that @xmath501 is @xmath2-pure for all @xmath17 , the @xmath15 assumption can be dropped from the above corollaries using the trick of . [ qtn : aribtrary_base ] does the compatibility of the relative canonical sheaf with base change stated in hold for singular @xmath19 ( with the adequate modification in the setup as in ) ? from the modular point of view , especially interesting would be the case of non - reduced @xmath19 . this case is open even in the projective case . it should be noted that the characteristic zero analogue of is known if @xmath356 is projective and @xmath19 is arbitrary ( * ? ? ? * theorem 7.9 ) . that is , the answer to the characteristic zero analogue of is positive when @xmath356 is projective . [ qtn : log_canonical ] can one replace sharply @xmath2-pure by log - canonical ( still assuming positive characteristic ) in the statement of ? this would also be important from the modular point of view , since sharply @xmath2-pure varieties can be deformed to log - canonical but not sharply @xmath2-pure varieties . : _ residues and duality _ , lecture notes of a seminar on the work of a. grothendieck , given at harvard 1963/64 . with an appendix by p. deligne . lecture notes in mathematics , no . 20 , springer - verlag , berlin , 1966 . mr0222093 ( 36 # 5145 ) : _ generalized divisors on gorenstein schemes _ , proceedings of conference on algebraic geometry and ring theory in honor of michael artin , part iii ( antwerp , 1992 ) , vol . 8 , 1994 , pp . 287339 . mr1291023 ( 95k:14008 ) : _ flips and abundance for algebraic threefolds _ , socit mathmatique de france , paris , 1992 , papers from the second summer seminar on algebraic geometry held at the university of utah , salt lake city , utah , august 1991 , astrisque no . 211 mr1225842 ( 94f:14013 ) : _ birational geometry of algebraic varieties _ , cambridge tracts in mathematics , vol . 134 , cambridge university press , cambridge , 1998 , with the collaboration of c. h. clemens and a. corti , translated from the 1998 japanese original . mr1658959 ( 2000b:14018 ) : _ contre - exemple au `` vanishing theorem '' en caractristique @xmath504 _ , c. p. ramanujam a tribute , tata inst . fund . studies in math . 8 , springer , berlin , 1978 , pp . 273278 . 541027 ( 81b:14011 )

we present the very long baseline interferometry ho maser monitoring observations of the red supergiant , pz cas , at 12 epochs from 2006 april to 2008 may . we fitted maser motions to a simple model composed of a common annual parallax and linear motions of the individual masers . the maser motions with the parallax subtracted were well modeled by a combination of a common stellar proper motion and a radial expansion motion of the circumstellar envelope . we obtained an annual parallax of 0.356 0.026 mas and a stellar proper motion of and mas yr eastward and northward , respectively . the annual parallax corresponds to a trigonometric parallax of kpc . by rescaling the luminosity of pz cas in any previous studies using our trigonometric parallax , we estimated the location of pz cas on a hertzsprung - russell diagram and found that it approaches a theoretically evolutionary track around an initial mass of . the sky position and the distance to pz cas are consistent with the ob association , cas ob5 , which is located in a molecular gas super shell . the proper motion of pz cas is close to that of the ob stars and other red supergiants in cas ob5 measured by the _ hipparcos _ satellite . we derived the peculiar motion of pz cas of 1.5 , 4.4 , and 4.4 km s . this peculiar motion has rather a large component , unlike those of near high - mass star - forming regions with negatively large motions . the uniform proper motions of the cas ob5 member stars suggest random motions of giant molecular clouds moving into local potential minima in a time - dependent spiral arm , rather than a velocity field caused by the spiral arm density wave .

1308.3580

recent very long baseline interferometry ( vlbi ) astrometry can determine distances to galactic objects on the kiloparsec scale . one of thse objects is a red supergiant ( rsg ) harboring a thick circumstellar envelope . since those sources radiate water masers in the radio frequency ( rf ) , we can measure trigonometric parallaxes by referring distant extragalactic radio sources using a phase - referencing technique ( e.g. , @xcite ) . if we can measure the trigonometric parallax distance of such a star , this enables us to evaluate the absolute magnitude . since the location of an rsg in an evolutionary track of a hertzsprung russell ( h - r ) diagram is not mainly dependent on an effective temperature , we can make a constraint on the initial mass from the absolute magnitude @xcite . in addition , because rsgs are sometimes seen in large ob associations in the milky way , they can become distance indicators for such ob associations . determining trigonometric rather than photometric distances to ob associations is important for the verification of the calibration method of the distance ladder . information about trigonometric parallaxes and proper motions together with radial velocities , also enables us to investigate the three - dimensional ( 3d ) motions of the stars in the milky way . because proper motions of the water masers around rsgs can be well fitted to a simple expanding flow even with some inhomogeneity in it , which is different from the case of star - forming regions , the proper motion of a mass center of an rsg can be rather easily estimated from the phase - referencing vlbi . @xcite obtained the 3d motion of an rsg , s per , and found it to have a relatively large deviation ( 15 km s@xmath4 ) from the motion expected from a flat galactic rotation curve . considering the age of s per is a few tens of million years , it is difficult to have a large peculiar motion that old stars often have . however , it is quite important to know whether or not rsgs motions are aligned with their accompanying ob associations for further discussions about milky way dynamics . @xcite pointed out that since s per is involved in the per ob1 association , such a large peculiar motion of an ob association member star may be related to a dynamics and mechanism of the spiral arm formation of the milky way @xcite . here we report on results of phase - referencing vlbi monitoring observations of the h@xmath13o masers associated with another rsg , pz cas , conducted over two years with the vlbi exploration of radio astrometry ( vera ) of the national astronomical observatory of japan ( naoj ) . these observations are described in section [ sec:02 ] . data reduction including astrometric analyses is presented in section [ sec:03 ] . astrometric results of pz cas are presented in section [ sec:04 ] . we discuss the results in section [ sec:05 ] . for sections [ sec:02][sec:04 ] , we adopt a solar motion of 20 km s@xmath4 @xcite relative to the local standard of rest ( lsr ) to the direction of ( @xmath14 , @xmath15)= ( @xmath16 , @xmath17 ) . hereafter this solar motion is referred to as the standard solar motion . with the standard solar motion , the systemic velocity of pz cas is @xmath18 km s@xmath4 , calculated from the heliocentric radial velocity of @xmath19 km s@xmath4 @xcite . on the other hand , we discuss the observational results on the basis of the solar motion reported by @xcite in section [ sec:05 - 03 ] . for the proper motion , we adopt the heliocentric coordinates . vlbi phase - referencing observations of pz cas at 22.2 ghz have been conducted at 12 epochs over two years with four vera antennas at mizusawa , iriki , ogasawara , and ishigakijima ( coded as miznao20 , iriki , ogasa20 , and ishigaki , respectively ) . the observing epochs are listed in table [ tbl:01 ] . the vera antennas have a dual beam receiving system each to observe a closely located pair of sources simultaneously @xcite . in our monitoring program , one beam ( beam a ) observed pz cas and another ( beam b ) observed a closely located continuum source , j233921.1 + 601011 ( hereafter , abbreviated to j2339 + 6010 ) , which was one of the vera 22 ghz calibrator survey sources @xcite @xmath20.7 away and used as a position reference . the observation duration was 810 hr . each observation was divided into several sessions for the simultaneous tracking of pz cas and j2339 + 6010 , each separated by short - term sessions of bright calibrators ( 3c 454.3 and j230043.0 + 165514 ) for 15 minutes in order to check the observing systems . we prepared one baseband channel ( bbc ) with a bandwidth of 16 mhz for pz cas and fourteen bbcs with a total bandwidth of 224 mhz for j2339 + 6010 in left - hand circular polarization . vlbi cross - correlation was carried out with the mitaka fx correlator at naoj , mitaka , japan , to produce vlbi fringe data with 512 spectral channels for pz cas for the central 8 mhz bandwidth , corresponding to a velocity spacing of 0.2107 km s@xmath4 for the h@xmath0o @xmath21 @xmath22 maser line . a typical frequency allocation after the cross - correlation is listed in table [ tbl:02 ] . to describe the phase referencing and imaging analysis process , we often refer to the results of a specific epoch observation ( epoch f on 2007 march 22 ) because we started the data analysis with this epoch data . we analyzed the vlbi data using the standard nrao data reduction software , astronomical image processing software ( aips ) , version 31dec10 . initial calibration tasks were common for both the sources . first , the amplitude calibration was carried out by using the system noise temperatures ( @xmath23 ) and gain calibration data . the measurements of @xmath23 often have abrupt time variations , as shown in figure [ fig:01 ] , especially when rain falls happened during the observation . we consider such rapid and abrupt time variations to be caused by the absorption effect of water puddles on the top of the vera antenna feedome . we carried out a polynomial fitting of the raw @xmath23 data after removing such abrupt changes . at epoch h , since @xmath23 was not recorded for either beams a or b of the ishigaki station . instead , @xmath23 of ogasa20 was used for ishigaki , because the atmospheric environment ( temperature and moisture ) is similar . note that minor amplitude calibration errors at this stage can be approximately corrected in the later self - calibration . the system noise temperatures for beams a and b were basically approximated to be identical , so the time variation of the system noise temperature must be mainly attributed to the common atmosphere . in the next step , the two - bit sampling bias in the analog - to - digital ( a / d ) conversion in vlbi signal processing was corrected . earth - orientation - parameter ( eop ) errors , ionospheric dispersive delays , and tropospheric delays were calibrated by applying precise delay tracking data calculated with an updated vlbi correlator model supplied by the naoj vlbi correlation center . the instrumental path differences between the two beams were corrected using the post - processing calibration data supplied by the correlation center @xcite . the phase tracking centers for pz cas and j2339 + 6010 of the updated correlator model were set to the positions given in table [ tbl:03 ] for all the observing epochs . after the initial calibration , an image of the reference source , j2339 + 6010 , was synthesized using the standard vlbi imaging method with the fringe fitting and self - calibration . we carried out the above process twice to make the image of j2339 + 6010 : the first process started with a single point source model and the second with the reference source image that had been made in the previous process . after imaging , the phase - referencing calibration solutions were directly calculated from the reference source fringe data : the calibrating phase , @xmath24 , at time @xmath25 was obtained from the following calculation with a vector average of all the spectral channels in all the bbcs : + @xmath26 & = & \sum^{n}_{n=1 } \sum^{m}_{m=1 } \mathrm{exp } \ { i [ \phi^{\mathrm{r}}_{\mathrm{raw}}(n , m , t ) \nonumber \\ & + & 2\pi [ \nu^{\mathrm{r}}_{n } + ( m-1)\delta \nu - \nu^{\mathrm{t}}_{0 } ] \delta\tau^{\mathrm{r}}_{\mathrm{g}}(t ) \nonumber \\ & - & \phi^{\mathrm{r}}_{\mathrm{v}}(u(t ) , v(t ) ) ] \ } , \nonumber \ ] ] + where @xmath27 is the raw data fringe phase of the reference source , and @xmath28 is the visibility phase calculated from the clean components of the reference source image for a baseline with @xmath29 and @xmath30 . @xmath31 is the bbc number , @xmath32 is the spectral channel number in the single bbc , @xmath33 is the rf frequency of the strongest h@xmath0o maser emissions in beam a , @xmath34 is the rf frequency at the lower band edge of the @xmath35th bbc for the reference source as listed in table [ tbl:02 ] , @xmath36 khz is the frequency spacing in the single bbc for the reference source , @xmath37 is the integer number indicating the order of the frequency channel in the single bbc , and @xmath38 is the third - order polynomial fitting result of the temporal variation of the group delay obtained from the preceded fringe fitting for the reference source . we refer to the phase calibration solutions obtained with this method as direct phase - transfer ( dpt ) solutions . the dpt solutions were then smoothed by making a running mean with an averaging time of 16 s in order to reduce the thermal noise . the smoothed dpt solutions , as well as the one obtained with an ordinary aips phase - referencing analysis ( aips cl table ) , are shown in figure [ fig:02 ] . those two results are basically consistent with each other . note that even when solutions could not be obtained with the fringe fitting or the following aips self - calibration because the signal - to - noise ratios ( s / n ) was below a cutoff value , the dpt solutions were created from the raw fringe data , as shown for the ogasa20 station in figure [ fig:02 ] . the final astrometric result with the smoothed dpt solutions is consistent with the aips cl table at epoch f. amplitude gain adjustment was performed using a third - order polynomial fitting for the amplitude self - calibration solutions for j2339 + 6010 . since s / n of j2339 + 6010 was unexpectedly low at epoch g , we could not obtain good solutions for j2339 + 6010 in the fringe - fitting . we therefore obtained the group delay solutions from 3c 454.3 and set the amplitude gain to the unity for all the observing time . the dpt solutions were created from the j2339 + 6010 fringe data together with the group delay solutions determined with 3c 454.3 . at epoch i , since iriki did not attend the observation , we used the remaining three stations in the data reduction . the doppler shift in pz cas spectra due to earth s rotation and revolution was corrected after the phase referencing . the observed frequencies of the maser lines were converted to radial velocities with respect to the traditional lsr using the rest frequency of 22.235080 ghz for the h@xmath13o @xmath22 transition . since the cross power spectra within the received frequency region showed a good flatness both in the phase and the amplitude , the receiver complex gain characteristics were not calibrated in this analysis . we inspected the cross power spectra for all the baselines in order to select a velocity channel of the maser emission which would be unresolved and not have rapid amplitude variation . at this step , fast phase changes with a typical timescale of a few minutes to a few tens of minutes can be removed by the phase referencing , but slow time variations including a bias component as shown in figure [ fig:03 ] can not . the maser image was distorted because of the residual time variations of the fringe phase on a timescale of several hours , which might be caused by uncertainties in the updated vlbi correlator model . we conducted the fringe fitting for the selected velocity channel with a solution interval of 20 minutes . we fitted the solutions to a third - order polynomial , as shown in figure [ fig:03 ] , and removed it from both of the fringe phases of pz cas and j2339 + 6010 . this procedure removes the residual fringe phase error from the pz cas data while it is transferred to the j2339 + 6010 data . as a result , this leads to an improvement of the pz cas image quality , but it also creates a distortion of the j2339 + 6010 image . this distortion of the reference source causes an astrometric error in the relative position between the sources , as later discussed in section [ sec:03 - 03 ] . figure [ fig:03 ] also shows the fringe phase residuals obtained with the aips cl table . although the fringe fitting results with the aips cl table can be almost identifiable to those with the dpt solutions , it is later noted that the image peak flux density at @xmath39 km s@xmath4 with the dpt solutions is 2% lower than that with the aips cl table at epoch f. in addition , the image root - mean - square noise resulting from the dpt solutions is 5% higher than that resulting from the aips cl table . we can also see that the residual phases resulting from the dpt solutions in the fifth and sixth observing sessions in figure [ fig:03 ] have a larger deviation than those resulting from the aips cl table . although the number of the calibrated visibilities after applying the dpt solutions was larger than that with the aips cl table , especially in the last two sessions , the visibilities with lower s / n might degrade the final image quality . we have to admit that the aips cl table may lead to a slightly higher quality in the image synthesis because the fringe fitting and self - calibration take off visibility with low s / n . the problem is that there were some observations whose data showed unexpectedly low s / n for certain baselines . for such cases , it was hard to create the aips cl table with the fringe fitting and self - calibration . for the following three reasons , we used the dpt solutions instead of the aips cl table for all the observing epochs : ( 1 ) the dpt is useful even when the fringe fitting fails because of the low s / n , ( 2 ) the final astrometric results between the two methods have little difference , and ( 3 ) it is preferable to use a unified analysis scheme for obtaining scientific results from a data set with the same ( @xmath40 , @xmath41 ) distribution . the dpt solutions and the long - term phase calibration data were applied to the remaining frequency channels to make image cubes for the lsr velocity range of @xmath42 to @xmath43 km s@xmath4 with a step of 0.2107 km s@xmath4 . image synthesis with 60 @xmath44as pixel for a region of @xmath45 pixels , roughly a @xmath46 square milliarcsecond ( mas ) region , was made by using imagr . to initially pick up maser emissions in the maps , we used the aips two - dimensional ( 2d ) gaussian component survey task , sad , in each velocity channel map in order to survey emissions . we visually inspected the surveyed gaussian components one by one . we stored the image pixels that were larger than a 5@xmath47 noise level for a visually confirmed maser emission . hereafter we define a maser `` spot '' as an emission in a single velocity channel , and a maser `` feature '' as a group of spots observed in at least two consecutive velocity channels at a coincident or very closely located positions . in the phase - referencing astrometric analysis , positions of the pz cas maser spots were obtained with respect to that of the j2339 + 6010 image . the positional error includes the effect of a peak position shift due to uncertainties in the vlbi correlator model ( atmospheric excess path lengths [ epls ] , antenna positions , eop , and so on ) for the sources . hereafter we refer to the relative position error of a pair of point sources using the phase referencing due to uncertainties in a vlbi correlator model as an astrometric error . we conducted monte carlo simulations of phase - referencing observations for the pair of pz cas and j2339 + 6010 to estimate the astrometric error . for generating simulated phase - referencing fringes , we used an improved version of a vlbi observation simulator , aris ( astronomical radio interferometer simulator ; * ? ? ? * ) , with new functions simulating station clocks @xcite and amplitude calibration errors . in the simulations , we assumed flux densities of pz cas and j2339 + 6010 of equivalently 10 jy and 0.2 jy for 15.6 khz and 224 mhz bandwidths , respectively , which were determined from our observations . the sources were assumed to be point sources . the observation schedule and observing system settings in the simulations , such as the recording bandwidth and a / d quantization level , were adopted from the actual vera observations . a tropospheric zenith epl error of 2 cm and other parameters were set as suggested for vera observations @xcite . for simplification of the simulations , reference source position errors were not considered . the simulation results from 200 trials are shown in figure [ fig:04 ] . they show that the 1@xmath47 errors in the relative position are 72 and 33 @xmath44as in right ascension and declination , respectively . we used these standard deviations as the astrometric error in the following analysis . at epoch i , the iriki station geographically located at the center of the vera could not attend the observation , so the astrometric accuracy may become worse with the remaining three stations . we conducted another simulation series for the three - station case , and the resultant astrometric accuracy is 101 and 39 @xmath44as for right ascension and declination , respectively . we therefore used this astrometric accuracy at epoch i in the following analysis . maser features of pz cas are spatially distributed over a region of 200 @xmath48 200 square mas . we identified 52 maser spots that are assembled into 16 maser features . those maser features were arranged in 11 maser groups , which are labeled _ a _ to _ k _ , as shown in figure [ fig:05 ] . the groups _ a _ to _ e _ , and _ g _ , _ h _ , and _ i _ have a radial velocity range between @xmath49 and @xmath50 km s@xmath4 . the groups _ f _ , _ j _ , and _ k _ , which have been detected at one epoch , have a radial velocity of @xmath51 , @xmath52 , and @xmath53 km s@xmath4 , respectively . the radial velocities of the identified maser spots are close to the stellar systemic velocity ( _ [email protected]@xmath10.68 km s@xmath4 ) . groups _ a _ and _ b _ located to the north and southeast of the spatial distribution , respectively , are bright and have complicated structures while the other groups ( _ c _ to _ k _ ) have a rather simple structure . for the following annual parallax analysis , we selected 16 maser spots among 52 using the following two criteria . ( 1 ) maser spots were detected with the image s / n greater than or equal to 7 . ( 2 ) the maser spot was detected at eight epochs or more . all the selected maser spots belong to either group _ a _ or _ b _ by chance . in the first step of the annual parallax analysis we performed a levenberg - marquardt least - squares model fitting of the maser motions in the same way as that described by @xcite using the peak positions of the selected maser spots . we adopted the astrometric error as mentioned in section [ sec:03 - 03 ] and a tentative value of a morphology uncertainty ( a positional uncertainty due to maser morphology variation ) of @xmath55as in the fitting , equivalent to 0.1 au by assuming the distance of 2 kpc to the source . figures [ fig:06 ] and [ fig:07 ] show the images and the fitting results of all the selected maser spots . the stellar annual parallax , @xmath56 , was determined by the combined fitting with the levenberg marquardt least - squares analysis for all the selected spots . the obtained stellar annual parallax was [email protected] mas at this stage . if we carefully look at figure [ fig:07 ] , the maser spots do not seem to be properly identified with the image peak positions because multiple maser spots are spatially blended . this is caused when the multiple maser spots , whose size is typically 1 au @xcite , are closely located comparing with the synthesized beam size of @xmath57 mas in this case , equivalent to 2.6 au at the distance of pz cas . to identify an individual maser spot properly , especially for blended structures , we performed aips jmfit for the 2d multiple gaussian component fitting of the emissions . here we refer to a maser spot that can be identified with the image peak position as an isolated maser spot , and to a maser spot that can not be identified with a single peak position because of the insufficient spatial resolution as a blended component . if we can identify 2d gaussian components using jmfit , instead of sad that automatically searches multiple spots , in the brightness distribution , we refer to such spots as identified spots . figure [ fig:08 ] provides an example of the annual parallax least - squares fitting for a maser spot using either the image peak position or the 2d gaussian peak . table [ tbl:04 ] lists our combined fitting results using the image peak only for the isolated spots , only for the blended components , and for all those involved in the two groups . we also tried to make another combined fitting using the gaussian peak only for the isolated spots , only for the blended but identified spots with jmfit , and all of them , as listed in the bottom row in table [ tbl:04 ] . the combined fitting results for the isolated spots show little difference between the cases of image peak and gaussian peak . however , the fitting result for the blended spots with jmfit peak positions is definitely changed from that with the image peak for the blended components and becomes closer to that obtained with the isolated spots . if a maser spot is recognized as an isolated spot , in other words , if a maser cloud is spatially resolved into multiple spots in the image , the image peak position can be treated as the maser position . because this can not be the case for the maser spots of pz cas , we adopted the gaussian peak positions in our least - squares model fitting . the model fitting results obtained from the individual spots are listed in table [ tbl:05 ] . from the combined fitting analysis by using all the selected maser spots , we obtained the stellar annual parallax of 0.356 @xmath1 0.011 mas with assumptions of the above astrometric error and morphology uncertainty at this stage . although h@xmath0o maser features around rsgs are good tracers for astrometry @xcite , the morphology of the maser spots observed with the high - resolution vlbi is no longer negligible for the precise maser astrometry . these monitoring observations of pz cas have been conducted so often for two years that we can investigate the contribution of the morphology uncertainty to the maser astrometry . figure [ fig:09 ] shows the position residuals of the 16 selected maser spots after removing the stellar annual parallax of 0.356 mas , proper motion , and initial position estimated for each of the spots . the standard deviation of the residuals is 93 and 110 @xmath44as for right ascension and declination , respectively . assuming that those errors have the statistical characteristics of a gaussian distribution , the morphology uncertainty is 59 ( = @xmath58 ) and 105 ( = @xmath59 ) @xmath44as for right ascension and declination , respectively , where 72 and 33 @xmath44as are the astrometric errors discussed in section 3.3 . it is unlikely that the morphology uncertainty has different values between right ascension and declination , so we adopted 105 @xmath44as as the morphology uncertainty to be on the safer side . based on the trigonometric parallax of 0.356 mas , 105 @xmath44as is 0.29 au at pz cas . we conducted the combined fitting with all the maser spots whose positions were determined with jmfit and obtained @xmath56 of @xmath60 mas at this stage . note that the above error of the annual parallax is a result of considering the random statistical errors from the image s / n and the randomly changing maser spot position , as well as a common positional shift to all the maser spots at a specific epoch due to the astrometric error . because the majority of the maser spots and features in the current parallax measurement are associated with the maser group _ a _ , the obtained stellar annual parallax could be affected by a time variation of the specific maser features at a certain level . assuming that all the 13 maser spots belonging to group _ a _ have an identified systematic error , the annual parallax error can be evaluated to be @xmath61 mas . however , if we assume that maser spots belonging to a certain maser feature have a common systematic error but that some of them have their own time variations , the annual parallax error should be estimated to be ranging somewhere between 0.011 and 0.040 mas . to evaluate the contributions of such a systematic error , we carried out monte carlo simulations introduced by @xcite for our astrometric analysis . first we prepared the imitated data set of the 16 maser spots with position changes added by considering the fixed stellar annual parallax and the maser proper motions . second we added an artificial positional offset due to the image s / n , the astrometric accuracy , and the morphology uncertainty to each of the maser spot positions . the positional offset caused by the image s / n was given to each of the maser spots at each epoch independently . the positional offset due to the astrometric error was common to all the maser spots at a specific epoch but randomly different between the epochs . in this paper , we assumed that the positional offset due to the morphology uncertainty was common to all the maser spots belonging to a specific maser feature at a specific epoch , but randomly different between the features and the epochs . we assumed the standard deviation of the maser morphology uncertainty was 105 @xmath44as , equivalent to the intrinsic size of 0.29 au . in a single simulation trial , we obtained the stellar annual parallax for the imitated data set . thirdly we carried out 200 trials in order to obtain the standard deviation of the stellar annual parallax solutions . the resultant standard deviation was 0.026 mas which can be treated as the annual parallax error in our astrometric analysis . the resultant estimate of the stellar annual parallax was 0.356 @xmath1 0.026 mas , corresponding to @xmath5 kpc . similar to the cases seen in other rsgs ( vy cma , @xcite ; s per , @xcite ; vx sgr , @xcite ; @xcite ; nml cyg , @xcite , @xcite ) , the maser features around pz cas are widely distributed over a @xmath62 square mas ( corresponding to @xmath63 square au at a distance of 2.81 kpc ) while the velocity of the features ( @xmath64 to @xmath65 km s@xmath4 ) is distributed more closely to the stellar systemic velocity than those found in other rsgs are . it is likely that the detected maser groups are located at the tangential parts of the shell - like circumstellar envelope , whose 3d velocity has a small line - of - sight velocity . we determined the proper motions of all detected maser spots were identified at least in two epochs . table [ tbl:06 ] lists maser spots whose proper motions and positions were calculated with the stellar annual parallax determined above . using the proper motion data listed in tables [ tbl:05 ] and [ tbl:06 ] , we then conducted the least - squares fitting analysis to obtain the position of pz cas and the stellar proper motion based on the spherically expanding flow model @xcite . the obtained stellar proper motion in right ascension and declination , @xmath663.7 @xmath1 0.2 mas yr@xmath4 and @xmath672.0 @xmath1 0.3 mas yr@xmath4 , respectively . the astrometric analysis result for pz cas is listed in table [ tbl:07 ] . the proper motions of the detected maser spots together with the stellar position indicated by the cross mark are shown in figure [ fig:10 ] . we obtained the trigonometric parallax to pz cas , @xmath68 kpc as described in section 4.1 . pz cas is thought to be a member of the cas ob5 association in the perseus spiral arm . it is of interest to compare the distance to pz cas determined in this paper to distances that have been accepted by other methods . @xcite estimated the trigonometric parallax to cas ob5 to be @xmath69 kpc from the median of trigonometric parallaxes of the member stars obtained with _ hipparcos_. since a typical statistical error of _ hipparcos _ annual parallaxes is @xmath57 mas , we can not fully rely on this statistical value for the object . another estimate of the distance to cas ob5 is a distance modulus of 12 , corresponding to the photometric parallax of 2.5 kpc @xcite . vlbi phase - referencing astrometry has revealed that its measured trigonometric parallaxes for rsgs are well consistent with photometric parallaxes of the related star clusters within @xmath70% ( s per , @xcite ; vy cma , @xcite ; nml cyg , @xcite ) . therefore , it is highly plausible that pz cas is involved in cas ob5 from the geometrical viewpoint . however , the difference between the distances derived from pz cas s annual parallax and the photometric parallax of cas ob5 is too large to be explained from the viewpoint of its extent in depth , compared with the distribution on the sky of the member stars . one of the plausible reasons for the inconsistency is the overestimation of @xmath71 to cas ob5 . if we attribute the inconsistency to the @xmath71 value , the difference is 0.16 mag , which is comparable to a typical statistical error of @xmath71 . the _ gaia _ mission is expected to yield measurements of accurate trigonometric distances for much more sampled stars than ever . the bolometric luminosity , @xmath72 , of pz cas was estimated to be @xmath73 , @xmath74 , and @xmath75 by @xcite , @xcite , and @xcite , respectively , with the distance modulus of cas ob5 of 12 @xcite . to derive the mass and age of pz cas , we first rescaled the luminosity values with our trigonometric parallax to @xmath76 , @xmath77 , and @xmath78 , respectively . @xcite estimated the effective temperature of pz cas to be 3600 k from the marcs stellar atmosphere models and the absolute spectro - photometry . using this effective temperature , we ploted pz cas in the h - r diagram reported by @xcite , as shown in figure [ fig:11 ] . in this figure the filled circles represent the re - estimated luminosities based on @xcite , @xcite , and @xcite , while the open circles represent the originally reported ones . the luminosity previously estimated by using the distance modulus yields the initial mass of around 20 @xmath7 while , using our trigonometric parallax , it is estimated to be @xmath7925 @xmath7 . according to evolutionary tracks shown in figures 1 and 2 of @xcite , the lifetime is estimated to be @xmath80 myr for an initial mass of 25 @xmath7 and @xmath81 myr for an initial mass of 20 @xmath7 . since the ages of four open clusters in cas ob5 have been reported to be @xmath82 myr for ngc 7790 @xcite , and 30.2 , 45.7 , and 13.2 myr for ngc 7788 , frolov 1 , and king 12 , respectively @xcite , it is expected that pz cas was born after the first generation in cas ob5 . the stellar proper motion and the radial velocity of pz cas enabled us to discuss its 3d motion in the milky way . in the analysis , we adopted the galactocentric distance to the sun of 8.5 kpc , and the flat galactic rotation curve with the rotation velocity of 220 km s@xmath4 . for the solar motion , we adopted the latest value obtained by @xcite , @xmath83 km s@xmath4 toward the direction of the galactic center ( @xmath84 component ) , @xmath85 km s@xmath4 toward the direction of the galactic rotation ( @xmath86 component ) , and @xmath87 km s@xmath4 toward the north pole ( @xmath88 component ) . since h@xmath0o maser features around rsgs are spherically distributed , it is hardly plausible that an uncertainty in the stellar proper motion was caused by the anisotropic distribution , as those in the case of outflows of star forming regions has to be considered ( e.g. , * ? ? ? therefore , we did not introduce an additional uncertainty of the pz cas stellar motion other than the statistical error described in section [ sec:04 - 02 ] in this error estimation . we followed the methods described by @xcite and @xcite to derive the peculiar motion and its error . the resultant peculiar motion of pz cas was ( @xmath11 , @xmath12 , @xmath89)@xmath90 ( @xmath91 , @xmath92 , @xmath93 ) km s@xmath4 . different from peculiar motions of galactic high - mass star - forming regions ( hmsfrs ) typically showing negative @xmath12 components ( @xmath94 to @xmath95 km s@xmath4 ; e.g. , @xcite ) , pz cas has a large peculiar motion not in the @xmath12 component but in @xmath11 . at first , in order to discuss whether pz cas s peculiar motion is due to the internal motion inside cas ob5 , we investigated proper motions of the member stars of cas ob5 using the _ hipparcos _ catalog @xcite . positions and velocities of cas ob5 member stars ( ob stars and rsgs ) cataloged in @xcite with compiled information from previous literatures @xcite are listed in table [ tbl:08 ] . figure [ fig:12 ] shows the positions and proper motions of the member stars listed in table [ tbl:08 ] as well as stars with annual parallax @xmath96 mas and proper motion @xmath97 mas yr@xmath4 in order to roughly exclude foreground stars . in figure [ fig:12 ] , the @xmath98 co contour map of a galactic plane survey @xcite taken from _ skyview _ ( internet s virtual telescope ) is superimposed . we obtained the unweighted average of those proper motions ( @xmath99 and @xmath100 mas yr@xmath4 ) which is well consistent with cas ob5 s peculiar motion obtained by @xcite ( @xmath101 and @xmath102 mas yr@xmath4 ) . figure [ fig:13 ] shows 3d views of the relative positions and peculiar motions of pz cas and the member stars of cas ob5 whose radial velocities can be found in previous literatures as listed in table [ tbl:08 ] . here we adopted pz cas s trigonometric distance for cas ob5 . the averaged peculiar motion of those member stars was ( @xmath103 , @xmath104 , @xmath105)@xmath90(@xmath106 , @xmath107 , @xmath108 ) km s@xmath4 whose @xmath40 component was prominent comparing with the other two . therefore , we concluded that the peculiar motion of pz cas with the large @xmath11 component reflects cas ob5 s peculiar motion . why does pz cas , or cas ob5 , have such a large peculiar motion in the @xmath40 component ? since the member stars including pz cas are located on the western side of a large shell - like structure molecular gas ( h@xmath109 super shell ) with the center position of a galactic coordinate of ( @xmath110 , @xmath111 ) , one may consider that supernovae in precedent giant molecular clouds triggered the formation of those stars and that they have an expanding motion @xcite . we investigated _ hipparcos _ proper motions of the member stars of cas ob5 , as well as those of cas ob4 and cas ob7 located on the eastern side of the shell - like structure , as shown in figure [ fig:12 ] . we also investigated 3d relative positions and peculiar motions of those of cas ob4 and cas ob7 , as listed in table [ tbl:09 ] . we adopted the photometric parallaxes for cas ob4 and cas ob7 reported by @xcite . figure [ fig:13 ] shows 3d views of the positions and proper motions of the memeber stars of cas ob5 as well as cas ob4 and cas ob7 . we could hardly find clear evidence of an expanding motion of the member stars , either in the proper motions or the 3d motions between those three ob associations . one of the member stars of cas ob7 , hd 4842 , has a much larger proper motion than those of the others , even if the _ hipparcos _ trigonometric parallax of @xmath112 mas is considered . therefore , this star could be a runaway star @xcite . vlbi astrometric observations of galactic maser sources on perseus spiral arm have revealed large peculiar motions of galactic hmsfrs . according to @xcite , 33 sources whose peculiar motions have been measured in the perseus and outer arms show that these peculiar motions are almost located in the fourth quadrant in the @xmath113 plane ( @xmath114 0 and @xmath115 0 ) . the peculiar motion of cas ob5 is located in the fourth quadrant while that of pz cas is located in the first quadrant . such a peculiar motion of the ob association may indicate the standard density wave theory @xcite . however , it is hard to explain the velocity dispersion of the peculiar motions of the hmsfrs from the density wave theory . such a large velocity dispersion in peculiar motions can be seen for the galactic ob associations within 3 kpc from the sun @xcite . the peculiar motion of cas ob5 hits upon an interesting idea of `` time - dependent potential spiral arms '' . @xcite reported on results of large @xmath116-body and smoothed particle hydrodynamics ( sph ) simulations that calculated the motions of multi - phase , self - gravity gas particles in self - induced stellar spiral arms in a galactic disk . in their simulations , spiral arms are developed by the self - gravity but can not keep their forms even in one galactic rotation evolution . other simulation results by @xcite suggest that a massive cloud falls into a spiral arm s potential from both sides , that is , there are flows of cold gas with various velocities . the flows converge into condensations of cold gas near the bottom of the potential well . the relative velocities of the gas to the stellar spirals are comparable to the random motion of the gas . another interesting point reported by @xcite is that supernovae feedback is not essential to change the velocity field of the gas near the spiral potential . each of the cold massive clouds and stars with ages younger than 30 myr may have a peculiar motion with a velocity dispersion of @xmath117 km s@xmath4 , which is not predicted by the density wave theory . although we have not had samples large enough to discuss the whole structure of the milky way , the self - consistent gravity spiral galaxy model can explain those peculiar motions of hmsfrs in the milky way . in other words , a live spiral structure model can fit to the recent sample of the hmsfrs spatial motions including cas ob5 . a deep understanding of the dynamics of the galactic spiral arms and the origin of the peculiar motions of hmsfrs will be possible in the near future when massive numerical simulations can be combined with much larger vlbi astrometric samples which only can be conducted toward the galactic plane obscured by dense gas . we have conducted phase - referencing vlbi monitoring observations for the h@xmath0o masers of the galactic rsg , pz cas . the distance to pz cas was determined to be @xmath118 kpc from the derived annual parallax in the maser spot motions . we estimated the initial mass of pz cas as @xmath6 @xmath7 using the re - scaled stellar luminosity with the annual parallax . together with the stellar radial velocity and obtained proper motion , we investigated the 3d motion of pz cas in the milky way , showing the non - circular motion with a rather large velocity toward the galactic center . we compared the proper motion and 3d peculiar motion of pz cas with those of the member stars of cas ob5 using the proper motions cataloged in _ hipparcos _ and their radial velocities . we found that the motion of pz cas in the milky way is close to that of the ob association . this peculiar motion is not aligned with those of the galactic hmsfrs . the observed peculiar motion supports a live spiral arm model recently revealed by large @xmath116-body / sph numerical simulations . the vera / mizusawa vlbi observatory is a branch of the national astronomical observatory of japan , national institutes of nature sciences . the authors acknowledge the use of nasa s skyview facility ( http://skyview.gsfc.nasa.gov ) located at nasa goddard space flight center . k. kusuno was financially supported by the center for the promotion of integrated science of the graduate university of advanced studies for this publication . abt , h. a , levy , s. g. , & gandet , t. l. 1972 , , 77 , 138 asaki , y. , deguchi , s. , imai , h. , et al . 2010 , , 721 , 267 asaki , y. , sudou , h. , kono , y. , et al . 2007 , , 59 , 397 baba , j. , asaki , y. , makino , j. , et al . 2009 , , 706 , 471 chen , x. , shen , z - q . , imai , h. , & kamohara , r. 2006 , , 640 , 982 choi , y. k. , hirota , t.,honma , m. , et al . 2008 , , 60 , 1007 dame , t. m. , hartmann , d. , & thaddeus , p. 2001 , , 547 , 792 evans , d. s. 1967 , in iau symp . 30 , determination of radial velocities and their applications , ed . a. h. batten & j. f. heard ( london : academic ) , 57 famaey , b. , jorissen , a. , luri , x. , et al . 2005 , , 430 , 165 fehrenbach , ch . , duflot , m. , genty , v. , & amieux , g. , 1996 , bicds , 48 , 11 fujii , m. s. , & zwart , s. p. 2011 , sci , 334 , 1380 garmany , c. d. , & stencel , r. e. 1992 , , 94 , 211 gupta , a. c. , subramaniam , a. , sagar , r. , & griffiths , w. k. 2000 , , 145 , 365 hachisuka , k. , brunthaler , a. , menten , k. m. , et al . 2006 , , 645 , 337 hg , e. , fabricius , c. , makarov , v. v. , et al . 2000 , , 355 , l27 honma , m. , fujii , t. , hirota , t. , et al . 2003 , , 55 , l57 honma , m. , hirota , t. , jike , t. , et al . 2010 , pnaoj , 13 , 57 honma , m. , kijima , m. , suda , h. , et al . 2008 , , 60 , 935 humphreys , r. w. 1970 , , 160 , 1149 humphreys , r. m. 1978 , , 38 , 309 imai , h. , omi , r. , kurayama , t. , et al . 2011 , , 63 , 1293 johnson , d. r. h. , & soderblom , d. r. 1987 , , 93 , 864 josselin , e. , blommaert , j. a. d. l. , groenewegen , m. a. t. , omont , a. , & li , f. l. 2000 , , 357 , 225 kamohara , r. , deguchi , s. , miyoshi , m. , & shen , z - q . 2005 , , 57 , 341 kerr , f. j. , & lynden - bell , d. 1986 , , 221 , 1023 kharchenko , n. v. , piskunov , a. e. , rser , s. , schilbach , e. , & scholz , r .- d . 2005 , , 438 , 1163 kraft , r. p. 1958 , , 128 , 161 levesque , e. m. , massey , p. , olesn , k. a. g. , et al . 2005 , , 628 , 973 lin , c. c. , & shu , f. h. 1964 , , 140 , 646 mauron , n. , & josselin , e. 2011 , , 526 , a156 melnik , a. m. , & dambis , a. k. 2009 , , 400 , 518 mermilliod , j. c. , mayor , m. , & udry , s. 2008 , , 485 , 303 meynet , g. , & maeder , a. 2003 , , 404 , 975 mnch , g. 1957 , , 125 , 42 nagayama , t. , takeda , k. , omodaka , t. , et al . 2008 , , 60 , 1069 petrov , l. , hirota , t. , honma , m. , et al . 2007 , , 133 , 2487 reid , m. j. , menten , k. m. , zheng , x. w. , et al . 2009 , , 700 , 137 reid , m. j. , & moran , j. m. 1981 , , 19 , 231 richards , a. m. s. , etoka , s. , gray , m. d. , et al . 2012 , , 546 , a16 rioja , m. , dodson , r. , asaki , y. , hartnett , j. , & tingay , s. 2012 , , 144 , 121 sakai , n. , honma , m. , nakanishi , h. , et al . 2012 , , 64 , 108 sanna , a. , reid , m. j. , moscadelli , l. , et al . 2009 , , 706 , 464 sato , m. , hirota , t. , honma , m. , et al . 2008 , , 60 , 975 schnrich , r. , binney , j. , & dehnen , w. 2010 , , 403 , 1829 valdes , f. , gupta , r. , rose , j. a. , singh , h. p. , & bell , d. j. 2004 , , 152 , 251 van leeuwen , f. 2007 , , 474 , 653 wada , k. , baba , j. , & saitoh , t. r. 2011 , , 735 , 1 wilson , r. e. 1953 , general catalogue of stellar radial velocities ( washington , dc : carnegie institute of washington ) zhang , b. , reid , m. j. , menten , k. m. , & zheng , x. w. 2012a , , 744 , 23 zhang , b. , reid , m. j. , menten , k. m. , zheng , x. w. & brunthaler , a. 2012b , , 544 , a42 in the _ hipparcos _ catalog . the blue and red circles represent ob stars and rsgs , respectively , which have been identified as cas ob5 members . the same goes for the cas ob4 and cas ob7 . the purple and orange circles represent ob stars and rsgs , respectively . the filled and open circles denote the member stars found and not found in the _ hipparcos _ catalog , respectively . the black points denote the stars whose proper motions are reported in the _ hipparcos _ catalog but not identified with cas ob5 , cas ob4 , and cas ob7 members . the arrows represent proper motions listed in the _ hipparcos _ catalog . the position of pz cas is located at the crossing point of the horizontal and vertical dotted lines . the green arrow represents the proper motion reported in this paper . the background contour is the co map reported by @xcite . bottom left : an enlarged picture of the area surrounded by the dashed square in the top panel . bottom right : same as the bottom left panel , but the average proper motion of the cas ob5 members was subtracted from each proper motion . , width=377 ] . the green point represents pz cas s peculiar motion obtained in this paper . the purple and orange points are ob stars and rsgs of the other two ob associations , respectively , listed in table [ tbl:09 ] . conicals in the enlarged right plots represent the peculiar motions of the member stars after removing the flat galactic rotation curve of 220 km s@xmath4 . a thick gray dashed line represents the position of the perseus spiral arm @xcite . , title="fig:",width=453 ] + . the green point represents pz cas s peculiar motion obtained in this paper . the purple and orange points are ob stars and rsgs of the other two ob associations , respectively , listed in table [ tbl:09 ] . conicals in the enlarged right plots represent the peculiar motions of the member stars after removing the flat galactic rotation curve of 220 km s@xmath4 . a thick gray dashed line represents the position of the perseus spiral arm @xcite . , title="fig:",width=453 ] clcp70 mm epoch & & time range + & & ( utc ) + a & 2006 apr 20 & 19:41 - 05:20 + b & 2006 jul 14 & 14:21 - 23:10 + c & 2006 aug 13 & 12:21 - 21:35 + d & 2006 oct 19 & 08:01 - 17:20 + e & 2006 dec 13 & 04:21 - 13:40 + f & 2007 mar 22 & 00:21 - 08:17 + g & 2007 may 2 & 18:51 - 04:40 + h & 2007 jul 25 & 13:20 - 23:04 + i & 2007 sep 19 & 10:20 - 20:10 + j & 2007 nov 7 & 07:20 - 17:10 + k & 2007 dec 22 & 04:25 - 14:15 + l & 2008 may 12 & 19:00 - 04:50 + c | l | l beam & & + source & & + base band number & & + & & 22.106 - 22.122 + & & 22.122 - 22.138 + & & 22.138 - 22.154 + & & 22.154 - 22.170 + & & 22.170 - 22.186 + frequency range & & 22.186 - 22.202 + ( ghz ) & 22.234 - 22.242 & 22.202 - 22.218 + & & 22.234 - 22.250 + & & 22.250 - 22.266 + & & 22.266 - 22.282 + & & 22.282 - 22.298 + & & 22.298 - 22.314 + & & 22.314 - 22.330 + & & 22.330 - 22.346 + frequency spacing & 15.625 khz ( 512 channels ) & 250.0 khz ( 64 channels ) + cccc spot identification method & blended component & isolated component & all + image peak & 0.402 @xmath1 0.021 & 0.371 @xmath1 0.013 & 0.380 @xmath1 0.011 + 2-d gaussian peak & 0.352 @xmath1 0.017 & 0.358 @xmath1 0.013 & 0.356 @xmath1 0.011 + ccccccc @xmath123 & @xmath124 & @xmath125 & @xmath126 & @xmath127 & @xmath128 & @xmath129 + @xmath130 & @xmath131 & @xmath132 & @xmath133 & @xmath134 & @xmath135 & @xmath136 + @xmath137 & @xmath138 & @xmath139 & @xmath140 & @xmath141 & @xmath142 & @xmath143 + @xmath144 & @xmath145 & @xmath146 & @xmath140 & @xmath147 & @xmath148 & @xmath149 + @xmath150 & @xmath151 & @xmath152 & @xmath153 & @xmath154 & @xmath142 & @xmath155 + @xmath156 & @xmath157 & @xmath158 & @xmath159 & @xmath160 & @xmath161 & @xmath162 + @xmath163 & @xmath151 & @xmath164 & @xmath165 & @xmath166 & @xmath167 & @xmath162 + @xmath168 & @xmath169 & @xmath164 & @xmath170 & @xmath171 & @xmath172 & @xmath173 + @xmath174 & @xmath175 & @xmath176 & @xmath177 & @xmath178 & @xmath179 & @xmath180 + @xmath181 & @xmath182 & @xmath183 & @xmath184 & @xmath185 & @xmath186 & @xmath187 + @xmath188 & @xmath189 & @xmath190 & @xmath170 & @xmath191 & @xmath172 & @xmath192 + @xmath193 & @xmath194 & @xmath195 & @xmath177 & @xmath196 & @xmath197 & @xmath198 + @xmath199 & @xmath200 & @xmath201 & @xmath202 & @xmath203 & @xmath204 & @xmath205 + @xmath206 & @xmath207 & @xmath208 & @xmath126 & @xmath209 & @xmath210 & @xmath211 + @xmath212 & @xmath213 & @xmath214 & @xmath133 & @xmath215 & @xmath216 & @xmath217 + @xmath218 & @xmath219 & @xmath220 & @xmath140 & @xmath221 & @xmath222 & @xmath223 ccccccc @xmath224 & @xmath225 & @xmath226 & @xmath50 & @xmath227 & @xmath228 + @xmath229 & @xmath225 & @xmath230 & @xmath231 & @xmath232 & @xmath233 + @xmath234 & @xmath235 & @xmath236 & @xmath126 & @xmath237 & @xmath238 + @xmath239 & @xmath240 & @xmath241 & @xmath133 & @xmath242 & @xmath243 + @xmath244 & @xmath245 & @xmath246 & @xmath140 & @xmath247 & @xmath248 + @xmath249 & @xmath250 & @xmath251 & @xmath50 & @xmath252 & @xmath253 + @xmath254 & @xmath255 & @xmath256 & @xmath231 & @xmath257 & @xmath258 + @xmath259 & @xmath260 & @xmath261 & @xmath153 & @xmath262 & @xmath263 + @xmath264 & @xmath265 & @xmath266 & @xmath133 & @xmath267 & @xmath268 + @xmath269 & @xmath270 & @xmath271 & @xmath177 & @xmath272 & @xmath273 + @xmath274 & @xmath275 & @xmath276 & @xmath170 & @xmath277 & @xmath278 + @xmath279 & @xmath280 & @xmath281 & @xmath184 & @xmath282 & @xmath198 + @xmath283 & @xmath284 & @xmath285 & @xmath184 & @xmath286 & @xmath287 + @xmath288 & @xmath289 & @xmath290 & @xmath291 & @xmath292 & @xmath293 + @xmath294 & @xmath295 & @xmath296 & @xmath297 & @xmath298 & @xmath299 + @xmath300 & @xmath301 & @xmath302 & @xmath50 & @xmath303 & @xmath304 + @xmath305 & @xmath306 & @xmath307 & @xmath231 & @xmath308 & @xmath309 + @xmath310 & @xmath311 & @xmath312 & @xmath153 & @xmath313 & @xmath314 + @xmath315 & @xmath316 & @xmath317 & @xmath133 & @xmath318 & @xmath319 + @xmath320 & @xmath321 & @xmath322 & @xmath140 & @xmath323 & @xmath324 + @xmath325 & @xmath326 & @xmath327 & @xmath153 & @xmath328 & @xmath329 + @xmath330 & @xmath331 & @xmath332 & @xmath165 & @xmath333 & @xmath334 + @xmath335 & @xmath336 & @xmath337 & @xmath170 & @xmath338 & @xmath339 + @xmath340 & @xmath341 & @xmath342 & @xmath177 & @xmath343 & @xmath344 + @xmath345 & @xmath346 & @xmath347 & @xmath348 & @xmath349 & @xmath350 + @xmath351 & @xmath352 & @xmath353 & @xmath354 & @xmath355 & @xmath356 + @xmath357 & @xmath358 & @xmath359 & @xmath133 & @xmath360 & @xmath361 + @xmath362 & @xmath363 & @xmath364 & @xmath140 & @xmath365 & @xmath366 + @xmath367 & @xmath368 & @xmath369 & @xmath153 & @xmath370 & @xmath371 + @xmath372 & @xmath373 & @xmath374 & @xmath159 & @xmath375 & @xmath376 + lcc annual parallax & + position ( j2000 ) & right ascension & declination + ( 2006 apr 20 ) & @xmath377 @xmath1 [email protected] & @xmath379 @xmath1 [email protected] + stellar proper motion & @xmath381 mas yr@xmath4 & @xmath382 mas yr@xmath4 + lcccccl object & @xmath383 & @xmath384 & @xmath385 & @xmath386 & @xmath41 & spectl type + & ( deg ) & ( deg ) & ( mas yr@xmath4 ) & ( mas yr@xmath4 ) & ( km s@xmath4 ) & + hd 223385 & @xmath387 & @xmath388 & @xmath389 & @xmath390 & @xmath391 & a3 i a+ + hd 223767 & @xmath392 & @xmath393 & @xmath394 & @xmath395 & @xmath396 & a4 i ab + hd 223960 & @xmath397 & @xmath398 & @xmath399 & @xmath400 & @xmath401 & a0 i a+ + bd + 61 2550 & @xmath402 & @xmath403 & @xmath404 & @xmath405 & @xmath406 & o9.5 ii + hd 224424 & @xmath407 & @xmath408 & @xmath409 & @xmath410 & @xmath411 & b1 i ab + bd + 62 2313 & @xmath412 & @xmath413 & @xmath400 & @xmath414 & @xmath415 & b3 i b + hd 224055 & @xmath416 & @xmath417 & @xmath404 & @xmath418 & @xmath419 & b1 i a + lsi + 61 106 & @xmath420 & @xmath421 & @xmath422 & @xmath423 & @xmath424 & b1 v + hd 225146 & @xmath425 & @xmath398 & @xmath426 & @xmath427 & @xmath428 & o9.7 i b + hd 225094 & @xmath429 & @xmath430 & @xmath431 & @xmath432 & @xmath433 & b3 i a + hd 108 & @xmath434 & @xmath435 & @xmath436 & @xmath437 & @xmath438 & o5f + @xmath439 cas & @xmath440 & @xmath441 & @xmath442 & @xmath443 & @xmath202 & f8 i ap + kn cas & @xmath444 & @xmath445 & @xmath446 & @xmath447 & @xmath448 & m1 i bep+b + tz cas & @xmath449 & @xmath450 & @xmath451 & @xmath452 & @xmath453 & m2 i ab + pz cas & @xmath454 & @xmath455 & @xmath456 & @xmath457 & @xmath458 & m4 i a + lcccccl object & @xmath383 & @xmath384 & @xmath385 & @xmath386 & @xmath41 & spectl type + & ( deg ) & ( deg ) & ( mas yr@xmath4 ) & ( mas yr@xmath4 ) & ( km s@xmath4 ) & + cas ob4 & @xmath459 & @xmath459 & @xmath424 & + hd 1544 & @xmath460 & @xmath461 & @xmath462 & @xmath463 & @xmath464 & b0.5 iii + hd 2451 & @xmath465 & @xmath466 & @xmath467 & @xmath468 & @xmath65 & b0.5 iv + hiltner 8 & @xmath469 & @xmath470 & @xmath471 & @xmath472 & @xmath473 & o9 iv + dl cas & @xmath474 & @xmath475 & @xmath476 & @xmath462 & @xmath477 & f7.5 i b + hd 236433 & @xmath478 & @xmath479 & @xmath480 & @xmath481 & @xmath482 & f5 i b + cas ob7 & @xmath459 & @xmath459 & @xmath424 & + hd 3940 & @xmath483 & @xmath484 & @xmath479 & @xmath485 & @xmath486 & a1 i a + hd 4694 & @xmath487 & @xmath488 & @xmath489 & @xmath403 & @xmath490 & b3 i a + hd 4841 & @xmath491 & @xmath492 & @xmath493 & @xmath437 & @xmath494 & b5 i a + hd 5551 & @xmath495 & @xmath496 & @xmath497 & @xmath498 & @xmath499 & b1.5 i a + hd 4842 & @xmath491 & @xmath500 & @xmath501 & @xmath502 & @xmath503 & m + hd 6474 & @xmath504 & @xmath505 & @xmath506 & @xmath507 & @xmath348 & g4 i a +

we show that for a general four derivative theory of gravity , only the holographic entanglement entropy functionals obey the second law at linearized order in perturbations . we also derive bounds on the higher curvature couplings in several examples , demanding the validity of the second law for higher order perturbations . for the five dimensional gauss - bonnet theory in the context of ads / cft , the bound arising from black branes coincides with there being no sound channel instability close to the horizon . repeating the analysis for topological black holes , the bound coincides with the tensor channel causality constraint ( which is responsible for the viscosity bound ) . furthermore , we show how to recover the holographic c - theorems in higher curvature theories from similar considerations based on the raychaudhuri equation .

1508.01658

in the context of ads / cft , entanglement entropy of the boundary quantum field theory can be calculated using the ryu - takayanagi @xcite prescription and its generalizations to higher curvature gravity theories . the derivation of this holographic entanglement entropy for two derivative einstein gravity was proposed by lewkowycz and maldacena in @xcite . this derivation allows one to obtain a surface equation ( for static situations ) for the entangling surface which minimizes an entropy functional . there were attempts to extend this calculation to find the entangling surface equation in higher curvature theories in @xcite and the corresponding entropy functionals were argued to be different from the wald entropy @xcite . in particular , for lovelock theories the entropy functional is the so - called jacobson - myers ( jm ) @xcite one while for general four derivative theories , it coincides with the fursaev - patrushev - solodukhin ( fps ) entropy functional @xcite . however , there still exist problems in finding the entangling surface using the lewkowycz - maldacena method and there are potential ambiguities related to higher order extrinsic curvature terms in the entropy functionals . moreover , it leads one to wonder if such entropy functionals could arise independent of ads / cft from different and perhaps more fundamental considerations . wald and collaborators @xcite had established the first law for black hole mechanics for any diffeomorphism invariant theory of gravity and proposed that the entropy of a black hole is a noether charge associated with the killing isometry generating the horizon . the wald entropy suffers from various ambiguities @xcite and therefore does not provide a unique answer for the horizon entropy . although , none of these ambiguities contribute to black hole entropy of a stationary killing horizon with regular bifurcation surface , the study of the second law of black hole mechanics beyond general relativity ( gr ) shows that these ambiguities need to be carefully included in the expression of black hole entropy to obtain an increase theorem similar to hawking area theorem in gr . in fact , for black holes in lovelock gravity , it is the jm functional which leads to an increase theorem for linearized perturbations @xcite . recently , it has been pointed out that the 2nd law for spherically symmetric black holes is satisfied at linear order in perturbations in general four derivative theories of gravity by the holographic entanglement entropy ( hee ) functionals @xcite . this has been also generalized beyond spherical symmetry and to generic perturbations up to linear order @xcite and a detailed construction of the entropy functionals has been proposed for any metric theory of gravity such that the linearized second law holds true . for theories with higher curvature terms , the construction produces the hee functionals . this is a remarkable result since it allows for a way independent of ads / cft of deriving entropy functionals these same entropy functionals therefore find applications in diverse settings . however , what still needs to be addressed is if only the hee functionals alone do this job . in this paper , we first show that at linear order in perturbations only the hee functionals can satisfy the second law . we start with an expression for horizon entropy with all possible ambiguous terms relevant for linear order in perturbation . we fix these terms by demanding the validity of the linearized second law and the final entropy coincides with hee functionals . to be clear , since our analysis is going to be perturbative , we will not be able to rule out higher order extrinsic curvature terms in the functionals ( e.g. @xmath0 in the fps functional)see section 5 . we will tackle the question keeping in mind two different motivations . first , we are interested in what happens in the case of asymptotically flat black holes . we will consider two different types of perturbations the first kind will be due to radially symmetric , slowly falling matter and the second kind being a shear perturbation . we will find that for the gauss - bonnet ( gb ) theory , if we go to second order in perturbations , the second law is automatically satisfied for regular horizons if the gb coupling is positive . if the gb coupling is negative then a certain lower bound on the horizon radius ensures that the second law is satisfied ; since this suggests that such black holes can not be formed from collapsing matter , one can take this as an indication that the negative coupling is disfavoured . if we consider a ricci - square theory then we find that similar extra conditions on the mass of the black hole may be needed for the second law to hold . second , we are interested in what happens in the context of ads / cft . we will find that for the gb theory , the second order analysis leads to a bound on the gauss - bonnet coupling . for black branes , this bound coincides with the absence of sound channel instabilities at the black hole horizon . a similar analysis for the ricci- square theory bounds the corresponding coupling constant . for gb topological black holes @xcite , for the zero mass case we find that the second law bound coincides with the tensor channel causality constraint @xcite . this bound is what leads to the lower bound on the ratio of shear viscosity to entropy density in the dual plasma and agrees with the lower bound on the @xmath1 ratio in 4d cfts @xcite with @xmath2 supersymmetry . we also investigate quadratic theories with @xmath3 and @xmath4 terms . in these theories , the causality constraints following from @xcite are trivial whereas the second law bound in certain examples we study is nontrivial . using the hee entropy functionals and the raychaudhuri equation , we then turn to the question of holographic c - theorems . we will argue that the holographic c - functions found in @xcite arise naturally from such a consideration provided the matter sector satisfies the null energy condition . in @xcite , a derivation for the c - functions in einstein gravity , found in @xcite , was given using the raychaudhuri equation . in @xcite an attempt was made to extend this to higher curvature theories using the iyer - wald prescription . unfortunately the resulting functions do not give the correct central charge ( namely the a - type anomaly coefficient in even dimensions ) . in @xcite it was observed that in all sensible holographic models , which can be constructed by demanding the absence of higher derivative terms in the radial direction ( to ameliorate the problem of ghosts ) allow for a simple c - function which at the fixed points coincide with the a - type anomaly coefficient in even dimensions . this c - function was monotonic under rg flow provided the matter sector satisfied the null energy condition . we will find that using the hee entropy functionals and considering the raychaudhuri equation naturally leads to the same c - functions as in @xcite . the organization of the paper as follows : in the next section we briefly describe the ambiguities in wald s noether charge construction . in the section [ second - set ] set up for proving second law has been described . next , we determine the coefficients of the ambiguous terms in wald s construction using linearized second law . in section [ beyond ] , we go beyond linear order in perturbation to study the gb theory and determine bound on the coupling parameter using ads black hole solution with different horizon topologies as background . we also consider various other examples like critical gravity theories and put bounds on the couplings in those theories . in section 6 , we derive holographic c - functions using the hee functionals and the raychaudhuri equation . we end this paper by discussing several follow - up questions . we use @xmath5 signature and set @xmath6 throughout the paper . let us start by describing the geometry of the horizon of a stationary black hole in @xmath7 dimensions . the event horizon is a null hyper - surface @xmath8 parameterized by a non - affine parameter @xmath9 . the vector field @xmath10 is tangent to the horizon and obeys non - affine geodesic equation @xmath11 with @xmath12 is the surface gravity of the horizon ( @xmath13 are bulk indices ) . all @xmath14 constant slices are co - dimension @xmath15 space - like surfaces and foliate the horizon . we construct another auxiliary null normal to the @xmath14 constant slices @xmath16 , and the inner product between @xmath17 and @xmath16 satisfies @xmath18 . the induced metric on any @xmath19 constant slice of the horizon is now constructed as @xmath20 . the horizon binormals are then given by @xmath21 . the change in the induced metric along these two null directions can be expressed as a sum of trace part and a trace - free symmetric part ( assuming null congruences are hypersurface orthogonal therefore have zero twist ) . the trace part measures the rate of change of the area of the horizon cross section along the null generators which is known to be the expansion and the other part measures the shear of the null geodesic congruences . we denote @xmath22 and @xmath23 to be the two expansion parameters in these two null directions and similarly we have two shears @xmath24 and @xmath25 . note that @xmath22 and @xmath24 typically vanish on a stationary horizon but @xmath26 and @xmath27 in general do not vanish . as mentioned in the introduction , the entropy of a stationary black hole in any general covariant theory is given by the noether charge associated with the boost symmetry generating killing vectors at the horizon @xcite . the wald entropy functional for any general covariant lagrangian @xmath28 in @xmath7 dimension takes the form : s_w = - 2 _ _ ab _ cd da,[wald_exp ] where @xmath29 denotes any horizon slice and @xmath30 is the area element . @xmath31 is the riemann tensor of bulk geometry and @xmath32 is the area of the @xmath33 dimensional cross - sections of horizon . this formula gives a unique expression for entropy so long one is confined to any stationary slice or to the bifurcation surface of the horizon . however , as pointed out in @xcite , the entropy expression constructed via noether charge approach has several ambiguities if one tries to apply it for nonstationary slices of the horizon . in fact it turns out , the wald entropy formula is just one of several possible candidates for the entropy . in @xcite , jacobson , kang and myers ( jkm ) identified three different types of ambiguities which may alter the wald construction . among these , the only relevant one for our purpose will be the one which gives us the freedom to add to the wald entropy a term of the form with arbitrary coefficients : s_a^(jkm)=-2_x.y da [ ambi ] where the integrand is invariant under a lorentz boost in a plane orthogonal to an arbitrary horizon slice @xmath29 but the terms @xmath34 and @xmath35 are not separately boost invariant @xcite . these ambiguities vanish for killing horizons but do not necessarily vanish on a nonstationary slice of the horizon . as a result the entropy functional for a nonstationary horizon may be expressed as and one spacelike @xmath36 normal such that , @xmath37 and @xmath38 and the extrinsic curvature @xmath39 for that codimension-2 surface with a induced metric @xmath40 defined as , @xmath41 trace of this is defined as @xmath42 . also we have @xmath43 and @xmath44 , where @xmath7 is the dimension of the bulk spacetime . ] = - 2 _ da . [ genn1]0.5 cm here we have used the notation @xmath45 . as can be seen from ( [ genn1 ] ) that the ambiguous terms in the entropy formula involve equal number of @xmath46 and @xmath47 subscripts which follows from the fact that these are the only boost invariant combinations that can appear in the entropy functional . also on a stationary slice the ambiguity terms vanish and then the expression coincides with the wald formula ( [ wald_exp ] ) . the first law of black hole mechanics @xcite does nt fix the coefficients @xmath48 and @xmath49 uniquely . so , to fix the coefficients of these terms one is thus forced to examine whether @xmath50 obeys a local increase law . recently in @xcite , the second law has been shown to hold for any arbitrary higher curvature theories if one allows only linear perturbations to a stationary black hole . this analysis also shows , the entropy functional proposed in the context of holographic entanglement entropy ( hee ) @xcite remarkably matches with the @xmath50 . however , it is not apparent from this analysis how just one condition ( entropy increases along @xmath9 ) fixes two coefficients in the entropy functional ! in this paper , we will show explicitly how this fixing happens for curvature squared gravity theories . note that the unknown coefficients are sitting in front of products of shear and expansion terms . since shear and expansion belong to different irreducible parts of a tensor , they will remain separated even when perturbation is turned on at the linearized level . consequently we will obtain two independent equations when we demand that entropy increase law holds for every slice of the horizon . this fixes all the coefficients of the entropy functional uniquely for generic quadratic curvature gravity . it is important to mention that we have only considered ambiguities which are quadratic order in expansion and shear . we may also have higher powers of such products added to the entropy functional . however , we can not fix such terms using linearized second law but in curvature squared theories these terms do not enter at linear order @xcite . our analysis matches with the result obtained in @xcite where the coefficients were fixed for ricci@xmath51 theory using a vaidya like solution in the linearized increase law . let us turn to the apparatus needed to verify the second law of black hole mechanics . the equation of motion for a generic higher curvature metric theory of gravity will be of the form , g_ab+h_ab=8t_ab[eqm ] , where @xmath52 is the einstein tensor coming from the einstein - hilbert part of the action and @xmath53 is the part coming from higher curvature terms in theories with a cosmological constant there will also be an additional term proportional to the metric which we can absorb into @xmath52 . @xmath54 is the energy momentum tensor which we will assume to obey the null energy condition ( nec ) : @xmath55 for some null vector @xmath17 . we will use the raychaudhuri equation for null geodesic congruence which describes the evolution of the expansion along the horizon generating parameter @xmath9 . in nonaffine parametrization this looks like = _ k --_k^2-r_kk.[ray1 ] our notation is @xmath56 and @xmath57 . now , we define an entropy density for any generic higher curvature theory as : = _ ( 1 + ) d^d-2x , [ entropyg ] where @xmath58 contains contribution from the higher curvature terms including ambiguities . for general relativity , @xmath59 . in the stationary limit @xmath60 will coincide with the wald expression ( [ wald_exp ] ) . now from this expression the change in entropy per unit area gives us the following expression of generalized expansion @xmath61 , = + _ k ( 1 + ) . [ theta ] the evolution of @xmath61 is governed by the following equation , - = & -&8t_kk-(1+)-_k^2 ( 1+)+_k + & + & h_kk+-r_kk-[ray2 ] , where we have used the equation of motion ( [ eqm ] ) and inserted eq . ( [ ray1 ] ) . we can now write eq , ( [ ray2 ] ) in a convenient form [ ray3 ] -=-8t_kk+e_kk , where [ eqn2 ] e_kk= h_kk+_k -r_kk+k^ak^b_a_b - ( + _ k^2)(1 + ) . we have used @xmath62 and @xmath63 next , consider a situation when a stationary black hole is perturbed by some matter flux obeying nec . the perturbation can be parametrized by some dimensionless parameter @xmath64 . note that @xmath65 is linear ( @xmath66 ) in perturbation and so as @xmath67 , @xmath68 and @xmath69 . now , to establish a linearized second law we ignore higher order terms in eq . ( [ ray3 ] ) . then , it is easy to see then eq . ( [ eqn2 ] ) reduces to , e_kk_k _ k - r_kk + h_kk[ekkl ] . we have already mentioned that @xmath65 is of order @xmath64 , so if the rest of the terms in ( [ ray3 ] ) are also collectively of higher order , i.e.,@xmath70 , then we obtain - = - 8 t_kk . the above equation implies @xmath71 , on every slice of the horizon . we assume that in the asymptotic future , the horizon again settles down to a stationary state , we must have @xmath72 in the future . this will imply that @xmath61 must be positive on every slice prior to the future and as a result the entropy given by ( [ entropyg ] ) obeys a local increase law . therefore , to establish the linearized second law , we only need to show that the linear order terms in @xmath73 exactly cancel each other . interestingly , this alone will be enough to obtain the values of both the coefficients @xmath48 and @xmath49 introduced in the entropy functional ( [ gen1 ] ) . in the next section , we will demonstrate this explicitly for the curvature squared gravity theories . we start with the most general second order higher curvature theory of gravity in five dimensions . the action of such a theory can be expressed as , s = d^5x ( r- 2 + r^2 + r_abr^ab + l_gb ) [ flag ] where @xmath74 is the gauss- bonnet ( gb ) combination and @xmath75 is the cosmological constant . it is reasonable to assume that such a theory admits a stationary black hole solution as in the case of gr . we also expect to have a non stationary black hole solution with in - falling matter by perturbing this solution . such a spacetime will be the counterpart of the vaidya solution in general relativity for spherically symmetric case and can be expressed as , [ vd1 ] ds^2=-f(r , v ) dv^2 + 2 dv dr + r^2d_3 ^ 2 @xmath76 can be any three dimensional space with positive , negative or zero curvature . we want to use this solution to investigate the issue of second law of black hole mechanics . note that , the location of the event horizon @xmath77 for this solution can be obtained by by solving the following equation , [ hor ] r= = . with appropriate boundary condition . note that , @xmath78 and @xmath79 denote respectively derivative with respect to @xmath80 and @xmath81 . the null generator of the event horizon is given by @xmath82 and the corresponding auxiliary null vector @xmath83 . the event horizon has nonzero expansion due to the perturbation caused by in falling matter . next , we will write the entropy associated with the horizon as , = - 2 _ da . [ gen1]we consider several choices of the coefficients @xmath84 , @xmath85 and @xmath86 etc and study the evolution of this horizon entropy . the aim is to fix the unknown coefficients @xmath87 and @xmath88 for different gravity theories by demanding the validity of the linearized second law . we will first study the local increase law for gauss - bonnet ( gb ) case . this corresponds to the choice @xmath89 in ( [ flag ] ) . the action is : s = d^5x . where we have also introduced a matter sector obeying the nec . @xmath90 is a coupling constant of dimension @xmath91 . the equation of motion for this theory is given by , [ eom ] g_ab+g_ab+ h_ab=8 g t_ab . with , @xmath92 -\frac{1}{2}g_{ab } { \cal l}_{gb}.\end{aligned}\ ] ] next we start with the expression of the entropy functional ( [ entropyg ] ) and evaluate the entropy density for the egb theory . this is given by sum of wald entropy density plus the ambiguity terms , & = & _ ( 1 + ) d^3x , + & = & _ ( 1 + 2(r + 4 r_kl - 2 r_klkl - p _ k_l - q _ k _ l ) ) d^3x , [ eee ] where @xmath93 is the contribution from the wald construction . also @xmath94 and @xmath95 the egb theory belongs to the general lovelock class of action functions which give rise to quasi linear equation of motions . it has already been shown in @xcite that black holes in all lovelock theories obey a linearized second law if one uses the jm entropy functional . for egb gravity , the jm entropy functional is the intrinsic ricci scalar @xmath96 of the horizon slice and using the null gauss - codazzi equation we can cast @xmath97 as , [ jmgb ] = r + 4r_kl - 2 r_klkl-43_k_l+2_k_l . we will now explicitly show below that in egb gravity the entropy functional which obeys the second law is indeed the jm entropy by fixing the coefficients @xmath48 and @xmath49 in ( [ eee ] ) . first , we proceed to evaluate the _ rhs _ of eq . ( [ ray3 ] ) to study the second law . note that , if we use the metric ( [ vd ] ) with the choice of @xmath98 to be a flat metric then the shear term in ( [ ray3 ] ) will vanish identically and @xmath49 will remain undetermined . this happens because isometries of the metric on a horizon slice essentially coincides with that of a sphere . so we will break the symmetry by adding a cross term and the metric on the horizon slice takes the form , ds^2=-f(r , v ) dv^2 + 2 dv dr + r^2 ( dx^2 + dy^2 + dz^2 + _ 1 h(r , v ) dx dy ) . we will assume that this shear mode @xmath99 will be balanced by some matter stress tensor still obeying the nec . also we do not require to find the explicit form of @xmath99 for our analysis . we will calculate @xmath73 order by order in @xmath100 and extract the coefficients of the linear order terms in @xmath64 from the evolution equation . setting those terms to zero will satisfy the linearized second law and in the process @xmath48 and @xmath49 will be determined . below we quote expressions for @xmath101 and @xmath102 for this solution . [ exp1 ] _ k=+(_1 ^ 2 ) , [ shear1 ] _ l=+(_1 ^ 2 ) and _ k_l=-+(_1 ^ 4 ) . now evaluating @xmath73 and extracting the zeroth order terms in @xmath103 ( and linear order in @xmath64 ) , we get the following equation , ( 4 - 3 p)=0 . from @xmath104 terms ( which is when @xmath99 makes its first appearance ) we get another equation , 2(-6 + 6p+q ) h(r , v)^2 + 3(2 - 3p - q)h(r , v)r(v)h(r , v)+(2+q)r(v)^2h(r , v)^2=0 . it is evident that , both of these equations are satisfied if @xmath105 this shows that linearized second law fixes all the quadratic ambiguities in entropy functional ( [ gen1 ] ) uniquely and it is the jm entropy which obeys the linearized second law for gb black holes ! although we have shown this calculation using the 5-dimensional lagrangian for concreteness , the same result holds in any dimension @xmath106 . let us repeat the above analysis for other curvature squared theories . first we take @xmath107 theory and we set @xmath108 and @xmath109 in ( [ flag ] ) . we will start with the following entropy functional , [ e1 ] = d^3 x ( 1 - 2(r_kl - p_k_l - q_k_l ) ) . when @xmath48 and @xmath49 are zero then this reduces to the corresponding wald entropy for the stationary case . again proceeding as before we will use the linearized second law to fix these coefficients . like the gauss - bonnet case we get two independent equations and solving them we get , @xmath110 as argued in @xcite , the shear part has not contributed to the entropy functional at linear order and the other coefficient of the entropy functional exactly matches with the one which has been shown to obey the linearized second law in @xcite . we now consider the case @xmath111 and @xmath112 in ( [ flag ] ) and the entropy functional now reads , [ e2 ] = d^3 x ( 1 + 2(r - p_k_l - q_k._l ) . it has already been argued in @xcite that @xmath113 theory obeys a linearized second law and the wald entropy functional itself does the job . it is clear , in this case setting @xmath48 and @xmath49 equal to zero will correspond to the wald entropy . if we again repeat the linearized second law analysis we find , @xmath114 so this is exactly what is expected from the earlier analysis in @xcite . therefore , considering all the individual terms in ( [ flag ] ) we have demonstrated that for any general curvature squared theory we can fix the entropy functional completely using only the linearized second law . in this section , we will consider local entropy increase law beyond linear order which means we are not allowed to neglect the terms in ( [ ray3 ] ) which have been thrown away in the previous section . now we will be considering all the relevant terms in equation ( [ ray3 ] ) and will determine the criteria such that second law holds non - perturbatively in the coupling . for this , we will assume a spherically symmetric vaidya like solution . this means , the metric in ( [ vd1 ] ) will now have the following form , [ vd ] ds^2=-f(r , v ) dv^2 + 2 dv dr + r^2d_3 ^ 2 . in fact for egb gravity one can obtain such a solution just setting the mass in the static spherically symmetric boulware - deser @xcite solution to be a function of the advanced time @xmath80 . beyond linearized level , one has to evaluate the full @xmath73 in the evolution equation ( [ ray3 ] ) . since the solution is spherically symmetric , shear is identically zero . in fact , for this particular case the evolution equation will be of the form , -=-8t_kk-^2_k , note that we must have @xmath115 to have a local entropy increase law . also , we have @xmath116 by nec . now consider a situation where the stationary black hole is perturbed by some matter flux and we are examining the second law when the matter has already entered into the black hole . in that case , the above evolution equation does not have any contribution from matter stress energy tensor and the evolution will be driven solely by the @xmath117 term . so we will have a equation of the form , -=-^2_k . in such a situation , if we demand the entropy is increasing , we have to fix the sign of the coefficient of @xmath117 term . we evaluate the coefficient in the stationary background and impose the condition that overall sign in front of @xmath118 is negative . this will give us a bound on the parameters of the theory under consideration . we start with the egb gravity . the event horizon is now a null surface whose equation is @xmath77 . calculating the r.h.s . of ( [ ray3 ] ) for the metric ( [ vd ] ) we obtain , -=- , where we have identified @xmath119 and introduced @xmath120 as the coefficient of the @xmath117 terms . the expression of @xmath120 is given by , [ zeta ] = . for @xmath121 , the coefficient reduces to @xmath122 which matches with gr . @xmath123 is the ricci scalar evaluated on the horizon . to satisfy the entropy increase law we now set , [ cond ] > 0.as discussed earlier , we will evaluate @xmath120 for different stationary backgrounds and determine bounds on the coefficient @xmath90 imposing the condition ( [ cond ] ) . now we consider the boulware - deser ( bd ) @xcite black hole as the background , for which , f(r)=1 + . also , [ eqn3 ] r_h^2 + 2=m determines the location of the horizon . existence of an event horizon demands @xmath124 . in this case horizon topology is a sphere . now , evaluating @xmath120 for the above background at the horizon @xmath125 , we get the following inequality < 0 , [ ineq ] which leads to @xmath126 to understand this better , note that we require @xmath127 to avoid the naked singularity of the black hole solution for @xmath128 . thus in this case for a spherically symmetric black hole @xmath120 will be positive and hence second law will be automatically satisfied . the condition of the validity of the second law is same as that for having a regular event horizon . also , for @xmath129 , it is possible to make @xmath130 as small as possible by tuning the mass @xmath131 . but when @xmath90 is negative ( a situation that appears to be disfavoured by string theory , see @xcite , @xcite and references therein ) , @xmath130 can not be made arbitrarily small and it would suggest that these black holes can not be formed continuously from a zero temperature set up . notice that we could have reached the conclusion without the second law if @xmath131 is considered to be positive however , our current argument does not need to make this assumption . due to this pathology , it would appear that the negative gb coupling case would be ruled out in a theory with no cosmological constant . next we consider the 5-dimensional ads black hole solution for egb gravity as the background . the function @xmath132 now becomes @xcite , f(r)=k+(1- ) . here @xmath47 is the length scale with the cosmological constant ( @xmath133 ) and @xmath46 can take values @xmath134 and @xmath135 corresponding to planar , spherical or hyperbolic horizons respectively . the intrinsic ricci scalar on the horizon is @xmath136 where , the horizon is at r_h^2=. first we will consider black brane solution for which @xmath137 and the horizon is planar . the intrinsic curvature of the codimension two slices of the event horizon vanishes . in this case the coefficient @xmath120 reads , [ betajm ] = ( 1 - 16 ) . introducing a rescaled coupling @xmath138 @xcite we get , = ( 1 - 8_gb ) . again demanding positivity of @xmath120 we get , _ gb < , so to satisfy the second law using @xmath50 at @xmath139 order we need to impose a bound on the gb coupling . incidentally the same has to be imposed to avoid instabilities in the sound channel analysis of the quasinormal mode for dual plasma . it was shown in @xcite that when @xmath140 the schroedinger potential develops a well which can support unstable quasinormal modes in the sound channel . it is interesting to see that the second law knows about this instability . next we consider @xmath141 case . this will give us a black hole with a hyperbolic horizon . we will also set @xmath142 , which corresponds to the zero mass solution . in this limit we find , = 5 - 4 . demanding @xmath143 we get , _ if we take the limit @xmath144 in the expression of @xmath120 with @xmath145 ( i.e. the case when the horizon sections becomes planar ) , we recover the bound on gb coupling @xmath146 for the @xmath137 case . however , this is a weaker bound on @xmath146 and the strongest bound on @xmath146 comes from the hyperbolic black hole in the massless limit . this is illustrated in figures 1a and 1b . we have checked that for the @xmath147 case other thermodynamic considerations ( e.g. , positive entropy , black hole being the correct phase ) do not lead to stronger bounds . [ cols="^,^ " , ] quite curiously we have recovered the tensor channel causality constraint @xcite which is stronger than the causality bounds for @xmath148 in @xcite . for flat case ( @xmath137 ) , the bound on gb coupling in @xmath149 dimensions can be expressed as @xmath150 . this means for any dimension the strongest bound arises when the black holes have hyperbolic horizon ( @xmath141 ) with @xmath151 . this would also suggest that in the large @xmath149 limit , @xmath152 contrary to @xmath153 obtained in @xcite . ] . the bound on @xmath146 in this channel leads to the lower bound on the ratio of shear viscosity ( @xmath154 ) to entropy density ( @xmath155 ) , @xmath156 in gb holography @xcite . the vector and scalar channels also lead to bounds on the coupling from causality constraints . the strongest bound in gb arises from the tensor and scalar channels @xcite . our finding coincides with the tensor channel constraint . it will be interesting to see if the other channels can also be reproduced using the second law analysis . we end this section with one last comment . instead of putting a bound on the gauss - bonnet coupling , one can try to modify the entropy functional @xmath50 itself so that all the @xmath139 terms will be canceled . one possible modification is to add a term like @xmath157 which vanishes on the stationary horizon and boost invariant in the null direction . so this type of higher order ambiguous term can occur in the wald derivation . we modify @xmath50 as , s_modify = d^3 x ( 1 + 2 + _ k^2_l^2 ) . @xmath120 in ( [ betajm ] ) will be modified as , [ modeq ] = so we can adjust @xmath158 such that the @xmath120 vanishes and second law automatically holds up to @xmath139 . one may also can add a term like @xmath159 at this order where @xmath160 is the covariant derivative with respect to the surface index . but these additional terms will generate further higher order terms starting from @xmath161 so one will need to add more terms to cancel them . it might possible to do this recursively . however , the coefficients of these terms will also depend on the background data . so it will be prudent not to modify the entropy functional . we have put a bound on the coupling instead of modifying the entropy functional at @xmath139 order such that second law continues to hold up to this order . it will be interesting to investigate what happens to this bound on @xmath146 when one goes beyond @xmath139 order . first consider the asymptotically flat black hole with spherically symmetric horizon in 5d . in this case , = 13(1++ f(r , v)+f(r , v)- f(r , v ) ) now , to extract a bound , we need an explicit form of a spherically symmetric static black hole solution in @xmath162 theory . we are not aware of any such solution except the 5d schwarzschild solution of gr which is also a solution of @xmath162 theory . so , as an example we take such a solution as the background and write , f(r)=1- . then the validity of the second law gives us the condition , > . when @xmath163 this is automatically satisfied whenever a horizon exist whereas for @xmath164 , a lower bound on the horizon size ensures the validity of the second law . + repeating this same analysis for 4d with the metric function @xmath165 we get , > , therefore , in both cases , there is a minimum horizon radius for the second law to hold when @xmath164 . again as discussed in [ bd - bnd ] , minimum @xmath130 scenario would appear pathological and may be taken as a reason to disfavour the @xmath164 sign . we repeat the same analysis of the previous section for @xmath107 theory but with the background as an asymptotically ads black brane in 5d . the metric function is then given by f(r)= f_(1-(r_h / r)^4 ) where @xmath166 and it satisfies the equation , 1-f_+_2 f_^2=0 . in this case we obtain , = ( + f(r , v)+- ) , define @xmath167 and after evaluating on the horizon with the above metric function as background , we get = - . from this , the bound becomes , _ 2 < . in this case if we repeat out analysis for a zero mass hyperbolic black hole we obtain a lower bound on @xmath168 . combining them we get , - < _ 2 if we repeat the same analysis for 4d and obtain the following bound , @xmath169 no analog of such a bound in the context of ads / cft is known in this case . this analysis also suggests that the second law bound arising from hyperbolic horizons is not necessarily connected to the causality constraints . the reason is that for these theories , in the analysis of @xcite , we only get @xmath170 . we will now consider bounds arising from the second law in quadratic curvature theories with a negative cosmological constant , which only involve terms like @xmath3 and @xmath4 . in particular , we will focus on the case of critical gravity @xcite . for any theory with only ricci@xmath51 or @xmath4 , the positive energy condition of hofman - maldacena @xcite does not yield any constraint ( see @xcite for general expressions ) except for the positivity of the two point function of stress tensors which follows from unitarity . in all the examples below , the positivity of black hole entropy leads to the same condition arising from demanding @xmath170 where @xmath171 is the coefficient in the two point function of cft stress tensors . the second law constraint will lead to further conditions as we will see . the action for the 4d version of critical gravity is given by , s = d^4x ( r+ + l^2(r_abr^ab - r^2 ) ) [ clag ] in general this theory posses a massive spin-2 mode in addition to the usual massless spin-2 degree of freedom . it has been shown in @xcite , at @xmath172 , which is called the critical point , the additional massive spin-2 mode becomes massless . also , at the critical point the entropy of the schwarzschild - ads black hole becomes zero . we will now use the second law to put a bound on the coupling @xmath84 . proceeding as before at the second order we get , = ( -+ l^2 f(r , v)- l^2 r(v ) f(r , v ) ) . in this case , we take the background as , f(r)= ( 1-()^3 ) . from this we get the bound on the coupling as , [ b1 ] . also , the black hole entropy has to be positive if it was zero then it must remain zero at all times . from this we have the following condition , 1 + 2>0 . so , [ b2 ] > - . combining ( [ b1 ] ) and ( [ b2 ] ) we get , [ b3 ] - < . now from the ads / cft perspective it is natural to assume the positivity of the coefficient @xmath173 of two point correlator of boundary stress tensor . that will give exactly the same condition as in ( [ b2 ] ) which is weaker than ( [ b3 ] ) . repeating the same analysis for a zero mass hyperbolic black hole we did not any further bounds . next we consider the critical gravity theory in 5d . the action is given by , s = d^5x ( r+ + l^2(r_abr^ab - r^2 ) ) [ clag ] the critical point now corresponds to @xmath174 @xcite . proceeding as before at the second order we get , = ( + + l^2 f(r , v)- l^2 r(v ) f(r , v ) ) . in this case , the background solution is taken as , f(r)= f_(1-()^4 ) , where the quantity @xmath175 satisfies the equation , @xmath176 the second law thus leads to , [ b4 ] - . again demanding the entropy to be positive we get , @xmath177 from which @xmath178 thus we have , [ b6 ] - < . if we repeat our analysis for a zero mass hyperbolic black hole , the resulting bound we obtain is weaker than this . we end this section by exploring the case of nmg theory in 3d . the action is given by , s = d^3x ( r+ + l^2(r_abr^ab - r^2 ) ) [ clag ] the critical point is now at @xmath179 for this case , = ( 1-+ l^2 f(r , v)- l^2 r(v ) f(r , v ) ) . the background solution is taken as , f(r)= f_(1-()^2 ) , where the quantity @xmath175 satisfies @xmath180 then we get the following bound , [ b7 ] . demanding the positivity of the black hole entropy we get , @xmath181 this gives , @xmath182 thus we have , -3 < . in this case , even if we consider a zero mass hyperbolic black hole we obtain the same bound . in this section we will construct a holographic @xmath183-function using the fps entropy functional for general curvature squared theories . here we will again use the evolution equation for @xmath61 ( [ ray3 ] ) and in the process we will have a geometric derivation for the @xmath183-function . for unitary lorentz invariant theory one can define a function in the space of couplings which decreases monotonically from uv to ir , parametrizing the rg flow . it coincides with the central charges of the corresponding cfts at the uv and ir fixed point . in @xmath184 and @xmath185 dimensions this has been established rigorously @xcite . now from the holographic point view this @xmath183 function plays a pivotal role in understanding nature of rg flow . a proposal for the @xmath183 function for arbitrary theories of gravity has been given in @xcite for arbitrary dimensions . in this section we will use the raychaudhuri equation as discussed in the previous sections and derive a @xmath183 function for a general curvature squared theory . for general higher curvature theories , we do not expect a c - theorem as there are bound to be problems with unitarity in such theories . thus some condition on the couplings , which presumably only explore a subset of conditions leading to unitarty as in @xcite , is expected . this is what we will find as well . at the uv and ir fixed points this function will coincide with the a - type anomaly coefficient for the curvature squared theory and it will also have to be a monotonically decreasing quantity along the rg flow . to prove the monotonicity we will use raychaudhuri equation . the basic setup is similar as before . to derive a @xmath183 function purely in terms of geometric quantities we will first start with a specific example . we will consider domain wall type of geometry in @xmath186 dimensions , the metric for which is given below , ds^2=dr^2+e^2 a(r)(-d^2+dx^2+dy^2 + dz^2 ) . the spacetime can be foliated by surfaces of codimension-2 at every constant time slice . the induced metric on such surfaces is given by , ds^2_surface = e^2 a(r)(dx^2+dy^2+dz^2 ) . next we will consider a hypersurface orthogonal null congruence . the tangent vector for this congruences is given by , k^a=(_)^a=\{-e^-a(r),e^-2 a(r),0,0,0}. it satisfies @xmath187 and @xmath188 we can construct another auxiliary null vector @xmath16 as , l^a=\{,,0,0,0 } such that , @xmath189 note that in this case @xmath17 and @xmath16 are not any kind of horizon null generators as compared to the previous sections . they simply correspond to a null congruence that converges along the light sheet projected out of the codimension-2 surface under consideration . also in this section we will use affine parametrization . next we will start with a guess for the @xmath183 function for curvature squared theory . the action for this theory is given by , s = d^4 x. the corresponding entropy functional is the fps functional . s_fps = d^3 x ( 1+)= d^3 x ( 1 + 2 ) . @xmath123 is the 3 dimensional ricci scalar intrinsic to the surface . now we will start with the following candidate , c()= where as usual , @xmath190 and @xmath191 is the affine parameter . this appears to be a natural guess since for the einstein theory it goes over to the c - function in @xcite . then we have to show two things , 1 . @xmath192 is monotonically decreasing under the flow along the null geodesic congruences . that is we have to check the sign of @xmath193 2 . further it has to coincide with the correct central charge ( a - type anomaly in four dimensions ) at the fixed points . now , @xmath194 using @xmath195 next we will replace @xmath196 by ( [ ray2 ] ) with @xmath197 and @xmath198 by raychaudhuri equation ( [ ray1 ] ) . we obtain , = + e_kk , where , @xmath199 also , @xmath200 and c()=. notice that in ( [ dcdt ] ) the term proportional to @xmath201 has no possibility of having a definite sign and it does not cancel with terms in @xmath73 so it is natural to set the coefficient to zero . now @xmath202 is positive and @xmath65 using the null energy condition is positive . with @xmath203 , we find e_kk= ( - ) . where we have used @xmath204 so we will have now , = + ( - ) , then we can define a effective quantity , |c()= c()+ such that , = 0 , next evaluating @xmath205 and multiplying both side by a factor of @xmath206 we get , ( ) 0 . now in terms of the radial coordinate @xmath81 this becomes , ( k^r_r + k^ _ ) ( ) 0 . from this , e^-2 a(r ) ( ) 0 . so , @xmath207 is a monotonically decreasing quantity along the rg flow , where the rg scale is determined by the radial coordinate @xmath81 . at the fixed point the geometry is ads and hence @xmath208 the @xmath209 coincides ( upto an inconsequential overall factor ) with the a - type anomaly coefficient at uv and ir ( see eg . @xcite for expressions in general higher curvature theories ) . in this construction , the starting point of the entropy functional played an important role in reaching the condition @xmath210 . in @xcite , this condition emerged by demanding the existence of a simple " c - function . our analysis gives another way of looking at the same condition . one may wonder if it is possible to come up with a scenario where we did not need to absorb @xmath73 into the c - function like what we have done . for example , instead of the hee functional , perhaps one could try to construct a functional that cancels the @xmath73 on the _ rhs _ of the raychaudhuri equation . this does not appear to be the case starting with the wald entropy functional instead of the hee functional does not naturally lead to the @xmath210 condition . in this paper , we considered four derivative gravity and subjected it to the condition that the second law of black hole thermodynamics is satisfied at linear and quadratic orders in perturbation . we showed that the entropy functional is uniquely fixed ( upto second order in extrinsic curvatures ) and coincides with the holographic entropy functionals . we also derived interesting bounds on the coupling which arises at next to leading order in perturbations . we conclude here with a brief discussion on possible future directions . * one should consider more generic perturbations to see what happens at linear and second order . this is important since the possibility exists that for generic perturbations the second law holds only for very specific cases like the lovelock theories . in @xcite , it was argued that causality constraints would require adding an infinite set of higher spin massive modes in a gb theory for consistency . it will be interesting to see if the second law knows about this in the perturbative sense used in the current work . * by considering asymptotically flat black holes and the gauss - bonnet theory , we found that for the negative gb coupling we get a minimum horizon radius for the second law to hold . taking this to be a pathology , this appears to disfavour this sign of the coupling . we further saw that the second law appears to know about the sound channel instablity in the context of gauss - bonnet holography . it will be interesting to see if @xmath211 which corresponds to the scalar channel instability @xcite can also be seen from the second law . * one should consider how our analysis will change if there is matter coupling to the higher curvature terms . in this case , it is not clear how to define the null energy condition . it seems natural that the matter couplings will get constrained by demanding the second law . * our analysis can be extended to general theories such as the quasi - topological theories @xcite . in order to get a sufficient number of conditions to fix all the extrinsic curvature terms , one will need to turn on generic perturbations . * in case of gravitational chern - simons terms , it will be interesting to see if and how the validity of second law can fix the entropy functionals @xcite . since the @xmath212 of the raychaudhuri equation does not know about topological terms , the linearized analysis will not be able to capture the effect of such terms . one may need to consider topology changing processes like black hole mergers @xcite to see the effect of such terms . * it will be interesting to compare our analysis with what arises from similar considerations in the fluid - gravity correspondence @xcite . demanding a positive entropy current for higher curvature theories will lead to constraints which are presumably going to be similar to what we have found . * the connection between the second law constraints and the positivity of relative entropy @xcite in the context of holography can also be explored . we found that when we consider the second law for topological black holes with zero mass , we recover the bound from the tensor channel causality constraint . why is this happening ? a plausible explanation is as follows . the topological black holes with zero mass are just ads spaces written in different co - ordinates and the finite wald entropy of these black holes can be interpreted as an entanglement entropy across a sphere @xcite in the dual field theory . thus the second law in this context may be related to the positivity of relative entropy due to a time dependent perturbation corresponding to the infalling matter we have considered . relative entropy is expected to be positive for a unitary field theory . hence there appears to be an interesting interplay between bulk causality and field theory unitarity . it will be interesting to make this connection more precise . it will also be interesting to understand the bounds in the context of c - theorems using the second law for causal horizons as advocated recently in @xcite . we thank shamik banerjee , sayantani bhattacharyya , menika sharma and aron wall for discussions . ab thanks iit - gandhinagar for hospitality during the course of this work . the research of ss is partially supported by iit gandhinagar start up grant no : ip / iitgn / phy / ss/ + 201415 - 12 . as acknowledges support from a swarnajayanti fellowship , govt . of india . 99 s. ryu and t. takayanagi , `` holographic derivation of entanglement entropy from ads / cft , '' phys . lett . * 96 * , 181602 ( 2006 ) [ hep - th/0603001 ] . + v. e. hubeny , m. rangamani and t. takayanagi , `` a covariant holographic entanglement entropy proposal , '' jhep * 0707 * , 062 ( 2007 ) [ arxiv:0705.0016 [ hep - th ] ] . h. casini , m. huerta and r. c. myers , `` towards a derivation of holographic entanglement entropy , '' jhep * 1105 * , 036 ( 2011 ) [ arxiv:1102.0440 [ hep - th ] ] . a. bhattacharyya , m. sharma and a. sinha , `` on generalized gravitational entropy , squashed cones and holography , '' jhep * 1401 * , 021 ( 2014 ) [ arxiv:1308.5748 [ hep - th ] ] . + a. bhattacharyya , a. kaviraj and a. sinha , `` entanglement entropy in higher derivative holography , '' jhep * 1308 * , 012 ( 2013 ) [ arxiv:1305.6694 [ hep - th ] ] . + a. bhattacharyya and m. sharma , `` on entanglement entropy functionals in higher derivative gravity theories , '' jhep * 1410 * , 130 ( 2014 ) arxiv:1405.3511 [ hep - th ] ] . + j. camps and w. r. kelly , `` generalized gravitational entropy without replica symmetry , '' jhep * 1503 * , 061 ( 2015 ) [ arxiv:1412.4093 [ hep - th ] ] . x. dong , `` holographic entanglement entropy for general higher derivative gravity , '' jhep * 1401 * , 044 ( 2014 ) [ arxiv:1310.5713 [ hep - th ] , arxiv:1310.5713 ] . j. camps , `` generalized entropy and higher derivative gravity , '' jhep * 1403 * , 070 ( 2014 ) [ arxiv:1310.6659 [ hep - th ] ] . r. x. miao and w. z. guo , `` holographic entanglement entropy for the most general higher derivative gravity , '' arxiv:1411.5579 [ hep - th ] . j. de boer , m. kulaxizi and a. parnachev , `` holographic entanglement entropy in lovelock gravities , '' jhep * 1107 * , 109 ( 2011 ) [ arxiv:1101.5781 [ hep - th ] ] . + l. y. hung , r. c. myers and m. smolkin , `` on holographic entanglement entropy and higher curvature gravity , '' jhep * 1104 * , 025 ( 2011 ) [ arxiv:1101.5813 [ hep - th ] ] . d. v. fursaev , a. patrushev and s. n. solodukhin , `` distributional geometry of squashed cones , '' phys . d * 88 * , no . 4 , 044054 ( 2013 ) [ arxiv:1306.4000 [ hep - th ] ] . r. m. wald , `` black hole entropy is the noether charge , '' phys . rev . * d48 * , 3427 - 3431 ( 1993 ) . [ gr - qc/9307038 ] . v. iyer and r. m. wald , some properties of noether charge and a proposal for dynamical black hole phys . d * 50 * , 846 ( 1994 ) [ arxiv : gr - qc/9403028 ] . t. jacobson , g. kang , r. c. myers , `` increase of black hole entropy in higher curvature gravity , '' phys . * d52 * , 3518 - 3528 ( 1995 ) . [ gr - qc/9503020 ] . s. sarkar and a. c. wall , `` generalized second law at linear order for actions that are functions of lovelock densities , '' phys . d * 88 * , 044017 ( 2013 ) [ arxiv:1306.1623 [ gr - qc ] ] . s. bhattacharjee , s. sarkar and a. wall , `` the holographic entropy increases in quadratic curvature gravity , '' arxiv:1504.04706 [ gr - qc ] . a. c. wall , `` a second law for higher curvature gravity , '' arxiv:1504.08040 [ gr - qc ] . r. g. cai , `` gauss - bonnet black holes in ads spaces , '' phys . d * 65 * , 084014 ( 2002 ) [ hep - th/0109133 ] . r. c. myers and a. sinha , `` holographic c - theorems in arbitrary dimensions , '' jhep * 1101 * , 125 ( 2011 ) [ arxiv:1011.5819 [ hep - th ] ] . + r. c. myers and a. sinha , `` seeing a c - theorem with holography , '' phys . d * 82 * , 046006 ( 2010 ) [ arxiv:1006.1263 [ hep - th ] ] . l. girardello , m. petrini , m. porrati and a. zaffaroni , `` novel local cft and exact results on perturbations of n=4 superyang mills from ads dynamics , '' jhep * 9812 * , 022 ( 1998 ) [ hep - th/9810126 ] . + d. z. freedman , s. s. gubser , k. pilch and n. p. warner , `` renormalization group flows from holography supersymmetry and a c theorem , '' adv . theor . math . phys . * 3 * , 363 ( 1999 ) [ hep - th/9904017 ] . s. sarkar and a. c. wall , `` second law violations in lovelock gravity for black hole mergers , '' phys . d * 83 * , 124048 ( 2011 ) [ arxiv:1011.4988 [ gr - qc ] ] . s. kar and s. sengupta , `` the raychaudhuri equations : a brief review , '' pramana * 69 * , 49 ( 2007 ) [ gr - qc/0611123 ] . e. poisson , a relativist s toolkit : the mathematics of black - hole mechanics " , cambridge university press ; 1 edition ( 5 november 2007 ) h. lu and c. n. pope , `` critical gravity in four dimensions , '' phys . rev . lett . * 106 * , 181302 ( 2011 ) [ arxiv:1101.1971 [ hep - th ] ] . + s. deser , h. liu , h. lu , c. n. pope , t. c. sisman and b. tekin , `` critical points of d - dimensional extended gravities , '' phys . d * 83 * , 061502 ( 2011 ) [ arxiv:1101.4009 [ hep - th ] ] . m. alishahiha , a. f. astaneh and m. r. m. mozaffar , `` entanglement entropy for logarithmic conformal field theory , '' phys . d * 89 * , no . 6 , 065023 ( 2014 ) [ arxiv:1310.4294 [ hep - th ] ] . + m. alishahiha and r. fareghbal , `` d - dimensional log gravity , '' phys . rev . d * 83 * , 084052 ( 2011 ) [ arxiv:1101.5891 [ hep - th ] ] . + d. grumiller , w. riedler , j. rosseel and t. zojer , `` holographic applications of logarithmic conformal field theories , '' j. phys . a * 46 * , 494002 ( 2013 ) [ arxiv:1302.0280 [ hep - th ] ] . + k. sen , a. sinha and n. v. suryanarayana , `` counterterms , critical gravity and holography , '' phys . d * 85 * , 124017 ( 2012 ) [ arxiv:1201.1288 [ hep - th ] ] . + m. alishahiha , m. m. qaemmaqami , a. naseh and a. shirzad , `` on 3d minimal massive gravity , '' jhep * 1412 * , 033 ( 2014 ) [ arxiv:1409.6146 [ hep - th ] ] . z. komargodski and a. schwimmer , `` on renormalization group flows in four dimensions , '' jhep * 1112 * ( 2011 ) 099 [ arxiv:1107.3987 [ hep - th ] ] . + z. komargodski , `` the constraints of conformal symmetry on rg flows , '' jhep * 1207 * ( 2012 ) 069 [ arxiv:1112.4538 [ hep - th ] ] . a. sinha , `` on higher derivative gravity , @xmath183-theorems and cosmology , '' class . * 28 * , 085002 ( 2011 ) [ arxiv:1008.4315 [ hep - th ] ] . + a. sinha , `` on the new massive gravity and ads / cft , '' jhep * 1006 * , 061 ( 2010 ) [ arxiv:1003.0683 [ hep - th ] ] . + j. t. liu , w. sabra and z. zhao , `` holographic c - theorems and higher derivative gravity , '' phys . d * 85 * , 126004 ( 2012 ) [ arxiv:1012.3382 [ hep - th ] ] . + i. gullu , t. c. sisman and b. tekin , `` c - functions in the born - infeld extended new massive gravity , '' phys . rev . d * 82 * , 024032 ( 2010 ) [ arxiv:1005.3214 [ hep - th ] ] . + m. f. paulos , `` new massive gravity extended with an arbitrary number of curvature corrections , '' phys . d * 82 * , 084042 ( 2010 ) [ arxiv:1005.1646 [ hep - th ] ] . r. c. myers and b. robinson , `` black holes in quasi - topological gravity , '' jhep * 1008 * , 067 ( 2010 ) [ arxiv:1003.5357 [ gr - qc ] ] . + j. oliva and s. ray , `` a new cubic theory of gravity in five dimensions : black hole , birkhoff s theorem and c - function , '' class . * 27 * , 225002 ( 2010 ) [ arxiv:1003.4773 [ gr - qc ] ] . + r. c. myers , m. f. paulos and a. sinha , `` holographic studies of quasi - topological gravity , '' jhep * 1008 * , 035 ( 2010 ) [ arxiv:1004.2055 [ hep - th ] ] . a. castro , s. detournay , n. iqbal and e. perlmutter , `` holographic entanglement entropy and gravitational anomalies , '' jhep * 1407 * , 114 ( 2014 ) [ arxiv:1405.2792 [ hep - th ] ] . + t. azeyanagi , r. loganayagam and g. s. ng , `` holographic entanglement for chern - simons terms , '' arxiv:1507.02298 [ hep - th ] . + w. z. guo and r. x. miao , `` entropy for gravitational chern - simons terms by squashed cone method , '' arxiv:1506.08397 [ hep - th ] . s. bhattacharyya , v. e. hubeny , s. minwalla and m. rangamani , `` nonlinear fluid dynamics from gravity , '' jhep * 0802 * , 045 ( 2008 ) [ arxiv:0712.2456 [ hep - th ] ] . + r. baier , p. romatschke , d. t. son , a. o. starinets and m. a. stephanov , `` relativistic viscous hydrodynamics , conformal invariance , and holography , '' jhep * 0804 * , 100 ( 2008 ) [ arxiv:0712.2451 [ hep - th ] ] . d. d. blanco , h. casini , l. y. hung and r. c. myers , `` relative entropy and holography , '' jhep * 1308 * , 060 ( 2013 ) [ arxiv:1305.3182 [ hep - th ] ] . + n. lashkari , m. b. mcdermott and m. van raamsdonk , `` gravitational dynamics from entanglement thermodynamics , '' jhep * 1404 * , 195 ( 2014 ) [ arxiv:1308.3716 [ hep - th ] ] . + t. faulkner , m. guica , t. hartman , r. c. myers and m. van raamsdonk , `` gravitation from entanglement in holographic cfts , '' jhep * 1403 * , 051 ( 2014 ) [ arxiv:1312.7856 [ hep - th ] ] . + s. banerjee , a. bhattacharyya , a. kaviraj , k. sen and a. sinha , `` constraining gravity using entanglement in ads / cft , '' jhep * 1405 * , 029 ( 2014 ) [ arxiv:1401.5089 [ hep - th ] ] . + s. banerjee , a. kaviraj and a. sinha , `` nonlinear constraints on gravity from entanglement , '' class . grav . * 32 * , no . 6 , 065006 ( 2015 ) [ arxiv:1405.3743 [ hep - th ] ] . + n. lashkari , c. rabideau , p. sabella - garnier and m. van raamsdonk , `` inviolable energy conditions from entanglement inequalities , '' jhep * 1506 * , 067 ( 2015 ) [ arxiv:1412.3514 [ hep - th ] ] . + j. bhattacharya , v. e. hubeny , m. rangamani and t. takayanagi , `` entanglement density and gravitational thermodynamics , '' phys . d * 91 * , no . 10 , 106009 ( 2015 ) [ arxiv:1412.5472 [ hep - th ] ] . + n. lashkari and m. van raamsdonk , `` canonical energy is quantum fisher information , '' arxiv:1508.00897 [ hep - th ] . + f. m. haehl , `` comments on universal properties of entanglement entropy and bulk reconstruction , '' arxiv:1508.00766 [ hep - th ] .

evidence is presented that 2mass j03202839 , a late - type dwarf with discrepant optical ( m8 : ) and near - infrared ( l1 ) spectral types , is an as - yet unresolved stellar / brown dwarf binary with late - type m dwarf and t dwarf components . this conclusion is based on low - resolution , near - infrared spectroscopy that reveals a subtle but distinctive absorption feature at 1.6 . the feature , which is also present in the combined light spectrum of the m8.5 + t6 binary scr 1845 , arises from the combination of feh absorption from an m8.5 primary and pseudo - continuum flux from a t5 secondary , as ascertained from binary spectral templates constructed from empirical data . the binary templates provide a far superior match to the overall near - infrared spectral energy distribution of 2mass j0320 than any single comparison spectra . laser guide star adaptive optics ( lgs ao ) imaging observations , including the first application of lgs ao aperture mask interferometry , fail to resolve a faint companion , restricting the projected separation of the system to less than 8.3 au at the time of observation . 2mass j0320 is the second very low mass binary to be identified from unresolved , low - resolution , near - infrared spectroscopy , a technique that complements traditional high resolution imaging and spectroscopic methods .

0803.0295

the optical and near - infrared spectral energy distributions of very low mass stars and brown dwarfs late - type m , l and t dwarfs are distinctly non - blackbody . overlapping molecular bands and strong line emission produce a rich array of spectral diagnostics for classification and characterization of physical properties . considerable effort is now being devoted toward decrypting the spectral fingerprints of late - type dwarfs to determine masses , ages , metallicities and other fundamental parameters ( e.g. , @xcite ) . in some cases , spectral peculiarities arise when an observed source is in fact an unresolved multiple system , with components of different masses , effective temperatures , and other spectral properties . while several classes of stellar multiples are recognized on the basis of their unusual spectral or photometric properties ( u geminorum stars , m dwarf + white dwarf systems , etc . ) , identifying such cases amongst late - type dwarfs is complicated by the influence of other physical effects . delineation of spectral peculiarities that arise purely from multiplicity as opposed to other physical effects is essential if we hope to unambiguously characterize the physical properties of the lowest luminosity stars and brown dwarfs . very low mass multiple systems are important in their own right , as they enable mass and occasionally radius measurements ( e.g. , @xcite ) , provide constraints for star / brown dwarf formation scenarios ( e.g. , @xcite ) and facilitate detailed studies of atmospheric properties ( e.g. , @xcite ) . of the roughly 90 very low mass multiple systems currently known , the majority have been identified through high angular resolution imaging , using the _ hubble space telescope _ ( _ hst _ ; e.g. , @xcite ) , ground - based adaptive optics systems ( e.g. , @xcite ) and more recently aperture masking interferometry ( e.g. , @xcite ) . however , as the vast majority of very low mass binaries have small separations ( @xmath390% have @xmath4 @xmath5 20 au ; @xcite ) , expanding the population of known binaries to greater distances requires either finer angular sampling or the identification of systems that are unresolved . the frequency of nearby , tightly - bound binaries is also essential for a complete assessment of the overall very low mass dwarf binary fraction , since imaging studies provide only a lower limit to this fundamental statistic . such systems are also more likely to eclipse , enabling radius measurements and fundamental tests of evolutionary models ( e.g. , @xcite ) . while searches for radial velocity variability via high resolution spectroscopy can be useful in this regime ( e.g. , @xcite ) , in many cases very low luminosity and/or distant late - type dwarfs are simply too faint to be followed up in this manner . recently , @xcite demonstrated that in certain cases the presence of an unresolved companion can be inferred directly from the morphology of a source s low - resolution near - infrared spectrum . in particular , it was shown that the spectrum of the peculiar l dwarf sdss j080531.84 + 481233.0 ( hereafter sdss j0805 + 4812 ; @xcite ) , which has highly discrepant optical and near - infrared spectral classifications , could be accurately reproduced as a combination of `` normal '' l4.5 + t5 components . indeed , the binary hypothesis provides a far simpler and more consistent explanation for the unusual optical , near - infrared and mid - infrared properties of sdss j0805 + 4812 than other alternatives ( e.g. , @xcite ) . the identification of unresolved multiples like sdss j0805 + 4812 by low - resolution near - infrared spectroscopy is a potential boon for low - mass multiplicity studies , as this method is not subject to the same physical or projected separation limitations inherent to high - resolution imaging and spectroscopic techniques . this article reports the discovery of a second unresolved very low mass binary system , 2mass j03202839@xmath00446358 ( hereafter 2mass j0320@xmath00446 ) , identified by the morphology of its low - resolution , near - infrared spectrum . the spectroscopic observations leading to this conclusion are described in @xmath6 2 , as are laser guide star adaptive optics ( lgs ao ) imaging observations aimed at searching for a faint companion . analysis of the spectral data using the binary template matching technique described in @xcite is presented in @xmath6 3 . @xmath6 4 discusses the viability of 2mass j0320@xmath00446 being a binary , with specific comparison to the known m dwarf + t dwarf system scr 1845@xmath06357 . we also constrain the projected separation of the 2mass j0320@xmath00446 system based on our imaging observations , and discuss overall limitations on the variety of unresolved m dwarf + t dwarf binaries that can be identified from composite near - infrared spectroscopy . conclusions are summarized in @xmath6 5 . 2mass j0320@xmath00446 was originally discovered by @xcite and @xcite in the two micron all sky survey ( 2mass ; @xcite ) , and classified m8 : ( uncertain ) and l0.5 on the basis of optical and near - infrared spectroscopy , respectively . the m8 : optical classification is uncertain because of the low signal - to - noise of the spectral data , and is not due to any specific spectral peculiarity . @xcite estimate a distance of 26@xmath24 pc for this source based on its classification and empirical @xmath7/spectral type relations . @xcite , using @xmath8-band plate data from the supercosmos sky survey ( sss ; @xcite ) , report a relatively high proper motion of 0@xmath968@xmath20@xmath904 yr@xmath10 at position angle 191@xmath11 for this source . figure [ fig_chart ] shows the field around 2mass j0320@xmath00446 imaged by @xmath12 and @xmath8 photographic plates . a faint source is seen in the 1955 palomar sky survey i @xcite @xmath12-band image roughly at the offset position indicated by the @xcite proper motion . by including this source position along with additional astrometry drawn from the sss and 2mass catalogs , an improved proper motion measurement of 0@xmath9562@xmath20@xmath9005 yr@xmath10 at position angle [email protected]@xmath11 was determined . this proper motion and the estimated distance indicates a rather large tangential space velocity for 2mass j0320@xmath00446 , @xmath13 = 69@xmath211 km s@xmath10 , suggesting that it could be an older disk star . none of the previous studies of 2mass j0320@xmath00446 report the presence of a faint companion . low resolution near - infrared spectral data for 2mass j0320@xmath00446 were obtained on 2007 september 16 ( ut ) using the spex spectrograph @xcite mounted on the 3 m nasa infrared telescope facility ( irtf ) . the conditions on this night were poor with patchy clouds , cirrus and average seeing ( 0@xmath98 at @xmath14-band ) , and 2mass j0320@xmath00446 was observed as a bright backup target ( @xmath14 = [email protected] ) . the 0@xmath95 slit was used to obtain 0.72.5 @xmath1 spectroscopy with resolution @xmath15 @xmath16 and dispersion across the chip of 2030 pixel@xmath10 . to mitigate the effects of differential refraction , the slit was aligned to the parallactic angle . six exposures of 90 s each were obtained in an abba dither pattern along the slit . the a0 v star hd 18571 was observed immediately after 2mass j0320@xmath00446 and at a similar airmass ( 1.21 ) for flux calibration . internal flat field and argon arc lamps were observed after both target and flux standard observations for pixel response and wavelength calibration . data were reduced with the idl spextool package , version 3.4 @xcite , using standard settings . a detailed description of the reduction procedures is given in @xcite . the near - infrared spectrum of 2mass j0320@xmath00446 is shown in figure [ fig_nirspec ] , compared to equivalent data for the optical spectral standards vb 10 ( m8 ; @xcite ) and 2mass j14392836 + 1929149 ( l1 , hereafter 2mass j1439 + 1929 ; @xcite ) . despite the poor observing conditions , the data for 2mass j0320@xmath00446 have exceptionally good signal - to - noise , @xmath17150 in the @xmath18 flux peaks and @xmath1950 in the bottom of the 1.4 and 1.8 @xmath1 h@xmath20o bands . color biases due to telluric cloud absorption do not appear to be present , as indicated by comparison of 2mass photometry and synthetic @xmath21 , @xmath22 and @xmath23 colors computed from the spectral data , which agree to within the photometric uncertainties . the morphology of the near - infrared spectrum of 2mass j0320@xmath00446 is typical of a late - type m or early - type l dwarf , with bands of tio and vo absorption at red optical wavelengths ( @xmath24 ) ; prominent h@xmath20o absorption at 1.4 and 1.8 @xmath1 ; feh absorption at 0.99 , 1.2 and 1.55 @xmath1 ; and line absorption in the 1.0 - 1.3 @xmath1 region ; weak lines at 2.2 @xmath1 ; and strong co bandheads at 2.32.4 @xmath1 . for the most part , the spectrum of 2mass j0320@xmath00446 is more consistent with that of 2mass j1439 + 1929 ; note in particular the similarities in the overall shape of the 1.01.35 @xmath1 @xmath14-band flux peak and the deep 1.4 @xmath1 h@xmath20o band . however , tio and vo bands are more similar to ( but weaker than ) those seen in the spectrum of vb 10 , while the weak 2.2 @xmath1 lines are rarely seen in l dwarf spectra ( e.g. , @xcite ) . the near - infrared spectrum of 2mass j0320@xmath00446 is also somewhat bluer than that of 2mass j1439 + 1929 , in line with their respective colors ( @xmath23 = [email protected] versus [email protected] ) . the similarities to 2mass j1439 + 1929 suggests an l1 near - infrared spectral type for 2mass j0320@xmath00446 , which is confirmed by examination of the spectral indices and index / spectral type relations defined by @xcite and @xcite . the average subtype for the four indices k1 ( measuring the shape of the @xmath25-band flux peak ; @xcite ) , h@xmath20o - a , h@xmath20o - b and [email protected] ( all measuring the strength of the 1.4 @xmath1 h@xmath20o band ) yields a near - infrared classification of l1 ( @xmath20.6 subtypes ) , consistent with the l0.5 near - infrared classification reported by @xcite . this classification is fully three subtypes later than the m8 : optical spectral type reported by @xcite . however , such discrepancies are not altogether uncommon amongst late - type dwarfs . @xcite and @xcite have found disagreements of up to 1.5 subtypes between optical ( based on the @xcite scheme ) and near - infrared classifications ( based on their own scheme ) for several l dwarfs . @xcite have discussed a subclass of unusually blue l dwarfs whose optical classifications are consistently 2 - 3 subtypes earlier than their near - infrared classifications . such discrepancies have been variously attributed to surface gravity , metallicity , condensate cloud or multiplicity effects ( e.g. , @xcite ) . the large @xmath13 of 2mass j0320@xmath00446 , indicating that this source may be somewhat older , suggests that high surface gravity and/or slightly subsolar metallicity could explain the discrepant optical and near - infrared spectral classifications . however , 2mass j0320@xmath00446 exhibits one unusual feature not seen in the comparison spectra in figure [ fig_nirspec ] , a slight dip at 1.6 @xmath1 , that suggests multiplicity is relevant in this case . the 1.6 @xmath1 feature is nearly coincident with the 1.57 - 1.64 @xmath1 feh absorption band commonly observed in l dwarf near - infrared spectra @xcite . yet its morphology is clearly different , with a cup - shaped depression as opposed to the flat plateau seen in the comparison spectra of figure [ fig_nirspec ] . more importantly , this feature has the same morphology and is centered at the same wavelength as the peculiar feature noted in the spectrum of sdss j0805 + 4812 @xcite . in that case , the 1.6 @xmath1 feature and other spectral peculiarities were attributed to the presence of a mid - type t dwarf companion . given the similar discrepancy in optical and near - infrared classifications for sdss j0805 + 4812 ( l4 and l9.5 , respectively ) , it is reasonable to consider whether 2mass j0320@xmath00446 also harbors a faint t dwarf companion . in an attempt to search for faint companions , 2mass j0320@xmath00446 was imaged on 2008 january 15 ( ut ) with the sodium lgs ao system @xcite and facility near - infrared camera nirc2 on the 10 m keck telescope . conditions were photometric with average / below - average seeing . the narrow field - of - view camera of nirc2 was utilized , providing an image scale of @xmath26 mas / pixel @xcite over a @xmath27 field of view . all observations were conducted using the mko @xmath28-band filter . the lgs provided the wavefront reference source for ao correction , while tip - tilt aberrations and quasi - static changes were measured contemporaneously by monitoring the @xmath29 mag field star usno - b1.0 0852@xmath00031783 @xcite , located 14 away from 2mass j0320@xmath00446 . the lgs , with an equivalent brightness of a @xmath30 mag star , was pointed at the center of the nirc2 field - of - view for all observations . 2mass j0320@xmath00446 was imaged using two different methods in order to probe the widest possible range of projected separations : ( 1 ) direct imaging and ( 2 ) aperture mask interferometry . in the first case , a series of 3 dithered 60-second images was obtained , offsetting the telescope by a few arcseconds between exposures , for a total integration time of 180 seconds . raw frames were reduced using standard procedures . normalized flat field frames were constructed from the differences of images of the telescope dome interior with and without continuum lamp illumination . a master sky frame was created from the median average of the bias - subtracted , flat - fielded images and subtracted from the individual exposures . individual frames were registered and stacked to form a final mosaic imaged . the observations achieved a point spread function ( psf ) full - width at half - maximum of 007 and a strehl ratio of 0.21 . with the exception of the primary target , no sources were detected in a @xmath31 region centered on 2mass j0320@xmath00446 . aperture mask observations were also obtained with the lgs ao+nirc2 instrumental setup . in this method , a 9-hole aperture mask is placed in a filter wheel near a re - imaged pupil plane within the nirc2 camera . the mask has non - redundant spacing , so each fourier component of the recorded image corresponds to a unique pair of patches on the keck primary mirror . the primary interferometric observables of squared visibility and closure - phase are therefore calibrated much better than images using the full aperture . this technique has a long history of achieving the full diffraction limit of a telescope ( e.g. @xcite ) and has been recently applied to natural guide star ao observations at keck @xcite . this is the first application of aperture mask interferometry to lgs ao observations that we are aware of . 2mass j0320@xmath00446 was observed in this setup using a two - point dither pattern , with five 50-second integrations at each dither position . the nearby field star 2mass j03381363@xmath00332508 , which has a similar @xmath28-band brightness and tip - tilt star asterism as 2mass j0320@xmath00446 , was contemporaneously observed to calibrate both instrumental closure phase and visibility . images of the interferograms formed by the mask were recorded by the nirc2 detector , and squared visibilities and closure - phases were extracted from the image fourier transforms . raw visibility amplitudes were @xmath190.05 on the longest baselines . the closure phases for this calibrator star were subtracted from those of 2mass j0320@xmath00446 , while the calibrator s squared visibilities were divided into those of 2mass j0320@xmath00446 . the one - sigma scatter in the calibrated closure phase was 5@xmath11 . using standard analysis techniques ( e.g. , @xcite ) , we found no evidence of a binary solution in the data . upper limits on the presence of a faint companion to 2mass j0320@xmath00446 were computed separately for the direct imaging and aperture mask observations . for the direct imaging data , upper limits were determined by first smoothing the final mosaic with an analytical representation of the psf s radial profile , modeled as the sum of multiple gaussians . we then measured the standard deviation of flux counts in concentric annuli out to 3@xmath32 in radius centered on the science target , normalized by the peak flux of the science target . we considered 10@xmath33 these values as the flux ratio limits for any companions , as visually verified by inserting fake sources into the image using translated and scaled versions of the science target . for the aperture mask data , detection limits at 99% confidence were calculated in three annuli spanning 0@xmath90200@xmath916 in separation ( the lower limit corresponding to the diffraction limit of the aperture mask ) using the monte - carlo approach described in @xcite . figure [ fig_ao ] displays the resulting flux ratio limits for a faint companion as a function of separation for both datasets . at separations @xmath340@xmath925 , the aperture mask data exclude any companions with @xmath35 3 mag for separations down to 0@xmath904 . note that better seeing , as opposed to longer integrations , would have provided greater improvement in sensitivity in this range . the direct imaging observations exclude any companions with @xmath35 7 mag at separations @xmath36 , with the floor set primarily by sky shot noise and detector read noise . these limits are discussed further in @xmath6 4.1 . as an alternative method to identify and characterize a possible companion to 2mass j0320@xmath00446 , a variant of the binary spectral template matching technique described in @xcite was applied to the near - infrared spectral data . in this method , the spectrum of a late - type source is compared to a large set of binary spectral templates constructed from empirical data for m , l and t dwarfs . the component spectra of each binary template were scaled according to empirical absolute magnitude / spectral type relations . to minimize systematic effects , source and template spectra are required to have the same resolution and wavelength coverage , which is facilitated in this case by using a sample of nearly 200 spex prism spectra of m5t8 dwarfs drawn from the literature and our own unpublished observations . spectral types for the sources in this sample were assigned according to published classifications , based either on the optical classification schemes of @xcite and @xcite for m5l8 dwarfs or the near - infrared classification scheme of @xcite for l9t8 dwarfs ( m and l dwarfs with only near - infrared classifications reported were not included here ) . the initial spectral sample was purged of low signal - to - noise data as well as spectra of those sources known to be binary or noted as peculiar in the literature ( e.g. , low surface gravity brown dwarfs , subdwarfs , etc . ) . this left a sample of 132 spectra of 125 sources , listed in table [ tab_templates ] . to ascertain whether an unresolved binary truly provides a better fit to the spectrum of 2mass j0320@xmath00446 , comparisons were first made to individual sources in the spex sample . all spectra were initially normalized to their peak flux in the 1.21.3 @xmath1 band . the statistic @xmath37 was then computed between the 2mass j0320@xmath00446 ( @xmath38 ) and template spectra ( @xmath39 ) , where @xmath40 ^ 2}{f_{\lambda}(0320)}\ ] ] ( see @xcite ) . the summation is performed over the wavelength ranges @xmath41 = 0.951.35 @xmath1 , 1.451.8 @xmath1 and 2.02.35 @xmath1 in order to avoid regions of strong telluric absorption . the denominator provides a rough estimate of shot noise in the spectral data , which is dominant in the highest signal - to - noise spectra , and therefore makes @xmath37 a rough approximation of the @xmath42 statistic.0446 and template data were taken , we chose not to include this term in our @xmath37 statistic . ] to eliminate normalization biases , each template spectrum was additionally scaled by a multiplicative factor in the range 0.51.5 to minimize @xmath37 . figure [ fig_single ] displays the four best single template matches , all having @xmath43 . the three best - fitting sources lehpm 1 - 6333 ( m8 ) , 2mass j1124 + 3808 ( m8.5 ) and lehpm 1 - 6443 ( m8.5)have optical spectral types consistent with the optical type of 2mass j0320@xmath00446 . the fourth - best fit , the l1 2mass j1493 + 1929 , was shown to provide an adequate match to the spectrum of 2mass j0320@xmath00446 in figure [ fig_nirspec ] . the lehpm sources have large proper motions ( @xmath444 yr@xmath10 ) , notably similar to 2mass j0320@xmath00446 . all four sources shown in figure [ fig_single ] provide reasonably good matches to the broad near - infrared spectral energy distribution of 2mass j0320@xmath00446 , but with two key discrepancies : an absence of the 1.6 @xmath1 feature ( inset boxes in figure [ fig_single ] ) and a shortfall in the peak spectral flux at 1.27 @xmath1 . in the first case , feh absorption bands are clearly seen in the comparison spectra but do not produce the distinct dip seen in the spectrum of 2mass j0320@xmath00446 . in the second case , the spectrum of 2mass j0320@xmath00446 is consistently brighter in the 1.21.35 @xmath1 range as compared to the ( appropriately scaled ) late - type m dwarf templates . as demonstrated below , both of these discrepancies can be resolved by the addition of a t dwarf component . binary spectral templates from the spex prism sample were constructed by first flux - calibrating each spectrum according to established absolute magnitude / spectral type relations . for m5-l5 dwarfs , the 2mass @xmath7/spectral type relation of @xcite was used . for l5-t8 dwarfs , both of the mko @xmath45/spectral type relations defined in @xcite were considered . the liu et al . relations are based on a sample of l and t dwarfs with measured pallaxes and mko photometry , but one relation ( `` bright '' ) was constructed after rejecting known ( resolved ) binaries while the other relation ( `` faint '' ) was constructed after rejecting all known and _ candidate _ binaries as described in that study . as illustrated in figure 3 of @xcite , these two relations envelope the @xmath45 values of currently measured sources ( including components of resolved binaries ) , but diverge by as much as @xmath191 mag for spectral types l8t5 . nevertheless , the @xcite relations represent our current best constraints on the absolute magnitude / spectral type relation across the l dwarf / t dwarf transition . in all cases , synthetic magnitudes to scale the data were calculated directly from the spectra . binary templates were then constructed by adding together the calibrated spectra of source pairs whose types differ by at least 0.5 subclasses , producing a total of 8248 unique combinations . the binary templates were then normalized to their peak flux in the 1.21.3 @xmath1 band and compared to the spectrum of 2mass j0320@xmath00446 in the same manner as the single source templates ; i.e. , with additional scaling to minimize @xmath37 . figure [ fig_double ] displays the best fitting binary templates constructed from the primaries shown in figure [ fig_single ] and using the `` faint '' @xmath45/spectral type relation of @xcite . for all four cases , the addition of a mid - type t dwarf secondary spectrum considerably improves the spectral template match . in particular , the 1.6 @xmath1 spectral dip is very well reproduced , while the flux peaks at 1.27 @xmath1 in the binary templates are more consistent with the spectrum of 2mass j0320@xmath00446 . even detailed alkali line and feh features in the 0.91.3 @xmath1 region are better matched with the binary templates . figure [ fig_double2 ] displays the best fitting binary templates using the `` bright '' @xmath45/spectral type relation of @xcite . there is a small degree of improvement in these fits over those using the `` faint '' @xmath45 relation , although the differences are very subtle due to the very small contribution of light by the t dwarf secondaries ( @xmath46 3.5 mag , depending on the components ) . this result is fortuitous , as it indicates that the better fits provided by the binary templates are only weakly dependent on the absolute magnitude relation assumed over a spectral type range in which such relations are currently most uncertain . besides the best - fit comparisons shown in figures [ fig_double ] and [ fig_double2 ] , there were many excellent matches ( @xmath47 ) found among binaries templates which had lehpm 1 - 6333 or 2mass j1124 + 3808 as primaries : 30 for the `` faint '' @xmath45/spectral type relation and 58 for the `` bright '' relation . the average primary and secondary spectral types for the combinations in this well - matched sample are [email protected] and [email protected] , respectively , with no significant differences between analyses using the `` faint '' or `` bright '' @xmath45/spectral type relations . the mean relative magnitudes of the primary and secondary components were @xmath48 mag , @xmath49 mag , @xmath50 mag for the `` faint '' relation and @xmath51 mag , @xmath52 mag , @xmath53 mag for the `` bright '' relation , as calculated directly from the flux - calibrated spectral templates . there is a large difference in the relative magnitudes between these two relations . if resolved photometry is eventually obtained for this system , such measurements could provide a means of distinguishing which of the absolute magnitude relations proposed in @xcite accurately characterize mid - type t dwarfs . the origin of the 1.6 @xmath1 feature in the spectrum of 2mass j0320@xmath00446 is clearly revealed in figures [ fig_double ] and [ fig_double2 ] : it is a combination of feh absorption in the m dwarf primary and ch@xmath54 absorption in the t dwarf secondary . specifically , the relatively sharp @xmath55-band flux peak in the spectrum of the t dwarf secondary blueward of the 1.6 @xmath1 ch@xmath54 band contributes light to the 1.551.6 @xmath1 spectrum of the composite system . this is on the blue end of the 1.55 - 1.65 @xmath1 feh absorption band , producing a distinct `` dip '' feature . similarly , the apparently brighter 1.21.35 @xmath1 flux in the spectrum of 2mass j0320@xmath00446 can be attributed to the t dwarf companion , which exhibits a narrow @xmath14-band peak between strong 1.1 @xmath1 and 1.4 @xmath1 h@xmath20o and ch@xmath54 bands . both spectral features are therefore unique to binaries containing late - type m and l dwarf primaries ( in which feh is prominent ) and t dwarf secondaries . it may be concluded from the analysis above that the near - infrared spectrum of 2mass j0320@xmath00446 , and in particular the subtle feature observed at 1.6 @xmath1 , can be accurately reproduced by assuming that this source is an unresolved m8.5 + t5 binary . but does this mean that 2mass j0320@xmath00446 actually is a binary ? our lgs ao imaging observations failed to detect any faint secondaries near 2mass j0320@xmath00446 to the limits displayed in figure [ fig_ao ] . based on the `` bright '' mko @xmath45/spectral type relation of @xcite and the @xmath28/@xmath25 filter transformations of @xcite , the measured upper limits rule out a t5 companion wider than a projected separation of @xmath190@xmath933 , or roughly 8.3 au at the estimated distance of 2mass j0320@xmath00446 ( see below ) . this is a relatively weak constraint given that less than 25% of known very low mass binaries have projected separations at least this wide @xcite . furthermore , 2mass j0320@xmath00446 could have been observed in an unfortunate geometry , as was originally the case for the l dwarf binary kelu 1 @xcite . on the other hand , if the physical separation of the 2mass j0320@xmath00446 system is significantly smaller than indicated by the imaging observations , high resolution spectroscopic monitoring could potentially reveal radial velocity signatures , although this depends critically on the component masses of this system . indeed , the determination of a spectroscopic orbit in combination with the component spectral types deduced here would provide both mass and age constraints for this system , making it a potentially powerful benchmark test for evolutionary models . an alternative test of the binary hypothesis for 2mass j0320@xmath00446 is to identify similar spectral traits in a comparable binary system . fortunately , one such system is known : the m8.5 + t6 binary scr 1845@xmath06357 @xcite . this nearby ( [email protected] pc ; @xcite ) , well - resolved binary ( angular separation of 1@xmath91 ) has individually classified components based on resolved spectroscopy @xcite . more importantly , the relative near - infrared magnitudes of this system ( @xmath56 mag , @xmath57 mag , @xmath58 mag ; @xcite ) are somewhat larger than but consistent with the estimated relative magnitudes of the putative 2mass j0320@xmath00446 system . figure [ fig_scr1845 ] displays the component spectra of this system , scaled to their relative @xmath55-band magnitudes ,- band portion of the spectrum of scr 1845@xmath06357a shown here is slightly reduced relative to the @xmath55- and @xmath28-band spectra as shown to figure 2 in @xcite . the relative flux calibration between spectral orders applied in that study did not account for missing data over 1.331.50 @xmath1 , slightly inflating the flux levels in the @xmath14-band . a recalibration of this spectrum was made by scaling each order by a constant factor to match the spex prism spectrum of 2mass j1124 + 3808 , which has a similar @xmath23 color ( [email protected] versus [email protected] for scr 1845@xmath06357 from @xcite ) and optical spectral type ( m8.5 ) . such recalibration is not necessary for the scr 1845@xmath06357b spectrum due to the strong 1.35 @xmath1 ch@xmath54 and 1.4 @xmath1 h@xmath20o bands in this source . the recalibration of the scr 1845@xmath06357a @xmath14-band spectrum does not affect the analysis presented here , which depends solely on the relative @xmath55-band scaling of the component spectra . ] as well as the sum of the component spectra . the composite spectrum shows a relative increase in spectral flux as compared to the primary in both the 1.21.35 @xmath1 and 1.551.6 @xmath1 regions . indeed , the latter gives rise to the same `` dip '' feature observed in the @xmath55-band spectrum of 2mass j0320@xmath00446 , particularly when the scr 1845@xmath06357ab data are reduced in resolution to match that of the spex prism data ( inset box in figure [ fig_scr1845 ] ) . the presence of this feature in the composite spectrum of a known m dwarf plus t dwarf binary lends some confidence to the conclusion that 2mass j0320@xmath00446 is itself an m dwarf plus t dwarf binary . assuming then that 2mass j0320@xmath00446 is a system with m8.5 and t5 dwarf components , it is possible to characterize the physical properties of these components in some detail based on the analysis in @xmath6 3.3 . synthetic component @xmath18 magnitudes on the mko system assuming the `` bright '' @xmath45/spectral type relation of @xcite were computed from the best - fitting binary templates ( @xmath47 ) and are listed in table [ tab_component ] . the m dwarf primary is only slightly fainter than the composite source , while the t dwarf companion is exceptionally faint , @xmath14 = [email protected] mag . the low luminosity of the secondary , @xmath59 = [email protected] dex based on its inferred spectral type @xcite , suggests that 2mass j0320@xmath00446 could have a relatively low system mass ratio ( @xmath60 m@xmath20/m@xmath61 ) . however , the mass ratio depends critically on the age of the system , for which the analysis presented above provides no robust constraints . using the evolutionary models of @xcite and component luminosities as listed in table [ tab_component ] , primary and secondary mass estimates for ages of 1 , 5 and 10 gyr were derived . if 2mass j0320@xmath00446 is an older system , as suggested by its large @xmath13 , its inferred mass ratio @xmath62 @xmath3 0.8 is consistent with the typical mass ratios of very low mass binaries in the field ( e.g. , @xcite ) . based on the primary s photometry and spectral type , and the @xmath7/spectral type relation of @xcite , a distance of 25@xmath23 pc is estimated for the 2mass j0320@xmath00446 system . the subtlety of the peculiar features present in the composite spectra of 2mass j0320@xmath00446 and scr 1845@xmath06357 is due entirely to the considerable difference in flux between their m and t dwarf components . yet in both cases the 1.6 @xmath1 feature , indicating the presence of a t dwarf companion , can be discerned . but for how early of an primary can a binary with a t dwarf companion be identified in this manner , and what variety of t dwarf companions can be discerned in such systems ? to examine these questions , figure [ fig_mtsim ] displays binary spectral templates for four primary types m7 , m8 , m9 and l0combined with t0t8 dwarf secondaries . for all cases , the 1.6 @xmath1 feature is most pronounced when the secondary is a mid - type t dwarf , spectral types t3t5 . this is due to a tradeoff in the sharpness of the @xmath55-band flux peak in this component ( i.e. , the strength of 1.6 @xmath1 ch@xmath54 absorption , which deepens with later spectral types ) and its brightness relative to the primary . not surprisingly , the 1.6 @xmath1 feature is more pronounced in binaries with later - type primaries , making it a useful multiplicity diagnostic for l dwarf + t dwarf systems ( such as sdss j0805 + 4812 ) but far more subtle in systems with m dwarf primaries . indeed , the spectra in figure [ fig_mtsim ] suggest that this feature is basically undetectable in binaries with m7 and earlier - type primaries . 2mass j0320@xmath00446 and scr 1845@xmath06357 probably contain the earliest - type primaries for which a t dwarf secondary could be identified solely from their composite near - infrared spectra . it is also important to consider the other prominent spectral peculiarity caused by the presence of a t dwarf companion , the slight increase in flux at 1.3 @xmath1 . this feature increases the contrast in the 1.4 @xmath1 h@xmath20o band , and therefore serves to bias h@xmath20o spectral indices toward later subtypes . this effect explains why the near - infrared classification of 2mass j0320@xmath00446 is so much later than its optical classification ( the t dwarf secondary contributes negligible flux in the optical ) . figure [ fig_mtsim ] shows that the 1.3 @xmath1 flux increase can be discerned for systems with early- and mid - type t dwarf companions . while it is again more pronounced for systems with later - type primaries , it is still present ( but subtle ) in the spectra of systems with m7 primaries . a source with unusually strong absorption at 1.35 @xmath1 , or equivalently with a near - infrared spectral type that is significantly later than its optical spectral type , may harbor a t dwarf companion . however , other physical effects , notably reduced condensate opacity ( e.g. , @xcite ) , can also give rise to this spectral peculiarity . hence , both the contrast of the 1.4 @xmath1 h@xmath20o band and the presence of the 1.6 @xmath1 dip should be considered together as indicators of an unresolved t dwarf companion . detecting the near - infrared spectral signature of a t dwarf companion need not be limited to low - resolution observations . while the dip feature at 1.6 @xmath1 is less pronounced in the higher - resolution composite spectrum of scr 1845@xmath06357ab from @xcite , individual ch@xmath54 lines may still be distinguishable amongst the many feh and h@xmath20o lines present in the same spectral region . it may also be possible to identify ch@xmath54 lins amongst the forest of h@xmath20o lines in the 1.30 - 1.35 @xmath1 region ( e.g. @xcite ) . such detections require significantly higher resolutions , of order @xmath15 @xmath63 20,000 or more , due to the substantial overlap of the many molecular features present at these wavelengths ( e.g. , @xcite ) . furthermore , an improved line list for the ch@xmath54 molecule may be needed @xcite . yet such observations have the potential to provide an additional check on the existence and characteristics of mid - type t dwarf companions in binaries with late - m / l dwarf primaries . relevant to the identification of late - type m dwarf plus t dwarf binaries from composite near - infrared spectra is the number of such systems that are expected to exist . as a rough estimate , we examined the results of the monte carlo mass function and multiplicity simulations presented in @xcite . using the baseline assumptions of these simulations a mass function that scales as @xmath64 , a component mass range of 0.01 @xmath65 m @xmath65 0.1 m@xmath66 , a flat age distribution over 10 gyr , the @xcite evolutionary models , and a binary mass ratio distribution that scales as @xmath67 ( see @xcite)we found that 12 - 14% of binaries with m8l0 primaries are predicted to contain a t3t5 secondary ; i.e. , detectable with composite near - infrared spectroscopy . these are primarily older systems whose components that just straddle the hydrogen burning minimum mass limit ( @xmath190.07 m@xmath66 ; @xcite ) . the overall binary fraction of very low mass stars and brown dwarfs has been variously estimated to lie in the 1035% range ( e.g. @xcite ) , and is thus currently uncertain by over a factor of three . however , within this range the monte carlo simulations predict that 1 - 5% of _ all _ m8l0 dwarfs harbor a t3t5 dwarf companion . while this percentage is small , in a given magnitude - limited survey there may be a similar number of t dwarf companions in these relatively bright systems as compared to faint , isolated t dwarfs . such companions , based on the analysis above , can be reasonably well - characterized without the need of resolved imaging . there are many other variables that must be considered if the binary spectral template technique described here and in @xcite is to be used to determine accurate binary statistics for very low mass stars and brown dwarfs . component peculiarities , such as unusual surface gravities or cloud variations ; intrinsic scatter in absolute magnitude / spectral type relations ; magnetic- or weather - induced photometric variability ; the detailed properties of the still poorly - constrained l dwarf / t dwarf transition ; and the possible presence of tertiary components all contribute in constraining the variety of systems that can be identified from composite near - infrared spectroscopy . furthermore , because brown dwarfs cool over their lifetimes , the detectability of binaries based on component spectral types does not map uniquely to the detectability of binaries based on their mass ratios and ages , resulting in complex selection biases . these issues will be addressed in a future publication . we have found that subtle peculiarities observed in the near - infrared spectrum of 2mass j0320@xmath00446 , in particular a characteristic bowl - shaped dip at 1.6 @xmath1 , indicate the presence of a mid - type t dwarf companion . this companion is unresolved in lgs ao imaging observations ( including the first application of aperture mask interferometry with lgs ao ) , indicating a maximum projected separation of 8.3 au at the time of observations . the binary scenario not only provides a simple and straightforward explanation for the 1.6 @xmath1 feature also present in the composite spectrum of the known m8.5 + t6 binary scr 1845@xmath06357but also resolves the discrepancy between the optical and near - infrared classifications of 2mass j0320@xmath00446 . furthermore , empirical binary templates composed of `` normal '' m dwarf plus t dwarf pairs provide a far superior match to the overall near - infrared spectral energy distribution of 2mass j0320@xmath00446 than any single comparison source . the hypothesis that 2mass j0320@xmath00446 is an unresolved binary is therefore compelling , and could potentially be verified through radial velocity monitoring observations . in addition , we estimate that roughly 1 - 5% of all late - type m dwarfs may harbor a mid - type t dwarf companion that could similarly be identified and characterized using low resolution near - infrared spectroscopy and binary spectral template analysis . the authors acknowledge telescope operator paul sears and instrument specialist john rayner at irtf , and al conrad , randy campbell , jason mcilroy , and gary punawai at keck , for their assistance during the observations . we also thank markus kasper for providing the spectral data for scr 1845@xmath06357 and sandy leggett , dagny looper and kevin luhman for providing a portion of the spex prism spectra used in the binary spectral template analysis . our anonymous referee provided a helpful and very prompt critique of the original manuscript . mcl and tjd acknowledge support for this work from nsf grant ast-0507833 and an alfred p. sloan research fellowship . this publication makes use of data from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center , and funded by the national aeronautics and space administration and the national science foundation . 2mass data were obtained from the nasa / ipac infrared science archive , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . this research has benefitted from the m , l , and t dwarf compendium housed at dwarfarchives.org and maintained by chris gelino , davy kirkpatrick , and adam burgasser ; the vlm binaries archive maintained by nick siegler at http://www.vlmbinaries.org ; and the spex prism spectral libraries , maintained by adam burgasser at http://www.browndwarfs.org/spexprism . the authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of mauna kea has always had within the indigenous hawaiian community . we are most fortunate to have the opportunity to conduct observations from this mountain . burgasser , a. j. , reid , i. n. , siegler , n. , close , l. m. , allen , p. , lowrance , p. j. , & gizis , j. e. 2007b , in planets and protostars v , eds . b. reipurth , d. jewitt and k. keil ( univ . arizona press : tucson ) , p. 427 wilson , j. c. , miller , n. a. , gizis , j. e. , skrutskie , m. f. , houck , j. r. , kirkpatrick , j. d. , burgasser , a. j. , & monet , d. g. 2003 , in brown dwarfs ( iau symp . 211 ) , ed . e. martn ( san francisco : asp ) , p. 197 llllcl sdss j0000 + 2554 & j00001354 + 2554180 & & t4.5 & [email protected] & * 1*;2 + 2mass j0034 + 0523 & j00345157 + 0523050 & & t6.5 & [email protected] & * 3*;1 + 2mass j0036 + 1821 & j00361617 + 1821104 & l3.5 & l4@xmath21 & [email protected] & * 4*;2,5,6 + hd 3651b & j0039191 + 211516 & & t7.5 & [email protected] & * 7*;8,9,10 + 2mass j0050@xmath03322 & j00501994@xmath03322402 & & t7 & [email protected] & * 11*;1,12 + 2mass j0103 + 1935 & j01033203 + 1935361 & l6 & & [email protected] & * 13*;6 + 2mass j0117@xmath03403 & j01174748@xmath03403258 & l2 : & & [email protected] & * 56*;14 + sdss j0119 + 2403 & j01191207 + 2403317 & & t2 & [email protected] & * 15 * + ipms 0136 + 0933 & j01365662 + 0933473 & & t2.5 & [email protected] & * 4*;16 + 2mass j0144@xmath00716 & j01443536@xmath00716142 & l5 & & [email protected] & * 4*;17 + sdss j0151 + 1244 & j01514155 + 1244300 & & t1 & [email protected] & * 3*;1,18 + 2mass j0205 + 1251 & j02050344 + 1251422 & l5 & & [email protected] & * 19*;6 + sdss j0207 + 0000 & j02074284 + 0000564 & & t4.5 & [email protected] & * 1*;18 + 2mass j0208 + 2542 & j02081833 + 2542533 & l1 & & [email protected] & * 4*;6 + sips j0227@xmath01624 & j02271036@xmath01624479 & l1 & & [email protected] & * 4*;20 + 2mass j0228 + 2537 & j02281101 + 2537380 & l0 : & l0 & [email protected] & * 4*;14,21 + gj 1048b & j02355993@xmath02331205 & l1 & l1 & & * 4*;22 + 2mass j0241@xmath01241 & j02415367@xmath01241069 & l2 : & & [email protected] & * 56*;14 + 2mass j0243@xmath02453 & j02431371@xmath02453298 & & t6 & [email protected] & * 3*;1,23 + sdss j0247@xmath01631 & j02474978@xmath01631132 & & [email protected] & [email protected] & * 15 * + so 0253 + 1625 & j02530084 + 1652532 & m7 & & [email protected] & * 4*;24,25 + denis j0255@xmath04700 & j02550357@xmath04700509 & l8 & l9 & [email protected] & * 1*;26,27 + 2mass j0310 + 1648 & j03105986 + 1648155 & l8 & l9 & [email protected] & * 28*;1,6 + sdss j0325 + 0425 & j03255322 + 0425406 & & t5.5 & [email protected] & * 15 * + 2mass j0328 + 2302 & j03284265 + 2302051 & l8 & l9.5 & [email protected] & * 4*;2,6 + lp 944@xmath020 & j03393521@xmath03525440 & m9 & & [email protected] & * 4 * + 2mass j0345 + 2540 & j03454316 + 2540233 & l0 & l1@xmath21 & [email protected] & * 29*;2,30,31 + sdss j0351 + 4810 & j03510423 + 4810477 & & [email protected] & [email protected] & * 15 * + 2mass j0407 + 1514 & j04070885 + 1514565 & & t5 & [email protected] & * 3*;1 + 2mass j0415@xmath00935 & j04151954@xmath00935066 & t8 & t8 & [email protected] & * 3*;1,23,32 + 2mass j0439@xmath02353 & j04390101@xmath02353083 & l6.5 & & [email protected] & * 28*;14 + 2mass j0510@xmath04208 & j05103520@xmath04208140 & & t5 & [email protected] & * 33 * + 2mass j0516@xmath00445 & j05160945@xmath00445499 & & t5.5 & [email protected] & * 4*;1,34 + 2mass j0559@xmath01404 & j05591914@xmath01404488 & t5 & t4.5 & [email protected] & * 1*;32,35 + 2mass j0602 + 4043 & j06020638 + 4043588 & & t4.5 & [email protected] & * 33 * + lehpm 2@xmath0461 & j06590991@xmath04746532 & m6.5 & m7 & [email protected] & * 4*;36,37 + 2mass j0727 + 1710 & j07271824 + 1710012 & t8 & t7 & [email protected] & * 11*;23,32 + 2mass j0729@xmath03954 & j07290002@xmath03954043 & & t8 & [email protected] & * 33 * + 2mass j0755 + 2212 & j07554795 + 2212169 & t6 & t5 & [email protected] & * 1*;23,32 + sdss j0758 + 3247 & j07584037 + 3247245 & & t2 & [email protected] & * 4*;1,2 + ssspm 0829@xmath01309 & j08283419@xmath01309198 & l2 & & [email protected] & * 38*;39,40 + sdss j0830 + 4828 & j08300825 + 4828482 & l8 & l9@xmath21 & [email protected] & * 4*;18,27 + sdss j0837@xmath00000 & j08371718@xmath00000179 & t0@xmath22 & t1 & [email protected] & * 33*;1,32,41 + 2mass j0847@xmath01532 & j08472872@xmath01532372 & l2 & & [email protected] & * 42*;14 + sdss j0858 + 3256 & j08583467 + 3256275 & & t1 & [email protected] & * 15 * + sdss j0909 + 6525 & j09090085 + 6525275 & & t1.5 & [email protected] & * 15 * + 2mass j0939@xmath02448 & j09393548@xmath02448279 & & t8 & [email protected] & * 1*;12 + 2mass j0949@xmath01545 & j09490860@xmath01545485 & & t2 & [email protected] & * 1*;12 + 2mass j1007@xmath04555 & j10073369@xmath04555147 & & t5 & [email protected] & * 33 * + 2mass j1010@xmath00406 & j10101480@xmath00406499 & l6 & & [email protected] & * 19 * + hd 89744b & j10221489 + 4114266 & l0 & l ( early ) & [email protected] & * 4*;43 + sdss j1039 + 3256 & j10393137 + 3256263 & & t1 & [email protected] & * 15 * + 2mass j1047 + 2124 & j10475385 + 2124234 & t7 & t6.5 & [email protected] & * 4*;1,32,44 + sdss j1048 + 0111 & j10484281 + 0111580 & l1 & l4 & [email protected] & * 4*;45,46 + sdss j1052 + 4422 & j10521350 + 4422559 & & t0.5@xmath21 & [email protected] & * 4*;15 + wolf 359 & j10562886 + 0700527 & m6 & & [email protected] & * 4 * + 2mass j1104 + 1959 & j11040127 + 1959217 & l4 & & [email protected] & * 3*;14 + 2mass j1106 + 2754 & j11061197 + 2754225 & & t2.5 & [email protected] & * 33 * + sdss j1110 + 0116 & j11101001 + 0116130 & & t5.5 & [email protected] & * 11*;1,18 + 2mass j1114@xmath02618 & j11145133@xmath02618235 & & t7.5 & [email protected] & * 11*;1,12 + 2mass j1122@xmath03512 & j11220826@xmath03512363 & & t2 & [email protected] & * 1*;12 + 2mass j1124 + 3808 & j11240487 + 3808054 & m8.5 & & [email protected] & * 3*;14 + sdss j1206 + 2813 & j12060248 + 2813293 & & t3 & [email protected] & * 15 * + sdss j1207 + 0244 & j12074717 + 0244249 & l8 & t0 & [email protected] & * 33*;1,45 + 2mass j1209@xmath01004 & j12095613@xmath01004008 & & t3 & [email protected] & * 3*;1,27 + sdss j1214 + 6316 & j12144089 + 6316434 & & t3.5@xmath21 & [email protected] & * 15 * + 2mass j1217@xmath00311 & j12171110@xmath00311131 & t7 & t7.5 & [email protected] & * 11*;1,32,44 + 2mass j1221 + 0257 & j12212770 + 0257198 & l0 & & [email protected] & * 4*;47 + 2mass j1231 + 0847 & j12314753 + 0847331 & & t5.5 & [email protected] & * 3*;1 + 2mass j1237 + 6526 & j12373919 + 6526148 & t7 & t6.5 & [email protected] & * 48*;1,32,44 + sdss j1254@xmath00122 & j12545393@xmath00122474 & t2 & t2 & [email protected] & * 3*;1,32,44 + 2mass j1324 + 6358 & j13243559 + 6358284 & & t2 & [email protected] & * 33 * + sdss j1346@xmath00031 & j13464634@xmath00031501 & t7 & t6.5 & [email protected] & * 11*;1,32,49 + sdss j1358 + 3747 & j13585269 + 3747137 & & t4.5@xmath21 & [email protected] & * 15 * + 2mass j1404@xmath03159 & j14044941@xmath03159329 & & t2.5 & [email protected] & * 33 * + lhs 2924 & j14284323 + 3310391 & m9 & & [email protected] & * 29 * + sdss j1435 + 1129 & j14355323 + 1129485 & & t2@xmath21 & [email protected] & * 15 * + 2mass j1439 + 1929 & j14392836 + 1929149 & l1 & & [email protected] & * 3*;31 + sdss j1439 + 3042 & j14394595 + 3042212 & & t2.5 & [email protected] & * 15 * + gliese 570d & j14571496@xmath02121477 & t7 & t7.5 & [email protected] & * 3*;1,32,50 + 2mass j1503 + 2525 & j15031961 + 2525196 & t6 & t5 & [email protected] & * 3*;1,32,51 + 2mass j1506 + 1321 & j15065441 + 1321060 & l3 & & [email protected] & * 28*;52 + 2mass j1507@xmath01627 & j15074769@xmath01627386 & l5 & l5.5 & [email protected] & * 28*;2,5,6 + sdss j1511 + 0607 & j15111466 + 0607431 & & t0@xmath22 & [email protected] & * 15 * + 2mass j1526 + 2043 & j15261405 + 2043414 & l7 & & [email protected] & * 3*;6 + 2mass j1546@xmath03325 & j15462718@xmath03325111 & & t5.5 & [email protected] & * 4*;1,23 + 2mass j1615 + 1340 & j16150413 + 1340079 & & t6 & [email protected] & * 33 * + sdss j1624 + 0029 & j16241436 + 0029158 & & t6 & [email protected] & * 11*;1,53 + 2mass j1632 + 1904 & j16322911 + 1904407 & l8 & l8 & [email protected] & * 28*;1,31 + 2mass j1645@xmath01319 & j16452211@xmath01319516 & l1.5 & & [email protected] & * 4*;54 + vb 8 & j16553529@xmath00823401 & m7 & & [email protected] & * 4 * + sdss j1750 + 4222 & j17502385 + 4222373 & & t2 & [email protected] & * 1*;2 + sdss j1750 + 1759 & j17503293 + 1759042 & & t3.5 & [email protected] & * 3*;1,18 + 2mass j1754 + 1649 & j17545447 + 1649196 & & t5 & [email protected] & * 4 * + sdss j1758 + 4633 & j17580545 + 4633099 & & t6.5 & [email protected] & * 11*;1,2 + 2mass j1807 + 5015 & j18071593 + 5015316 & l1.5 & l1 & [email protected] & * 4*;14,21 + 2mass j1828@xmath04849 & j18283572@xmath04849046 & & t5.5 & [email protected] & * 3*;1 + 2mass j1901 + 4718 & j19010601 + 4718136 & & t5 & [email protected] & * 3*;1 + vb 10 & j19165762 + 0509021 & m8 & & [email protected] & * 3 * + 2mass j2002@xmath00521 & j20025073@xmath00521524 & l6 & & [email protected] & * 4*;55 + sdss j2028 + 0052 & j20282035 + 0052265 & l3 & & [email protected] & * 3*;45 + lhs 3566 & j20392378@xmath02926335 & m6 & & [email protected] & * 3 * + 2mass j2049@xmath01944 & j20491972@xmath01944324 & m7.5 & & [email protected] & * 3 * + sdss j2052@xmath01609 & j20523515@xmath01609308 & & t1@xmath21 & [email protected] & * 4,15 * + 2mass j2057@xmath00252 & j20575409@xmath00252302 & l1.5 & l1.5 & [email protected] & * 3*;14,46 + 2mass j2107@xmath00307 & j21073169@xmath00307337 & l0 & & [email protected] & * 3*;14 + sdss j2124 + 0100 & j21241387 + 0059599 & & t5 & [email protected] & * 15*;1,2 + 2mass j2132 + 1341 & j21321145 + 1341584 & l6 & & [email protected] & * 59*;55 + 2mass j2139 + 0220 & j21392676 + 0220226 & & t1.5 & [email protected] & * 1*;56 + hn peg b & j21442847 + 1446077 & & t2.5 & [email protected] & * 10 * + 2mass j2151@xmath02441 & j21512543@xmath02441000 & l3 & & [email protected] & * 56*;55,57 + 2mass j2151@xmath04853 & j21513839@xmath04853542 & & t4 & [email protected] & * 4*;1,58 + 2mass j2154 + 5942 & j21543318 + 5942187 & & t6 & [email protected] & * 33 * + 2mass j2212 + 1641 & j22120345 + 1641093 & m5 & & [email protected] & * 3 * + 2mass j2228@xmath04310 & j22282889@xmath04310262 & & t6 & [email protected] & * 3*;1,34 + 2mass j2234 + 2359 & j22341394 + 2359559 & m9.5 & & [email protected] & * 3 * + sdss j2249 + 0044 & j22495345 + 0044046 & l3 & [email protected] & [email protected] & * 4*;2,18,45 + 2mass j2254 + 3123 & j22541892 + 3123498 & & t4 & [email protected] & * 3*;1,23 + 2mass j2331@xmath04718 & j23312378@xmath04718274 & & t5 & [email protected] & * 3*;1 + 2mass j2339 + 1352 & j23391025 + 1352284 & & t5 & [email protected] & * 1*;23 + lehpm 1@xmath06333 & j23515012@xmath02537386 & m8 & m8 & [email protected] & * 4*;36,40,55 + lehpm 1@xmath06443 & j23540928@xmath03316266 & m8.5 & m8 & [email protected] & * 4*;36,40 + 2mass j2356@xmath01553 & j23565477@xmath01553111 & & t5.5 & [email protected] & * 1*;23 + lccc spectral type & [email protected] & [email protected] & + @xmath68 ( mag ) & [email protected] & [email protected] & [email protected] + @xmath69 ( mag ) & [email protected] & [email protected] & [email protected] + @xmath70 ( mag ) & [email protected] & [email protected] & [email protected] + @xmath59 & [email protected] & [email protected] & [email protected] + @xmath71 ( @xmath32 yr@xmath10 ) & [email protected] & & + @xmath72 ( @xmath11 ) & [email protected] & & + @xmath73 ( pc ) & 25@xmath23 & & + @xmath13 ( km s@xmath10 ) & 67@xmath28 & & + @xmath4 ( au ) & @xmath58.3 ( @xmath50@xmath933 ) & & + mass ( m@xmath66 ) at 1 gyr & 0.081 & 0.035 & 0.44 + mass ( m@xmath66 ) at 5 gyr & 0.086 & 0.068 & 0.79 + mass ( m@xmath66 ) at 10 gyr & 0.086 & 0.074 & 0.86 +

we report on spectropolarimetric observations of a near - ir line of located at 15262.702 whose intensity and polarization profiles are very sensitive to the presence of hyperfine structure . a theoretical investigation of the magnetic sensitivity of this line to the magnetic field uncovers several interesting properties . the most important one is that the presence of strong paschen - back perturbations due to the hyperfine structure produces an intensity line profile whose shape changes according to the absolute value of the magnetic field strength . a line ratio technique is developed from the intrinsic variations of the line profile . this line ratio technique is applied to spectropolarimetric observations of the quiet solar photosphere in order to explore the probability distribution function of the magnetic field strength . particular attention is given to the quietest area of the observed field of view , which was encircled by an enhanced network region . a detailed theoretical investigation shows that the inferred distribution yields information on the average magnetic field strength and the spatial scale at which the magnetic field is organized . a first estimation gives g for the mean field strength and a tentative value of.45 " for the spatial scale at which the observed magnetic field is horizontally organized .

astro-ph0612389

most part of the solar surface is covered by quiet regions that appear non - magnetic in typical solar magnetograms . however , the investigation of such apparently non - magnetic regions is of great scientific interest , and it is important to determine how much photospheric magnetic flux remains hidden from view . the reason is that it might have a significant impact on the magnetic coupling to the outer atmosphere and on chromospheric and coronal heating ( e.g. , the recent reviews by * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? there are at least three possibilities for obtaining empirical information on the magnetism of the quiet " sun ( e.g. , stenflo 1994 ) : the polarization induced by the zeeman effect , the hanle effect and the zeeman broadening of the intensity profiles . these techniques have , of course , their advantages and disadvantages , but the important point is that they can be suitably complemented in order to obtain far richer information on solar surface magnetism than that provided by a high - resolution magnetogram . the detection of polarization produced by the zeeman effect implies the presence of a magnetic field . the main problem with the polarization of the zeeman effect as a diagnostic tool is that it is insensitive to magnetic fields that are tangled on scales too small to be resolved . the solar photosphere is certainly expected to have highly tangled field lines with resulting mixed magnetic polarities on very small spatial scales , well below the current spatial resolution limit ( e.g. , * ? ? ? * ; * ? ? ? ? * ; * ? ? ? * ; * ? ? ? consistently , the analysis of high - resolution observations of the polarization induced by the zeeman effect has led to the conclusion that the portion of the resolution element filled with a non - tangled magnetic field that produces the observed net flux is of the order of 2% ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? although there is a general consensus with this conclusion , lack of agreement exists in the exact probability distribution function of the magnetic field strength associated with such fields that produces a measurable non - zero net flux . the inferred distribution function is different in different works @xcite . this point remains one of the key problems to clarify in the near future because such @xmath1 of the magnetism of the quiet sun might still carry a significant fraction of the total magnetic energy if the probability distribution function ( pdf ) turns out to present an important contribution at kg fields ( as concluded by * ? ? ? * ; * ? ? ? the contradictory results obtained over the last few years suggest that it is fundamental to use a large number of spectral lines when analyzing the magnetic field in unresolved structures ( e.g. , * ? ? ? for example , it has been shown recently @xcite that the information encoded in , for instance , the pair of lines at 630 nm may not be enough to constrain the thermodynamical and magnetic properties of the plasma . this helps to explain why the probability distribution function obtained by several authors give different results . @xcite have pointed out that the exact value of the magnetic field strength inferred from the observations critically depends on the rest of poorly known thermodynamical parameters . as a consequence , it is mandatory to include additional constraints ( ideally obtained from observables ) to obtain reliable magnetic and thermodynamical information from the observations . following this strategy , @xcite have recently considered the polarization of some lines to investigate the magnetism of the quiet sun . apart from thorium , all the elements of the periodic table present at least one isotope with a non - zero nuclear spin . selected species like , and present strong observable perturbations in the stokes profiles produced by the presence of hyperfine structure . since the energy separation between hyperfine levels is much smaller than that between fine structure levels , strong perturbations take place in the line profiles when a magnetic field is present . these perturbations are a consequence of the transition to the intermediate paschen - back regime . @xcite investigated in detail the possibility of detecting hyperfine features in several lines of , locating a spectral line at 5537 that is of diagnostic interest . the advantage of these lines with hyperfine structure is that a purely morphological inspection of the stokes profiles can be used to infer which is the dominant magnetic field strength in the resolution element . in the case of the 5537 line , these features allow to distinguish field strengths above and below @xmath0900 g. although again limited by cancellation effects ( i.e. , magnetic fields tangled at scales below the resolution element can not be detected ) , this diagnostic tool has the advantage that it can be used to get information on the magnetic field strength instead of simply the magnetic flux through the resolution element . obviously , the investigation of the magnetism of the quiet sun can not be done by using only the spectral line polarization induced by the zeeman effect . we need to complement it with other diagnostic tools based on physical mechanisms whose observable signatures are sensitive to the presence of tangled magnetic fields at sub - resolution scales . these are the hanle effect and the zeeman broadening of the intensity profiles . the hanle effect is the modification of the atomic - level polarization ( and of the ensuing observable effects on the emergent stokes profiles ) caused by the action of a magnetic field _ inclined _ with respect to the symmetry axis of the pumping radiation field ( e.g. , the recent reviews by * ? ? ? * ) . the basic formula to estimate the magnetic field intensity , @xmath2 ( measured in g ) , sufficient to produce a sizable change in the atomic level polarization results from equating the zeeman splitting with the natural width ( or inverse lifetime ) of the energy level under consideration , thus @xmath3 . in this expression , @xmath4 and @xmath5 stand , respectively , for the land factor and the level s lifetime ( in seconds ) , which can be either the upper or the lower level of the chosen spectral line . this formula provides a reliable estimation only when radiative transitions dominate the atomic excitation , but its application to the upper and lower levels of typical spectral lines is more than sufficient to understand that the hanle effect may allow us to diagnose solar and stellar magnetic fields having intensities between milligauss and a few hundred gauss , i.e. , in a parameter domain that is very hard to study via the zeeman effect alone . for instance , the sensitivity range of the sr i 4607 line lies between 2 and 200 g , approximately . this means that a volume - filling microturbulent field of 200 g would produce an amount of depolarization similar to that caused by a microturbulent magnetic field of 1000 g. it is the application of the hanle effect in atomic and molecular lines which led @xcite to conclude that there must be a vast amount of hidden magnetic energy and unsigned magnetic flux localized in the ( intergranular ) downflowing regions of the quiet solar photosphere , carried mainly by tangled fields at subresolution scales with strengths between the equipartition field values and @xmath6 kg ( see also * ? ? ? * for understanding the contraints they used for establishing this upper limit of @xmath01 kg ) . the zeeman broadening of the intensity profiles is proportional to the squared modulus of the magnetic field strength , @xmath7 , so that it can also give us information on the presence of tangled magnetic fields at subresolution scales . the problem is that it is extremely difficult to disentangle the zeeman line broadening from that due to the thermal and convective motions . nevertheless , @xcite applied it to hundreds of visible iron lines and could establish via a statistical regresion analysis an upper limit of about 100 g for the case of a volume - filling and single value microturbulent field . an attractive possibility to enhance the sensitivity of this diagnostic tool is to use lines with larger wavelengths . for this reason , asensio ramos & trujillo bueno ( 2006 ; in preparation ) have theoretically investigated the zeeman line broadening technique using the lines at 1.56 @xmath8 m , pointing out that plausible distributions of relatively strong tangled magnetic fields in the intergranular regions of the quiet solar photosphere would produce a measurable zeeman broadening signature in the red wing of the @xmath9 line . the previous summary of the advantages and disadvantages of the various techniques that have been proposed suggests that it would be of great diagnostic interest to identify spectral lines whose intensity profiles show still more sensitivity to the zeeman line broadening effect and whose polarization profiles could be used to distinguish more easily between weak and strong fields . this is precisely the motivation which led us to search for near - ir lines of manganese located at wavelengths that can be observed with the tenerife infrared polarimeter . the aim of this paper is to report on the interesting properties presented by the 15262.702 line in the near - ir . the paschen - back perturbations are so large that it is possible to detect perturbations not only in the circular polarization profiles but also in the intensity profile . since the perturbation of the intensity profile is sensitive to the magnetic field strength , it does not suffer from cancellation effects . consequently , this spectral line can be used to diagnose the net flux and the average magnetic field strength in the resolution element . the comparison of both measurements in a magnetically enhanced region of the quiet sun sheds some light on the complex magnetism of the quiet sun . almost all the elements of the periodic table present an isotope with non - zero nuclear angular momentum @xmath10 . such nuclear angular momentum couples with the sum of the orbital and spin angular momentum @xmath11 . therefore , the fine structure levels characterized by their value of @xmath11 present a splitting due to the precesion of @xmath11 around @xmath10 . the hyperfine splitting is usually much smaller than the fine structure splitting . when a magnetic field is present , its interaction with the atom generates a splitting of the magnetic sublevels @xmath12 belonging to each hyperfine level @xmath13 . since the hyperfine splitting is small , the presence of a magnetic field of weak strength is sufficient to produce zeeman splittings that are of the order of the energy level separation between consecutive @xmath13 levels . as a consequence , non - linear interactions among the magnetic sublevels arise produced by non - diagonal coupling terms in the total hamiltonian . this regime of intermediate paschen - back effect and the total electronic angular momentum @xmath11 is known as the back - goudsmit effect . however , in the physics and astrophysics literature this effect is almost always termed `` the paschen - back effect for the hyperfine structure '' . for this reason , we will follow the standard terminology in the rest of this paper . ] leads to strong perturbations on the zeeman patterns , which have an important impact on the emergent stokes profiles . several lines of species presenting hyperfine structure are located in the near - ir . we have focused on a line that is specially interesting . this line is situated at a wavelength of 15262.702 and is produced by the transition between the fine structure levels @xmath14 . manganese has only one stable isotope , @xmath15mn , which has a nuclear angular momentum @xmath16 . consequently , all the manganese lines present signatures produced by the hyperfine structure @xcite . for zero magnetic field both @xmath11-levels present six @xmath13 levels that arise due to the coupling between @xmath11 and @xmath10 . this follows from the standard rule for angular momentum addition , that gives @xmath17 . the upper fine structure level presents hyperfine levels from @xmath18 to @xmath19 , while the lower fine structure level presents hyperfine levels from @xmath20 to @xmath21 . the energy splitting of these @xmath13 levels with respect to the original @xmath11 level ( the fine structure level without hyperfine structure ) is given with very good approximation by @xcite : @xmath22 where @xmath23 the energy splitting is represented in @xmath24 when the constants @xmath25 and @xmath26 are given in @xmath24 . these constants are the magnetic - dipole and electric - quadrupole hyperfine structure constants , and are characteristic of a given fine structure level . the line under investigation in this work has both levels with strong hyperfine interactions with coupling constants @xmath27 @xmath24 and @xmath28 @xmath24 @xcite . the electric - quadrupole constants are not known with sufficient precision and we have decided to set them to zero . in any case , their influence is always much smaller than that of the magnetic - dipole interaction . in the presence of a magnetic field , these levels suffer from a rapid transition to the intermediate paschen - back regime . therefore , the numerical diagonalization of the total hamiltonian ( hyperfine+magnetic ) turns out to be fundamental . the energy level separation between consecutive fine structure levels is very large in comparison to the typical zeeman splitting produced by the magnetic fields we are interested in . then , it is enough to focus only on the coupling between the hyperfine and magnetic interactions . the total hamiltonian is block - diagonal and each block can be written as ( e.g. , * ? ? ? * ) : @xmath29 where @xmath30 is the bohr magneton , @xmath26 is the magnetic field strength and @xmath31 is the land factor of the level in l - s coupling . according to the nist database , the lower level has @xmath32 and it is known to be in l - s coupling . there is no information about the land factor for the upper level , but according to @xcite , the land factor is @xmath33 , similar to the value given in l - s coupling . if we assume that both levels are well described by l - s coupling , the effective land factor of the line is @xmath34 . when hyperfine structure is neglected , the line presents a weak magnetic sensitivity due to the small value of the effective land factor . however , we will see below that the presence of hyperfine structure produces anomalous stokes profiles of great diagnostic potential . the total hamiltonian is diagonal in @xmath12 , so that it remains a good quantum number even in the presence of a magnetic field . this is not the case with @xmath13 , that looses its meaning as a good quantum number because the total hamiltonian mixes levels with different values of @xmath13 . after a numerical diagonalization of the hamiltonian , the eigenvalues are associated with the energies of the @xmath12 magnetic sublevels . the transition between the upper and lower fine structure levels produce many allowed transitions following the selection rules @xmath35 . the strength of each component can be obtained by evaluating the squared matrix element of the electric dipole operator @xcite : @xmath36 where @xmath37 and @xmath38 are the eigenvectors of the hamiltonian . the symbol @xmath39 is used for labelling purposes since @xmath13 is not a good quantum number ( e.g. , * ? ? ? figure [ fig : splitting ] shows the splitting of the upper and lower levels of the transition at 15262.702 when the magnetic field is increased . note the presence of the six @xmath13 levels for both fine structure levels . when the field is smaller than 100 g , all the hyperfine levels are in the linear zeeman regime , where the splitting is roughly proportional to the magnetic field strength . since the hyperfine levels are so close in energy , interferences among the magnetic sublevels arise and all the hyperfine levels enter into the intermediate paschen - back regime . as a result , the @xmath13 quantum number looses its meaning and the magnetic sub - levels start to interact . the coupling introduced by the magnetic hamiltonian only acts between levels with the same value of @xmath12 belonging to different @xmath13 levels . due to the special structure of the hamiltonian , they anticross having @xmath40 can be different from zero . it can be demonstrated that the elements of the upper and lower diagonals are never zero , so that the matrix has non - degenerate eigenvalues ( e.g. , * ? ? ? * ) . this shows that the energy levels with equal value of @xmath12 can never cross , but anticross . ] . the levels with different values of @xmath12 are those that cross at the values of the magnetic field shown in fig . [ fig : splitting ] . when the field strength is sufficiently strong ( stronger than the ones shown in the figure ) , the sublevels enter into the complete paschen - back regime , where a linear relation between the splitting of the levels and the magnetic field strength is again encountered . the zeeman patterns for two different magnetic field strength are shown in fig . [ fig : zeeman_patterns ] . these figures indicate the splitting and strength of each of the transitions permitted by the selection rules between the two fine structure levels . since the @xmath12 levels of each @xmath13 level have the same energy for @xmath41 g , the zeeman pattern is very simple . when a magnetic field is present , this degeneracy is broken and a large number of components following the @xmath35 selection rules appear . these figures clearly indicate the complexity of the zeeman pattern . the large number of level crossings that we find for magnetic field strengths typical of the solar atmosphere translate into strong perturbations in the observed polarization spectrum . in order to clarify this behavior , we have calculated the synthetic profiles emerging from a milne - eddington atmosphere in which a constant vertical magnetic field pointing away from the observer is present . this field topology is chosen for an easier comparison between the synthetic profiles and the observations we present below . the results are shown in fig . [ fig : milne ] for four values of the magnetic field strength that cover approximately the expected range in the quiet sun : 100 , 600 , 900 and 1300 g. the thermodynamical parameters of the milne - eddington atmosphere are chosen so that the stokes @xmath10 profile observed in weak magnetized regions is approximately fitted . the figure clearly shows that the presence of hyperfine structure and level crossings produce strong perturbations in the emergent stokes profiles . the intensity profile for zero magnetic field contains two lobes . the ratio between the relative absorptions in the two lobes is close to 2 . when the field increases the relative absorption in the two lobes tends to be similar . if the field increases even more , the wavelength distance between both lobes starts to increase . this great sensitivity of the stokes @xmath10 profile to the magnetic field strength is produced by two effects . firstly , the line is in the incomplete paschen - back regime in a range of field strengths typical of the solar atmosphere . this produces perturbations in the zeeman pattern that translates into perturbations in the emergent stokes parameters . secondly , the zeeman splitting @xmath42 is proportional to @xmath43 , while the thermal broadening of the line is proportional to @xmath44 . therefore , since the zeeman broadening increases faster than the doppler broadening , it is possible to detect the separation of the zeeman components in these near - ir lines for relatively weak fields . concerning the stokes @xmath45 profile , we detect the presence of two peaks in the blue lobe of the line when the magnetic field is very weak . additionally , the blue lobe is systematically broader and shallower than the red lobe . this behavior is contrary to the one found from the correlation between stronger fields ( those that produce broader profiles ) and redshifted material ( in the intergranular regions ) . when the field increases , the two peaks in the blue lobe change their relative strength . for fields of the order of @xmath0700 g , the two peaks have the same amplitude and it is impossible to detect them for fields above @xmath01200 g because of the broadening of the profile . for stronger fields , the stokes @xmath45 profile has the typical antisymmetric shape . the amplitude of the stokes @xmath45 lobes saturates at around 800 g , indicating that the line is surely in the weak field regime for fields well below 800 g. for a spectral line without hyperfine structure and in the weak field regime , the amplitude of the stokes @xmath45 signal is proportional to the magnetic flux density @xmath46 , where @xmath47 is a magnetic filling factor and @xmath48 is the longitudinal component of the magnetic field . in the case of a line with hyperfine structure , the intrinsic morphological changes produced by the presence of hyperfine structure can help us distinguish the value of the magnetic field strength itself . it is interesting to analyze the variation of several intrinsic properties of the line for different values of the magnetic field strength . these simple diagnostic tools illuminate the possible diagnostic capabilities of this line . we have focused on two different properties : peak ratios and peak separations . peak ratios are the most conspicuous properties of the line because they present a large variation with the magnetic field strength ( see fig . [ fig : milne ] ) . it is possible to define two different ratios , one for stokes @xmath45 and another one for stokes @xmath10 . their behavior are shown in fig . [ fig : peak_ratio ] . the ratio defined for the blue lobe of stokes @xmath45 is : @xmath49 where @xmath50 is the wavelength of the hyperfine feature in the blue lobe of stokes @xmath45 profile that has a larger wavelength , while @xmath51 is the wavelength of the hyperfine feature that presents a shorter wavelength . this ratio is close to 2 when the field is weak and decreases monotonically when the field increases . as already noted above , the ratio becomes 1 for fields of the order of @xmath0700 - 800 g. the left panel of figure [ fig : peak_ratio ] presents the ratio for fields below 1000 g because the two peaks can not be identified above this value and the ratio is no more reliable . concerning stokes @xmath10 , we have defined the following ratio : @xmath52 this ratio , shown in the right panel of fig . [ fig : peak_ratio ] rapidly decreases when the field increases until arriving at a saturation above @xmath0700 g. stronger fields present a ratio that goes again above 1 . the curve exhibits a minimum , so that the ratio @xmath53 is only reliable for fields below 600 g. assuming a linear functional form for the calibration curve , we can estimate the magnetic field strength using @xmath54 . it is of interest to consider what happens when the magnetic field within the resolution element is chaotic and presents random inclinations and can be considered random or deterministic . ] . in this case , the peak ratio @xmath53 remains quite similar to the one calculated for a vertical magnetic field and shown in the right panel of fig . [ fig : peak_ratio ] . in order to obtain this result , the propagation matrix of the stokes - vector transfer equation has to be averaged over all directions ( see 9.25 of * ? ? ? the non - diagonal terms of the propagation matrix vanish because of the isotropy of the magnetic field and because the absorption coefficient for the intensity is equivalent to that obtained with a magnetic field with an inclination equal to the van vleck angle ( @xmath55 ) . therefore , we use the curve shown in fig . [ fig : peak_ratio ] to estimate the magnetic field from the observations . finally , we have investigated the peak separation in the stokes @xmath10 and stokes @xmath45 profiles . the separation in between the two peaks in the stokes @xmath10 profile is shown in the left panel of fig . [ fig : peak_separation_stokesi ] . this separation remains constant for fields below @xmath0600 g and then increases almost linearly with the magnetic field . this behavior can be understood because the centers of gravity of the @xmath56 and @xmath57 components are linear functions of the magnetic field strength even in the case that hyperfine structure is present ( see 3.5 of * ? ? ? this strong sensitivity of the peak separation to the magnetic field strength can be used to discard between the two possible solutions that are consistent with similar values of the ratio between the peaks of the stokes @xmath10 profile . a similar behavior is found for stokes @xmath45 and it is plotted in the right panel of fig . [ fig : peak_separation_stokesi ] . this is the so - called strong field regime of the zeeman effect in which the peak separation of the two lobes of the stokes @xmath45 profile increases . the plot is limited to fields above @xmath0900 g because the peak separation can not be defined for weaker fields due to the appearance of the hyperfine features . the stokes @xmath45 peak separation also increases linearly . we have performed exploratory spectropolarimetric observations of the spectral region around 15260 with the tenerife infrared polarimeter ( tip ; * ? ? ? * ) mounted on the vacuum tower telescope ( vtt ) at izaa ( tenerife ) . the observations were taken during june 8@xmath58 2002 . we performed a scanning of a magnetically enhanced region of the quiet solar photosphere containing an enhanced network region of circular shape with an internetwork region inside . the seeing conditions were not especially good and the spatial resolution was of the order of @xmath01.4 " . the region was situated very close to the disk center ( @xmath59 ) and three small pores were also in the field of view . the left panel of figure [ fig : totalv ] shows an image taken in the local continuum of the 15262.702 line . the right panel of the figure shows the map of the integrated absolute value of the circular polarization signal . the annular enhanced network region is clearly seen in white . the polarimetric signals of the annular region are quite strong in the line . except for very few points , the linear polarization signals could not be detected . however , we detected stokes @xmath45 profiles produced by the longitudinal component of the magnetic field vector . outside the annulus , the circular polarization signals were very weak . the value of the signal - to - noise in the observed internetwork region is poor . in order to increase it , we have carried out a de - noising procedure based on the pca decomposition of the stokes @xmath45 data . after having obtained the eigenvectors along the directions of maximum covariance , the stokes @xmath45 data has been projected on a manifold of reduced dimension using the first 8 eigenvectors ( those that carry the statistically most relevant information ) . we have verified that this approach highly reduces the noise without introducing strong perturbations in the stokes profiles with respect to the original ones . even after applying the above - mentioned noise reduction technique , the observations do not allow us to investigate in detail the stokes @xmath45 profiles pixel by pixel . therefore , we have focused on a statistical approach by means of a stokes @xmath45 profile classification . in order to perform this classification , we have used a self - organizing map ( som ; * ? ? ? * ) , also known as kohonen s network . the som is a special kind of neural network that is usually applied for visualization and classification purposes ( e.g. , * ? ? ? * ; * ? ? ? it can be considered as a nonlinear mapping between the high - dimensional input space of stokes parameters ( in fact the space spanned by the first 8 eigenvectors of the pca decomposition ) and a regular two - dimensional grid . it consists of a two - dimensional grid of neurons where each neuron is associated with a feature in the original high - dimensional input space ( the stokes profiles observed for all the pixels in our field of view ) , with nearby neurons containing features that present similarities . in a way similar to a pca classification , the som is an unsupervised classification method and it is capable of clustering data in the input space parameter . the pca classification can only search for linear features in the high - dimensional input space . in this sense , the som classification produces much better results because it can detect nonlinearities in the input space . since the classification is performed in an iterative scheme with initially random profiles in all the neurons , the final result typically depends on the number of neurons and on the initial configuration . this is produced by local minima in the learning process . therefore , it is fundamental to carry out the learning process a sufficiently large number of times and verify that the same solution is obtained in a large fraction of such runs . a more detailed description is presented in appendix [ sec : app_som ] . the stokes @xmath45 profiles are first normalized so that each profile has unit length if considered as a vector . we have carried out two different classifications , one with a 4@xmath604 map and another one with a 6@xmath606 map , obtaining quantitatively similar results . for clarity purposes , we only show the results of the 4@xmath604 map . the resulting profiles assigned to each neuron of the map are shown in fig . [ fig : classes_4x4 ] . it presents profiles with the typical antisymmetric shape in one corner , while the profiles become more and more distorted when moving towards the opposite corner . the profile associated with each neuron that the som classification produces may not have physical sense . in order to overcome this problem , we have calculated the stokes @xmath45 profile corresponding to each neuron by averaging all the pixels that have been associated with each class . the results are shown in fig . [ fig : average_profile_4x4 ] . it is immediate to verify that the som has generated a classification in which profiles are ordered according to the field strength ( see fig . [ fig : milne ] ) . in each row , profiles associated with stronger fields ( profiles in which the two peaks in the blue lobe are absent or hardly present ) are located more to the right . the magnetic field decreases for the profiles located to the left because the two peaks in the blue lobe present a value of @xmath53 that increases monotonically . one can argue that this may be produced by the presence of noise . as an argument in favor of the physical significance of the som classification , we can say that it has been carried out without any knowledge of the physical ingredients that produce these hyperfine signatures . as a result , since the final classification shows profiles that follow the trend obtained from the theory ( fig . [ fig : milne ] ) , we are inclined to think that these features are real and not produced by noise . the ratio of the two peaks is smaller than 1 for large fields ( above @xmath0900 g ) and they become larger than 1 for weaker fields . of course , such a one - to - one relation between peak ratio and magnetic field strength is only possible in the case that only one field strength is present within the resolution element . in the more general situation where a combination of magnetic fields is present in the resolution element , this association is not possible . in spite of this , the presence of the hyperfine signature in the blue lobe of the stokes @xmath45 profiles indicates the presence of a field predominantly weaker than @xmath0700 g within the resolution element . in this sense , the 15262.702 line presents a diagnostic potential similar to that of the lines investigated by @xcite . the profiles in the top left part of the map are associated with weak fields in which noise and the presence of mixed polarities plays an important role . for this reason , these profiles are highly distorted . for a fixed column in the som , the vertical variation of the profiles appears to be also related to the strength of the field , with the field decreasing when moving upwards in each column . for comparison purposes , fig . [ fig : classes_4x4 ] indicates the percentage of points in the observed map that are associated with each neuron of the som . a large portion of the field - of - view is covered by the enhanced network region , so that the most abundant profiles are associated with those of stronger fields . in order to better understand the classification carried out by the som , we indicate the location of each class in the observed map in fig . [ fig : location_4x4 ] . the background images in the figures represent ( in grayscale ) the class of each point as obtained with the som . each panel represents one of the rows of the classification network . each color is associated with the profiles belonging to the columns of each row . the neurons of the fourth row of the map are clearly associated with the points belonging to the magnetically enhanced region that include the three pores . the neurons of the first row are clearly associated with the internetwork regions . the neurons belonging to the second and third row are associated with regions located between the network and the internetwork . the classification has detected the presence of ring - like regions that have been classified in the same row suggesting some kind of segregation of the magnetic field strength and a relation between the network and in the internetwork . inside each ring - like structure , it is possible to find a distribution of magnetic fields whose distribution clearly shows smaller values for the inner parts than for the outer parts . a magnetic field can be approximately associated with the profiles of the classes with the aid of the milne - eddington calculation and the calibration curve presented in fig . [ fig : peak_ratio ] . obviously , this can only be done for those profiles that present the hyperfine signature in the blue lobe of the stokes @xmath45 profile . the values used in this work are indicated in table [ tab : peak_ratio_4x4 ] . the profiles for which the hyperfine features can not be easily identified because the profile shows a clear antisymmetric shape are assigned an arbitrary field larger than 1 kg . we do not assign any magnetic field strength to the noisy profiles . assuming only one magnetic component in the resolution element , it is possible to plot maps of the magnetic field strength . figure [ fig : magnetic_field_map ] shows the map of field strength resulting from associating each class with a magnetic field strength . the kg fields have been saturated to 1000 g ( they can be larger , but not smaller ) , while the noisy profiles are set to 0 g. it is interesting to note that the strong field strengths profiles are only placed in the magnetically enhanced region , specially in the pores and the surrounding regions . when we move towards the center of the internetwork region through the network - internetwork separation layer , the field strength rapidly decreases to the sub - kg regime . although the majority of the profiles coming from the network region have been classified as kg , it is possible to clearly identify the shape and position of the peaks of the stokes @xmath45 profiles . consequently , we have calculated the field strength associated with each profile measuring the wavelength separation between the peaks of the blue and red lobes and using the calibration curve presented in the right panel of fig . [ fig : peak_separation_stokesi ] . the results are shown in the right panel of fig . [ fig : magnetic_field_map ] . the strongest fields are associated with the pores , with field strengths above 1600 g and with 2000 g at some points . an estimation of the filling factor can be carried out assuming two components with the same thermodynamics , one magnetic and the other non - magnetic . the amplitude of the peak in the strong field regime arrives to 10 % for a filling factor of 1 ( see fig . [ fig : milne ] ) . the amplitude of the strongest stokes @xmath45 signals ( those of the pores ) is @xmath03 % . this suggests that the maximum filling factor of the kg elements in the pores might be of the order of 30 % . the presence of hyperfine structure allows us to use the intensity profile of the 15262.702 line as a powerful diagnostic tool . according to the milne - eddington results , the line presents two peaks whose ratio is close to 2 for zero magnetic field . this ratio can be easily measured in the observations . the most important source of problems is that the ratio critically depends on the exact value of the continuum . figure [ fig : tip_vs_fts ] shows the comparison between one of the profile in the internetwork regions and the fts atlas . the peak ratio given by the fts atlas is @xmath01.75 , equivalent to a field of the order of @xmath0250 g using the calibration curve shown in fig . [ fig : peak_ratio ] . a comparison between the observed tip spectrum and the fts atlas indicates that the observed stokes @xmath10 spectrum contains several spurious features . these features make it difficult to fix a value for the continuum . following a conservative approach , we have selected two different values of the continuum . these values are indicated with vertical lines in the upper panel of fig . [ fig : ratio_peak_intensity ] . the two values of the continuum are chosen so that they represent an upper and lower limit of the continuum . the weaker continuum ( vertical dotted line ) is chosen so that the largest value of the ratio is smaller than 2 . the reason is that ratios larger than 2 are not compatible with the results presented in fig . [ fig : milne ] . as a consequence , the magnetic field strengths obtained from these ratios represent a lower boundary for the true magnetic field strength distribution . the larger value of the continuum ( vertical dashed line ) , chosen as an upper limit of the continuum , will give smaller ratios and , consequently , larger values of the magnetic field strength . this might be considered as an upper limit for the true magnetic field strength distribution . the upper panel of figure [ fig : ratio_peak_intensity ] shows the value of the ratio of the two peaks of the intensity profile for all the points in the field - of - view for the two values of the continuum . it is clearly seen that the smaller values of the ratios are correlated with the magnetically enhanced region , as expected from the previous considerations . due to the presence of two solutions for peak ratios below @xmath01.14 , the points that fulfill this criterion are indicated in the map with contours . fortunately , this criterion is only fulfilled in those regions of the map that coincide with the pores , where we expect the strongest fields . the smooth variation of the ratio between these strong field regions and the internetwork regions suggests a smooth variation in the distribution of fields . this smooth behavior also discards the noise as the responsible for the variations of the peak ratio . additionally , we have verified that the observed peak separation in stokes @xmath10 is constant over the whole field - of - view , even for the enhanced network and the pores . the profiles of these points present peak ratios close to 1 , which is indicative of moderately strong fields ( of the order or above @xmath0500 g ) . however , since we do not detect any peak separation from the zero - field case ( @xmath61 ) , we have to conclude that the average magnetic field that we are measuring with stokes @xmath10 can not be much stronger than @xmath0600 g. we apply the calibration curve to transform the observed ratios into magnetic field strength . since this measurement of the magnetic field strength has been carried out with stokes @xmath10 , it does not suffer from cancellation effects . therefore , it represents an average value of unknown nature of the magnetic field per resolution element ( see [ sec : flux_cancellation ] ) . the lower panels of fig . [ fig : ratio_peak_intensity ] show the magnetic field strength maps . the maximum value of the magnetic field strength shown in the maps is 700 g. stronger fields are represented with the same color coding . this maximum value are obtained at the points where the peak ratio is smaller than @xmath01.14 . the previous analysis allows us to calculate some statistical properties of the magnetic field traced with the stokes @xmath10 profiles . the left panels of fig . [ fig : histograms ] represent the histogram of the ratios obtained for the whole field - of - view using the two assumed values of the continuum . the behavior is very well represented by an exponential for ratios below @xmath01.6 . these histograms clearly show that the maximum ratio obtained for the smallest value of the continuum is 2 , while the maximum value of the ratio turns out to be smaller when the second value of the continuum is chosen . we point out that almost all the points with ratios below 1.6 ( @xmath030% of the field of view ) are located in the most active region . the points in the internetwork region have systematically ratios above 1.6 . in order to see this behavior in more detail , we have separated the points belonging to the network and those belonging to the internetwork . a pixel is assigned to the network when the maximum of the @xmath62 profile is above 1.6@xmath6010@xmath63 . we have verified that this value provides a good isolation of the network region . the histograms of the network points are shown in figure [ fig : histograms ] as dotted lines while the histograms of the internetwork points are shown as dashed lines . the histogram of the internetwork points peaks at @xmath01.75 and it rapidly falls when moving towards higher ratios . the contribution of the internetwork to the points with ratios smaller than 1.6 is also small . the histogram of the internetwork region closely resembles a gaussian centered at a peak ratio of @xmath01.75 , equivalent to a gaussian centered at a magnetic field strength of @xmath0200 g. concerning the network , the histogram has a less gaussian shape , with a peak at a ratio of @xmath01.5 , equivalent to a magnetic field strength of @xmath0350 g. in an effort to understand the previous histograms , we have separated the internetwork histograms in granules and intergranules . our observations were carried out with quite bad seeing conditions , so that the spatial resolution is about 1.4 " . consequently , the separation of the granular and intergranular regions is not very reliable . the granules are obtained as the set of points whose continuum intensity is larger than 1@xmath56 above the mean continuum intensity of the whole internetwork . the intergranular regions are obtained as the set of pixels whose continuum is smaller than 1@xmath56 below . the percentage of points assigned to granules is 15.8% while 15.3% of the points are classified as intergranules . the results are similar to those obtained by @xcite , who find 18% of granules and 17% of intergranules . the histograms of the peak ratios for these points are plotted in fig . [ fig : histograms_granule_intergranule ] , indicating no apparent difference between granules and intergranules . two possible explanations can be given . firstly , the poor resolution produces a contamination between the light coming from the granular and the intergranular material , thus masking any possible intrinsic difference between them . secondly , it is possible that there is no intrinsic difference and the background field that we measure with the stokes @xmath10 profile is not spatially structured . the only way of investigating this in more detail is by performing observations with much better spatial resolution . from the previous discussion , it is clear that the stokes @xmath45 profiles that emerge from the network regions are associated with a quite strong field . we find profiles in the pores and nearby regions that we associate with fields above 1 kg due to the lack of hyperfine features in the blue lobe and that we confirm to be kg fields due to the peak separation in stokes @xmath45 . however , the ratio of stokes @xmath10 peaks clearly indicates the presence of weak fields with an upper limit of @xmath0700 g. furthermore , we do not detect enhanced stokes @xmath10 peak separations in the pore with respect to what we see in the internetwork regions . an explanation might be the presence of relatively weak fields in the network ( and in the pore ) that have a large filling factor in comparison with the kg fields that are detected by the stokes @xmath45 profile . is it possible to retrieve the underlying probability distribution function of the magnetic field strength from the stokes @xmath10 peak ratio measurements ? is the weak - field tail that is detected both in the internetwork and network histogram real ? does the measurement introduce any bias ? in order to investigate these questions , a numerical experiment has been carried out . we have generated a set of @xmath64 realizations of stokes profiles with different values of a vertical magnetic field . the value of the magnetic field for each realization is selected following a given probability distribution function ( pdf ) . technically , the field strength for each one of the 30000 realizations is obtained by picking a random number @xmath65 following a uniform distribution in the interval @xmath66 $ ] . by an appropriate variable transformation @xmath67 , the resulting numbers @xmath68 turn out to be distributed according to the desired distribution . motivated by the results found in different works @xcite , we have chosen to use exponential and maxwellian pdfs in this numerical experiment . a me synthetic profile is obtained for each value of the field and the ratio between stokes @xmath10 peaks is calculated . this procedure is also repeated by averaging the stokes profiles emerging from @xmath69 consecutive realizations . consequently , we end up with @xmath70 average profiles for which the ratio is obtained . this mimics the loss of spatial resolution in the observation . we synthesized 30000 stokes profiles with a vertical field . the > > strength of the magnetic field strength is chosen to follow > > several probability distribution functions . in our case , we > > focus on an exponential and maxwellian pdfs . technically , the > > field strength for each one of the 30000 realizations is obtained > > by picking a random number x following a uniform distribution in > > the interval [ 0,1 ] . by an appropriate transformation y = f(x ) , the > > resulting numbers y are distributed according to the desired > > distribution . for instance , for an exponential pdf , the number > > of synthetic stokes profiles with weak fields is far larger than > > the number of stokes profiles with strong fields . summarizing , > > the 30000 realizations are in fact representative of the > > underlying pdf that we have assumed . we will focus on the exponential pdfs of the form @xmath71 . in such an exponential pdf , the probability of finding a point with a weak field is far larger than the probability of finding a strong fields . in the averaging process , the weak fields have more weight and the resulting stokes @xmath10 profile presents systematically larger values of the ratio . as a consequence , the pdf retrieved from the ratio will have the fewer strong fields the larger the value of @xmath69 . this is indeed observed in the experiment , as shown in the left panels of fig . [ fig : experiment ] for an exponential pdf with @xmath72 g and different values of @xmath69 . with @xmath73 , we are able to reproduce the slope of the original pdf and the exponential behavior for small fields . when @xmath69 increases , the strong field tail looses weight and moves towards weaker fields , increasing its slope . thus , the slope of the recovered pdf combines information about the average field strength and the resolution . a similar effect is also detected for the weak field tail , which moves towards higher fields when the spatial resolution gets worse . the explanation for this is the following . only for extremely weak fields the peak ratio is close to 2 . larger fields always present smaller ratios . therefore , after averaging several profiles , the ratio will always be smaller than 2 unless all the averaged pixels have @xmath41 . a reduction of the number of profiles showing ratios close to 2 appears and the weak field tail moves towards stronger fields . this behavior dominates when @xmath69 increases because we average more points . it also dominates when @xmath74 increases because it reduces the amount of very weak fields in the averaging process . the two processes together ( reduction of the strong and weak field tails ) tend to transform the retrieved pdf into a gaussian when the spatial resolution is not perfect . this behavior is also found in the experiment carried out with the maxwellian and seems to be a general property of the pdf when the resolution decreases . in the limit of a perfect resolution , it is possible to correctly recover the pdf . in the limit of an infinitely poor resolution , in which we only have one resolution element covering the whole observed area , the retrieved pdf is a dirac delta function centered at the average field of the original pdf . in an intermediate regime , the retrieved pdf tends to a gaussian that finally converges to a dirac delta . fortunately , the average field given by the distorted pdf is still the average field of the original pdf . this is a particularly important conclusion . according to the previous calculations , the histograms presented in fig . [ fig : histograms ] indicate that the field distribution in the internetwork has a pdf whose average is in the range @xmath75 g for the two values of the continuum . the gaussian shape of the observed pdf suggests ( at the light of the previous numerical experiments ) that we are not resolving the fields . if we assume that the true pdf of the internetwork is an exponential , a direct comparison with the numerical experiment tends to suggest that @xmath76 is a plausible value for our observations . striking is the fact that the inferred internetwork pdf is very similar to the one obtained in the numerical experiment for @xmath72 g and @xmath76 . not only the mean value of the field but the individual details of the recovered pdf are quite similar . if we assume that the field distribution does not change with height , the previous result implies that the horizontal scale at which the field is organized in the internetwork has to be of the order of @xmath77 times smaller than our resolution element . for our resolution elements of the order of 1.4``-1.5 '' , this yields a value between 0.44``-0.48 '' . concerning the network field distribution , the peak gives an estimation of @xmath78 g for the mean field strength . in the network , the shape of the pdf has a less gaussian shape than for the internetwork . the fields between 350 g and 550 g still present an exponential decay . this might imply that the scales of the magnetic field in the network are better resolved than in the internetwork . note that the slope of the strong field tail of the network distribution is consistent with what is found in the numerical experiment using @xmath76 for an exponential pdf with @xmath79 g. the previous numerical experiments have been also carried out with a maxwellian pdf . however , the observational results are not so favourably compared with the results obtained with such a maxwellian pdf . in this case , when the spatial resolution worsens , the field distribution converges to a gaussian centered at the mean field of the maxwellian given by @xmath80 . are the previous results in agreement with those obtained by @xcite via analysis of the hanle effect in and c@xmath81 ? with the investigation of the hanle effect in the 4607 line of , @xcite have concluded that , assuming an exponential pdf that does not distinguish between granules and intergranules , the best fit between the synthetic and observed linear polarization profiles is obtained for @xmath82 g. as it can be seen in fig . 1 of trujillo bueno et al . ( 2004 ) , this pdf with @xmath82 g gives a good fit to the inferred hanle depolarization at @xmath83 ( which corresponds to an atmospheric height of @xmath0300 km ) . however , in fig . 1 of @xcite it can also be seen that for explaining the observations at @xmath84 ( which corresponds to an atmospheric height of @xmath0200 km ) , the exponential pdf should have @xmath85 g. our observations are close to disk center and , because the is weak , we expect it to be formed deeper than the line . furthermore , the observations were performed in very quiet regions close to the north solar limb . contrarily , our results are representative of an internetwork area that was encircled by an enhanced network region . therefore , our results support those of @xcite . one of the most interesting diagnostic capabilities of this line is that stokes @xmath10 and stokes @xmath45 are sensitive to different aspects of the magnetic field distribution inside the resolution element . stokes @xmath10 is sensitive to the total field distribution , and gives an idea of the total unsigned flux and the mean field strength in the resolution element . on the contrary , stokes @xmath45 is sensitive to the part of the magnetic field distribution that is not cancelled out , that is , to the net flux . as a consequence , we can estimate the ratio between the net flux and the total flux , thus giving an idea of the amount of flux that cancels out . the net flux in the internetwork region has been obtained with the aid of the som classification . a net flux is assigned to each class of the som classification through the application of the weak - field approximation . we do not take into account those profiles marked as `` noisy '' in the som classification . therefore , the flux cancellation @xmath86 can be estimated directly . on the contrary , since the stokes @xmath45 profiles in the network region are entering into the saturation regime , the amplitude of the @xmath45 profile is no longer proportional to the net flux . however , a rough estimation of the filling factor can be obtained by taking into account that the 15262.702 line is weak . in this case , the emergent stokes parameters are proportional to the absorption coefficients , so that @xmath87 and @xmath88 ( see , e.g. , * ? ? ? * ) , with @xmath89 being the continuum intensity . making the additional assumption that the intensity profile is approximately equal to that obtained for the zero - field case ( no enhanced peak separation is found even in the strong magnetized regions ) , we find that the emergent intensity at line center @xmath90 fulfills @xmath91 . when stokes @xmath45 is approximately described in the strong field regime , its value in one of the lobes can be written as @xmath92 . finally , we obtain the following relation for the filling factor @xcite : @xmath93 where @xmath94 is the field inclination and @xmath90 is assumed to be not modified by the magnetic field . in spite of the strong simplifications assumed for the development of eq . ( [ eq : ff_strong ] ) , we have verified via me synthetic profiles that the previous approximate formula holds approximately for magnetic field strength values greater than @xmath0 1 kg . from the previous estimation of the filling factor , it is possible to calculate the net magnetic flux @xmath95 per resolution element as : @xmath96 where @xmath97 is the area of the resolution element , in our case of the order of 1.4``@xmath601.4 '' . the magnetic field strength @xmath98 has been obtained from the peak separation of the stokes @xmath45 profile ( represented in the the right panel of fig . [ fig : magnetic_field_map ] ) . the left panel of fig . [ fig : net_total_flux ] presents the flux in logarithmic scale . the obtained values are in accordance with the typical values found in the network and in small pores ( e.g. , * ? ? ? these results can be compared against those obtained from the peak ratio in stokes @xmath10 . assuming that the value of the magnetic field obtained from the peak ratio ( we call it @xmath99 ) is representative of a volume filling field ( with filling factor unity ) , the ratio @xmath100 gives us information about the cancellation of magnetic flux in those regions where eq . ( [ eq : ff_strong ] ) can be applied ( i.e. , the network , pores , etc . ) . the ratio between the net and total fluxes is shown in the right panel of fig . [ fig : net_total_flux ] . the network presents ratios always above @xmath010% . this indicates that the cancellation of flux in the network is always below 90% . the variation of the flux cancellation in the network is smooth . the pores present values of the ratios close to 100% , giving the indication that almost all the field in these strongly magnetized regions is in the form of a field with a well established direction . in these regions , the magnetic field strength diagnosed with stokes @xmath10 and stokes @xmath45 coincides . the cancellation ratio gets close to zero when moving towards the internetwork and it presents patches of equal value of the flux density ( always below 10 mx @xmath101 ) . the contour shown in fig . [ fig : net_total_flux ] indicates the separation between the points that belong to the network and those that belong to the internetwork . it is important to remind that , as already mentioned , the net flux in these regions has been obtained following a different approximation . however , it is interesting to point out that a smooth transition between the network and internetwork regions is found . the flux cancellation in the internetwork is such that the ratio between the net flux and the total flux is below 10% for more than 90% of the points . in view of the approximations we have been forced to use due to the low stokes @xmath45 signal of the line , it will be of great help to observe simultaneously this line together with other more magnetically sensitive lines so that the net flux can be estimated with more accuracy . we have presented the results of a theoretical and observational investigation of a line located in the near - ir , which shows very attractive properties for magnetic field diagnostics . the combined effects of the larger zeeman splitting present in the near - ir and the presence of strong perturbations on the zeeman patterns due to the paschen - back effect , makes this line ideal for the diagnostic of magnetic fields in unresolved structures . the hyperfine structure produces modifications in the intensity profile of the line when a magnetic field is present , with a response that covers a large range of magnetic field strengths . the advantage over other zeeman diagnostic tools is that it does not suffer from cancellation effects . we have shown that the analysis of the ratio of the two peaks of the stokes @xmath10 profiles measured in the quietest area of the observed field of view leads to a pdf of gaussian shape that is centered at 250 - 350 g , being unable to distinguish between granules and intergranules with our spatial resolution . this is a value that agrees with typical equipartition values in the photosphere and supports the hanle - effect conclusion of @xcite that there is a substantial amount of hidden magnetic energy and unsigned flux in the quiet sun . a theoretical investigation assuming different pdfs shows that the center of the inferred pdf gives information about the real pdf and that its gaussian shape is related to a degradation of the spatial resolution . although higher signal - to - noise ratio observations are needed for obtaining better stokes @xmath45 profiles , we have shown for the first time a map of how the flux cancels in unresolved magnetic elements . in addition to the presence of pores and a magnetically active network , we must emphasize that the field of view of our observation contains an internetwork region where the net flux seldom amounts to 10% of the total flux . while the spatially average net flux found in very quiet internetwork regions ( e.g. , * ? ? ? * ; * ? ? ? * ) is very close to zero ( see , however , * ? ? ? * ; * ? ? ? * ) , we have a unipolar net flux in our internetwork region in accordance with the polarity of the surrounding network . the numerical experiments on turbulent dynamos and magnetoconvection carried out by cattaneo and coworkers @xcite are initialized by a seed magnetic field that is then tangled up producing a mixed - polarity field with a distribution of field strengths such that the magnetic energy density is a significant fraction of the kinetic energy density . the field topology conserves any imbalance of net flux already present in the initial magnetic seed @xcite . in very quiet regions one can assume a random seed field resulting in a zero net magnetic flux . our observations match , on the other hand , a scenario with an initial seed field with a non - zero net magnetic flux . since the surrounding network of the observed field of view is magnetically enhanced and mostly unipolar , it is apparent that such a unipolar field may be ubiquitous through the encircled internetwork area and will play the role of a seed magnetic field with a non - zero net flux . such a seed field might then be tangled up through dynamo action and/or magnetoconvection into the mixed - polarity fields that we find to constitute more than 90% of the total flux found in the observed internetwork region . the high level of mixing ( 90% at least ) can be now qualitatively compared with simulations of magnetoconvection . particularly instructive is fig . 2 of @xcite where their case 0 ( whose seed field is a completely random field ) results in a 100% cancellation , while their case 4 ( with a seed field with a net flux equivalent to 200 g ) results in a level of mixing smaller than 75% . the measured level of 90% would therefore translate into a seed field of a few g in apparent contradiction with the enhanced field found all around the internetwork area in our data . such an apparent contradiction can be explained in two different ways . the first one is that the actual seed field through the internetwork is actually of just a few g and , therefore , results in a cancellation of 90% as predicted by simulations . the second possibility , and the one we prefer , is that solar magnetoconvection is more successful in mixing fields than present simulations , perhaps just because today s numerical experiments are made with insuficient spatial resolution corresponding to relatively low reynolds number as compared to the real values of the solar photosphere , at least in what concerns the magnetic description . another interesting conclusion is that the uncancelled flux reaches very high percentages ( approaching 100% ) in the pores and other magnetic concentrations over the network . in these regions stokes v pointed toward kg fields , but stokes i unveiled a greater contribution of weak fields under the 700 g threshold . since both components appear to share the same polarity and there is not much cancellation , our analysis suggests that weak fields coexist with kg fields ( carrying most part of the flux ) even in these magnetically - enhanced regions . the potential of this line has to be improved by simultaneously observing it with other lines , like the lines at 525 nm , 630 nm and 1.56 @xmath8 m and the line at 5537 . furthermore , other lines with hyperfine structure can be found in the red and near - ir that can be of great diagnostic potential if they can be observed simultaneously . finally , this ir line may be of great interest for diagnosing magnetic fields in other stars apart from the sun . the line ratio we have developed can be obtained provided one observes with sufficiently high spectral resolution observation and for sufficiently slow rotators . with our line ratio technique , it should be possible to estimate the average magnetic field strength of the star . we thank r. manso sainz for helpful discussions . this article is based on observations taken with the vtt telescope operated on the island of tenerife by the kiepenheuer - institut fr sonnenphysik in the spanish observatorio del teide of the instituto de astrofsica de canarias . this research has been partly funded by the ministerio de educacin y ciencia through project aya2004 - 05792 . self - organizing maps ( som ) are one of the multiple variants of artificial neural networks . they are trained using unsupervised learning techniques and they produce a low - dimensional representation of the training sample . one of the most important characteristics is that the low - dimensional representation tries to preserve the topological properties of the input space ( training samples ) , so that nearby samples are mapped into nearby neurons @xcite . they have been mostly applied to the visualization of high - dimensional data . the self - organizing map consists on a single layer feedforward neural network . the output layer is arranged in a low - dimensional ( usually 2d or 3d ) grid of @xmath102 neurons . the input layer is formed by a set of @xmath103 neurons , where @xmath103 is the dimensionality of the input training space . the network is fully connected , in the sense that every neuron of the output layer is connected to every neuron of the input layer . each neuron of the output layer is associated with a weight vector @xmath104 , i.e. , with the dimensionality of the input space . motivated by how sensory information is handled in different parts of the brain , the goal of the unsupervised learning algorithm is to associate different parts of the map to different parts of the input space . therefore , similar parts of the map will produce larger response when several input data that share a given property is shown to the neural network . the iterative scheme is started by initializing the weights with small random numbers ( it is also possible to initialize by uniformly sampling the subspace generated by the first two largest principal components of the input space ) . the iteration number is accounted for by the discrete index @xmath105 . each iteration needs to repeat the following update rule for each input data @xcite : @xmath106,\ ] ] where @xmath107 is the weight associated with each neuron @xmath39 at iteration @xmath108 , @xmath109 represents the input vector , while @xmath110 is the so - called neighborhood function . the neighborhood function is built in the following way . first of all , the euclidean distances between all the neuron s weights and the input vector are calculated . the neuron with the smallest distance is known as the _ best - matching node _ , whose distance with the input vector is @xmath111 . the neighborhood function is used for propagating the information of the input vector to the surrounding neurons , and it is usually written as : @xmath112 where @xmath113 is the euclidean distance between the weight of neuron @xmath39 an the input vector . the functions @xmath114 ( learning rate ) and @xmath115 ( width of the kernel ) are some monotonically decreasing function of the index @xmath105 that control the spread of the neighborhood kernel . for the initial iterations , the large width of the kernel leads to a convergence in the global scale . when the width of the kernel is reduced , the convergence tends to be more local . after a large number of iterations , convergence of the map is obtained . the weights of the neurons tend to be associated with _ patterns _ in the input data , with similar patterns being located in nearby neurons . finally , it is possible to use the map to classify any input vector , whether it was in the learning set or not . the euclidean distance is calculated between the input vector and all the neuron s weights . the neuron whose weight lies closest to the input vector will give the location in the map of the group in which the input vector has been classified . profile & profile & profile & ( 457 g ) + 0.9895 & 1.0425 & 1.129 & 1.219 + ( 592 g ) & ( 623 g ) & ( 671 g ) & ( 717 g ) + 0.747 & 0.757 & @xmath1161 kg & @xmath1161 kg + ( 421 g ) & ( 429 g ) & & + 0.814 & 0.843 & @xmath1161 kg & @xmath1161 kg + ( 474 g ) & ( 496 g ) & & +

in peripheral relativistic heavy ion collisions at the large hadron collider , jetjet and photonjet final states can be produced when a photon from the virtual photon field surrounding the nucleus interacts with a parton in the opposite nucleus ( direct production ) . the virtual photon may also fluctuate into states with multiple gluons and pairs ( resolved production ) , opening more channels for jet photoproduction . we compare the rates for direct and resolved jetjet and photonjet production to explore the sensitivity to the nuclear and photon parton distribution functions . we calculate the transverse momentum distributions of both partonic jets and leading hadrons produced by jet fragmentation .

hep-ph0407298

we discuss measurements of the nuclear parton distributions via dijet ( jet@xmath0jet ) and @xmath2jet photoproduction in ultraperipheral heavy ion collisions at the large hadron collider ( lhc ) . in these collisions , the accelerated ion is surrounded by a cloud of almost real photons of virtuality @xmath3 where @xmath4 is the nuclear radius . the virtuality , less than @xmath5 for nuclei with @xmath6 , can be neglected . since photon interactions are long range , they can interact with partons in the opposite nucleus even when the nuclei themselves do not interpenetrate . because the photon energies are less than those of the nucleons , these photonuclear interactions have a smaller average center of mass energy than hadronic parton - parton collisions . however , even though the energy is smaller , the coherent photon beams have a flux proportional to the square of the nuclear charge , @xmath7 , enhancing the rates relative to those of photoproduction in @xmath8 collisions at the same energy . thus photoproduction rates in heavy ion collisions can be high . photoproduction occurs two ways in heavy ion collisions : `` direct '' and `` resolved '' production . `` direct '' photoproduction occurs when a photon emitted from one nucleus interacts with a parton from the other nucleus . the photon can also fluctuate into states with multiple @xmath9 pairs and gluons , _ i.e. _ @xmath10 . one of these photon components can interact with a quark or gluon from the target nucleus ( `` resolved '' production ) @xcite . the photon components are described by parton densities similar to those used for protons except that no useful momentum sum rule applies to the photon @xcite . the quark and gluon constituents of the photon open up more channels for jet photoproduction and could , in principle , lead to larger rates for resolved production in certain regions of phase space . in central collisions at rhic , leading particles in jets are easier to detect above the high charged particle multiplicity background than the jets themselves since these high transverse momentum , @xmath11 , particles can be tracked through the detector @xcite . in peripheral collisions , especially at lhc energies , jets should be easier to isolate and may be observed directly using standard high energy jet detection techniques . we thus discuss the @xmath11 distributions of both jets and leading particles . we work at leading order , lo , to avoid any ambiguities such as jet reconstruction and cone size . the ratio of the next - to - leading order , nlo , to lo jet cross sections appears to be relatively constant at high @xmath11 @xcite . we discuss the fragmentation of jets and present the transverse momentum distributions of charged pions , charged kaons and protons / antiprotons . the first hadronic reaction we study is @xmath12 where @xmath13 and @xmath14 are the photon and nucleon four - momenta . the diagrams for direct and resolved dijet hadroproduction are given in fig . [ dijetdia ] . the two parton - level contributions to the jet yield in direct photoproduction are @xmath15 and @xmath16 where also @xmath17 . the produced partons are massless , requiring a minimum @xmath11 to keep the cross section finite . at lo , the jet yield is equivalent to the high @xmath11 parton yield . the jet @xmath11 distribution is modified for photoproduction from _ e.g. _ refs . @xcite , @xmath18 \ , \ , , \end{aligned}\ ] ] where @xmath19 is the fraction of the initial hadron momentum carried by the interacting parton and @xmath20 is the momentum scale of the interaction . recall that the extra factor of two on the right - hand side of eq . ( [ jetdireq ] ) arises because both nuclei can serve as photon sources in @xmath21 collisions . the partonic cross sections are @xmath22 \delta(s + t + u ) \ , \ , , \label{gamapart } \\ s^2 \frac{d^2\sigma_{\gamma q \rightarrow gq}}{dt du } & = & - \frac{8}{3 } \pi \alpha_s(\mu^2 ) \alpha e_q^2 \bigg [ \frac{s^2 + t^2}{s t}\bigg ] \delta(s + t + u ) \ , \ , . \label{qcdcompt}\end{aligned}\ ] ] the first is the photon - gluon fusion cross section , the only contribution to massive @xmath23 photoproduction @xcite , while the second is the qcd compton process . at lo , the partonic cross section is proportional to @xmath24 , where @xmath25 is the strong coupling constant to one loop , @xmath26 is the electromagnetic coupling constant , and @xmath27 is the light quark charge , @xmath28 and @xmath29 . the partonic invariants , @xmath30 , @xmath31 , and @xmath32 , are defined as @xmath33 , @xmath34 , and @xmath35 . in this case , @xmath36 where @xmath37 is the lorentz boost of a single beam and @xmath38 is the proton mass . since @xmath13 can be a continuum of energies up to @xmath39 , we define @xmath40 , analogous to the parton momentum fraction in the nucleon where @xmath41 is the nucleon four momentum . for a detected parton in a nucleon - nucleon collision , the hadronic invariants are then @xmath42 , @xmath43 , and @xmath44 . the produced parton rapidities are @xmath45 and @xmath46 . the jet parton rapidity is related to the invariant @xmath47 by @xmath48 . at lo , @xmath49 and @xmath50 . we calculate @xmath51 and @xmath19 as in an @xmath52 collision and then determine the flux in the lab frame for @xmath53 , equivalent to the center of mass frame in a collider . the photon flux is exponentially suppressed for @xmath54 , corresponding to a momentum fraction @xmath55 . the maximum @xmath56 center of mass energy , @xmath57 , is much lower than the hadronic @xmath58 . the equivalent hadronic invariants can be defined for photon four momentum @xmath13 as @xmath59 , @xmath60 , and @xmath61 @xcite . the partonic and equivalent hadronic invariants for fixed @xmath13 are related by @xmath62 , @xmath63 , and @xmath64 . the photon flux is given by the weizscker - williams method . the flux from a charge @xmath7 nucleus at a distance @xmath65 is @xmath66 \ , \ , \label{wwr}\ ] ] where @xmath67 and @xmath68 and @xmath69 are modified bessel functions . the photon flux decreases exponentially above a cutoff energy determined by the nuclear size . in the lab frame , the cutoff is @xmath70 . in the rest frame of the target nucleus , the cutoff is boosted to @xmath71 . table [ gamfacs ] shows , for @xmath21 collisions at the lhc , the nucleon - nucleon center of mass energies , @xmath58 , the beam energies , @xmath72 , lorentz factors , @xmath37 , maximum photon energies in the lab and rest frames , @xmath73 and @xmath74 respectively , as well as the corresponding maximum @xmath75 center of mass energy , @xmath76 , for single photon interactions with protons . we have also included the @xmath21 luminosities in table [ gamfacs ] to aid the calculation of rates . the total photon flux striking the target nucleus is the integral of eq . ( [ wwr ] ) over the transverse area of the target for all impact parameters subject to the constraint that the two nuclei do not interact hadronically @xcite . this must be calculated numerically . however , an analytic approximation for @xmath21 collisions is given by the photon flux integrated over distances @xmath77 , @xmath78 \ , \ , \label{analflux}\ ] ] where @xmath79 . the difference between the numerical calculation and the analytic results is typically less than 15% except for photon energies near the cutoff . we use the more accurate numerical calculations . the nuclear parton densities @xmath80 can be factorized into @xmath81 and @xmath82 independent nuclear density distributions , position and nuclear - number independent nucleon parton densities , and a shadowing function @xmath83 that describes the modification of the nuclear parton distributions in position and momentum space . then @xcite @xmath84 where @xmath85 is the parton density in the nucleon . we use the mrst lo parton distributions @xcite at @xmath86 . in the absence of nuclear modifications , @xmath87 . the nuclear density distribution , @xmath88 , is a woods - saxon with parameters determined from electron scattering data @xcite . although most models of shadowing predict a dependence on the parton position in the nucleus , we neglect any impact parameter dependence here so that @xmath89 . the nuclear modification , collectively referred to as shadowing here , encompasses three @xmath81 regions : low @xmath81 shadowing where @xmath90 , @xmath91 ; antishadowing where @xmath92 , @xmath93 and the emc region where again @xmath90 but @xmath94 . we use two parameterizations of nuclear shadowing which cover extremes of low @xmath81 gluon shadowing . the eskola _ parameterization , eks98 , is based on the grv lo @xcite parton densities . at the minimum scale , @xmath95 , valence quark shadowing is identical for @xmath32 and @xmath96 quarks . likewise , the shadowing of @xmath97 and @xmath98 quarks are identical at @xmath95 . although the light quark shadowing ratios are not constrained to be equal at higher scales , the differences between them are small . shadowing of the heavier flavor sea , @xmath99 and higher , is calculated separately at @xmath95 . the shadowing ratios for each parton type are evolved to lo for @xmath100 gev and are valid for @xmath101 @xcite . interpolation in nuclear mass number allows results to be obtained for any input @xmath102 . the parameterization by frankfurt , guzey and strikman , fgs , combines gribov theory with hard diffraction @xcite . it is based on the cteq5 m @xcite parton densities and evolves each parton species separately to nlo for @xmath103 gev . although the given @xmath81 range is @xmath104 , the sea quark and gluon ratios are unity for @xmath105 and @xmath95 . the eks98 valence quark shadowing ratios are used as input at @xmath95 since gribov theory does not predict valence shadowing . the parameterization is available for four different values of @xmath102 : 16 , 40 , 110 and 206 . figure [ shadcomp ] compares the two parameterizations for @xmath106 and @xmath107 , 100 and 400 gev . some care must be taken when applying these parameterizations to high @xmath11 since the upper limit of their fit range is 100 gev . the valence ratios , identical for eks98 and fgs at @xmath95 , are somewhat different at high scales because the evolution is different , lo for eks98 and nlo for fgs . the sea quark ratios are rather dissimilar . the eks98 ratios have no antishadowing but , instead , show a peak at @xmath108 when @xmath109 , decreasing smoothly at larger @xmath81 , the emc region . the fgs parameterization has an antishadowing peak at @xmath110 after which @xmath111 for @xmath105 , as at @xmath95 . thus quark shadowing in the emc @xmath81 region will be stronger for eks98 . at the lowest @xmath81 and @xmath20 shown , the fgs gluon ratio shows stronger shadowing and larger antishadowing than eks98 . by @xmath112 gev , the results are rather similar . it is notable that at @xmath95 the fgs gluon ratio is @xmath113 for @xmath105 and decreases below 1 at higher scales , unlike the sea quark ratios , exhibiting an emc - like effect . at high @xmath11 , therefore , the eks98 and fgs gluon ratios should not be significantly different . as we have already pointed out , although the upper limit on the range of both shadowing parameterizations is 100 gev , the parameterizations do not strongly deviate from the trends at lower scales . thus the results can be taken as indicative of the expected behavior . there are a few photon parton distributions available @xcite . the data @xcite can not definitively rule out any of these parton densities . as expected , @xmath114 flavor by flavor because there are no `` valence '' quarks in the photon . the gluon distribution in the photon is less well known . we use the grv - g lo set @xcite . at @xmath115 gev , the difference in results due to the photon parton densities is minimal . the direct jet photoproduction @xmath11 distributions are given in fig . [ jetdir ] for @xmath21 interactions at the lhc . for pb+pb collisions at @xmath116 tev , we show the @xmath11 distributions of the produced quarks , antiquarks and gluons separately , as well as their sum . for ar+ar collisions at @xmath117 tev and o+o collisions at @xmath118 tev , we show only the total @xmath11 distributions . all the results are shown in the rapidity interval @xmath119 . quarks and antiquarks are produced in greatest abundance , with only a small difference at high @xmath11 . photon - gluon fusion alone produces equal numbers of quarks and antiquarks . the quark excess arises from the qcd compton diagram which also produces the small final - state gluon contribution . the @xmath120 contribution grows with @xmath11 since the valence quark distributions eventually dominate production , as shown in fig . [ jetdir](b ) where the @xmath121 contribution is compared to the total . at low @xmath11 , the @xmath121 contribution is @xmath122% of the total , dropping to @xmath123% at @xmath124 gev . at the large values of @xmath81 needed for high @xmath11 jet production , @xmath125 . thus the qcd compton process eventually dominates dijet production , albeit in a region of very low statistics . the @xmath121 contribution is larger for the lighter nuclei since the higher energies reduce the average @xmath81 values . the direct dijet photoproduction cross section is significantly lower than the hadroproduction cross section . some of this reduction is due to the different couplings . the photoproduction rate is reduced by a factor of @xmath126 . there are also fewer diagrams available for jet photoproduction relative to all @xmath127 scatterings in hadroproduction . in addition , the @xmath128 hadroproduction process , with its high parton luminosity , has no direct photoproduction equivalent . since the typical scales of jet production are large , the effects of shadowing , reflected in @xmath129/dp_t)/(d\sigma[s_a^i=1]/dp_t)$ ] , are rather small , see fig . [ jetdir](c ) and ( d ) , because the average @xmath81 is high . the differences between the two shadowing parameterizations are on the few percent level . at low @xmath11 , the produced quarks and antiquarks are mainly from gluons . the produced gluons only come from quarks . the peak for the produced quarks and antiquarks in @xmath130 between @xmath131 gev is due to gluon antishadowing . the total @xmath130 for all produced partons in pb+pb collisions is dominated by the @xmath121 contribution . the maximum value of @xmath132 in the antishadowing region is @xmath133 for eks98 and @xmath134 for fgs , reflecting the high @xmath82 behavior of the shadowing parameterizations . the eks98 ratios for the produced quarks and antiquarks in fig . [ jetdir](c ) follow @xmath130 for the total rather closely over all @xmath11 . the quark and antiquark ratios are slightly above the total at low @xmath11 due to the small @xmath135 contribution . they continue to follow the total at high @xmath11 since all the eks98 ratios exhibit similar behavior at large @xmath81 . the produced gluon ratio follows the quark ratios for @xmath136 gev . the large @xmath11 contribution arises from the valence quarks . some antishadowing is observed at low @xmath11 due to the valence quark contribution . the total ratios for the lighter ions are closer to unity for all @xmath11 due to their smaller @xmath102 . the results are similar for fgs , shown in fig . [ jetdir](d ) , but there are some subtle differences . the ratio @xmath130 for produced gluons , arising from the @xmath135 contribution , exhibits a larger antishadowing effect on @xmath130 because @xmath137 is higher for this parameterization , see fig . [ shadcomp ] . the fgs antiquark shadowing ratio goes to unity for @xmath138 , flattening @xmath130 for antiquarks ( dotted curve ) due to the contribution from @xmath139 . the fgs valence quark ratio is taken from eks98 , resulting in the similarity of @xmath130 in figs . [ jetdir](c ) and ( d ) at high @xmath11 . recall that some care must be taken when applying these parameterizations to high @xmath11 since the upper limit of their fit range is 100 gev . while no extraordinary effects are seen in their behavior beyond this scale , the results should be taken as indicative only . finally , we remark that we have only considered the range @xmath119 . including contributions from all rapidities would increase the effect of shadowing since smaller @xmath81 values can be reached when @xmath140 is large . in fig . [ jet_dir_rap ] , we present the rapidity distributions with two different values of the minimum @xmath11 , @xmath141 gev on the left - hand side and 100 gev on the right - hand side . the results , given by the solid curves , are shown without nuclear shadowing effects . note that , in this case , the photon comes from the left . there is a symmetric case where the photon comes from the right , the factor of two on the @xmath11 distribution in eq . ( [ jetdireq ] ) . in this case , the @xmath45 distribution is reflected around @xmath142 . with the 10 gev cut , the distributions are rather broad at negative @xmath45 where the photon has small momentum and , hence , large flux . at large @xmath143 , corresponding to small @xmath81 for the nucleon momentum fractions and high photon momentum , the distributions fall rapidly since at high photon momenta , the photon flux is cut off as @xmath144 . the distributions with the 100 gev cutoff are narrower because the edge of phase space is reached at lower values of @xmath45 . the rapidity distributions are broader in general for the lighter systems due to the higher @xmath58 . figure [ jet_dir_shad ] gives the ratio @xmath145/dy_1)/ ( d\sigma[s_a^i = 1]/dy_1)$ ] for the two @xmath11 cuts . the ratios reflect the direction of the photon , showing an antishadowing peak at @xmath146 , an emc region at @xmath147 and a shadowing region for @xmath148 for @xmath149 gev , the left - hand side of fig . [ jet_dir_shad ] . the shadowing effect is not large , @xmath150% at @xmath151 for pb+pb collisions and decreasing with @xmath102 . the antishadowing peak is higher for fgs while its shadowing effect is larger at positive @xmath45 , as also noted in the @xmath11-dependent ratios in fig . [ jetdir ] . a comparison of the average effect around @xmath152 with the @xmath11 ratios shown in fig . [ jetdir ] , are in good agreement . even though @xmath19 is smaller for the lighter systems , the shadowing effect is also reduced . since shadowing also decreases with @xmath82 , the effect is even smaller for @xmath153 gev , only @xmath154% at @xmath155 , as shown on the right - hand side of fig . [ jet_dir_shad ] . here the rise at @xmath156 is the fermi motion as @xmath157 . at @xmath158 , the antishadowing region is reached . the total dijet photoproduction cross sections without shadowing and the corresponding rates in a one month , @xmath159 s , lhc run are shown in table [ jetdirrates ] . the rates are based on the luminosities of table [ gamfacs ] . all results are given in the rapidity interval @xmath152 . extended rapidity coverage , corresponding to _ @xmath160 for the cms barrel and endcap systems , could increase the rates by a factor of @xmath161 . ( the increase in rate with rapidity acceptance is not linear in @xmath140 because the rapidity distributions are asymmetric around @xmath142 and increasing @xmath11 narrows the rapidity distribution . the effect of changing the @xmath45 cut is closer to linear at low @xmath11 and larger at high @xmath11 because the peak is at @xmath162 for large @xmath11 , as seen in the @xmath45 distributions on the right - hand side of fig . [ jet_dir_rap ] . ) note that by @xmath163 gev , only a few events are expected per month . however , while high @xmath11 jets are rare , they should be easily observable in a clean environment like photoproduction . there is a difference of @xmath164 in the pb+pb and o+o cross sections at @xmath165 gev , decreasing to less than a factor of four at @xmath166 gev . the difference decreases with @xmath11 due to the larger phase space available at high @xmath11 for the higher @xmath58 systems . note that the rates are nearly the same for all systems because even though @xmath102 is larger for pb+pb , higher luminosities and higher @xmath58 compensates for the lower @xmath102 in lighter systems . we point out that the jet hadroproduction rates are much higher because they are multiplied by @xmath167 for hard processes which increase with the number of binary collisions in @xmath21 collisions . ( the relation is not exact due to shadowing . ) the integration over all impact parameters leads to @xmath167 scaling while there is only a factor of @xmath102 in the dijet photoproduction rate since the photon flux is already integrated over impact parameter for @xmath168 . this , combined with the lower effective energy and fewer channels , considerably reduces the photoproduction rates relative to hadroproduction . we now turn to a description of the final - state particle production in the hadronization of jets . the particle with the highest @xmath11 is called the `` leading '' particle . the corresponding leading particle @xmath11 distribution is @xcite @xmath169 \ , \ , \nonumber\end{aligned}\ ] ] where the @xmath170 on the left - hand side includes all final - state hadrons in addition to @xmath171 but @xmath172 on the right - hand side denotes the unobserved final - state parton . the subprocess cross sections , @xmath173 , are related to @xmath174 in eq . ( [ jetdireq ] ) through the momentum - conserving delta function @xmath175 and division by @xmath176 . the integrals over rapidity have been replaced by an integral over center - of - mass scattering angle , @xmath177 , corresponding to a given rapidity cut . here @xmath178 and @xmath179 covers the full rapidity range while @xmath180 and @xmath181 roughly corresponds to @xmath152 . the fraction of the final hadron momentum relative to that of the produced parton , @xmath182 , appears in the fragmentation function , @xmath183 , the probability to produce hadron @xmath171 from parton @xmath184 . the fragmentation functions are assumed to be universal , independent of the initial state . the produced partons are fragmented into charged pions , charged kaons and protons / antiprotons using lo fragmentation functions fit to @xmath185 data @xcite . the final - state hadrons are assumed to be produced pairwise so that @xmath186 , @xmath187 , and @xmath188 . the equality of @xmath189 and @xmath190 production obviously does not describe low energy hadroproduction well . as energy increases , this approximation may become more reasonable . the produced hadrons follow the parent parton direction . we have used the lo kkp fragmentation functions @xcite . the kkp scale evolution is modeled using @xmath191 data at several different energies and compared to @xmath192 , @xmath193 and @xmath194 data . after some slight scale modifications @xcite all the @xmath195 data could be fit . however , there are significant uncertainties in fragmentation when the leading hadron takes most of the parton momentum @xcite , as is the case here . we assume the same scale in the parton densities and the fragmentation functions , @xmath196 . a larger scale , @xmath197 , has sometimes been used in the parton densities . at high @xmath11 , where @xmath182 is large , the difference in the results for the two scales is small . we have not included any intrinsic transverse momentum broadening in our calculations @xcite . this `` cronin '' effect can be important when @xmath11 is small but becomes negligible for transverse momenta larger than a few gev . the largest contribution to the total final - state charged particle production is charged pions , followed by kaons . the proton contribution is smallest even though at rhic the @xmath198 ratio approaches unity in central and 0.4 in peripheral au+au collisions @xcite for @xmath199 gev . this discrepancy is due to the poor knowledge of the fragmentation functions at large @xmath182 , particularly for baryons . hopefully by the time the lhc begins operation , updated fragmentation functions incorporating @xmath8 data from rhic will be available , allowing better estimates of leading particle production at higher @xmath182 . the corresponding hadron distributions from direct jet photoproduction are shown in fig . [ jethaddir](a ) for @xmath21 collisions . note that the leading hadron cross sections are lower than the partonic jet cross sections , compare figs . [ jetdir](a ) and [ jethaddir](a ) . several factors can account for this . the maximum @xmath57 is a factor of five or more less than @xmath58 for @xmath21 collisions . the reduced number of processes available for direct dijet photoproduction is a significant contribution to the decrease . note also that the @xmath11 distribution is steeper for leading hadrons than for the jets , as may be expected since the effective @xmath11 of the hadron is higher than than of the produced parton . the average @xmath182 for direct photoproduction of high @xmath11 particles is @xmath200 for particles with @xmath201 gev , increasing to @xmath202 for @xmath153 gev . the lower @xmath182 values correspond to lighter ion collisions . in this @xmath182 region , the fragmentation functions are not very well known . as pointed out in ref . @xcite , a small change in the fragmentation function fits can produce significant changes at large @xmath182 . this region is not well constrained by the @xmath185 data used in the fits . the effect of fragmentation on the production channels is shown in fig . [ jethaddir](b ) where we present the fraction of leading hadron production from the @xmath121 channel for all charged hadrons . the ratios are rather similar to those of the partonic jets although the @xmath121 fraction is somewhat smaller due to the larger average @xmath81 of hadron production with respect to jets , as we discuss later . the shadowing ratios for charged pions produced in pb+pb collisions by quarks , antiquarks , gluons and the total from all partons , are shown for the eks98 and fgs parameterizations in fig . [ jethaddir](c ) and ( d ) . the ratios for pion production in ar+ar and o+o collisions are also shown . the high @xmath11 fgs antiquark ratios flatten out relative to the eks98 ratio because the @xmath203 channel dominates gluon production at high @xmath11 . the flattening begins at lower @xmath11 here because the @xmath81 for hadron production is higher than that for the jets . the ratio of pions arising from produced gluons follows the valence ratio , as expected . the ratios decrease with increasing @xmath11 due to the emc effect for @xmath105 when @xmath153 gev . we now discuss the relative values of the nucleon momentum fraction , @xmath81 for parton and hadron production . on the left - hand side of fig . [ avexjet ] , we compare the average @xmath81 values for produced gluon jets ( upper plot ) and for pions produced by these gluons ( lower plot ) . we have chosen to compute the results for produced gluons alone to better compare with resolved jet photoproduction , discussed next . since we are interested in produced gluons , we only consider the qcd compton contribution , @xmath204 . this channel is biased toward larger momentum fractions than @xmath205 since the gluon distribution is largest at small @xmath81 while the valence quark distribution in the proton is peaked at @xmath206 . the average @xmath81 for a gluon jet is @xmath207 at @xmath208 gev , increasing to @xmath209 at 50 gev . the smallest @xmath81 is from the highest energy o+o collisions . the average @xmath81 increases with @xmath11 , to @xmath210 at @xmath211 gev . when final state pions are considered , in the lower left - hand plot , at low @xmath11 , the average @xmath81 is larger than for gluon jets . at @xmath201 gev , @xmath212 while at 50 gev , @xmath213 . at high @xmath11 , however , the average @xmath81 becomes similar for jet and hadron production as @xmath214 increases to @xmath215 at @xmath211 gev . we now turn to resolved production . the hadronic reaction , @xmath216 , is unchanged , but in this case , prior to the interaction with the nucleon , the photon splits into a color singlet state of @xmath1 pairs and gluons . on the parton level , the resolved lo reactions are _ e.g. _ @xmath217 where @xmath81 is the fraction of the photon momentum carried by the parton . the lo processes for resolved photoproduction , illustrated in fig . [ dijetdia](c)-(h ) , are the same as those for lo @xmath218 hadroproduction except that one parton source is a photon rather than a nucleon . the resolved jet photoproduction cross section for partons of flavor @xmath219 in the subprocess @xmath220 in @xmath221 collisions is , modified from refs . @xcite , @xmath222 \ , \ , \label{sigjetres } \end{aligned}\ ] ] where @xmath223 , @xmath224 , and @xmath225 . the @xmath218 minijet subprocess cross sections , @xmath226 , given in the review article by owens @xcite , are related to @xmath227 through the momentum - conserving delta function @xmath228 . the sum over initial states includes all combinations of two parton species with three flavors while the final state includes all pairs without a mutual exchange and four flavors ( including charm ) . the factor @xmath229 accounts for identical particles in the final state . the resolved jet results , shown in fig . [ pjetres ] , are independent of the photon parton densities for @xmath115 gev . along with the total quark , antiquark and gluon cross sections in pb+pb collisions , we also show the individual partonic contributions to the jet @xmath11 distributions . the produced gluon contribution dominates for @xmath230 gev but , by 50 gev , quark and antiquark production becomes larger due to the increase of the @xmath231 channel relative to the @xmath128 channel . we also show the total @xmath11 distributions for ar+ar and o+o collisions . for lighter nuclei , the crossover between gluon and quark / antiquark dominance occurs at higher @xmath11 due to the higher collision energy . note that the resolved dijet photoproduction contribution is larger than the direct by a factor of two to three for @xmath232 gev , despite the lower effective center of mass energy of resolved production . resolved production opens up many more channels through the parton components of the photon . indeed , now all the @xmath218 channels for lo jet hadroproduction are available . in addition , the quark and antiquark distributions in the photon are the same . these distributions are large at high momentum fractions , higher than the quark and antiquark distributions in the proton . thus the quark and antiquark channels are enhanced relative to hadroproduction . the largest difference between the quark and antiquark production rates is due to the difference between the valence and sea quark distributions in the nucleus . where the valence and sea quark contributions are similar , as for @xmath119 , the difference is rather small . if all rapidities were included , the relative quark and antiquark rates could differ more . the direct and resolved rapidity distributions are compared in fig . [ jet_dir_rap ] for the two @xmath11 cuts , 10 and 100 gev . while the @xmath119 resolved contribution is a factor of two to three larger than the direct at @xmath233 gev , a comparison of the @xmath45 distributions over all rapidities shows that the resolved contribution can be considerably larger , a factor of @xmath234 at @xmath156 for @xmath235 gev . at @xmath153 gev , the resolved contribution is still equivalent to or slightly larger than the direct at @xmath236 but drops below at larger rapidities . thus , going to higher @xmath11 can separate direct from resolved production , especially at forward rapidities . recall that the produced gluons dominate resolved production at @xmath230 gev while they are only a small contribution to direct production . the largest gluon production channels are typically @xmath128 and @xmath237 . as @xmath45 becomes large and negative , the photon @xmath81 decreases while @xmath19 of the nucleon increases , leading to the dominance of the @xmath238 channel . the photon gluon distribution is largest as @xmath81 decreases . the valence quark distribution of the proton is also important at high @xmath11 , causing the resolved to direct ratio to flatten for @xmath239 when @xmath240 gev . in fig . [ jet_dir_shad ] , we compare the direct and resolved shadowing ratios , @xmath241 . a smaller shadowing effect is observed for the resolved component due to the larger @xmath19 for resolved production . the difference in the direct and resolved shadowing ratios is reduced for larger @xmath11 . to measure the nuclear parton densities most directly , it is preferable for direct production to be dominant . however , fig . [ jet_dir_rat ] shows that a @xmath11 cut is not very effective for dijet production , as previously mentioned , even at forward rapidity . the resolved to direct production ratios are all larger than unity for @xmath141 gev , even for large , positive @xmath45 . while the ratio is less than 1 for @xmath239 and @xmath242 gev , it is only @xmath243 for pb+pb , increasing to 0.8 for o+o . thus , although clean separation is possible at @xmath153 gev , precision parton density measurements are not possible at these values of @xmath11 due to the low rate . other means of separation must then be found . resolved processes will not be as clean as direct in the direction of the photon due to the breakup of the partonic state of the photon . the multiplicity in the photon fragmentation region will be higher than in direct production where the nucleus should remain intact . a cut on multiplicity in the photon direction may help separate the two so that , although there should be a rapidity gap for both direct and resolved photoproduction , the gap may be less prominent for resolved production . the total resolved dijet photoproduction cross sections without shadowing and the corresponding monthly ( @xmath159 s ) lhc rates are given in table [ jetresrates ] . at low @xmath11 , the cross sections and rates are a factor of @xmath244 higher than for direct dijet production . the largest increase is for the lightest nuclei since the lowest @xmath81 values are probed . however , with increasing @xmath11 , the phase space is reduced . the average photon momentum is increased and , at large photon momentum , the flux drops faster . the average momentum fractions probed by the nuclear parton densities grows large and only valence quarks contribute . the lower effective energy of resolved relative to direct photoproduction reduces the high @xmath11 phase space for resolved production . thus , at the highest @xmath11 , the rate is reduced relative to direct by a factor of @xmath245 with the smallest decrease for the lightest system due to the higher effective @xmath58 . since resolved production has a narrower rapidity distribution than direct production , increasing the rapidity coverage would not increase the rate as much as for direct photoproduction . we show the individual partonic shadowing ratios for pb+pb collisions with the eks98 parameterization in fig . [ pjetres](b ) . the quark and antiquark shadowing ratios are very similar although the quark ratio becomes larger for higher @xmath11 ( higher @xmath81 ) values of @xmath81 due to the valence distribution . now the gluon ratio shows larger antishadowing since gluon production is dominantly through the @xmath246 and @xmath247 channels rather than @xmath135 in direct production , compare fig . [ jetdir ] . the fgs parameterization gives similar results , fig . [ pjetres](c ) . however , since the small fgs gluon antishadowing is stronger , @xmath130 is larger for @xmath248 gev . we now turn to leading particles from resolved jet photoproduction . the leading particle @xmath11 distributions are @xmath249 \frac{d_{h / k}(z_c,\mu^2)}{z_c } \ , \ , . \label{sighadres } \end{aligned}\ ] ] the subprocess cross sections , @xmath226 , are related to @xmath250 in eq . ( [ sigjetres ] ) through the momentum - conserving delta function @xmath251 and division by @xmath252 . the drop in rate between jets and high @xmath11 hadrons is similar to that in direct photoproduction , as can be seen by comparison of figs . [ pjetres ] and [ jethadres ] relative to figs . [ jetdir ] and [ jethaddir ] . now that gluon fragmentation makes a larger contribution , the relative pion contribution is larger than in direct photoproduction while the relative proton contribution is significantly reduced . the smaller effective center of mass energy for resolved photoproduction lowers the phase space available for fragmentation . baryon production is then reduced compared to light mesons . the reduction in phase space for leading hadrons relative to fast partons can be seen in the comparison of the average @xmath81 values for resolved photoproduction of jets and leading hadrons , shown on the right - hand side of fig . [ avexjet ] for gluons and pions from gluons respectively . at @xmath201 gev , the average @xmath81 of the gluon jet is @xmath253 , increasing to @xmath254 at @xmath255 gev , higher than for direct photoproduction , as expected . the @xmath81 values for hadron production are larger still , @xmath256 at low @xmath11 while @xmath257 at @xmath255 gev . the shadowing ratios in fig . [ jethadres ] also reflect the increasing @xmath81 . now the antishadowing peak is shifted to @xmath258 gev since the average @xmath81 values are in the emc region , even at low @xmath11 . the values of @xmath130 at high @xmath11 are somewhat lower than for direct production due to the higher @xmath81 . the average @xmath182 of the fragmentation functions is also somewhat larger for resolved production , @xmath259 at @xmath124 gev . since the resolved jet cross section is larger than the direct at low @xmath11 , it is more difficult to make clean measurements of the nuclear gluon distribution unless the two contributions can be separated by other methods . instead , the large valence quark contribution at high @xmath11 suggests that jet photoproduction can probe the nuclear valence quark distributions at larger @xmath82 than previously possible . at @xmath260 gev , more than half of direct jet production is through the @xmath135 channel . unfortunately , the rates are low here , making high precision measurements unlikely . however , the events should be very clean . a clean method of determining the quark distribution in the nucleus at lower @xmath11 is the process where a jet is produced opposite a photon in the final state , compton scattering in direct production . the cross sections are reduced relative to the jet@xmath0jet process since the coupling is @xmath261 in the coupling rather than @xmath262 , as in dijet production . in addition , the quark distributions are lower than the gluon , also reducing the rate . the diagrams for @xmath2jet production are shown in fig . [ comptdia ] . we now discuss the jet and leading particle distributions for direct and resolved @xmath263jet photoproduction . the hadronic process is @xmath264 . the only partonic contribution to the @xmath2jet yield in direct photoproduction is @xmath265 ( or @xmath266 ) where the produced quark is massless . we now have @xmath267 \ , \end{aligned}\ ] ] where the partonic cross section for the compton process is @xmath268 \delta(s + t + u ) \ , \ , .\end{aligned}\ ] ] the extra factor of two on the right - hand side of eq . ( [ compdireq ] ) again arises because both nuclei can serve as photon sources in @xmath21 collisions . the kinematics are the same as in jet@xmath0jet photoproduction , described in the previous section . the direct @xmath263jet photoproduction results are given in fig . [ compdir ] for @xmath21 interactions at the lhc . we show the transverse momentum , @xmath11 , distributions for all produced quarks and antiquarks in pb+pb , ar+ar and o+o collisions for @xmath119 . there is a drop of nearly three orders of magnitude between the dijet cross sections in fig . [ jetdir ] and the @xmath2jet cross sections in fig . [ compdir ] . most of this difference comes from the relative couplings . the rest is due to the reduced number of channels available for direct @xmath2jet production since more than half of all directly produced are gluon - initiated for @xmath269 gev , see fig . [ jetdir](b ) . we have not distinguished between the quark and antiquark initiated jets . however , the quark - initiated jet rate will always be somewhat higher due to the valence contribution . when @xmath269 gev , the quark and antiquark jet rates are very similar since @xmath81 is still relatively low . at higher @xmath11 , the valence contribution increases so that when @xmath270 gev , the quark rate is 1.5 times the antiquark rate . since the initial kinematics are the same for @xmath263jet and jet+jet final states , the average momentum fractions for @xmath263jet production are similar to those shown for the @xmath204 channel in fig . [ avexjet ] . the shadowing ratios shown in fig . [ compdir](b ) are dominated by valence quarks for @xmath153 gev . the fgs ratio is slightly higher because the eks98 parameterization includes sea quark shadowing . the effect is similar to the produced gluon ratios , at the same values of @xmath81 in fig . [ jetdir](c ) and ( d ) , since the final - state gluons can only come from quark and antiquark induced processes . we next present the rapidity distributions for the same two @xmath11 cuts used for dijet photoproduction in fig . [ compt_rap ] . note that the rapidity distribution for @xmath141 gev is broader at negative @xmath45 than the dijet distributions in fig . [ jet_dir_rap ] because direct dijet production is dominated by @xmath205 at these @xmath11 while the valence distribution entering the @xmath271 does not drop as rapidly at large @xmath19 as the gluon distribution . when the turnover at large negative @xmath45 occurs , it is steeper than for the dijets . however , it drops even more steeply at forward @xmath45 because the quark distribution is smaller than the gluon at low @xmath19 . when @xmath153 gev , the @xmath272jet @xmath45 distribution is narrower than the dijets since the quark distributions drop faster with increasing @xmath19 at high @xmath11 . the shadowing ratios as a function of @xmath45 are shown in fig . [ compt_shad ] . they exhibit some interesting differences from their dijet counterparts in fig . [ jet_dir_shad ] because of the different production channels . at @xmath141 gev , the antishadowing peak is lower at @xmath273 and the shadowing is larger at @xmath274 . although this may seem counterintuitive , comparing the valence and sea quark shadowing ratios in fig . [ shadcomp ] can explain this effect . valence antishadowing , the same for eks98 and fgs , is not as strong as that of the gluon . the sea quarks have either no antishadowing ( eks98 ) or a smaller effect than the valence ratios ( fgs ) . thus antishadowing is reduced for direct @xmath2jet production . at large @xmath45 , the @xmath19 values , while smaller than those shown in fig . [ jet_dir_shad](a ) for @xmath275 , are still moderate . since the evolution of the gluon distribution is faster with @xmath82 , sea quark shadowing is actually stronger than gluon shadowing at @xmath141 gev and low @xmath19 . when @xmath153 gev , the fermi momentum peak is not as prominent because the sharp increase in the valence and sea shadowing ratios appears at higher @xmath19 than for the gluons , muting the effect , particularly for the lighter systems . the cross sections and rates in a one month , @xmath159 s , run , shown in table [ comdirrates ] , are lower than those for hadroproduction . direct @xmath263jet photoproduction proceeds through fewer channels than hadroproduction where the lo channels are @xmath276 and @xmath277 , the same diagrams for resolved @xmath2jet photoproduction . this , along with the lower effective energy and correspondingly higher @xmath81 , reduces the photoproduction cross sections relative to hadroproduction . the lower @xmath102 scaling for photoproduction also restricts the high @xmath11 photoproduction rate . we now turn to a description of final - state hadron production opposite a photon . the leading particle @xmath11 distribution is @xcite @xmath278 \ , \ , \nonumber\end{aligned}\ ] ] where @xmath170 on the left - hand side includes the final - state gluon . on the partonic level , both the initial and final state partons are identical so that parton @xmath279 fragments into hadron @xmath171 according to the fragmentation function , @xmath280 . the subprocess cross sections , @xmath173 , are related to @xmath174 in eq . ( [ compdireq ] ) through the momentum - conserving delta function @xmath175 and division by @xmath176 . our results , shown in fig . [ comphaddir ] , are presented in the interval @xmath152 . the cross section for @xmath2hadron production are , again , several orders of magnitude lower than the dijet calculations shown in fig . [ jethaddir](a ) . at the values of @xmath182 and @xmath81 important for dijet production , the final state is dominated by quarks and antiquarks which fragment more frequently into charged hadrons than do gluons . while @xmath205 produces quarks and antiquarks with identical distributions , the contribution from the @xmath281 channel makes _ e.g. _ pion production by quarks and antiquarks asymmetric . we also note that for @xmath269 gev , 60% of the dijet final state particles are pions , @xmath282% kaons and @xmath283% protons . as @xmath11 increases , the pion and proton contributions decrease slightly while the kaon fraction increases . in the case of @xmath263hadron final states , there is no initial state gluon channel . thus the valence quarks dominate hadron production and the relative fraction of produced pions increases to 66% . the kaon and proton fractions are subsequently decreased to @xmath284% and @xmath285% respectively . the shadowing ratios , shown in fig . [ comphaddir](b ) and ( c ) for produced pions , kaons and protons separately for pb+pb as well as the total ratios for ar+ar and o+o collisions , reflect the quark - initiated processes . we show the results for all charged hadrons here since we do not differentiate between quark and antiquark production . the ratios , almost identical for produced pions , kaons and charged hadrons , are quite different from the ratios shown for pion production by quarks and antiquarks in fig . [ jethaddir](c ) and ( d ) since these pions originate from initial - state gluons and thus exhibit antishadowing . the results are similar to pions from gluon jets in fig . [ jethaddir ] . however , in this case the ratios are slightly higher due to the relative couplings . the proton ratios are lower than those for pions and kaons due to the nuclear isospin . the dominance of @xmath96 valence quarks in nuclei reduces the proton production rate since @xmath96 quarks are only half as effective at producing protons as @xmath32 quarks in the kkp fragmentation scheme @xcite . this lower weight in the final state reduces the effectiveness of proton production by the initial state , decreasing the produced proton shadowing ratios relative to pions and kaons . valence quarks dominate the observed final state shadowing at these larger values of @xmath81 , as in fig . [ avexjet ] . now we turn to resolved production of @xmath263jet final states . the resolved jet photoproduction cross section for partons of flavor @xmath219 in the subprocess @xmath286 in @xmath221 collisions is modified from eq . ( [ sigjetres ] ) so that now @xmath287 \ , \ , . \label{sigcompres } \end{aligned}\ ] ] the resolved diagrams are those for hadroproduction of direct photons , @xmath288 and @xmath289 . the @xmath218 minijet subprocess cross sections are @xcite @xmath290 \delta ( \hat s + \hat t + \hat u ) \ , \ , \\ \label{qqbtogamg } \hat s^2 \frac{d^2\sigma_{q \overline q}}{d\hat t d\hat u } & = & \frac{8}{9 } \pi \alpha_s \alpha e_q^2 \bigg [ \frac{\hat t^2 + \hat u^2}{\hat t \hat u } \bigg ] \delta(\hat s + \hat t + \hat u ) \ , \ , .\end{aligned}\ ] ] note that there is no factor @xmath229 , as in eq . ( [ sigjetres ] ) , since there are no identical particles in the final state . the resolved @xmath2jet results are shown in fig . [ compjetres ] using the grv lo photon parton densities . along with the total partonic rates in pb+pb collisions , we also show the individual partonic contributions to the jet @xmath11 distributions in fig . [ compjetres](a ) . the total yields are slightly higher for the resolved than the direct contribution where only one channel is open and the coupling is smaller . quark and antiquark production by the @xmath247 process is dominant for @xmath291 gev but , at higher @xmath11 , gluon production dominates from the @xmath1 channel . the large values of @xmath81 again makes the valence quark contribution dominant at higher @xmath11 . the total @xmath11 distributions for ar+ar and o+o collisions are also shown . the strong antishadowing in the produced quark and antiquark ratios in fig . [ compjetres](b ) and ( c ) comes from the @xmath247 channel . the antiquark ratio is higher because the @xmath247 parton luminosity peaks at higher @xmath81 than the @xmath292 luminosity and at lower @xmath81 the gluon antishadowing ratio is larger . the difference between the quark and antiquark ratios increases with @xmath11 since the average @xmath81 and thus the valence quark contribution also grow with @xmath11 . at high @xmath11 , the flattening of the fgs quark and antiquark ratios is due to the flattening of the gluon parameterization at @xmath105 . the final - state gluon ratio shows little antishadowing since it arises from the @xmath1 channel . the antishadowing in the eks98 ratio is due to the valence quarks while the higher ratio for fgs reflects the fact that the antiquark ratios also show antishadowing for @xmath293 . the ratio for the total is essentially the average of the three contributions at low @xmath11 , where they are similar , while at high @xmath11 , where the @xmath1 channel dominates , the total ratio approximates the produced gluon ratio in both cases . the resolved rapidity distributions are also shown in fig . [ compt_rap ] for the two @xmath11 cuts . the resolved distribution is not as broad at negative @xmath45 as that of the dijet process in fig . [ jet_dir_rap ] due to the smaller relative gluon contribution and the reduced number of channels available for the @xmath2jet process . note that the relative resolved to direct production is reduced here and the direct process is actually dominant at @xmath294 for @xmath141 gev and for all @xmath45 at @xmath153 gev . the antishadowing peak is higher for resolved production , shown in fig . [ compt_shad ] , thanks to the gluon contribution to resolved production . finally , we show the resolved to direct ratio in fig . [ compt_rat ] . the direct rate alone should be observable at @xmath295 for pb+pb , @xmath273 for ar+ar and @xmath296 for o+o and @xmath141 gev . direct production dominates over all @xmath45 by a large factor when @xmath153 gev . although the rates are lower than the dijet rates , the dominance of direct @xmath2jet production implies than the nuclear quark distribution can be cleanly studied . the resolved @xmath263jet rates are shown in table [ comresrates ] . these rates are only slightly larger than the direct rates for @xmath297 gev but drop below the direct rates at higher @xmath11 since there is no large growth in the number of available production channels , as in dijet production . in addition , the lower effective energy plays an important role here as well . the leading particle @xmath11 distributions of jets from @xmath2jet production are @xmath298 \frac{d_{h / k}(z_c,\mu^2)}{z_c } \ , \ , . \label{sigcomphadres } \end{aligned}\ ] ] the subprocess cross sections , @xmath226 , are related to @xmath250 in eq . ( [ sigcompres ] ) through the momentum - conserving delta function @xmath251 and division by @xmath252 . the resolved @xmath11 distributions for hadrons are shown in fig . [ comphadres](a ) . note that the resolved cross section for leading hadron production is similar to direct production , shown in fig . [ comphaddir](a ) . the same effect is seen for dijet production in figs . [ jethadres ] and [ jethaddir ] . the shadowing ratios are shown in fig . [ comphadres ] . the difference between the shadowing ratios for pions produced by quarks and antiquarks is rather large and reflects both gluon antishadowing at low @xmath11 as well as the relative valence to sea contributions for quark and antiquark production through @xmath299 . in the fgs calculations , the antiquark ratio reflects the flattening of the antiquark and gluon ratios at @xmath105 . since pions produced by gluons come from the @xmath300 channel alone , only a small effect is seen , primarily in the emc region . now the total pion rates follow those for quark and antiquark production of final - state pions instead of those of the gluon . although our @xmath11-dependent calculations have focused on the midrapidity region of @xmath119 , we have shown that extending the rapidity coverage could lead to greater sensitivity to the small @xmath19 region and larger contributions from direct photoproduction , especially at low @xmath11 . thus @xmath263jet production is a good way to measure the nuclear quark distribution functions . direct photoproduction is dominant at central rapidities for moderate values of @xmath11 . final - state hadron production is somewhat larger for direct production so that , even if the rates are low , the results will be relatively clean . there are a number of uncertainties in our results . all our calculations are at leading order so that there is some uncertainty in the total rate , see refs . some uncertainty also arises from the scale dependence , both in the parton densities and in the fragmentation functions . the fragmentation functions at large @xmath182 also introduce uncontrollable uncertainties in the rates . hopefully more data will bring the parton densities in the photon , proton and nucleus under better control before the lhc begins operation . the data from rhic also promises to bring the fragmentation functions under better control in the near future . while the photon flux is also an uncertainty , it can be determined experimentally . the hadronic interaction probability near the minimum radius depends on the matter distribution in the nucleus . our calculations use woods - saxon distributions with parameters fit to electron scattering data . this data is quite accurate . however , electron scattering is only sensitive to the charge distribution in the nucleus . recent measurements indicate that the neutron and proton distributions differ in nuclei @xcite . this uncertainty in the matter distribution is likely to limit the photon flux determination . the uncertainty in the photon flux can be reduced by calibrating it with other measurements such as vector meson production , @xmath301 . studies of well known two - photon processes , like lepton production , can also help refine the determination of the photon flux . with such checks , it should be possible to understand the photon flux in @xmath302 relative to @xmath21 to better than 10% , good enough for a useful shadowing measurement . we would like to thank s. r. klein and m. strikman for discussions . this work was supported in part by the division of nuclear physics of the office of high energy and nuclear physics of the u.s . department of energy under contract no . de - ac-03 - 76sf00098 . c. adler _ et al . _ [ star collaboration ] , phys . * 90 * ( 2003 ) 082302 [ arxiv : nucl - ex/0210033 ] . m. chiu [ phenix collaboration ] , nucl . a * 715 * ( 2003 ) 761 [ arxiv : nucl - ex/0211008 ] . k. j. eskola and x. n. wang , int . j. mod . phys . a * 10 * ( 1995 ) 3071 . v. emelyanov , a. khodinov , s. r. klein and r. vogt , phys . c * 61 * ( 2000 ) 044904 [ arxiv : hep - ph/9909427 ] . k. j. eskola , k. kajantie and j. lindfors , nucl . b * 323 * ( 1989 ) 37 . k. j. eskola and k. kajantie , z. phys . c * 75 * ( 1997 ) 515 [ arxiv : nucl - th/9610015 ] . s. r. klein , j. nystrand and r. vogt , phys . rev . c * 66 * ( 2002 ) 044906 [ arxiv : hep - ph/0206220 ] . l. frankfurt , v. guzey and m. strikman , arxiv : hep - ph/0303022 . h. l. lai _ et al . _ [ cteq collaboration ] , eur . phys . j. c * 12 * ( 2000 ) 375 [ arxiv : hep - ph/9903282 ] . m. glck , e. reya , and a. vogt , phys . d * 46 * ( 1992 ) 1973 . h. plothow - besch , ` pdflib : proton , pion and photon parton density functions , parton density functions of the nucleus , and @xmath303 calculations ' , user s manual - version 8.04 , w5051 pdflib , 2000.04.17 , cern - ett / tt . b. a. kniehl , g. kramer and b. ptter , nucl . b * 582 * ( 2000 ) 514 [ arxiv : hep - ph/0010289 ] . b. a. kniehl , g. kramer and b. ptter , nucl . b * 597 * ( 2001 ) 337 [ arxiv : hep - ph/0011155 ] . x. f. zhang , g. fai and p. lvai , phys . * 89 * ( 2002 ) 272301 [ arxiv : hep - ph/0205008 ] . m. gyulassy , p. levai and i. vitev , phys . b * 538 * ( 2002 ) 282 [ arxiv : nucl - th/0112071 ] . i. vitev , m. gyulassy and p. levai , arxiv : nucl - th/0204019 . t. sakaguchi [ phenix collaboration ] , nucl . phys . a * 715 * ( 2003 ) 757 [ arxiv : nucl - ex/0209030 ] . j. f. owens , rev . * 59 * ( 1987 ) 465 . r. vogt , heavy ion phys . * 17 * ( 2003 ) 75 [ arxiv : hep - ph/0207359 ] . a. trzcinska , j. jastrzebski , p. lubinski , f. j. hartmann , r. schmidt , t. von egidy and b. klos , phys . lett . * 87 * ( 2001 ) 082501 . .pertinent parameters and kinematic limits for @xmath21 collisions at the lhc @xcite . we first give the luminosities and the @xmath52 collision kinematics , the nucleon - nucleon center of mass energies , @xmath58 , the corresponding beam energies , @xmath72 , lorentz factors , @xmath37 . we then present the photon cutoff energies in the center of mass frame , @xmath304 , and in the nuclear rest frame , @xmath74 , as well as the equivalent maximum photon - nucleon center of mass energies , @xmath57 . [ cols="^,^,^,^,^,^,^,^",options="header " , ]

we describe a new microscopic approach for analyzing interacting electron systems with local moments or , in principle , any local order parameter . we specialize attention to the doped mott insulator phase of the hubbard model , where standard weak coupling perturbation methods fail . a rotationally invariant stratonovich hubbard field is introduced to decouple the static spin components of the interaction at each site . charge degrees of freedom are then treated by `` slave '' hartree fock in the presence of a spatially varying random spin field . the effective action reduces to a classical heisenberg model at half filling , insuring that the system has ( i ) finite range order at with an exponentially diverging correlation length and ( ii ) a one electron mott hubbard gap in the presence of disorder . away from half filling properties are determined by strongly non gaussian fluctuations in the amplitude and orientation of the local spin fields . the saddle point equations of the theory at zero temperature reduce to inhomogeneous hartree fock , so that disordering of domain walls at finite temperature is in principle included . we present preliminary small system results for the intermediate and large interaction regimes obtained by monte carlo simulation of random spin field configurations . # 1 0= -.025em0 - 0 .05em0 - 0 -.025em.0433em0

cond-mat0010480

in this paper we describe a new approach to perturbation theory for systems with local moments or , in principle , any local order parameter . we begin by summarizing the present status of weak coupling perturbation theories for such systems , with the hubbard model as a primary example . over the course of the last decade weak coupling perturbation theories have been used extensively to investigate the physics of the hubbard model and its extensions . the simplest perturbative approach to the hubbard model is just hartree fock theory for one body correlation functions . the conserving approximation implied for two body correlations is then the random phase approximation , or rpa . in the large@xmath1 limit at half filling the static spin response is unstable within hartree fock , and the system exhibits a broken symmetry ground state with infinite range antiferromagnetic order @xcite . furthermore a gap of order @xmath1 appears in the one body density of states for the ordered phase . if the hartree fock equations are extended to finite temperature , long range antiferromagnetic order onsets at a temperature of order @xmath1 in the large@xmath1 limit . the zero temperature behavior of the hartree fock approximation away from half filling has also been investigated , and incommensurate states , as well as vertical and diagonal domain walls , have been studied @xcite . however , the appropriate treatment for finite temperatures has remained unclear , because the energetic competition of the various states is delicate . in fact the hartree fock approximation at finite temperatures provides only a crude description of the physics in a two ( or one ) dimensional system . in accordance with the mermin wagner theorem @xcite , infinite range magnetic order is forbidden for @xmath0 in two dimensions . the magnetic correlation length @xmath2 instead diverges exponentially in @xmath3 for @xmath4 at half filling , and the quantitative behavior of @xmath5 , as well as the zero temperature ordered moment , is renormalized by quantum fluctuations . nevertheless the half filled model exhibits a gap ( or _ pseudogap _ , since we do not mean to imply a rigorously zero density of states ) at finite temperature in the spin disordered phase . with the intention of modeling strong spin fluctuations in a disordered system , the present authors proposed the fluctuation exchange , or flex , approximation @xcite as a plausible starting point . the self consistent treatment of spin fluctuations in the one body self energy suppresses the magnetic ordering transition to zero temperature . nevertheless the new approximation is itself flawed : it does not produce a sharp gap at half filling and instead smears out the density of states , leaving a finite weight at zero energy for all temperatures @xcite . furthermore the spin response function diverges with decreasing temperature in a way which only roughly mirrors the exact behavior of the model . the failure at half filling calls into question results obtained for large @xmath1 at small doping , and in the end the flex approximation fails to describe the pseudogap regime . the inadequacy of the flex approximation has motivated various attempts to improve the theory . kampf and schrieffer have calculated the electron spectral density near an antiferromagnetic transition to lowest order in an effective spin fluctuation interaction , and their results exhibit a pseudogap when the spin correlation length is large @xcite . vilk and tremblay @xcite have introduced a factorization approximation for the electron four point function which is constrained to be exact when all space time coordinates coincide . others @xcite have argued that perturbative calculations based on an unrenormalized electron propagator are justified by a cancellation between vertex corrections and self energies , leading to the appearance of a pseudogap in some parameter regimes . in addition efforts have been made to extend the flex approximation within a scheme based on infinite order ( or `` parquet '' ) vertex renormalization @xcite . quantitative studies of parquet renormalization have had the goal of establishing the validity or inadequacy of perturbation theory resummations for systems with a gap , but without infinite range order . why would one assume that such resummations have any chance of succeeding ? for at least one nontrivial model , the anderson impurity in a metallic host , it has been known since the mid1980 s that zero temperature perturbation theory for thermodynamic properties has an infinite radius of convergence in @xmath1 @xcite . this fact , as well as some suggestive results in the low density limit , initially encouraged hope that an infinite order parquet summation for the doped mott insulator might succeed if the technical difficulties impeding its implementation could be overcome . in fact this hope is not justified . full numerical solutions of the minimal parquet equations for the two dimensional hubbard model in the large@xmath1 limit @xcite yield the following results : ( i ) no gap appears in the converged solution at half filling . ( ii ) the spectral density at low energies is actually larger than that in a flex calculation . we interpret these results to mean that conventional perturbation theory @xcite fails to describe the physics of a mott hubbard gap or pseudogap in a spin disordered system . furthermore the parquet results , which consistently incorporate vertex renormalization to infinite order , bring into question whether the basic problem can be circumvented by the constrained factorization approximation or a cancellation between parts of perturbative diagrams . rather we believe the failure of perturbation theory around a uniform ( i.e. , fermi liquid ) starting point is due to a qualitative change in the saddle point structure of the action for the hubbard model at temperatures of order @xmath1 in the large@xmath1 limit . this change is induced by the appearance of local moments ( or , more generally speaking , a local order parameter field ) in the theory . conventional perturbation theory becomes unreliable below a temperature scale which depends on both the interaction @xmath1 and the chemical potential @xmath6 . below this scale one needs a revised perturbation theory which incorporates `` anomalous '' feynman graphs , but does so without inducing infinite range order . these seemingly contradictory objectives may be accomplished by introducing a spatially varying static field with the full symmetry of the associated classical order parameter . one way to introduce such a field is via a stratonovich - hubbard ( s h ) transformation @xcite of the interaction . in particular we employ below a spin rotation invariant s h field to decouple the static @xmath7 component of the hubbard interaction . about thirty years ago wang et al . @xcite showed for the magnetic alloy problem that the effective static free energy induced by a time varying s h field provides a physical picture of the crossover from band like to local moment behavior . evenson et al . @xcite analyzed the coupling between isolated moments which arises within this approach and noted that the simplest scheme generates an ising , rather than heisenberg , interaction . below we show that when a vector s h field is introduced , the effective action reduces to a classical heisenberg form at half filling . this means no infinite range order at @xmath0 , an exponentially diverging correlation length @xmath2 , _ and _ a mott hubbard gap in the spin disordered phase . furthermore the approximation scheme should provide a systematic way to study doping of the mott insulator . in the doped case both orientation and amplitude fluctuations of the field become important . in particular the effective action for the static field is strongly non gaussian @xcite . since the action arises directly from integrating out the fermion degrees of freedom , its form reflects both the underlying spin and charge structure of the system . in this way domain wall stripe fluctuations can occur naturally if they are present in the model . we describe this new approximation scheme for the hubbard model in section 2 , then derive expressions for specific electron correlation functions in section 3 . in section 4 we provide a brief description of our monte carlo sampling scheme for integration over the fluctuating local fields . in section 5 we present some illustrative results obtained on a work station for a @xmath8 system . we outline in section 6 the form of generalized approximations which extend the present approach in the same sense that flex extends hartree fock . finally we close in section 7 with some speculations on possible implications for high temperature superconductivity and with an outline of the next steps in this work . in this section we show how to decouple the the static spin interaction in the repulsive hubbard model by introducing a spin rotation invariant s h field . this field may be interpreted as a classical heisenberg order parameter which generalizes the saddle point in conventional temperature dependent hartree fock calculations . it is conventional to decouple the interaction term by introducing independent fields at each point in imaginary time . when this is done , some freedom remains in choosing the form of the interaction , due to the equal time fermion anticommutation relations . one may write equivalently un_in_i & = & u(n_i^2 - m_iz^2 ) [ intone1 ] + & = & u(n_i - m_iz^2 ) [ intone2 ] + & = & un_i -u(m_ix^2 + m_iy^2 ) , [ intone3 ] where n_i & = & n_i + n_i + _ i & = & _ c^_i_ c^_i . [ opsone ] there is in fact considerably more freedom allowed , since the interaction is rotationally invariant : the z component of the moment operator in eqns . [ intone1 ] and [ intone2 ] may be replaced by the x or y component , and likewise different pairwise combinations of moment operators may be substituted in eqn . [ intone3 ] . one is free to consider in addition arbitrary linear combinations of these expressions , weighted properly to reproduce the original interaction . this was realized by several authors around 1970 during early studies of the single impurity anderson model and the hubbard model . in particular wang et al . employed eqn . [ intone2 ] @xcite and hamann employed eqn . [ intone1 ] @xcite as the starting points for equal time stratonovich hubbard decoupling schemes . note that both these expressions imply an ising like s h field governing the spin degrees of freedom , while the third expression implies an xy like field . all three expressions result in actions which break the spin rotational symmetry of the theory , and the symmetry is only restored by the integration over field configurations . as long as all configurations are included there is no difficulty , but problems arise if approximations are introduced for any system beyond a single impurity . this was realized by evenson et al . @xcite in large@xmath1 studies of the two site hubbard model within a static approximation : the effective spin spin coupling assumes an ising , rather than heisenberg , form when eqn . [ intone2 ] is used as a starting point . it would clearly be desirable to introduce a symmetrical heisenberg decoupling , but this is difficult for the following reason : no matter how the spin quantization axis is chosen , the three equivalent components of the spin spin interaction are divided up between the two particle hole channels ( longitudinal and transverse ) of the electron vertex , which `` overlap '' completely at equal time . a heisenberg decoupling can be introduced artificially if , for example , the interaction is written as an equal weight linear combination of eqn . [ intone1 ] for the three choices of moment operator , i.e. , un_in_i = un_i^2 -u _ [ intone4 ] this procedure introduces spurious factors of @xmath9 in the saddle point equations when a direction is chosen for the broken symmetry spin axis , however , and so offers an unsatisfactory solution . we may restate the problem as follows : a conventional equal time s h transformation , followed by a static approximation within the resulting theory , violates the crossing symmetry of the hubbard interaction vertex and prevents a spin rotation invariant treatment of fluctuations . to insure an equivalent static treatment of all three spin components , we propose the introduction of a zero frequency , rather than equal time , s such a decoupling scheme is possible because the static components of the direct and exchange channels in the hubbard model have essentially zero overlap at low temperatures . this allows the simultaneous treatment of interactions in both crossed channels and leads to a unique prescription for the decoupling scheme independent of the expression employed for the interaction in eqns . [ intone1]-[intone4 ] . the same procedure can be applied in more general situations when it is necessary to decouple multiple interactions in crossed channels . it is possible to carry out this transformation directly within an operator based approach , but the manipulations are more transparent with anticommuting c numbers , and we follow the latter approach . we consider a discretized representation of the partition function z = tre^-h , [ partition ] where h = -t_ic^_i+,c^ _ i + u_ic^_ic^_ic^_i c^_i - _ ic^_ic^ _ i . [ hamone ] let _ & = & + & = & /l , [ tau ] with @xmath10 the number of time slices . we write the anticommuting c numbers for time @xmath11 , site @xmath12 , and spin @xmath13 as @xmath14 , @xmath15 . it is convenient to use the frequency transformed variables @xmath16 , @xmath17 , where ^i _ & = & _ n =- l/2^l/2 - 1e^i_n _ ^i_n + c^i _ & = & _ n =- l/2^l/2 - 1e^-i_n _ c^i_n [ fourier ] with _ n = ( 2n+1)t [ matsubara ] for integer @xmath18 . in order to keep track of discretization effects , we define i_n^ 1-e^-i_n , [ discrete ] which approaches @xmath19 for @xmath20 . the discretized fermion action is s & = & -_in(i_n^+)^i_n c^i_n + _ in(-t)^i+,_n c^i_n + & & + u_imnn^i_n+m ^i_-n-1 c^i_-n-1c^i_n+m . [ actionone ] the interaction term is represented in figure 1 . ( note that the integer label @xmath21 corresponds to frequency @xmath22 . ) equivalently one may write s = s_0 + s_int , [ actiontwo ] with s_0 = _ ijn ^i_n_ijc^j_n , [ actionthree ] where the matrix in brackets has spatial indices , and _ ij & = & _ ij + ( * h_0*)_ij & = & [ mats ] now note that the interaction can be rewritten in two equivalent ways to emphasize the particle hole channels ( see figures 2a b ) : s_int & = & -u_imnn ^i_n+mc^i_n ^i_nc^i_n+m + & = & u_imnn ^i_n+m c^i_n^i_nc^i_n+m . [ crossing ] at this stage we restrict attention to terms with @xmath23 , i.e. , the static component of the particle hole interactions . for @xmath23 the two sums in eqn . [ crossing ] are essentially independent : the only terms which appear in both sums are restricted to @xmath24 . therefore the static particle hole interaction may be written s^m=0_int = -u_i + u_i[actionfour ] where we have dropped from the double sums an @xmath24 correction term , which becomes negligible at low temperatures @xcite . now all terms in eqn . [ actionfour ] may be reduced to sums of perfect squares in preparation for the introduction of s h fields . first note that _ n^i_nc^i_n & = & _ n ^i_n(_x+i_y)_c^i_n + _ n^i_nc^i_n & = & _ n ^i_n(_x - i_y)_c^i_n . [ sumsone ] using these identifications one has = , [ sumstwo ] where m^i_x0 & = & _ n^i_n(_x)_ c^i_n + m^i_y0 & = & _ n^i_n(_y)_ c^i_n . [ sumsthree ] likewise , _ n^i_nc^i_n & = & ( n^i_0+m^i_z0 ) + n^i_nc^i_n & = & ( n^i_0-m^i_z0 ) , [ sumsfour ] where n^i_0 & = & _ n(^i_nc^i_n+^i_nc^i_n ) + m^i_z0 & = & _ n^i_n(_z)_ c^i_n . [ sumsfive ] in this case one identifies = . [ sumssix ] thus s^m=0_int = u_i(n^i_0)^2 - u_i(^i_0^i_0)^2 . [ actionsix ] the repulsive charge interaction is retained for treatment by perturbation theory ( i.e. , saddle point plus fluctuations ) . the attractive spin interaction in eqn . [ actionsix ] may be decoupled using an independent static s h field at each site @xmath12 : & & = + & & = _ i ( ) ^3/2 d^i_xd^i_yd^i_z . + & & [ stratonovich ] the s h fields @xmath25 may be interpreted as classical magnetic fields which couple to the local electron spin polarization . the hubbard action may be rewritten as s = _ i|^i|^2 + s_0 + s^m=0_charge + s^m 0_int , [ actionseven ] where s_0 = _ ijn^i_n\{-i^ _ n__ij + } c^j_n [ actioneight ] and s^m=0_charge = u_i(n^i_0)^2 . [ actionnine ] the interaction contributions to the action for @xmath26 take precisely the form they would in a conventional perturbative treatment , _ except _ that all vertices related to the zero frequency spin vertex by crossing are absent . note that because we have decoupled only the @xmath23 spin components , we have left open the possibility of higher order perturbative expansions about a static approximation . we consider now the static charge interaction @xmath27 in eqn . [ actionseven ] . this term could be decoupled in parallel with @xmath28 by introducing a real valued s h field @xmath29 . since the charge interaction is repulsive , however , the fermion bilinear coupling to @xmath29 would be pure imaginary and unsuitable for monte carlo simulation . it is straightforward to show that a saddle point approximation for the @xmath29 fields in the presence of a fixed @xmath25 configuration yields the same result as a spatially non uniform hartree fock . for this reason we do not introduce @xmath29 fields , but reserve @xmath27 for treatment by perturbation theory ( i.e. , hartree fock at lowest order ) . within such a hartree fock approximation for fixed @xmath25 , s^m=0_charge u_inn^i _ ^i_nc^i_n - u_in^i_^2 , [ actionten ] where @xmath30 is the mean occupancy at site @xmath12 . it is essential to retain the `` constant '' term , since it depends on the local field configuration @xmath31 . from this point we use the notation @xmath32 to denote the expectation value of the fermion observable @xmath33 in the presence of a fixed field configuration . the static @xmath34dependent action assumes the form s ( ) = s_0 ( ) + s_f(,,c ) , [ actioneleven ] where s_0 ( ) & = & _ i|^i|^2 - u_i n^i_^2 + s_f(,,c ) & = & _ ijn^i_n\ { -i^_n_ij_ + } c^j_n . [ actiontwelve ] the fermion contribution to the action may now be integrated out exactly , since it takes a one body form ( albeit a form non diagonal in spin ) : ddc = _ r , [ fermion ] where the @xmath35 are eigenvalues of the one body hamiltonian h ( ) = -t_ic^_i+,c^_i - _ ic^_i_ c^_i + _ i c^_ic^_i . [ hamtwo ] thus the final expression for the effective action after the fermion integration is s_eff ( ) = s_0 ( ) - _ r . [ newaction ] note that for fixed @xmath6 and @xmath25 a hartree fock iteration is required to adjust the non uniform charge field @xmath36 to satisfy the condition n^i & = & _ rf(_r)_|r|i|^2 + f(_r ) & = & 11+e^_r . [ scfone ] in this equation @xmath37 is the one body eigenvector corresponding to eigenvalue @xmath38 ; and the @xmath39 constitute a complete set of basis vectors corresponding to spin @xmath40 on site @xmath12 . note that for an @xmath41 real space lattice , the size of the eigenvector space is @xmath42 . the form of the charge interaction hartree fock employed here provides a template for the family of baym kadanoff @xcite ( or flex @xcite ) generalizations of the present approach discussed in section 6 . the charge field is here viewed as a nonsingular `` slave '' to the spin field : it is fixed to its saddle point , which changes when the @xmath34 configuration changes . no further approximation to the action @xmath43 is now possible without explicitly breaking spin rotational and translational invariance . in conventional hartree fock treatments it is assumed that the static field @xmath25 assumes a rigid configuration . at half filling the mean field ansatz is just & = & e^iqr_iq= ( , ) + n^i & = & 1 . [ ansatz ] a saddle point evaluation of @xmath44 within this restricted field space reproduces the conventional hartree fock equations for a commensurate antiferromagnet . furthermore inhomogeneous hartree fock solutions @xcite for @xmath45 follow by assuming a spin space form ( collinear , spiral , etc . ) for the order parameter and solving the resulting saddle point equations . we propose instead to evaluate electron correlation functions using @xmath44 as the action in a classical field monte carlo . the monte carlo is classical since all finite frequency contributions to the action have been set aside for treatment by perturbation theory : such contributions , when included , alter the form of the static field action , but are not themselves statistically sampled . by this means it is possible to insure that higher order approximations have a positive definite weighting factor @xmath46 , avoiding the sign problem inherent in a full quantum monte carlo . it is important to note that this approach has been explicitly designed to treat the low temperature regime in which local moments exist , i.e. , in which the most probable configurations of @xmath25 have non zero amplitude ( at least for some sites @xmath12 ) . we do not address in this paper the problem of local moment formation per se . it is clear that at high temperatures the `` free '' @xmath25 fluctuations have variance @xmath47 , and it is not appropriate to use @xmath43 as a starting point for perturbation theory in this regime . instead our purpose is to construct an approximation scheme which builds in the presence of disordered local moments at temperatures satisfying 4t^2/u t u [ condition1 ] in three dimensions and t u [ condition2 ] in two dimensions . it is also important to note that this approximation omits quantum effects associated with `` tunneling '' of the order parameter fields , i.e. , possible @xmath48dependent instantons . as a result the kondo effect is outside the present description . we restrict our attention here to the implications of a classical ( i.e. , @xmath23 ) order parameter description and do not discuss additional quantum effects . we next examine the behavior of @xmath44 in the large@xmath1 limit , then comment on the action s relationship to other work in the literature . in the large@xmath1 limit the hubbard model may be mapped to a quantum spin@xmath49 heisenberg model with near neighbor antiferromagnetic exchange integral @xmath50 . we show below that in this limit the approximate action @xmath44 describes a classical heisenberg model with exchange integral @xmath50 . this insures that in a two dimensional system the approximation obeys the mermin wagner theorem , i.e. , infinite range spin ordering is absent down to @xmath51 . however this approach does not include the quantum fluctuations which renormalize the zero temperature moment and reduce the finite temperature correlation length . it is essential to include finite frequency fluctuations in order to address these points . nevertheless the static action @xmath44 achieves the goal of reproducing the exact classical critical behavior within a well defined zeroth order approximation . to demonstrate the mapping to the classical heisenberg model , consider first the local one site action ( @xmath52 ) in the large@xmath1 limit for @xmath53 : s_eff^one - site = ||^2 - _ r . [ onesite ] in this case there are only two eigenvalues @xmath35 . to find @xmath35 , choose the spin quantization axis along the direction of @xmath34 . then the field dependent term in @xmath54 is just -(c^_ic^_i - c^_i c^_i ) , [ intern ] with eigenvalues _ = -,=1 . [ eigs ] minimization of the action with respect to @xmath5 gives = _ e^1+e^ , [ minimiz ] i.e. , = m_z , [ solution ] since m_z = _ = 111+e^- . [ mzz ] for large @xmath1 the solution has m_z & & 1 + & & /2 . [ limits ] as expected , the one site gap is just _ - _ = u-(-u ) = u . [ gapp ] note that this result has been obtained without breaking the rotational invariance of the one site hamiltonian , since the direction of @xmath34 is arbitrary , and the boltzmann weight must be integrated over @xmath34 . thus to zeroth order in @xmath55 , the half filled system consists of fluctuating fields @xmath25 of fixed length @xmath56 and uncorrelated directions . the first non zero terms in perturbation theory enter at @xmath57 , and describe the process of one electron hopping to a neighboring site , then back to its starting point . physically the hopping terms should induce correlations in the field directions at neighboring sites to lower the kinetic energy . let the field directions at neighboring sites @xmath12 and @xmath58 be @xmath59 and @xmath60 . suppose also that @xmath59 may be obtained by rotating @xmath61 in a right handed sense by angle @xmath62 about axis @xmath63 . then the eigenstates with spin projection @xmath64 along @xmath59 can be written |n^i , & = & |z , + ( -i_x+_y)|z , + [ sumsquares ] the one site solution gives || = = n^i^i , [ meanfield ] so that @xmath65 will be occupied and @xmath66 will be empty ( and higher in energy by @xmath1 ) . now apply second order perturbation theory to the electron ground state for the frozen field configuration with direction vectors @xmath67 . the matrix element for hopping from |n^i,|n^j , |n^j,|n^j , [ hoppingsites ] is ( -t)n^j,|n^i , = ( -t ) ( -i_x-_y)^ * , [ hoppingmatrix ] where @xmath60 is obtained from @xmath59 by a right handed rotation by @xmath68 about axis @xmath69 . the energy lowering of the ground state due to the virtual hop from @xmath70 and back is then = -t^2u^22 , [ lowering ] since _ x^2+_y^2 = 1 . [ sumup ] the hopping from @xmath71 supplies the same energy , so the total perturbation energy lowering in this particular field configuration is -2t^2u_ij^2(_ij/2 ) & = & t^2u_ij ( _ ij-1 ) + & = & 4t^2u_ij + & = & 4t^2u_ij , [ heisen ] where @xmath72 is the electron spin expectation value in configuration @xmath31 for @xmath52 . note finally that the effective action at half filling reduces to s_eff ( ) = _ i|^i|^2 - _ r . [ actionthirteen ] in the large@xmath1 limit _ r , [ prodd ] with @xmath73 the perturbed ground state energy , since all other occupancies result in an energy higher by at least @xmath1 . since @xmath74 is a constant in this limit , the part of the action dependent on the fluctuating fields is precisely @xmath75 , with h_eff = 4t^2u_ij(n^in^j - ) , [ hamthree ] establishing the mapping to the classical heisenberg model . the behavior of the action away from half filling in the large@xmath1 limit is not so easy to deduce , since the hartree fock correction from the static charge interaction is crucial . if one assumes @xmath76 sites are singly occupied and @xmath77 sites are empty , the one site saddle point equation implies @xmath78 on the singly occupied sites and @xmath79 on the empty sites . a degeneracy then arises , since the low lying one electron states on the singly occupied sites pick up a hartree fock density contribution @xmath80 , while the states on the empty sites have no hartree fock correction . the two sorts of sites then have degenerate low lying states in the absence of @xmath81 : the occupied state on each singly occupied site has energy @xmath82 , and the two states on each unoccupied site also have energy 0 . as expected , the hopping becomes crucial for determining @xmath36 and the favored @xmath25 configurations . thus in the doped case there are important amplitude , as well as orientation , fluctuations in the s h field configurations arising from @xmath44 . we next comment briefly on the relationship of our approximation to other work in the literature . we have argued in section 1 that perturbation theory in @xmath1 fails to resum properly at low temperatures in the doped mott insulator phase due to a fundamental change in the system s saddle point structure . nevertheless a number of authors @xcite@xcite have observed that gaps or pseudogaps in the one electron density of states for the positive@xmath1 and negative@xmath1 hubbard models may be obtained within approximations which resemble perturbation theory about a non interacting ground state . such approximations have in common the feature that the one electron self energy @xmath83 may be written in the form ( k , i_n ) = t_qm v(q , i_m)g_0(k - q , i(_n- _ m ) ) , [ perturb ] where @xmath84 is a momentum and frequency dependent effective potential and @xmath85 is a bare electron propagator . crucial points to note are that ( i ) the electron propagator is left bare ; and ( ii ) higher order perturbation theory in @xmath84 leads to a deterioration , rather than an improvement , in the approximation for the one electron density of states . we argue that the reasons for this behavior are as follows : the local moments which underlie the mott hubbard gap generate anomalous diagrams in perturbation theory , i.e. , terms in the one electron self energy off diagonal in momentum and spin . at zero temperature and half filling the moments have true long range staggered order , so that a momentum space description becomes possible . if the static off diagonal self energy is denoted @xmath86 , with @xmath87 , then the effective diagonal self energy becomes ( k , i_n ) = ^2_qg_0(k- ) , [ afmhartree ] generating a hartree fock antiferromagnetic gap of @xmath88 in the density of states . in this case , as is well known , @xmath85 enters eqn . [ afmhartree ] rather than @xmath89 to avoid double counting . thus approximations of the form in eqn . [ perturb ] may be viewed as generalizations of eqn . [ afmhartree ] which `` smear '' the off diagonal self energy @xmath90 over a range in momentum and frequency space @xcite . in this sense these approximations are quite different from perturbation theory in @xmath84 , and they do not provide support for the convergence of perturbation theory in the doped mott insulator . we would argue that such approximations model only roughly the physics of this system @xcite . in the doped phase the order parameter field has strong non gaussian fluctuations in both amplitude and orientation , and these are coupled to the charge degrees of freedom . furthermore , as discussed in reference @xcite , this general approach has difficulties addressing rare large amplitude fluctuations that play an essential role . we comment also on the relationship of the current approach to the phenomenological spin fermion ( sf ) model . in recent years a number of perturbative studies @xcite have been carried out for this model , which describes a system of electrons coupled to a gaussian distributed vector spin field @xmath91 . the microscopic action @xmath44 differs from @xmath92 in two crucial ways : ( i ) @xmath92 omits coupling to charge degrees of freedom , while @xmath44 describes a highly nonlinear coupling between spin and charge . the form of this coupling is dictated by the hubbard interaction itself , and its omission prevents a proper description of the doped mott insulator phase . ( ii ) the finite frequency corrections to the actions differ significantly , since we have introduced only a static s h field to circumvent the problems associated with an equal time decoupling . finally , our point on the non gaussian nature of fluctuations in @xmath44 applies to @xmath92 as well , i.e. , in the interesting parameter regime for this model we expect a conventional perturbative treatment to be impossible due to the formation of local moments . electron correlation functions for the action @xmath93 in eqn . [ actioneleven ] have been calculated using the metropolis algorithm for statistical sampling of @xmath25 configurations.the monte carlo algorithm adopted relies on single site updating and resembles a hybrid of glauber and kawasaki dynamics @xcite for the ising or classical heisenberg models . the algorithm is described more precisely in section 4 . electron correlation functions have the same general form as in full quantum monte carlo calculations with continuous valued s h fields . let @xmath94 be an electron observable written in anticommuting c number notation , and let d _ i ( ) ^3/2d^i_xd^i_yd^i_z . [ measure ] then a & = & dddc a(,c ) dddc + & = & 1zdz_f ( ) , + & & [ expect ] where z_f ( ) & = & ddc + & = & _ r , [ expect2 ] and the last line in eqn . [ expect ] follows by multiplying and dividing by @xmath95 within the integral . the expression in square brackets is just an electron thermal average in the presence of the background @xmath25 configuration . thus finally a & = & 1zd a(,c ) _ + & & a(,c ) _ , [ expect3 ] where s_eff ( ) & = & s_0 ( ) - z_f ( ) + & = & s_0 ( ) - _ r . [ expect4 ] the simple form for @xmath44 means that approximations which incorporate finite frequency fluctuations , i.e. , which generalize @xmath96 , may in principle be developed to extend the approach worked out in detail here ( see section 6 ) . we derive below the form of several one and two body electron correlation functions @xmath97 in the presence of the background fields . as stated previously , n^i _ & = & _ g_(ii;=0 ^ - ) _ + & = & _ rf(_r)|r|i|^2 , [ density ] where g_(ij ; ) _ = -c^i()^j(0 ) _ [ gg ] in terms of anticommuting c numbers ; and the @xmath37 are eigenstates of the spin dependent hamiltonian @xmath54 in eqn . [ hamtwo ] . likewise , the spin polarization at site @xmath12 is ^i _ & = & _ ^i _ c^i _ + & = & _ _g_(ii;=0 ^ - ) _ + & = & _ rf(_r)_r|i i|r , [ polarization1 ] in terms of the @xmath98basis states @xmath99 . writing out the vector components explicitly gives m^i_x _ & = & _ rf(_r ) + m^i_y _ & = & _ rf(_r ) + m^i_z _ & = & _ rf(_r ) . [ polarization2 ] note that these last results generalize to give the one body propagator g_i , j(i_n ) _ & = & _ 0^de^i_n + & = & -c^i_n^j_n _ + & = & _ r i|rr|ji_n-_r . [ propagator ] the real axis spectral density may be calculated directly by analytic continuation before the field averaging is performed . thus the one particle spectral density for spin @xmath62 at site @xmath12 after field averaging is n_i ( ) & = & -1img_(ii;+i0^+ ) _ + & = & _ r|i|r|^2(-_r ) . [ rhoa ] the total one particle spectral density for spin @xmath62 is n _ ( ) = _ in_i ( ) . [ rhotot ] ( note that in a properly equilibrated calculation translational and spin rotational invariance are restored by field averaging , and the quantity @xmath100 is site and spin independent . ) the momentum resolved spectral density @xmath101 is only slightly more difficult to calculate . note again that translational invariance is only restored by field averaging , so that the propagator in a single field configuration is not diagonal in momentum . after averaging we have a_k ( ) = _ r(-_r)_ije^-ik(r_i - r_j)i|rr|j all the spectral densities may be calculated conveniently by histogram binning , i.e. , by incrementing a bin centered on frequency @xmath102 whenever an eigenvalue @xmath35 falls in the range @xmath103 . in this way spectral densities may be obtained which integrate exactly to unity within machine precision . in some cases it may be desirable to compare imaginary time propagators @xmath104 with the results from a full quantum monte carlo simulation . the necessary expressions within the present approach are g_k ( ) = [ gtau ] where f^r_k = 1_ije^-ik(r_i - r_j)i|r r|j . [ ffactor ] a quantity of particular interest in the local moment regime is the average double occupancy d = 1_id^i _ , [ double1 ] where d^i _ = n^in^i _ . [ double2 ] in a general @xmath25 configuration with broken spin rotational invariance the appropriate expression for @xmath105 is not @xmath106 . instead d^i _ & = & _ ^i ( ) c^i()^i()c^i( ) _ + & = & n^i_n^i _ + c^i^i_^ic^i _ . [ double3 ] note ^ic^i _ & = & + c^i^i _ & = & - , [ double4 ] while n^i _ & = & + n^i _ & = & . [ double5 ] thus d^i _ = . [ double6 ] the double occupancy assumes an explicitly spin rotation invariant form as required . the expression in eqn . [ double6 ] should in principle be supplemented by zero and finite frequency rpa corrections to insure that eqn . [ double3 ] reproduces the @xmath107 limit of the general two body correlation function . ( as noted below in the discussion of response functions , rpa corrections appear at zero frequency because the charge interaction is being treated within hartree fock . there is no zero frequency rpa in the spin channel , since that portion of the interaction is being treated exactly . ) rpa corrections in equal time expectation values @xmath108 are actually problematic in any perturbation theory for the following reason : the pauli principle implies an exact identity n^i()n^i ( ) = n^i . [ double7 ] in perturbation theory , this identity can only be preserved exactly in approximations with a crossing symmetric two body vertex . such a vertex appears in order by order perturbation theory ( and in parquet theory ) , but not in a standard rpa , whether about the unperturbed vacuum or about broken symmetry states as in the present case . the rpa can be augmented to make it crossing symmetric by incorporating the missing contributions in the crossed channel ( see figure 3 ) , but the result is no longer conserving in the baym kadanoff sense . the rpa corrections to the double occupancy are in any event expected to be small in the present case . since they are also problematic , we omit them from the calculation of @xmath109 reported here . we next consider the static spin and charge susceptibilities @xmath110 and @xmath111 . as mentioned above , one expects the appearance of rpa corrections in the static charge response , since the charge interaction is being treated by `` slave '' hartree fock , but naively one might expect no corrections in the spin response . this is not correct : an average must be performed over the symmetry breaking field @xmath25 , and in the presence of @xmath25 spin and charge excitations are mixed . for this reason the repulsive charge interaction makes its presence felt weakly in the spin channel , as well as in the charge channel . the static spin response function takes the form ^ij _ m^i _ _ /h^j_|_h=0 , [ chi1 ] where the assumed local coupling to the external field @xmath112 is h = -h^j_c^_j_ c^_j . [ chi2 ] temporarily neglecting rpa corrections , the field derivative yields ^ij _ = -1tm^i _ _ m^j _ _ + 1 t _ n^a_n^_c^i_n _ n^i_n^_c^j_n _ . [ chi3 ] the first term on the right hand side vanishes for zero applied field ( since the system has no spontaneous magnetization at finite temperature ) , and the second term ( see figure 4 ) reduces to ^ij _ = 1tm^i _ _ m^j _ _ + -t_n_ g_(ji;i_n)_g_(ij;i_n ) _ ^_^_ , [ chi4 ] i.e. , formally an `` anomalous '' and a `` normal '' contribution . the anomalous contribution may be evaluated using previous expressions for the site spin polarizations , while the normal term may be written as the configurational average of _ m^ij_,(i=0 ) _ ^_^_ , [ chi5 ] where m^ij__1_2,_1_2(i ) _ -_rs j_1|rr|i_1i_2|s s|j_2 f(_r)-f(_s ) _ r + i-_s . [ chi6 ] ( note that for @xmath113 , the ratio involving the fermi functions must be replaced by @xmath114 . ) the static charge response function is ^ij n^i _ /^j|_=0 , [ chi7 ] where in this case the assumed coupling to the external field is h = -^j_c^_jc^_j . [ chi8 ] in complete analogy with the spin response calculation the result before the inclusion of rpa corrections is ^ij & = & 1 t + & & + _ m^ij_ , (i=0)___ , [ chi9 ] with the normal `` bubble '' @xmath115 defined as in eqn . [ chi6 ] . as emphasized above , rpa corrections necessarily appear when an interaction is treated perturbatively by hartree fock . in the present case , where spin and charge excitations mix before the configurational average is performed , rpa charge vertex corrections appear in both the static charge and spin response functions . the calculation of rpa corrections for _ finite frequency _ response functions is trickier than in conventional perturbation theory because the zero frequency spin vertices have been explicitly removed from the theory . this affects response functions at @xmath116 by partially removing vertices which carry zero frequency in a crossed particle hole channel . the analytic continuation to real energies is also complicated by this feature of the theory . we defer discussion of spin and charge response at finite frequency and concentrate in the present paper on rpa corrections to the static susceptibilities . the repulsive rpa charge vertex produces a geometric series of corrections as in conventional perturbation theory . due to the lack of symmetry in individual configurations , the series can only be summed in matrix form . writing v^ij__1_2,_1_2 = u_ij__1_2 _ _ 2_1 , [ chi10 ] the complete rpa corrected expressions for the spin and charge susceptibilities are given by ^ij _ & = & 1tm^i__m^j _ _ + _ ^ij_,(i=0)_^ _ ^_ + & & [ chi11 ] and ^ij & = & 1 t + & & + _ ^ij_,(i=0)__ _ . [ chi12 ] finally the momentum resolved spin susceptibility takes the form _ ( q ) = 1_ije^iq(r_i - r_j)^ij _ , [ chi13 ] and similarly for the charge susceptibility . ( measurements of the rpa corrected susceptibilities require the inversion of small matrices , but the effect on overall calculational efficiency is negligible . ) the effective action in eqn . [ expect4 ] has been simulated using the metropolis algorithm . in comparison with a classical heisenberg spin model , the present theory has several complications : ( i ) the theory has three degrees of freedom ( orientation on the unit sphere and amplitude ) at each site , rather than two . ( ii ) local field updates induce global changes in the action through the fermi partition function @xmath96 . ( iii ) a hartree fock convergence step is necessary after each spin field update to keep the `` slave '' charge fields @xmath30 on the saddle point . ( iv ) single site updating ( i.e. , glauber dynamics ) is insufficient to achieve efficient equilibration for @xmath117 . an analog of kawasaki dynamics , or spin swapping , is required , since the doping @xmath118 plays a role similar to a conserved order parameter . we consider each complication in turn . ( i ) the natural physical interpretation of field fluctuations in terms of amplitude and orientation suggests an optimal representation for the functional integral over field configurations . at half filling it is expected that the only important fluctuations for @xmath119 correspond to magnons , i.e. , slow modulations of the field orientation for fixed amplitude . a cartesian representation of the fields @xmath120 with independent component updates leads to a small acceptance ratio in the low temperature limit . therefore we write d^i_xd^i_yd^i_z = _ 0^(r^i)^2dr^i_-1 ^ 1 d(^i)_0 ^ 2d^i , [ measure1 ] choosing as independent variables @xmath121 , @xmath122 , and @xmath123 . variations in the measure @xmath124 at each site are incorporated into the metropolis updating factor , and moves are restricted , in a way consistent with detailed balance , to lead to @xmath125 . proposed changes to @xmath121 are restricted to a maximum scale @xmath126 , while large scale changes are permitted in @xmath127 and @xmath123 . \(ii ) for the initial monte carlo calculations reported here we have simply diagonalized the effective hamiltonian matrix @xmath54 , which has row dimension @xmath42 , for each proposed @xmath34 configuration . the time for a single site update thus grows as the cube of the system size , and the time for a lattice sweep as the fourth power of the system size . we believe that a hybrid molecular dynamics / monte carlo algorithm can be developed to significantly reduce the time needed for the simulation of large lattices @xcite . \(iii ) the hartree fock treatment of the static charge interaction is an integral component of the present approach . more sophisticated treatments of fluctuations would also in principle involve a self consistent calculation in the presence of symmetry breaking background fields . for the parameter sets studied in the present paper , the hartree fock iteration loop generally converges in three to four steps . since each step requires a matrix rediagonalization , overall calculational time scales with the average number of hartree fock iterations . \(iv ) at half filling the hartree fock charge field convergence becomes trivial , and @xmath36 remains equal to unity at all sites in all configurations . this just reflects the particle hole symmetry of the model for this special parameter set . away from half filling the situation is quite different . typical configurations contain a few sites with small fields , while the majority of sites have fields with amplitudes close to the half filled value . conventional single spin updating tends to leave `` holes '' trapped at isolated locations , since moving a hole requires passing through high energy intermediate configurations . in this sense the number of hole @xmath118 functions like a conserved order parameter , preventing the equilibration of the system by glauber dynamics . to combat this effect we have introduced a second type of updating move , viz . , a swap of the spin fields @xmath34 at two sites chosen at random . when the spin fields are swapped , the `` slave '' charge fields are dragged along , and the system equilibrates to a uniform set of site occupancies @xmath128 , even though the hole distribution in individual configurations is highly non uniform . the results for @xmath8 periodic lattices reported below were obtained on an ibm powerpc work station and a sun3000 server . for all runs reported here , a total of 100 monte carlo warmup sweeps and 5000 sampling sweeps were carried out . runs on @xmath129 and @xmath130 lattices are now in progress on a sun10000 work station array at the san diego supercomputer center . the total density of states for spin @xmath62 , @xmath131 , is plotted in figure 5 for @xmath132 and in figure 6 for @xmath133 , both at temperature @xmath134 . the chemical potential @xmath135 is lowered from @xmath136 ( half filling ) as holes are doped into the insulating state . the mean site occupancy @xmath137 and double occupancy @xmath109 are summarized in table 1 . note that the double occupancy at half filling is reduced by factors of order two and six from the value 0.25 for a non interacting system . a prominent mott hubbard gap , or pseudogap , appears in the density of states for both parameter sets at half filling ( figures 5a and 6a ) . the remnants of the discrete quantum energy levels in the non interacting system remain visible . note again that these systems are not magnetically ordered : the spin up and spin down densities of states are identical within statistical sampling errors , and the spin susceptibility is isotropic , as discussed below . as the chemical potential is pulled below @xmath138 , the site occupancy changes slowly at first , particularly for the large charge gap which appears when @xmath133 ( figures 6b c ) . this is just what one would expect for rigid band doping in a semiconductor , and such behavior has been observed in previous lanczos @xcite and quantum monte carlo @xcite studies of the hubbard model on small lattices @xcite . note that the rigid band picture begins to break down as soon as the site occupancy drops appreciably below unity ( figure 5b ) . in the doped system with @xmath132 and @xmath139 , spectral weight has been drawn from the lower edge of the upper hubbard band into the gap . due to this transfer of weight , the gap feature rapidly collapses with doping , though remnants remain for a wide range of fillings ( figures 5c d ) . by the time @xmath135 reaches 0.4 for @xmath132 , the shape of the non interacting band has begun to re emerge , though levels are still significantly broadened by the fluctuating spin background . this broadening is especially evident in the evolution of the k resolved spectral density for points near the fermi surface . recall that a special symmetry of the @xmath8 lattice induces an equivalence between the generally distinct k points @xmath140 and @xmath141 . this means all points on the half filled fermi surface have the same @xmath101 for the small systems discussed here . to contrast the behavior of points near and far from the fermi surface , we plot @xmath101 for @xmath142 and @xmath143 at @xmath132 in figures 7 and 8 . while the @xmath144 point at the top of the band ( figure 8) has a sharp quasiparticle like spectral density whose profile changes only slightly with doping , the @xmath140 point behaves quite differently . at half filling @xmath145 for this point is symmetrically split between the upper and lower hubbard band edges , just as one would expect in an antiferromagnetic hartree fock solution @xcite . away from half filling , the spectral density remains exceptionally broad , reflecting the continued presence of strong short range antiferromagnetic correlations between the fluctuating spins in the doped system . as noted above in the discussion of @xmath146 , remnants of the half filled gap remain for a wide range of fillings . the spectral broadening decreases for fillings far from unity , consistent with the restoration of the non interacting band profile ( and presumably the recovery of fermi liquid behavior ) . a broadening of the spectral density is also present in perturbative studies , including self consistent field methods such as flex @xcite ; however , the details of the broadening mechanism are quite different in the present case . flex and phenomenological spin fluctuation models @xcite assume a small amplitude scattering mechanism which may be treated by perturbation theory about a non interacting ( i.e. , fermi liquid ) background . in contrast the present approach treats scattering by large amplitude background distortions , with highly non gaussian static fluctuations in both amplitude and orientation ; these distortions evolve smoothly into the ordered antiferromagnetic background at half filling and zero temperature . the behavior of the spectral density at the @xmath140 point for @xmath133 is illustrated in figure 9 . as expected , the spectral weight is split symmetrically at half filling , and the two peaks are separated by an energy of order @xmath1 . in contrast with the @xmath132 result , the weight is distributed broadly throughout the upper and lower hubbard bands . signs of a more narrow coherent feature in the doped lower band are present for @xmath147 ( figure 9c ) . the momentum dependence of the static charge and spin susceptibilities is plotted in figures 10 and 11 for @xmath132 and 8 at temperature @xmath134 . note that the uniform charge susceptibility ( or compressibility ) is exponentially small at half filling , as expected for a system with a charge gap . in addition the compressibility increases very rapidly with doping : note particularly the increase for @xmath133 ( figure 11a ) . for general fillings @xmath111 is relatively small ( note the difference in scales for the charge and spin susceptibility ) and structureless at this temperature . this is expected in such a small system . possible manifestations of charge ordering within the fluctuating spin background may become apparent , however , in larger systems . in contrast the spin susceptibility is large and strongly peaked at the antiferromagnetic wave vector @xmath144 . the peak is expected to move off @xmath144 for doped systems on larger lattices . for the parameter sets illustrated here , the spin susceptibility is dominated by the contributions of the `` anomalous '' term _ ije^iq(r_i - r_j ) m^i _ _ [ anom ] the @xmath148 component of @xmath149 is actually plotted in figures 10 and 11 . for 5000 lattice sweeps the agreement between the @xmath150 , @xmath151 , and @xmath152 components of the @xmath144 spin susceptibility is at the level of the statistical error bars , i.e. , a few percent . ( the agreement can be arbitrarily improved by increasing the monte carlo sampling time . ) the present approach consists of simulation of the action for a set of local static fields coupled to electron spin degrees of freedom . when the action is constrained to be at its temperature dependent global saddle point ( if such a point indeed exists in the most general case ) , these static fields are determined by inhomogeneous hartree fock equations and reduce to the anomalous fields of conventional diagrammatic perturbation theory . just as hartree fock theory with an anomalous self energy may be extended using the baym kadanoff prescription @xcite , there is no formal barrier preventing such an extension in the present case . the computational cost of carrying out self consistent field ( scf ) perturbation theory in a broken symmetry background is quite large , however , since total momentum and spin are no longer good quantum numbers . in this section we sketch out the steps necessary for a generalized approach and note several of the technical pitfalls which must be studied and overcome . the form for the scf action must be determined separately for each configuration of the @xmath25 fields . in full analogy with baym kadanoff theory for a uniform system , the action takes the form s^scf ( ) = _ i|^i|^2 - tr(-g^scf _ ) ^-1 + , [ actionfourteen ] where ( ) = -tr(g^scf ) + [ freeenergy ] and = _ (xx ) . [ phi ] in these formal expressions @xmath153 is the discrete time propagator ( retained to insure a proper high frequency regularization of the @xmath154 term ) ; x ( r^i , ) , [ xx ] and @xmath155 is the scf generating functional ( e.g. , a sum of ring and ladder diagrams in flex ) . since the scf action may be interpreted as @xmath156 , with @xmath157 a well behaved free energy , the statistical weighting factor @xmath158 is positive definite . this is in sharp contrast to the behavior in imaginary time quantum monte carlo simulations , where the weighting factor can be positive or negative with nearly equal probability . what is not so clear is the nature of approximations which insure the stability of @xmath159 , particularly for exponentially rare field configurations . the interaction which enters @xmath160 is not the original hubbard interaction , since the zero frequency spin components have been removed for special treatment . this means that there are no obvious zero frequency instabilities left to appear in @xmath159 . behavior at finite frequencies must also be addressed , however , and we are not prepared to comment on the stability of general baym kadanoff approximations ( and in particular flex ) in the present paper . as an example of the baym kadanoff prescription in practice , note that when the static charge interaction is treated within hartree (xx ) = un^i__ii_ _ . [ sighf ] the appropriate generating functional is = u_in^i_^2 . [ phii ] furthermore ( g^scf ) = 2[trace ] and so & = & - + & = & -u_in^i_^2 . [ calf ] note that this form for the hartree fock `` constant '' in the action is exactly that which has appeared previously in eqn . [ actiontwelve ] . we believe the present hubbard model analysis , based on monte carlo simulation of a classically disordered local spin field , incorporates several desirable features which have eluded previous perturbative studies . most important the present theory has classical critical behavior at half filling and describes the formation of a mott hubbard gap ( or pseudogap ) in a system with a finite spin correlation length . in addition the spectral densities for the doped systems have the features suggested by exact methods . the present static field analysis can be extended to significantly larger systems . since the computations involve monte carlo field sampling and have reasonably small memory requirements , it would be natural to carry out studies on multiple processors with different random seeds , then to merge the equilibrated results from numerous short runs . it is almost unavoidable to speculate on the implications for charge density wave or stripe formation @xcite and d wave singlet pairing in the cuprate superconductors . since the present approach reduces to standard inhomogeneous hartree fock for @xmath4 , hartree fock results previously obtained for stripe formation should re emerge @xcite . by allowing sampling of configurations removed from the zero temperature saddle point in a well defined way , the present approach provides a framework to describe the destruction of striped order by thermal fluctuations . to study potential d wave superconductivity it seems essential to incorporate the exchange of low frequency spin fluctuations about the disordered background configurations . the low energy fluctuations which become magnons in the directions transverse to an ordered spin state ( i.e. , the goldstone modes of the order parameter field ) must evolve into a broadened spectrum of fluctuations in the important disordered configurations . if low frequency spin fluctuations are important , the pair order parameter should exhibit retardation effects , in contrast with the spin order parameter . in addition the disordered background has important implications for the onset of pairing . presumably the pair field appears initially with a highly non uniform distribution in a small subset of @xmath34 configurations . such behavior would imply superconducting precursor effects in the density of states well above the transition temperature for global phase coherence . the effects would be more akin to the behavior of a system of superconducting grains with variable grain size than to conventional kosterlitz thouless phase fluctuations in a clean two dimensional superconductor . in any case the scenario described above is at this stage purely speculative . the next steps in the development of the present approach consist of ( i ) studies of larger systems ; ( ii ) analysis of the spin and charge response at finite frequencies , with particular care given to the treatment of potential instabilities in rare field configurations ( see section 6 ) ; and ( iii ) analysis of pairing eigenvalues and eigenvectors from exchange of fluctuations in sets of non uniform configurations . we gratefully acknowledge conversations with s. haas , w. hanke , s.a . kivelson , h. monien , s. moukouri , r.t . scalettar , r.l . sugar and s.c . zhang during the course of this work . neb acknowledges support from the national science foundation under grant no . dmr9811381 , and djs acknowledges support from the u.s . department of energy under grant no . de fg0385er45197 . the behavior of high order perturbation theory for the anderson impurity model seems intimately connected with the re emergence of a fermi liquid description for @xmath174 . see , e.g. , a.c . hewson , phys . @xmath175 , 4007 ( 1993 ) . we suggest that perturbation theory in @xmath1 fails to resum properly for the anderson impurity model at temperatures @xmath176 , where a stiff local moment exists . the @xmath24 correction is present because equivalent components in the order parameter field appear in crossed scattering channels . in contrast no such correction appears when one decouples a general instantaneous pairing interaction or a holstein phonon interaction . while the @xmath24 term must be retained to properly regain the high temperature limit , it seems clear that it has no special physical implications at low temperatures . further the correction may at least formally be treated exactly by introducing a `` strong coupling '' ( but still @xmath23 ) generalization of the @xmath25 fields . it is possible to provide a heuristic derivation of a one electron self energy with the general form in eqn . [ perturb ] starting from the full theory for @xmath44 in section 2 . the steps are as follows : ( i ) write down the expression for the one electron propagator dictated by eqn . [ expect ] . ( ii ) expand the expression for @xmath184 within the @xmath25 integral as a power series in @xmath185 , noting that the odd order terms integrate to zero . ( iii ) resum the series within the integral as @xmath186^{-1}$ ] , where @xmath187 takes the form of a self energy second order in @xmath25 . ( iv ) approximate the resulting field integral as @xmath188^{-1}$ ] , i.e. , average the second order self energy , rather than the propagator . the field averaged self energy assumes exactly the form of eqn . [ perturb ] , with @xmath84 proportional to the full static spin susceptibility of the system . the electron propagator which appears in the self energy is bare for the reason discussed in the text . step ( iv ) is of course not justifiable , since the action is strongly non gaussian , but it provides the phenomenological link between some previous theories and the current work .

we study the local behavior of gravitational lensing near fold catastrophes . using a generic form for the lensing map near a fold , we determine the observable properties of the lensed images , focusing on the case when the individual images are unresolved , i.e. , microlensing . allowing for images not associated with the fold , we derive analytic expressions for the photometric and astrometric behavior near a generic fold caustic . we show how this form reduces to the more familiar linear caustic , which lenses a nearby source into two images which have equal magnification , opposite parity , and are equidistant from the critical curve . in this case , the simplicity and high degree of symmetry allows for the derivation of semi - analytic expressions for the photometric and astrometric deviations in the presence of finite sources with arbitrary surface brightness profiles . we use our results to derive some basic properties of astrometric microlensing near folds , in particular we predict for finite sources with uniform and limb darkening profiles , the detailed shape of the astrometric curve as the source crosses a fold . we find that the astrometric effects of limb darkening will be difficult to detect with the currently planned accuracy of the _ space interferometry mission ( sim ) _ for galactic bulge sources ; however , this also implies that astrometric measurements of other parameters , such as the size of the source , should not be compromised by an unknown amount of limb darkening . we verify our results by numerically calculating the expected astrometric shift for the photometrically well - covered galactic binary lensing event ogle-1999-bul-23 , finding excellent agreement with our analytic expressions . our results can be applied to any lensing system with fold caustics , including galactic binary lenses and quasar microlensing . # 1equation ( [ # 1 ] ) # 1eq . [ # 1 ]

astro-ph0112531

gravitational lensing has proven to be an exceptional tool for studying a diverse set of astrophysical phenomena . its utility is due , at least in part , to the fact it operates in a number of qualitatively different regimes . the term strong lensing , or macrolensing , is usually applied when a distant source ( typically cosmological ) is lensed into multiple , resolved images by an intervening mass , such as a foreground cluster or a galaxy . weak lensing is used to refer to the case when multiple images are not created , and the gravitational field of the intervening matter serves only to slightly distort the image of the source . for most applications of both strong and weak lensing , the source , lens , and observer can be considered static . the term microlensing is often used to describe the case when multiple images are created , but are not resolved . typically the separation of the images created by a gravitational microlens are of order the einstein ring radius , @xmath0 here @xmath1 is the mass of the lens , @xmath2 is defined by , @xmath3 , and @xmath4 , @xmath5 , and @xmath6 are the distances between the observer - source , observer - lens , and lens - source , respectively . in cosmological contexts , angular diameter distances should be used . when @xmath7 is less than the resolution , individual images in general can not be distinguished . due to the small scale of @xmath7 , it is typically not a good approximation to assume that the source , lens , and observer are static . therefore the lensing properties can be expected to change on timescales of order the einstein ring crossing time , @xmath8 where @xmath9 is the transverse speed of the lens relative to the observer - source line - of - sight . the standard observables in gravitational microlensing are therefore the time rate of change of the total magnification and center - of - light ( centroid ) of all the microimages . there are two different regimes where microlensing has been discussed : quasar microlensing @xcite and microlensing in the local group @xcite . in the local group , gravitational microlensing occurs whenever a massive , compact object passes close to our line of sight to a more distant star . microlensing was originally suggested as a method to detect baryonic dark matter in the halo of our galaxy @xcite , but has been developed and applied as an important tool in studying a number of astrophysical topics , including the stellar mass function @xcite , extrasolar planets @xcite , stellar atmospheres @xcite , and stellar multiplicity @xcite . the only microlensing effect currently observable is the magnification of the background source as function of time . this is because , for typical distances in the local group , the angular einstein ring radius is @xmath10 and therefore too small to be resolved with current instruments . the timescale for a microlensing event is @xmath11 . in general , it is much easier to determine the center - of - light of an image than it is to resolve it . thus future interferometers , such as the _ space interferometry mission ( sim ) _ , although still not able to resolve separations of @xmath12 , should be able to measure the centroid of all the images to much better than this , perhaps even down to @xmath13 in the case of _ sim_. such accuracy is sufficient to easily detect the motion of the centroid of the images created in a microlensing event , which is also of order @xmath7 . this regime is typically referred to as astrometric microlensing , as opposed to photometric microlensing when only the total magnification is observable . astrometric microlensing has a number of important applications . by combining ground - based photometry of microlensing events with photometry and astrometry from an astrometric satellite on an earth - like orbit , the masses of microlenses can routinely be measured @xcite , allowing for the determination of the compact object mass function in the bulge , including stellar remnants @xcite . astrometric information alone allows for the precise ( few % ) measurement of the masses of nearby stars @xcite . finally , for a subset of events , it will be possible to obtain precision measurements of angular diameters of stars in the galactic bulge using astrometric information @xcite . binary microlenses have proven to be enormously useful in photometric microlensing studies . this is primarily because binary lenses exhibit caustics : closed curves on which the mapping from the lens plane to light source plane becomes critical , and the point - source magnification becomes formally infinite . regions near caustics exhibit large , rapidly changing ( with respect to source position ) magnification , and are therefore useful not only for providing a large source flux , but also high angular resolution . however , binary lenses have also proven to be difficult to study both theoretically and observationally . this is partially because the lens equation , which describes the mapping from the lens plane to the light source plane , is equivalent to a fifth - order complex polynomial in the source position @xcite , and therefore is not analytically solvable in general . furthermore , care must be taken when considering finite source effects near caustics due to the divergent magnification . however , considerable progress can be made when one realizes that the smooth arcs ( away from cusps ) of caustics that arise in nearly equal - mass binary lenses are well approximated as simple linear fold catastrophes , which have generic , simple , and most importantly , _ analytic _ behavior . thus the caustics of binary lenses can be analyzed analytically or semi - analytically without reference to the global ( and non - analytic ) topology of the general binary lens . in particular , a simple equation for the magnification of a source near a fold exists @xcite , which has been used in a number of important applications including binary - lens fitting @xcite , stellar atmospheres @xcite , and caustic - crossing predictions @xcite . in contrast to the astrometric behavior of single lenses , which is analytic and has been quite well studied @xcite , there have been only a few preliminary studies of the astrometric properties of binary gravitational lenses @xcite . it is known that astrometric binary - lens curves exhibit complex behavior , including instantaneous @xmath14 jumps in the image centroid trajectory that occur when a point source crosses a binary - lens caustic and two highly - magnified images appear in a position unrelated to the position of the centroid of the other three binary - lens images . the generic behavior of these centroid jumps , or how they are altered by finite source effects , is not understood . as is the case for photometric microlensing , the astrometric behavior of binary lenses will likely prove quite useful for several applications . the usefulness of binary lenses is primarily related to the complex image centroid trajectories and the large centroid jumps . although , in general , these properties do not allow one to measure any additional parameters over the single lens case ; they do allow one to measure these parameters much more easily . in particular , @xcite have shown that lens mass measurements can be made to a given accuracy with 1 - 3 orders of magnitude fewer photons with caustic - crossing binary - lens events than with single lens events , thus greatly reducing the resources required to achieve one of the primary proposed science goals of astrometric microlensing . caustic crossing binary - lens events are also enormously useful for measuring the angular radii of microlensing source stars in the bulge , for two reasons . first , the expected ratio of binary - to - single lens events for which the source star is resolved is a factor of @xmath15 for giant sources and @xmath16 for main sequence sources . furthermore , the large and complex centroid shifts expected for caustic - crossing binary - lens events makes the requisite astrometric measurements easier . @xcite have shown that , by combining accurate ground - based photometry with a handful of precise astrometric measurements , caustic - crossing binary - lens events should yield @xmath17 stellar radius measurements with reasonable expenditure of resources . thus , given the importance of caustic crossing binary - lens events , an analytic study of the generic behavior of astrometric microlensing near folds would prove quite useful . in quasar microlensing , the separate macroimages of a quasar that is multiply - imaged by a intervening galaxy or cluster also feel the combined , non - linear effect of individual point masses ( i.e. stars ) in the macrolensing object that are near the macroimage position . the individual macroimages are in fact composed of many , unresolved microimages with separations of order the einstein radius of a @xmath18 object at cosmological distances , @xmath19 . the typical timescale for the source to cross an angle of @xmath7 is @xmath20 ; however microlensing light curves should show structure on much smaller timescales due to the combined effects of many individual microlenses . since it was first discussed by @xcite , cosmological microlensing has been studied theoretically by numerous authors ( see @xcite and references therein ) , and detected in at least two systems ( q2237 + 0305 , @xcite , b1600 + 434 , @xcite ) . observations have been used to place constraints on , e.g. , the size of the emitting region of quasars @xcite and the mass function of microlenses @xcite . quasar microlensing differs markedly from microlensing in the local group in that the surface mass density in units of the critical density for lensing ( the `` optical depth '' ) is of order unity , rather than @xmath21 for the local group . in the high optical depth regime the lensing effects of the individual microlenses add nonlinearly , resulting in a complex caustic network . due to this nonlinear behavior and the large number of lenses typically involved , calculation of the observable properties of such a lensing system is difficult and time consuming . although in the high optical depth regime the caustics often exhibit considerably more complicated global behavior than the caustics of binary lenses in the local group , it is still the case that the smooth arcs ( away from cusps ) of the caustics are locally well - approximated by generic fold catastrophes . this fact , combined with a simple formula for the magnification near folds , has been exploited by numerous authors to quickly and efficiently calculate various observable properties of quasar microlensing @xcite . the observable effects of quasar microlensing have been limited to the relative magnifications of the various macroimages as a function of time . as with local group microlensing , astrometric effects should also be present . the centroid of the individual macroimages should vary as a function of time , particularly when new images are created or destroyed when the source crosses a caustic . this effect has been studied by @xcite and @xcite . in particular , @xcite predict that magnitude of the centroid shift for the q2237 + 0305 system ( apparent magnitude @xmath22 ) can be as large as @xmath23 , and thus potentially observable with _ sim_. they also note that the magnitude of the centroid shift is often correlated with the magnitude of the change in total magnification . the analytic results presented in this paper may prove useful for this application . here we study the generic , local behavior of microlensing near fold catastrophes . in 2 we present an analytic study of the photometric and astrometric behavior near folds . we begin with the equations that describe the mapping near a fold caustic in 2.1 , and use these to derive the behavior of a point source near a fold . we extend this analysis to finite sources in 2.2 , and limb - darkened sources in 2.3 . in 2.4 we show how and when our generic parabolic fold form reduces to the more familiar linear caustic . in 2.5 we use our analytic results to derive some generic results about the astrometric behavior near folds . we verify the applicability of our results in 3 by numerically calculating the photometric and astrometric behavior of one well - observed binary - lens event . we find excellent agreement with our analytic formulae . finally , we summarize and conclude in 4 . our goal is to provide a thorough , comprehensive study of gravitational microlensing near fold caustics . although our study is interesting in its own right , the primary utility of the results presented here is their potential application to the topics mentioned in the previous paragraphs . a prescription for how specifically our results can be applied to these topics is beyond the scope of this paper , but we will make general comments along these lines over the course of the paper , and more specific comments in 2.5 . we are currently preparing a complementary , similarly detailed study of microlensing near cusps . combined with this study , we will have a reasonably thorough and complete understanding of the local behavior of microlensing observables near all stable gravitational lensing singularities . we note that some of the results derived here , particularly the results on the photometric behavior near folds , have been presented elsewhere ( see , e.g. @xcite , @xcite , and @xcite ) . we include those results here for the sake of completeness . for a general gravitational lens , the lensed images @xmath24 of a source @xmath25 , are given by the solutions of the lens equation , which is the mapping @xmath26 : @xmath27 where @xmath28 , and @xmath29 is the projected newtonian potential of the lens , @xmath30 where @xmath31 , @xmath32 is the surface density of the lens , and @xmath33 is the critical surface density for lensing . note that @xmath34 . we are assuming that @xmath35 and @xmath36 , where @xmath37 and @xmath38 are the proper vector positions in the lens and light source planes , respectively . the mapping @xmath39 can produce multiple images of the source @xmath25 . the magnification of lensed image @xmath40 is @xmath41| } , \label{eqn : magoneimage}\ ] ] where @xmath42 is the jacobian matrix of the lensing map @xmath39 . the critical curve is the set of positions @xmath43 such that the determinant of the jacobian matrix vanishes , i.e. where @xmath44=0 , \label{eqn : jdef}\ ] ] and the caustics are @xmath45 . in the case of microlensing , the individual images are by definition unresolved , and thus it is useful to define the total magnification @xmath46 where the sum is over all images . the center - of - light , or centroid @xmath47 of the images is simply the magnification weighted sum of the image positions , @xmath48 for simplicity , we will focus primarily on the quantity @xmath49 . however , it is important to note that , in general , centroid measurements will be made with respect to the unlensed source position , thus the observable is @xmath50 also , while the angular variables we will be working with will be in units of @xmath7 , the astrometric observables are in physical units , such as arcseconds . to convert to observable quantities , all angular quantities must be multiplied by @xmath7 . for example , the observable centroid shift is given by @xmath51 . we now derive the generic behavior of the photometric and astrometric properties of gravitational lensing near folds . we will present our derivations in some detail , in order to document the approximations and simplifying assumptions that are implicit in the final analytic expressions . in figure 1 , we provide an illustrative example of the basic properties of lensing near a fold . we will refer to this figure repeatedly during the course of the derivations . suppose that the lensing map @xmath39 sends the origin to itself ( which can always be accomplished by appropriate translations ) and a fold caustic curve passes through the origin . by taylor expanding the gravitational potential @xmath29 about the origin , one can find an orthogonal change of coordinates that is the same in the lens and light source planes such that the lensing map @xmath39 can be approximated by the following mapping in a neighborhood of the origin ( petters , levine , & wambsganss 2001 , pp . 341 - 353 ; schneider , ehlers , & falco 1992 , p. 187 ) : [ eq - le - fold ] u_1 = a _ 1 + ^2_2 + c _ 1 _ 2 , u_2 = ^2_1 + b _ 1 _ 2 + ^2_2 , where @xmath52 and @xmath53 denote the respective coordinates in the lens and light source planes , and a = 1 - _ 11 ( ) 0 , b = -_122 ( ) , c = - _ 112 ( ) , d = - _ 222 ( ) 0 . here the subscripts refer to the partial derivatives of @xmath29 with respect to the original global cartesian coordinates of the lensing map . for the example in figure 1 , we have adopted @xmath54 , @xmath55 , and @xmath56 . the jacobian matrix of is a = . the critical curve is given by j a = ( a + c _ 2)(b _ 1 + d _ 2 ) - ( c _ 1 + b _ the tangent line to the critical curve at the origin is given by [ eq - gradj ] 0 = j ( ) = ab _ 1 + ad _ 2 , that is , [ eq - tanline - cc ] _ 2 = - _ 1 since @xmath57 and @xmath58 . substituting @xmath59 into yields u_1 = a _ 1 + ( b^2 - cd ) ^2_1 a _ 1 , u_2 = ( cd - b^2 ) ^2_1 . note that in the expression for @xmath60 , the term @xmath61 dominates @xmath62 near the origin . inserting @xmath63 into @xmath64 above , we see that the tangent line at the origin of the critical curve is mapped into a parabola @xcite : c ( u_1,u_2 ) 2a^2d(u_2 -e u_1 [ eq - c ] where we have introduced the combination of local derivatives of @xmath29 : e . [ eq - curvature - e ] we show in [ sec : simple ] that @xmath65 is one - half the curvature of the caustic at the origin . thus when @xmath66 , the caustic can be approximated as @xmath67 , i.e. the @xmath60-axis . for the example shown in figure 1 , @xmath68 . since the tangent line ( ) approximates the critical curve near the origin , the parabola ( ) approximates the caustic near the origin . see figure 1(a , b ) . now multiplying @xmath64 in by @xmath69 and substituting @xmath63 yields [ eq - c-2 ] c ( u_1,u_2 ) = ( b u_1 + ad _ 2)^2 . hence , if a light source is located at a position @xmath70 where @xmath71 , then there is no lensed image locally , while a source with @xmath72 has at least one image . for an @xmath73-point mass lens , we have @xmath74 . consequently , if @xmath75 , the parabolic caustic lies in the lower - half plane locally . in addition , the region above the parabola is such that sources lying there have double images locally ( fig . if @xmath76 , then parabola is in the upper - half plane with the region below the parabola yielding double images locally . in other words , the caustic is locally convex ( see petters et al . 2001 , sec 9.3 for a detailed treatment ) . let us determine the images for @xmath72 . equation ( [ eq - c-2 ] ) is equivalent to a quadratic equation in @xmath77 , 0 = [ a^2 d^2 ] _ 2 ^ 2 + [ 2abd u_1 ] _ 2 + [ b^2 u_1 ^ 2 - c(u_1,u_2 ) ] . the solutions are in front of the square root since @xmath78 . ] [ eq - image-2 ] _ 2 = . the expression for @xmath60 in yields [ eq - th1 ] _ 1 = ( u_1 - ^2_2 ) . ignoring terms of order 3 or higher , we obtain _ 1 = - u_1 _ 2 - _ 2 ^ 2 , [ eq - th1 - 2 ] where the approximation @xmath79 was employed . substituting into ( [ eq - th1 - 2 ] ) and keeping only terms that are linear in @xmath60 and @xmath64 , it follows that [ eq - image-1 ] _ 1 = . hence , a source with @xmath80 has two ( opposite parity ) images given by [ eq - image - positions ] _ ( _ , 1 , _ , 2 ) = ( d u_1 - b u_2 , -b u_1 ) . figure 1(a , b ) illustrates the mapping from source to images near a fold . consider a point @xmath81 on the caustic , i.e. , @xmath82 . locally there is one image of @xmath81 , located on the critical curve , and close to the origin , @xmath60 dominates @xmath83 , and thus the image position is @xmath84 , which is indeed a point on the critical line ( see ) . ] at the position _ c = ( d u_c,1 - b u_c,2 , -b u_c,1 ) . [ eq - image - cc ] now consider a source at @xmath85 inside the caustic ( i.e. , @xmath86 ) and let @xmath87 be the point on the caustic with the same horizontal position as @xmath25 , i.e. , @xmath88 and @xmath89 . the vertical separation of the source from the caustic is then the difference between the @xmath64-coordinates of @xmath25 and @xmath90 : [ eq - vsepartion ] u_2 - e ( u^*_c,1)^2 = . note that @xmath91 if and only if @xmath75 ( since the source is located where @xmath92 ) . using equations ( [ eq - image - positions ] ) and ( [ eq - image - cc ] ) , we find that the distances of the images @xmath93 from the point @xmath43 on the critical line are the same , [ eq - image - distance ] =| _ ( u_1,u_2 ) - _ c | = ^1/2 . near the origin , the first term in brackets dominates over the second , and thus , [ eq - absimage - distance ] = . using the earlier tangent line approximation to the critical curve at the origin , we taylor expand the jacobian determinant @xmath94 about the origin to first order : j(_1 , _ 2 ) = j__1 ( ) _ 1 + j__2 ( ) _ 2 = ab _ 1 + ad _ 2 , which is equal to the right - hand - side of ( [ eq - gradj ] ) . inserting the expressions for @xmath61 and @xmath77 from , we obtain j(_1 , _ 2 ) = -u_2 . keeping only the lowest order terms in @xmath60 and @xmath64 gives , the caustic ( ) , the magnification of the images are given locally by [ eq - mag ] _ ( _ ) = = , where @xmath95 . thus the two images have the same magnification if the source is sufficiently close to the caustic . the total magnification @xmath96 of the two images is simply _ f = , [ eq - muf ] which agrees with the expressions derived by both schneider , ehlers & falco ( 1992 , p190 ) and @xcite . equation ( [ eq - muf ] ) is the well - known result that the magnification varies inversely as the square - root of the distance of the source from the caustic . it is not often appreciated however , that this property holds for the general parabolic fold form as well , provided that the _ vertical distance _ is used , rather than , for example , the minimum distance between the source and the fold caustic . rigorously , the vertical distance is the distance between the source and the caustic in the direction perpendicular to the tangent line of the caustic at the origin , where the origin is defined as the point around which the potential @xmath29 is taylor expanded . in the limit of a straight fold ( @xmath66 ) , the vertical distance and the minimum distance are equivalent . write the source position @xmath25 in terms of the fold local coordinates , i.e. , @xmath85 . there are no local images for source a position with @xmath97 . from equations ( [ eqn : cl ] ) and ( [ eq - image - positions ] ) , and since @xmath98 , for sources inside the caustic the centroid is given locally as [ eq - centroidf ] _ f ( ) = ( _ + + _ - ) = ( d u_1 - b u_2 , - b u_1 ) , c ( ) 0 . we express the rectilinear motion of the source as follows : [ eq - trajectory ] ( t ) = _ c + ( t - t_c ) , where @xmath81 is the position at which the source intersects the caustic at time @xmath99 and @xmath100 is the constant angular velocity vector of the source , = ( , ) . recall that @xmath101 , where @xmath9 is the transverse speed of the lens relative to the observer - source line - of - sight . here @xmath102 is the angle of the source s trajectory with respect to the @xmath60-axis . note that the @xmath60-axis does not necessarily coincide with the caustic , and thus @xmath102 is not necessarily the angle of the trajectory with respect to the tangent to the caustic at @xmath81 . we shall assume that @xmath103 corresponds to the source s trajectory lying in the double - image region . in that region , the centroid follows a straight line locally : [ eq - centroidf2 ] _ f ( t ) = _ f , c + ( t - t_c ) _ ( f , c ) , t t_c where @xmath104 and @xmath105 ( both constant vectors ) . note that since the slope of the centroid line @xmath106 is , @xmath107 , where @xmath108 is the angle between the centroid line and the @xmath61-axis , it follows that @xmath109 can be expressed as _ ( f , c ) = ( _ c , 1 ) . fold caustics do not , of course , exist in isolation . they are tied to the global properties of the lens in consideration . for practical purposes , we therefore consider images created by the lens that are not associated with the fold under consideration . we define @xmath110 to be the total magnification of all the images not associated with the fold , and we define @xmath111 as the centroid of all these images . we will assume that there is locally only one fold caustic , and that all the other image magnifications and positions are only slowly varying functions of the source position . the total magnification is then @xmath112 see figure 1(c ) . here @xmath113 is the heaviside unit step function ( i.e. , @xmath114 for @xmath115 and @xmath116 if @xmath117 ) ; it accounts for the fact that sources below the caustic ( with @xmath118 ) have no images locally ( i.e. , near the critical curve ) . since @xmath119 , we have , [ eqn : genbxcl ] = 1 [ _ f _ f + _ 0 _ 0 ] where _ f ( ) = ( d u_1- b u_2,- b u_1 ) ( ) . we can calculate the dependence of the observables @xmath120 and @xmath49 on time by assuming a rectilinear source trajectory and replacing @xmath25 in equations ( [ eqn : mutotexp ] ) and ( [ eqn : genbxcl ] ) by . we first taylor expand @xmath121 and @xmath110 about the time of the caustic crossing , @xmath122 , keeping terms of first order in @xmath123 : _ 0 & = & _ 0,c + ( t- ) _ ( 0,c ) , [ eqn : btfvt ] + _ 0 & = & _ 0,c + ( t- ) _ ( 0,c ) , where @xmath124 and @xmath125 . from equations ( [ eq - vsepartion ] ) and ( [ eq - trajectory ] ) , @xmath126 is given by = . the magnification as a function of time is then , @xmath127^{-1/2 } \theta ( t-\tcc ) \nonumber \\ & & \hspace{3.5 in } + \ \mu_{0,c } \ + \ \dot{\mu}_{(0,c ) } ( t-\tcc ) , \nonumber \\ \label{eqn : muvtime}\end{aligned}\ ] ] where @xmath128 is the effective rise time of the caustic crossing , and @xmath129 defines the characteristic rise length as before . notice that when the curvature of the caustic is small , i.e. , when @xmath66 , the magnification associated with the fold reduces to the more familiar form @xmath130^{-1/2}\theta(t - t_c)$ ] . the centroid of all images is @xmath131 + [ \mu_{f}\dot{\bt}_{(f , c)}\theta ( t-\tcc ) + \mu_0 \dot{\bt}_{0,c } ] ( t-\tcc ) \right\ } , \label{eqn : bxclvtime}\ ] ] where @xmath132 and @xmath109 are defined below . figure 1(e , f ) illustrates the behavior of the two components of @xmath133 as a function of time , whereas figure 1(g ) shows @xmath134 versus @xmath135 . the results of the previous section assumed a pointlike source . this results in an astrometric curve that exhibits an instantaneous jump from @xmath136 to @xmath132 at @xmath122 . all real sources will have a finite extent which will smooth out the discontinuous jump . for a finite source , the magnification is the surface brightness weighted magnification integrated over the area of the source , @xmath137 where @xmath138 is the disc - shaped region of the source and @xmath139 is the surface brightness of the source . let @xmath140 be the average surface brightness of the source , @xmath141 . here @xmath142 , is the angular source radius @xmath143 in units of @xmath7 . the denominator in is then simply @xmath144 . define @xmath145 to be the normalized surface brightness . also define a new set of source plane coordinates such that = , where @xmath146 is the position of the center of the source . then for a point @xmath25 inside the disc source @xmath138 , we have @xmath147 . equation ( [ eqn : fsgen ] ) becomes ^fs= _ d d ( ) ( ) . [ eq - mufsscale ] all of the preceding results apply to generic parabolic fold catastrophes . however , in order to continue making significant progress analytically , we must make the following simplifying assumption : we furthermore assume that @xmath66 , and thus the caustic coincides with the @xmath60-axis and @xmath148 . this considerably simplifies the form for the fold magnification @xmath149 . let @xmath150 be such that @xmath151 is the vertical separation of the center @xmath146 of the source from the caustic , i.e. , @xmath152 . the source @xmath138 is in the upper - half plane ( i.e. , @xmath153 ) if and only if @xmath154 . if @xmath155 , then @xmath138 just touches the caustic , while for @xmath156 a portion of @xmath138 is below the caustic with the center of @xmath138 on the caustic for @xmath157 . the center of @xmath138 lies below the caustic for @xmath158 , with a portion of @xmath138 still above the caustic for @xmath159 and @xmath138 completely below the caustic for @xmath160 . since the fold magnification @xmath149 is nonzero only for points @xmath25 in @xmath138 that lie in the upper - half plane , we get [ eq - mufp ] _ f ( ) = ( ) ^1/2 . by equations ( [ eq - mufsscale ] ) and ( [ eq - mufp ] ) , we find for a fold caustic and arbitrary surface brightness profile , _ f^fs ( z ) = ( ) ^1/2 . [ eqn : mufsgen ] for a uniform source , i.e. , @xmath161 , this simplifies to @xcite , @xmath162 where @xmath163 note that @xmath164 can be expressed as an elliptic integral . figure [ fig : fig2 ] shows @xmath165 for @xmath166 . for small source sizes @xmath167 , the magnification @xmath110 of the images not associated with fold is a slowly varying function of @xmath25 over the source , and thus @xmath168 therefore , the total finite source magnification is just _ tot^us=_f^us + _ 0,cn , [ eq - mtfs ] where @xmath169 in analogy to the point source case ( 2.1 ) , we can expand @xmath110 about the time of the second caustic crossing . for a source with rectilinear motion , the total finite source magnification in the time domain is then , @xmath170 where @xmath128 ( effective rise time of the caustic crossing ) and @xmath171 is the time scale of the caustic crossing , i.e. , the time between when the source first touches the caustic and when it straddles the caustic . note that @xmath172 . figure 1(d ) illustrates the behavior of @xmath173 . for an extended source , the image centroid is the position of the images weighted by both the surface brightness and magnification , integrated over the area of the source : @xmath174 where the sum is over all the microimages . the denominator is simply the total flux , @xmath175 . again , for @xmath167 , the centroid @xmath176 of the images not associated with the fold varies slowly over the source . this yields @xmath177 ( since @xmath178 is also slowly varying over the source ) . equation ( [ eqn : tclfsgen ] ) then separates into two terms : @xmath179 \ + \ { \mu_{0,cn } \over \mu_{tot}^{fs } } \ \bt_{0 , { cn } } \label{eqn : tclfs1 } \\ & = & { \mu_{f}^{fs } \over \mu_{tot}^{fs } } \ \bt_{f}^{fs } + { \mu_{0 , { cn } } \over \mu_{tot}^{fs}}\ \bt_{0 , { cn } } , \label{eqn : tclfs2}\end{aligned}\ ] ] where @xmath180 . the first term in ( [ eqn : tclfs2 ] ) is the contribution from the two images associated with the fold caustic , while the second term is the contribution from all unrelated images . for convenience , we defined @xmath181 to be the factor within the brackets in . the term @xmath181 will now be evaluated . since @xmath182 , we obtain _ f()=_f(_cn)+ _ * _ f( ) . define @xmath183 so @xmath184 vanishes when the disc source lies completely below the fold caustic . when @xmath185 , and thus @xmath186 is not defined for @xmath158 . this apparent discrepancy can be alleviated by simply assuming that @xmath187 is a function that is defined for all @xmath188 , i.e. , @xmath189 for all @xmath188 . ] using ( [ eq - mufp ] ) , we find a simple formula for the finite source image centroid , [ eq - theta - fs - f ] _ f^fs(t ) = _ f , cn + _ f ( _ s_n , _ s_n ) , where we have defined _ s_n ( z ) & = & _ max(-z,-1)^1 u_2 _ -^ u_1 u_1 ( u_1,u_2 ) , [ eq - ch ] + _ s_n ( z ) & = & _ max(-z,-1)^1 u_2 u_2 _ -^ u_1 ( u_1,u_2 ) . [ eq - ck ] in the case of a uniform source , we obtain for any profile with the symmetry @xmath190 . ] @xmath191 and @xmath192 , where _ n ( z ) ^-1/2 ( n+1 ) ! ( n + 1/2 ) ! _ max(-z,-1)^1 dx x ( 1-x^2)^n+1/2 ( x+z)^1/2 ( 1+z ) . [ eqn : kfuncn ] for a uniform source , the finite source centroid of the two images associated with the fold is then , [ eq - fscentroid - us - s ] _ f^us = _ f , cn - _ 0 ( z ) , where @xmath193 . inserting the definition of @xmath194 , this can also be written in the alternate form , [ eq - fsscentroid - us - alt ] _ f^us = _ f , cn - . figure 2 shows the functions @xmath195 and @xmath196 , while figure 3 depicts @xmath197 figure 1(g ) illustrates the behavior of the image centroid ( ) for a finite source . in this section we consider the effect of non - uniform sources on the magnification and centroid shift near folds . the effect of generic surface brightnesses can be evaluated using the general integral forms for the magnification ( ) , and the two components of the centroid shift ( eqs . [ eq - ch ] and [ eq - ck ] ) . we will concentrate on a specific form for @xmath139 applicable to stellar sources , namely @xmath198\right\ } , \label{eqn : sbp}\ ] ] where @xmath146 is the center of the source . here @xmath199 is the limb darkening parameter , which may be wavelength dependent . this form was originally introduced by @xcite , and it has the desirable property that there is no net flux associated with the limb darkening term . inserting this form for the surface brightness into , we recover the result of @xcite for the limb darkened magnification @xmath200 , \label{eqn : mufld}\ ] ] where @xmath201 is defined in , and @xmath165 , @xmath202 , and @xmath203 are shown in figure 2 . since this form of the surface brightness profile is symmetric , @xmath190 , the integral @xmath204 vanishes . inserting into the general form for @xmath205 , we find @xmath206 , and thus , [ eq - fscentroid - ld - s ] _ f^ld = _ f , cn - [ _ 0 ( z)+(_1/2-_0 ) ] , which can also be written , [ eq - fscentroid - ld - sv2 ] _ f^ld = _ f , cn - . figure 2 shows the functions @xmath195 , @xmath207 , and @xmath208 , while figure 3 depicts the term in brackets above for several values of @xmath199 . the majority of the results presented in [ sec : analytic ] ( with the exception of the assumption made in [ sec : fsources ] and [ sec : limb ] that the caustic is coincident with the @xmath60-axis ) are applicable for the general case of a parabolic fold catastrophe . in some cases , however , it is possible to simplify the expressions considerably and recover the more familiar linear fold form . to see how this limit may be reached , consider the form for the general fold mapping in . in this form , the source and image plane coordinates have all been normalized by @xmath7 , and thus are dimensionless , as are the coefficients @xmath209 ( since @xmath29 is dimensionless ) . it is clear that the caustic will ( in general ) be appreciably curved when one considers order unity variations in @xmath188 ( i.e. , on absolute angular scales of order @xmath7 ) . in other words , in the general case one can expect that all of the coefficients ( @xmath209 ) to be of the same order of magnitude . now consider variations on some smaller scale @xmath210 . renormalizing the angular source and image plane variables such that @xmath211 and @xmath212 , the fold mapping can be recast in the form [ eq - le - fold-2 ] u_1 = a _ 1 + ^2_2 + c _ 1 _ 2 , _ 2 = ^2_1 + b _ 1_2 + ^2_2 , where the relation between the coefficients are @xmath213 , @xmath214 , @xmath215 , and @xmath216 . therefore , on scales of order @xmath217 , we can expect @xmath218 to be larger than @xmath219 and @xmath138 by a factor @xmath220 . in the galactic microlensing case , for example , one is typically concerned with variations on the scale of the source , i.e. , scales of order @xmath221 . in this case @xmath222 . let us consider the curvature of the fold caustic on scales of @xmath223 . in general , for a twice continuously differentiable function @xmath224 on an open interval of @xmath225 , the magnitude of the curvature at a point @xmath226 on the graph of @xmath227 is given by [ eq - curvature ] |(x_0)| = . in the coordinates of ( [ eq - le - fold-2 ] ) , i.e. on scales of @xmath223 , equation ( [ eq - c ] ) yields that the fold caustic can be expressed as a graph @xmath228 where f(_1 ) = ^2_1 , . by ( [ eq - curvature ] ) , |(_1)| = . but = ( ) e , _ 1 = ( ) u_1 . hence , the magnitude of the curvature of the fold caustic can be expressed as [ eq - curvature - f ] |(_1)| = ( ) . equation ( [ eq - curvature - f ] ) yields that , at the origin and on scales comparable to @xmath217 , the magnitude of curvature scales as @xmath229 , and is given by @xmath230 . note that the radius of curvature at the origin is @xmath231 . in galactic microlensing @xmath232 and @xmath233 . therefore it is generally the case that @xmath234 , i.e. , the curvature of the fold caustic is negligible on scales of order @xmath143 , and the fold can be treated as linear . we therefore ask what happens to the expressions derived in the previous sections in the limit that @xmath235 . we find that @xmath236 , and thus the caustic , which is defined by @xmath237 , collapses to the @xmath60-axis ( linear fold ) . the critical curve is still given by @xmath238 . there are images whenever @xmath239 , which are located at _ = ( , ) , [ eq - sf - mag ] with magnifications , _ the centroid of the two images therefore takes on the simple form , _ f = ( , 0 ) . including the additional images , and assuming a rectilinear trajectory , the total magnification can be written , _ tot ( t ) = ( ) ^1/2(t- ) + _ 0,c + _ ( 0,c ) ( t- ) , [ eqn : muvtime - sf ] the centroid of all the images has the same form as before ( ) , with _ f , c=(,0)_(f , c)=(,0 ) . for finite sources , the centroid in ( [ eq - fscentroid - us - s ] ) reduces to @xmath240 the finite source centroid then becomes ^fs_cl = _ f^fs _ tot^fs _ f , cn + _ 0 , cn _ tot^fs _ 0,cn , [ eq - samemag ] this is the same result as for point sources ( see ) if the point source magnification is replaced with the finite source magnification . equations ( [ eq - sf - mag])-([eq - samemag ] ) should hold whenever ( 1 ) @xmath241 , ( 2 ) one is not too close to a higher - order catastrophe , i.e. , a cusp , and ( 3 ) there are no other nearby folds . we can use the results from the previous sections to derive some generic results about the astrometric behavior near folds . we first consider the magnitude of the astrometric jump when the source crosses a caustic . it is clear that for a point source , the maximum centroid shift is @xmath242 , where @xmath132 is the point on the critical curve where the images merge . although @xmath243 ultimately depends on the source trajectory and the topology of the lens , we can typically expect that @xmath243 is @xmath14 . for the finite source case , the difference between the centroid position just before ( @xmath244 ) and just after ( @xmath245 ) the caustic crossing is @xmath246 . for a uniform source , this is @xmath247^{-1 } \dtj . \label{eqn : fsjump}\ ] ] adopting typical parameters , @xmath248 , @xmath249 , and @xmath250 , we find the fractional change from the point source case to be @xmath251 . we now consider the magnitude of limb darkening effect on the magnification and centroid shift relative to the uniform source case . from equations ( [ eq - mtfs ] ) and ( [ eqn : mufld ] ) , the fractional difference between the limb - darkened and uniform - source magnification is , @xmath252^{-1}. \label{eqn : deltamu}\ ] ] early in the caustic crossing ( @xmath253 ) , the factor @xmath254 is small compared to unity , and thus @xmath255 . the magnitude of @xmath256 is @xmath257 for the majority ( @xmath258 ) of the caustic crossing . near the end of the caustic crossing ( @xmath259 ) the term in brackets begins to dominate as @xmath260 , and thus @xmath261 , which goes to zero as @xmath262 . thus for typical values of @xmath263 , and @xmath264 , the fractional difference from a uniform source is @xmath265 for the majority of the caustic crossing . see figure 3 . the difference in the centroid due to limb darkening is _ cl^ld = _ cl^ld - _ cl^us = _ ld ( _ f - _ 0 ) . [ eqn : deltatld ] figure 3(c ) shows the prefactor @xmath266 for @xmath267 , @xmath268 , and several values of @xmath269 . to assess the detectability of the deviation of the centroid shift from the uniform source case due to limb darkening in galactic bulge microlensing events , we now make a crude estimate for the maximum magnitude of @xmath270 by adopting typical parameters . at @xmath122 , we have that @xmath271 , and is maximized . we will assume that @xmath272 . we have just argued that @xmath273 . near the end of the caustic crossing ( where @xmath274 is maximized ) , @xmath275 . inserting these values into , we find @xmath276 for typical values of @xmath277 in the optical , @xmath278 , which is only a factor of @xmath279 larger than the sensitivity expected from _ sim_. note that this is the _ maximum _ centroid shift ; inspection of figure 3(c ) reveals that @xmath280 is considerably smaller than this maximum for the majority of the caustic crossing . therefore , detection of the astrometric effects of limb - darkening will likely be challenging , at least for galactic bulge lensing events , with the currently planned accuracy for sim . on the other hand , the fact that the astrometric effects of limb darkening are small implies that limb - darkening can generally be ignored . thus predictions for the signatures of other effects , such as the finite source effect itself , are robust . in other words , measurements generally should not be compromised by an unknown amount of limb - darkening . consider two observers that are not spatially coincident . the source trajectories as seen by the two observers will be displaced relative to each other by an amount @xmath281 , the magnitude of which is @xmath282 , where @xmath283 is the component of the separation between the two observers perpendicular to the line - of - sight , and @xmath284 is the einstein ring radius projected to the observer plane . if @xmath281 is sufficiently large , then both the photometric and astrometric behavior of the event will be measurably different between the two observers , an effect commonly known as parallax . since @xmath283 is known , this observed difference can be used to infer @xmath285 , providing additional physical constraints on the properties of the lens . for example , when combined with a measurement of @xmath7 from the astrometric centroid shift itself , the mass of the lens can be determined , @xmath286 . for a given @xmath287 , parallax effects are typically largest near caustics , because the magnification and the centroid vary rapidly with respect to source position . thus caustic crossings are ideal for use in measuring parallax effects @xcite . if @xmath287 is sufficiently small that the behavior near the fold caustic crossing for both observers can be described by the same local expansion of the lens mapping , than the results derived here can be used to fit the astrometric and photometric behavior for both observers and derive the parallax effects without regard to the global behavior of the lens . we note that , for generic fold caustics , any significant displacement @xmath281 , regardless of the its orientation relative to the caustic , will result in a difference in the observable behavior . however , for linear fold caustics ( [ sec : simple ] ) , only displacements perpendicular to the caustic will result in significant differences in the magnification and centroid . thus , for generic linear fold caustics , only a projection of @xmath285 is measurable from the local caustic behavior @xcite . during a fold caustic crossing the source is resolved , altering both the photometric and astrometric behavior with respect to a point source ( [ sec : fsources ] ) . the photometric behavior near a caustic crossing depends on @xmath269 , the source size in units of @xmath7 , whereas the astrometric behavior depends on actual angular size of the source @xmath288 .. ] thus , while fitting to the photometric data near a caustic crossing does not yield the angular size of the source , fitting to the astrometric data near a caustic crossing does . therefore it is possible , in principle , to measure the angular radius of the source star of a caustic - crossing binary - lens event by fitting a few astrometric measurements taken during the caustic crossing to the expressions we have derived for the local astrometric behavior ( see [ sec : fcen ] and eq . [ eqn : tclfs2 ] ) . in practice , however , this is complicated by the fact that , for linear fold caustics , only the degenerate combination @xmath289 can be measured , where @xmath102 is the angle of the trajectory with respect to the caustic . for general fold caustics , the degeneracy between @xmath143 and @xmath290 is broken ( see [ sec : fcen ] and eq . [ eq - fsscentroid - us - alt ] ) , however , for many cases the simple linear fold will be applicable . thus , is in order to determine @xmath143 separately , the global geometry must be generally specified , which can be accomplished using the global photometric light curve . in this section we numerically calculate the expected astrometric behavior for a binary - lens fold caustic crossing , and compare this with the analytic results from the previous sections . we do this in order to verify our expressions and also to explore the accuracy with which our ( necessarily ) approximate results reproduce the exact behavior . for definiteness , we will calculate the expected astrometric behavior for the photometrically well - observed caustic - crossing binary lens event @xcite . this has the advantage that , up to an orientation on the sky and subject to small errors in the inferred parameters , the astrometric behavior can be essentially completely determined from the photometric solution , including the size of @xmath7 and the effects of limb - darkening . for a system of @xmath291 point masses located at positions @xmath292 , and no external shear , the lens equation ( ) takes the form @xmath293 where @xmath294 is the mass of the @xmath295-th lens in units of the total mass . note that angles in are normalized to the @xmath7 for the total mass of the system . for @xmath296 , the lens equation is equivalent to a fifth - order polynomial in @xmath24 , thus yielding a maximum of five images . all of the image positions for a given point on the source plane can be found numerically using any standard root finding algorithm . then the individual magnifications , total magnification and centroid of these images can be found using equations ( [ eqn : magoneimage ] ) , ( [ eqn : mutot ] ) , and ( [ eqn : cl ] ) . for a finite source size , it is necessary to integrate over the area of the source ( ) . this can be difficult to do numerically in the source place near the caustics , due to the divergent magnification . a more robust method is inverse ray shooting . this works as follows . the image plane is sampled uniformly and densely , and at each @xmath24 , is used to find the corresponding @xmath297 . the local ratio of the density of rays in @xmath24 to the density of rays in @xmath25 is the local magnification . thus one can create a map of @xmath298 , the magnification as a function of @xmath25 . similarly , one can determine the astrometric deviation by sampling in @xmath24 , using to determine @xmath297 , and then summing at each @xmath25 the values of @xmath299 . the astrometric deviation at @xmath25 is then the summed values of @xmath299 , weighted by the local magnification . thus one creates two astrometric maps , for each direction . in practice , inverse ray shooting requires one to bin the rays in the source plane , with the resolution of the maps being determined by the size of the bin , and the accuracy determined by surface density of rays in the image plane relative to the ( unlensed ) surface density of rays in the source plane . the advantage of inverse ray shooting is that the procedure conserves flux , therefore the maps can be convolved with any source profile to produce the finite source photometric and astrometric behavior for arbitrary source size and surface brightness profile . to apply this method to predict the detailed photometric and astrometric behavior of the caustic crossings for , it is essential that the resolution of the astrometric and photometric maps be considerably smaller than the source size @xmath269 . we will be using resolutions of @xmath300 , which corresponds to @xmath301 for . this is sufficient to accurately resolve the source . we sample the image plane with a density of @xmath302 , corresponding to an poisson error per resolution element of @xmath303 . since there are @xmath304 resolution elements per source size , the total poisson error is always @xmath305 , considerably smaller than any of the effects we will be considering . before studying the detailed behavior near the photometrically well - covered ( second ) caustic crossing of , we first analyze the global astrometric behavior of the entire event . we specify the binary - lens topology and source trajectory using the parameters of the best - fit solution with limb darkening ( see table 2 of @xcite ) . in figure 4 we show various aspects of the inferred lensing system and the event itself . the best - fit lens system has a wide topology , with two well - separated caustic curves , one near the position of each lens . the mass ratio of the two lenses is @xmath306 , and they are separated by @xmath307 . the solution has the source crossing the caustic associated with the least massive lens , which we will call the secondary caustic . figure 4(a ) shows the photometric light curve centered on the event . there are two fold caustic crossings , separated by @xmath308 days ; the second crossing was densely covered photometrically by @xcite , allowing them to determine not only the source size @xmath269 , but also limb - darkening coefficients @xmath199 in each of two different photometric bands : @xmath309 ( @xmath310-band ) and @xmath311 ( @xmath312-band ) . this allows us to predict the astrometric behavior including finite source and limb - darkening effects . furthermore , by combining a measurement of @xmath313 , with a determination of the angular size of the source @xmath143 from its color and magnitude , @xcite measured the angular einstein ring radius of the lens to be @xmath314 . therefore we can determine the absolute scale of the astrometric features and assess their detectability by comparing them with the expected accuracy of upcoming interferometers . figure 4(b ) shows the caustics and critical curves , as well as the trajectory of the images of the source . one image is always near the most massive lens ; this image has little effect on the resulting astrometric deviation other than a small net offset along the binary axis . figure 4(c ) shows a close - up of the region near the secondary caustic , along with the image centroid @xmath49 . the components parallel and perpendicular to the binary axis are shown in figures 4(e , d ) . finally , figure 4(f ) shows @xmath315 , the centroid relative to the unlensed source position @xmath25 . in figures 4(c - f ) , the large , @xmath14 jumps that occur when the source crosses the caustic are evident . we now focus on the photometric and astrometric behavior near the second caustic crossing for . in figure 5(a ) we show the caustic geometry and source trajectory near this crossing . the source passed @xmath316 from a cusp . on the scale of the source size , @xmath317 , the curvature of the caustic can be neglected . in figure 5 we show the behavior near the second caustic crossing . figure 5(c ) shows the photometric behavior for two days centered on the caustic crossing , for the assumption of a uniform source , and a limb - darkened source with surface brightness profile given in , with @xmath309 and @xmath311 . note the similarity of the shapes of the uniform source and limb - darkened light curves with the analytic forms @xmath164 and @xmath318 presented in figure 2(a ) . we show our prediction for the astrometric behavior as a function of time for the components parallel and perpendicular to the binary axis in figures 5(d ) and ( e ) , respectively [ compare with figure 1(e , f ) ] . we show both the instantaneous discontinuous jumps for a point - source centroid , along with the continuous centroid curves for finite sources . the limb - darkened and uniform source astrometric curves are extremely similar . in figure 5(f ) we show the predicted total astrometric behavior for the same time span . notice how well the form of figure 5(f ) compares with figure 1(g ) , which was obtained from our analytic forms . figure 6 shows the same curve as figure 5(f ) , except we have rotated the axes by @xmath319 , shifted the origin to the image of the caustic crossing , and stretched the axes for visibility . in figure 6 , notice how the predicted shaped for the finite source centroids are smoothly rounded off near the bottom and has a sharp turn near the top . although the difference between the finite source and point source curves is quite substantial , the difference between the uniform source and limb darkened sources is very small . this is more clearly illustrated in figure 7(a , b ) , where we show the two components of this difference , @xmath320 , for about four source radius crossing times centered on the crossing , i.e.@xmath321 . the form of @xmath270 is very similar to the analytic expectation [ see and figure 3(c ) ] . furthermore , the two components of @xmath270 are essentially perfectly ( anti- ) correlated , implying that the difference is essentially one - dimensional . this can be seen best in figure 7(c ) , where we plot @xmath322 versus @xmath323 . panel ( d ) is the same as panel ( c ) , except we have rotated and stretched the axes . the maximum deviation is @xmath324 , in satisfying agreement with the rough expectation ( ) . we have presented a detailed study of gravitational lensing near fold catastrophes , concentrating on the regime where the individual images are unresolved , i.e. microlensing . by taylor expanding the scalar potential @xmath29 in the neighborhood of a fold up to third order in the image position , one can obtain a generic form for the lensing map near a fold . beginning with this mapping , we derive the local lensing properties of a source in the vicinity of the fold caustic . approximating the critical curve by its tangent line at the origin , we find that the caustic is locally a parabola . on one side of the parabola , the fold lenses a nearby source into two images ; on the other side of the parabola , there are no images . we derive the image positions and magnifications as a function of the position of the source . we find that the magnifications of the two images are equal , and recover the well - known result that the magnification is inversely proportional to the square root of the distance to the caustic . we show how this holds for parabolic caustics ( as well as linear caustics ) , provided that the ` vertical ' distance from the caustic is used . assuming a rectilinear source trajectory , and allowing for the existence of slowly- and smoothly - varying images not associated with the fold caustic , we derive analytic expressions for the total magnification and image centroid ( center - of - light ) as a function of time . we then consider how the photometric and astrometric behavior is altered in the presence of a finite source size . we derive semi - analytic expressions for the magnification and centroid as a function of time for both a uniform source , and limb - darkened source . along the way we derived expressions that can be used to evaluate the photometric and astrometric behavior near a fold for a source with arbitrary surface brightness profile . we then show how and under what conditions the generic parabolic fold reduces to the more familiar linear fold . we derive simplified expressions for the individual and total image positions and magnifications near a linear fold . we used some of our analytic results to derive a few generic properties of microlensing near folds . in particular , we derive and evaluate expressions for the magnitude of the centroid jump that occurs when a finite source crosses a fold relative to the point source jump , and the magnitude of the effect of limb darkening on both the photometric and astrometric behavior . notably , we predict , for galactic bulge lensing events , the shape of the centroid due to finite sources with uniform and limb darkening surface brightness profiles . we also find for galactic bulge lensing that the effect of limb darkening on the image centroid near a fold is quite similar to the uniform source case , making the limb darkening effect difficult to detect by the currently planned accuracy for the instrumentation of _ we discussed how our formulae can be used to fit both photometric and astrometric data sets near fold caustic crossings and thus used to derive such properties as the angular size of the source and the microlensing parallax . finally , we numerically calculate expected astrometric behavior of the photometrically well - observed galactic bulge binary lensing event @xcite , finding excellent agreement with our analytic predictions . caustics are ubiquitous in gravitational lenses , and the most common type of caustic is the fold . caustics play an especially important role in microlensing , as the rapid time variability of the total image magnification allows the possibility of detailed studies of the source and lens . in the future , we can expect that time - series photometric measurements will be supplemented by time - series _ astrometric _ measurements of the center - of - light of microlens systems . this paper presents the most thorough and comprehensive study of the photometric and astrometric behavior of gravitational microlensing near fold caustics to date . the results should prove useful to those studying microlens systems with caustics : the analytic expressions derived here can be used to fit fold caustic crossings observed both photometrically and astrometrically , gain some insight into more complicated numerical studies , and establish predictions for the feasibility of future observations . we would like to thank the referee , eric agol , for several helpful suggestions that led to a much improved manuscript . b.s.g . was supported in part by nasa through a hubble fellowship grant from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , under nasa contract nas5 - 26555 . a.o.p . was supported in part by an alfred p. sloan research fellowship and nsf career grant dms-98 - 96274 .

finding the node with the largest label in a labeled network , modeled as an undirected connected graph , is one of the fundamental problems in distributed computing . this is the way in which _ leader election _ is usually solved . we consider two distinct tasks in which the largest - labeled node is found deterministically . in _ selection _ , this node has to output 1 and all other nodes have to output 0 . in _ election _ , the other nodes must additionally learn the largest label ( everybody has to know who is the elected leader ) . our aim is to compare the difficulty of these two seemingly similar tasks executed under stringent running time constraints . the measure of difficulty is the amount of information that nodes of the network must initially possess , in order to solve the given task in an imposed amount of time . following the standard framework of _ algorithms with advice _ , this information ( a single binary string ) is provided to all nodes at the start by an oracle knowing the entire graph . the length of this string is called the _ size of advice_. the paradigm of algorithms with advice has a far - reaching importance in the realm of network algorithms . lower bounds on the size of advice give us impossibility results based strictly on the _ amount _ of initial knowledge outlined in a model s description . this more general approach should be contrasted with traditional results that focus on specific _ kinds _ of information available to nodes , such as the size , diameter , or maximum node degree . consider the class of-node graphs with any diameter , for some integer . if time is larger than , then both tasks can be solved without advice . for the task of _ election _ , we show that if time is smaller than , then the optimal size of advice is , and if time is exactly , then the optimal size of advice is . for the task of _ selection _ , the situation changes dramatically , even within the class of rings . indeed , for the class of rings , we show that , if time is , for any , then the optimal size of advice is , and , if time is ( and at most ) then this optimal size is . thus there is an _ exponential _ increase of difficulty ( measured by the size of advice ) between selection in time , for any , and selection in time . as for the comparison between election and selection , our results show that , perhaps surprisingly , while for small time , the difficulty of these two tasks on rings is similar , for time the difficulty of election ( measured by the size of advice ) is exponentially larger than that of selection . * keywords : * election , selection , maximum finding , advice , deterministic distributed algorithm , time . 0.18 in 0.0 in 0.2 in

1411.1319

* background . * finding the node with the largest label in a labeled network is one of the fundamental problems in distributed computing . this is the way in which _ leader election _ is usually solved . ( in leader election , one node of a network has to become a _ leader _ and all other nodes have to become _ non - leaders _ ) . in fact , to the best of our knowledge , all existing leader election algorithms performed in labeled networks choose , as leader , the node with the largest label or the node with the smallest label @xcite . the classic problem of leader election first appeared in the study of local area token ring networks @xcite , where , at all times , exactly one node ( the owner of a circulating token ) has the right to initiate communication . when the token is accidentally lost , a leader is elected as the initial owner of the token . * model and problem description . * the network is modeled as an undirected connected graph with @xmath0 labeled nodes and with diameter @xmath3 at most @xmath2 . we denote by @xmath10 the diameter of graph @xmath11 . labels are drawn from the set of integers @xmath12 , where @xmath13 is polynomial in @xmath0 . each node has a distinct label . initially each node knows its label and its degree . the node with the largest label in a graph will be called its _ largest node_. we use the extensively studied @xmath14 communication model @xcite . in this model , communication proceeds in synchronous rounds and all nodes start simultaneously . in each round , each node can exchange arbitrary messages with all of its neighbours and perform arbitrary local computations . for any @xmath15 and any node @xmath16 , we use @xmath17 to denote the knowledge acquired by @xmath16 within @xmath18 rounds . thus , @xmath17 consists of the subgraph induced by all nodes at distance at most @xmath18 from @xmath16 , except for the edges joining nodes at distance exactly @xmath18 from @xmath16 , and of degrees ( in the entire graph ) of all nodes at distance exactly @xmath18 from @xmath16 . hence , if no additional knowledge is provided _ a priori _ to the nodes , the decisions of @xmath16 in round @xmath18 in any deterministic algorithm are a function of @xmath17 . we denote by @xmath19 the set of labels of nodes in the subgraph induced by all nodes at distance at most @xmath18 from @xmath16 . the _ time _ of a task is the minimum number of rounds sufficient to complete it by all nodes . it is well known that the synchronous process of the @xmath14 model can be simulated in an asynchronous network . this can be achieved by defining for each node separately its asynchronous round @xmath20 ; in this round , a node performs local computations , then sends messages stamped @xmath20 to all neighbours , and waits until it gets messages stamped @xmath20 from all neighbours . to make this work , every node is required to send at least one ( possibly empty ) message with each stamp , until termination . all of our results can be translated for asynchronous networks by replacing `` time of completing a task '' by `` the maximum number of asynchronous rounds to complete it , taken over all nodes '' . we consider two distinct tasks in which the largest node is found deterministically . in _ selection _ , this node has to output 1 and all other nodes have to output 0 . in _ election _ , all nodes must output the largest label . note that in election all nodes perform selection and additionally learn the identity of the largest node . both variations are useful in different applications . in the aforementioned application of recovering a lost token , selection is enough , as the chosen node will be the only one to get a single token and then the token will be passed from node to node . in this case , other nodes do not need to know the identity of the chosen leader . the situation is different if all nodes must agree on a label of one of the nodes , e.g. , to use it later as a common parameter for further computations . then the full strength of election is needed . likewise , learning the largest label by all nodes is important when labels carry some additional information apart from the identities of nodes , e.g. , some values obtained by sensors located in nodes . our results also remain valid in such situations , as long as the `` informative label '' can be represented as an integer polynomial in @xmath0 . our aim is to compare the difficulty of the two seemingly similar tasks of selection and election executed under stringent running time constraints . the measure of difficulty is the amount of information that nodes of the network must initially possess in order to solve the given task in an imposed amount of time . following the standard framework of _ algorithms with advice _ , see , e.g. , @xcite , this information ( a single binary string ) is provided to all nodes at the start by an oracle knowing the entire graph . the length of this string is called the _ size of advice_. the paradigm of algorithms with advice has a far - reaching importance in the realm of network algorithms . establishing a tight bound on the minimum size of advice sufficient to accomplish a given task permits to rule out entire classes of algorithms and thus focus only on possible candidates . for example , if we prove that @xmath4 bits of advice are needed to perform a certain task in @xmath0-node graphs , this rules out all potential algorithms that can work using only some linear upper bound on the size of the network , as such an upper bound could be given to the nodes using @xmath21 bits by providing them with @xmath22 . lower bounds on the size of advice give us impossibility results based strictly on the _ amount _ of initial knowledge outlined in a model s description . this more general approach should be contrasted with traditional results that focus on specific _ kinds _ of information available to nodes , such as the size , diameter , or maximum node degree . * our results . * consider the class of @xmath0-node graphs with any diameter @xmath1 , for some integer @xmath2 . first observe that if time is larger than @xmath3 , then both tasks can be solved without advice , as nodes learn the entire network and _ learn that they have learned it_. thus they can just choose the maximum of all labels seen . for the task of _ election _ , we show that if time is smaller than @xmath3 , then the optimal size of advice is @xmath4 , and if time is exactly @xmath3 , then the optimal size of advice is @xmath5 . here our contribution consists in proving two lower bounds on the size of advice . we prove one lower bound by exhibiting , for any positive integers @xmath23 , networks of size @xmath24 and diameter @xmath25 , for which @xmath26 bits of advice are needed for election in time below @xmath3 . to prove the other lower bound , we present , for any positive integer @xmath2 , networks of diameter @xmath1 for which @xmath27 bits of advice are needed for election in time exactly @xmath3 . these lower bounds are clearly tight , as even for time 0 , @xmath28 bits of advice are enough to provide the largest label in the network , and , for time @xmath3 when nodes know the entire network , but they _ do not know that they know it _ @xmath29 bits of advice are enough to give the diameter @xmath3 to all nodes and thus reassure them that they have seen everything . in this case , they can safely choose the maximum of all labels seen . hence , a high - level statement of our results for election is the following . if time is too small for all nodes to see everything , then no more efficient help in election is possible than just giving the largest label . if time is sufficient for all nodes to see everything , but too small for them to realize that they do , i.e. , the time is exactly @xmath3 , then no more efficient help in election is possible than providing @xmath3 , which supplies nodes with the missing certainty that they have seen everything . it should be noted that @xmath0 could be exponential in @xmath2 , as in hypercubes , or @xmath2 can even be constant with respect to arbitrarily large @xmath0 . thus , our results for election show that , for some networks , there is an exponential ( or even larger ) gap of difficulty of election ( measured by the size of advice ) between time smaller than @xmath3 and time exactly @xmath3 . another such huge gap is between time @xmath3 , when advice of size @xmath5 is optimal , and time larger than @xmath3 , when 0 advice is enough . these gaps could be called _ intra - task _ jumps in difficulty for election with respect to time . for the task of _ selection _ , the situation changes dramatically , even within the class of rings . indeed , for the class of rings , we show that , if time is @xmath6 , for any @xmath7 , then the optimal size of advice is @xmath5 , and , if time is @xmath8 ( and at most @xmath3 ) then this optimal size is @xmath9 . here our contribution is three - fold . for selection in time @xmath6 , for any @xmath7 , we exhibit , for any positive integer @xmath2 , a class of rings with diameter at most @xmath2 which requires advice of size @xmath27 . as before , this lower bound is tight , even for time 0 . further , for selection in time at most @xmath30 , for @xmath31 , we construct a class of rings with diameter @xmath1 which requires advice of size @xmath32 . at first glance , it might seem that this lower bound is too weak . indeed , providing either the diameter or the largest label , which are both natural choices of advice , would not give a tight upper bound , as this may require @xmath33 bits . however , we use a more sophisticated idea that permits us to construct a very compact advice ( of matching size @xmath34 ) and we design a selection algorithm , working for all rings of diameter @xmath1 in time at most @xmath30 , for which this advice is enough . thus there is an _ exponential _ increase of difficulty ( measured by the size of advice ) between selection in time @xmath6 , for any @xmath7 , and selection in time @xmath8 . as in the case of election , another huge increase of difficulty occurs between time @xmath3 and time larger than @xmath3 . these gaps could be called _ intra - task _ jumps of difficulty for selection with respect to time . as for the comparison between election and selection , our results show that , perhaps surprisingly , while for small time , the difficulty of these two tasks on rings is similar , for time @xmath8 the difficulty of election ( measured by the size of advice ) is exponentially larger than that of selection , even for the class of rings . while both in selection and in election the unique leader having the maximum label is chosen , these tasks differ in how widely this label is known . it follows from our results that , if linear time ( not larger than the diameter ) is available , then the increase of difficulty ( in terms of advice ) of making this knowledge widely known is exponential . this could be called the _ inter - task _ jump of difficulty between election and selection . figure [ summarytable ] provides a summary of our results . ] * related work . * the leader election problem was introduced in @xcite . this problem has been extensively studied in the scenario adopted in the present paper , i.e. , where all nodes have distinct labels . as far as we know , this task was always solved by finding either the node with the largest or that with the smallest label . leader election was first studied for rings . a synchronous algorithm based on label comparisons and using @xmath35 messages was given in @xcite . it was proved in @xcite that this complexity is optimal for comparison - based algorithms . on the other hand , the authors showed an algorithm using a linear number of messages but requiring very large running time . an asynchronous algorithm using @xmath35 messages was given , e.g. , in @xcite , and the optimality of this message complexity was shown in @xcite . deterministic leader election in radio networks has been studied , e.g. , in @xcite , as well as randomized leader election , e.g. , in @xcite . in @xcite , the leader election problem was approached in a model based on mobile agents for networks with labeled nodes . many authors @xcite studied leader election in anonymous networks . in particular , @xcite characterize message - passing networks in which leader election can be achieved when nodes are anonymous . characterizations of feasible instances for leader election were provided in @xcite . memory needed for leader election in unlabeled networks was studied in @xcite . providing nodes or agents with arbitrary kinds of information that can be used to perform network tasks more efficiently has previously been proposed in @xcite . this approach was referred to as _ algorithms with advice_. the advice is given either to nodes of the network or to mobile agents performing some network task . in the first case , instead of advice , the term _ informative labeling schemes _ is sometimes used , if ( unlike in our scenario ) different nodes can get different information . several authors studied the minimum size of advice required to solve network problems in an efficient way . in @xcite , given a distributed representation of a solution for a problem , the authors investigated the number of bits of communication needed to verify the legality of the represented solution . in @xcite , the authors compared the minimum size of advice required to solve two information dissemination problems using a linear number of messages . in @xcite , it was shown that advice of constant size given to the nodes enables the distributed construction of a minimum spanning tree in logarithmic time . in @xcite , the advice paradigm was used for online problems . in @xcite , the authors established lower bounds on the size of advice needed to beat time @xmath36 for 3-coloring cycles and to achieve time @xmath36 for 3-coloring unoriented trees . in the case of @xcite , the issue was not efficiency but feasibility : it was shown that @xmath37 is the minimum size of advice required to perform monotone connected graph clearing . in @xcite , the authors studied radio networks for which it is possible to perform centralized broadcasting in constant time . they proved that constant time is achievable with @xmath38 bits of advice in such networks , while @xmath39 bits are not enough . in @xcite , the authors studied the problem of topology recognition with advice given to nodes . to the best of our knowledge , the problems of leader election or maximum finding with advice have never been studied before . notice that in order to perform election in a graph @xmath11 in time larger than its diameter , no advice is needed . indeed , after time @xmath40 , all nodes know the labels of all other nodes , and they are aware that they have this knowledge . this is because , in round @xmath40 , no messages containing new labels are received by any node . so , it suffices for all nodes to output the largest of the labels that they have seen up until round @xmath40 . our first result shows that , if election time is no more than the diameter of the graph , then the size of advice must be at least logarithmic in the diameter . this demonstrates a dramatic difference between the difficulty of election in time @xmath10 and election in time @xmath40 , measured by the minimum size of advice . [ diam ] consider any algorithm @xmath41 such that , for every graph @xmath11 , algorithm @xmath41 solves election within @xmath42 rounds . for every integer @xmath43 , there exists a ring of diameter at most @xmath2 for which algorithm @xmath41 requires advice of size @xmath27 . fix any integer @xmath43 . we will show a stronger statement : at least @xmath2 different advice strings are needed in order to solve election within @xmath10 rounds for rings @xmath11 with diameter at most @xmath2 . the high - level idea of the proof is to first construct a particular sequence of @xmath2 rings of increasing sizes , each with a different largest label . with few advice strings , two such rings get the same advice . we show that there is a node in these two rings which acquires the same knowledge when executing algorithm @xmath44 , and hence has to elect the same node in both rings , which is a contradiction . to obtain a contradiction , assume that @xmath45 different advice strings suffice . consider a ring @xmath46 of diameter @xmath2 whose node labels form the sequence @xmath47 ( see figure [ rings ] ) . for each @xmath48 , define @xmath49 to be the ring obtained from @xmath46 by taking the subgraph induced by the nodes at distance at most @xmath50 from node @xmath51 and adding an edge between nodes @xmath52 and @xmath53 ( see figure [ rings ] ) . first , note that , for each @xmath48 , the diameter of @xmath54 is @xmath50 , and , the largest node in @xmath54 has label @xmath53 . the correctness of @xmath41 implies that , when @xmath41 is executed at a node in @xmath54 , it must halt within @xmath50 rounds and output @xmath53 . next , by the pigeonhole principle , there exist @xmath55 with @xmath56 such that the same advice string is provided to nodes of both @xmath57 and @xmath58 when they execute @xmath41 . when executed at node 1 in @xmath57 , algorithm @xmath41 halts in some round @xmath59 and outputs @xmath60 . we show that , when executed at node 1 in @xmath58 , algorithm @xmath41 also halts in round @xmath18 and outputs @xmath60 . indeed , the algorithm is provided with the same advice string for both @xmath57 and @xmath58 , and , @xmath61 in @xmath57 is equal to @xmath61 in @xmath58 . this contradicts the correctness of @xmath41 since , for the ring @xmath58 , there is a node with label @xmath62 . to conclude , notice that , since at least @xmath2 different advice strings are needed for the class of rings of diameter at most @xmath2 , the size of advice must be @xmath27 for at least one of these rings . and @xmath54 , as constructed in the proof of theorem [ diam ] ] note that the lower bound established in theorem [ diam ] is tight . indeed , to achieve election in time @xmath10 for any graph @xmath11 , it is enough to provide the value of @xmath10 to the nodes of the graph and have each node elect the node with the largest label it has seen up until round @xmath10 . hence we have the following corollary for the class of graphs of diameter at most @xmath2 . [ cor1 ] the optimal size of advice to complete election in any graph in time at most equal to its diameter is @xmath5 . we next consider accomplishing election in time less than the diameter of the graph . one way to do it is to provide the maximum label to all nodes as advice . this yields election in time 0 and uses advice of size @xmath28 , where @xmath0 is the size of the graph , since the space of labels is of size @xmath13 polynomial in @xmath0 . the following result shows that this size of advice can not be improved for election in _ any _ time below the diameter . this result , when compared to corollary [ cor1 ] , again shows the dramatic difference in the difficulty of election ( measured by the minimum size of advice ) but now between times @xmath63 and @xmath10 . consider any positive integers @xmath64 . there exists @xmath65 such that , for any election algorithm @xmath66 in which every execution halts within @xmath45 rounds , there exists an @xmath0-node graph of diameter @xmath2 for which the size of advice needed by @xmath66 is @xmath67 . the high - level idea of the proof is the following . we first construct a family of `` ring - like '' graphs . for a given number of advice strings , we obtain a lower bound on the number of such graphs for which the same advice is given . on the other hand , an upper bound on this number is obtained by exploiting the fact that no node can see the entire graph within @xmath45 rounds . comparing these bounds gives the desired bound on the size of advice . let @xmath0 be the smallest integer greater than @xmath68 that is divisible by 2d ( and note that @xmath65 ) . consider a family @xmath69 of @xmath70 pairwise disjoint sets , each of size @xmath71 . in particular , let @xmath72 , where @xmath73 . we construct a family @xmath74 of @xmath0-node graphs . each graph in @xmath74 is obtained by first choosing an arbitrary sequence of @xmath75 sets from @xmath69 , say @xmath76 . the nodes of the graph are the elements of these sets ( which are integers ) , and this induces a natural labeling of the nodes . next , for each @xmath77 , add all edges between pairs of elements of the set @xmath78 , as well as all edges between every element in @xmath78 and every element in @xmath79 ( where the indices are taken modulo @xmath75 ) . in other words , each graph in @xmath74 is a `` fat ring '' , as illustrated in figure [ fatring ] . we uniquely identify each graph in @xmath74 by its sequence of sets @xmath76 , where the node @xmath80 with the smallest label belongs to the set @xmath81 , and @xmath82 contains the smallest neighbour of @xmath80 outside of @xmath81 . the size of @xmath83 is calculated in the following claim . and @xmath84 . ] * claim 1 * @xmath85 . to prove the claim , first note that the number of sequences @xmath76 consisting of @xmath75 distinct sets from @xmath69 is @xmath86 . to count the number of such sequences belonging to @xmath87 , we first divide this integer by @xmath75 to eliminate those sequences in which @xmath81 does not contain the smallest label . then , we divide the result by 2 to eliminate those sequences in which the labels in @xmath88 are smaller than those in @xmath82 . this completes the proof of the claim . next , let @xmath89 be the maximum number of advice bits provided to @xmath41 , taken over all graphs in @xmath74 . by the pigeonhole principle , there exists a family @xmath90 of at least @xmath91 graphs in @xmath74 such that the algorithm receives the same advice string when executed on each graph in @xmath90 . the following claim will be used to find an upper bound on the size of @xmath90 . * claim 2 * _ consider two graphs from @xmath90 , say @xmath92 and @xmath93 . suppose that , for some @xmath94 , algorithm @xmath66 elects a node from set @xmath95 when executed on @xmath96 and elects a node from set @xmath97 when executed on @xmath98 . if @xmath99 for each @xmath100 , then @xmath101 . _ to prove the claim , let @xmath102 ) . since @xmath99 for all @xmath100 , it follows that each node @xmath103 is also an element of @xmath104 , and , moreover , for each such @xmath16 , knowledge @xmath105 in @xmath96 is equal to knowledge @xmath105 in @xmath98 . the nodes of @xmath106 output some label @xmath107 at the end of the execution of @xmath66 on @xmath96 . since we assumed that every execution of @xmath44 halts within @xmath45 rounds , and the same advice is given for @xmath96 and @xmath98 , the nodes of @xmath104 also output label @xmath107 at the end of execution of @xmath66 on @xmath98 . since @xmath66 elects a node from @xmath95 when executed on @xmath96 and elects a node from @xmath97 when executed on @xmath98 , it follows that @xmath107 is a label that appears in both @xmath95 and @xmath97 . as the sets in @xmath69 are pairwise disjoint , it follows that @xmath101 . this concludes the proof of the claim . using claim 2 , we now obtain an upper bound on the size of @xmath90 . in particular , for each @xmath94 , consider the subfamily of graphs @xmath76 in @xmath90 such that algorithm @xmath66 elects a node from @xmath108 . by the claim , for each choice of the @xmath109 sets @xmath110 , there is exactly one set @xmath108 such that @xmath76 belongs to @xmath90 . the number of such choices is bounded above by @xmath111 . since this is true for all @xmath75 possible values for @xmath20 , we get that @xmath112 . comparing this upper bound to our lower bound on @xmath113 , it follows from claim 1 that @xmath114 . re - arranging this inequality , we get that @xmath115 , and hence @xmath116 . hence we have the following corollary . the optimal size of advice to complete election in any graph in time less than its diameter is @xmath117 . in this section , we study the selection problem for the class of rings . it turns out that significant differences between election and selection can be exhibited already for this class . as in the case of election , and for the same reasons , selection in time larger than the diameter can be accomplished without any advice . hence , in the rest of the section , we consider selection in time at most equal to the diameter . for any ring @xmath46 and any selection algorithm @xmath118 , denote by @xmath119 the advice string provided to all nodes in @xmath46 when they execute algorithm @xmath118 . denote by @xmath120 the node that outputs 1 in the execution of algorithm @xmath118 on ring @xmath46 . we first look at selection algorithms working in time at most @xmath121 , for any ring @xmath46 and any constant @xmath122 . we start with the lower bound on the size of advice needed by any selection algorithm working in time equal to the diameter of the ring . this lower bound shows that , for any positive integer @xmath2 , there exists a ring with diameter at most @xmath2 for which such an algorithm requires @xmath32 bits of advice . of course , this implies the same lower bound on the size of advice for selection in any smaller time . the following theorem provides our first lower bound on the size of advice for selection . [ lb1 ] consider any selection algorithm @xmath118 such that , for any ring @xmath46 , algorithm @xmath118 halts within @xmath123 rounds . for every positive integer @xmath2 , there exists a ring @xmath46 of diameter at most @xmath2 for which algorithm @xmath118 requires advice of size @xmath32 . at a high level we consider a `` rings - into - bins '' problem , in which each bin represents a distinct advice string . we recursively construct sets of rings , such that the rings constructed at a given recursion level can not be put into the same bin as previously - constructed rings . we continue the construction long enough to run out of bins . with few bins , the number of levels of recursion is sufficiently small to keep the diameters of the constructed rings bounded by @xmath2 . the following claim will be used to show that a particular ring that we construct will cause algorithm @xmath118 to fail . we will use the following _ chopping _ operation in our constructions . for any selection algorithm @xmath118 and any ring @xmath46 of odd size , we define the _ chop _ of @xmath46 , denoted by @xmath124 , to be the path obtained from @xmath46 by removing the edge between the two nodes at distance @xmath123 from @xmath120 ( see figure [ chopandglue](a ) ) . * claim 1 * _ consider any selection algorithm @xmath118 such that , for any ring @xmath46 , algorithm @xmath118 halts within @xmath123 rounds . consider two disjoint rings @xmath125 of odd size such that @xmath126 . for any ring @xmath127 that contains @xmath128 and @xmath129 as subgraphs such that @xmath130 , two distinct nodes in @xmath127 output 1 when executing @xmath118 . _ in order to prove the claim , consider the execution of @xmath118 by the nodes of @xmath127 . using the definition of @xmath128 , it can be shown that knowledge @xmath131 in @xmath127 is equal to knowledge @xmath131 in @xmath132 . therefore , the execution of @xmath118 at node @xmath133 in @xmath127 halts in round @xmath134 , and @xmath133 outputs 1 . similarly , the execution of @xmath118 at node @xmath135 in @xmath127 halts in round @xmath136 , and @xmath135 outputs 1 . since @xmath132 and @xmath137 are disjoint , we have that @xmath138 . hence , two distinct nodes in @xmath127 output 1 when executing @xmath118 . this proves the claim . it is enough to prove the theorem for sufficiently large @xmath2 . fix any integer @xmath139 and set the label space to be @xmath140 . to obtain a contradiction , assume that @xmath141 different advice strings suffice . form a family @xmath142 of @xmath143 disjoint sequences of integers , each of size 3 . more specifically , let @xmath144 . let @xmath145 . note that @xmath146 is a set of integer labels , each of which is larger than all labels that belong to sequences in @xmath142 . to verify that we have enough labels to define these sets , note that the largest label @xmath147!)$ ] . using the inequality @xmath148 , we get that @xmath149 . when @xmath139 , one can verify that @xmath150{\log d}$ ] , and @xmath151{(\log d)^5 } + \sqrt[3]{\log d } < \log d$ ] , so @xmath152 . it follows that @xmath153 . next , we construct a special family @xmath154 of rings of diameter at most @xmath2 . we will add rings to @xmath154 by following a procedure that we will describe shortly . each new ring that we add to @xmath154 will be the result of at most one ` gluing ' operation , denoted by @xmath155 , that takes two disjoint odd - sized rings @xmath156 and forms a new odd - sized ring @xmath127 . more specifically , the @xmath157 gluing operation takes two disjoint odd - sized rings @xmath132 and @xmath137 and forms the new ring defined as follows : construct paths @xmath128 and @xmath129 , add a new edge between the leaf of @xmath128 with smaller label and the leaf of @xmath129 with smaller label , add a new node with label @xmath158 , and add edges from this new node to the two remaining leaves . the gluing operation is illustrated in figure [ chopandglue](b ) . the additional node is introduced so that the resulting ring has an odd number of nodes . note that this additional node s label comes from @xmath146 , which ensures that the new ring formed by a gluing operation does not contain duplicate labels . further , note that , due to the dependence of the additional node s label on @xmath20 , no two gluing operations introduce additional nodes with the same label . from node @xmath159 is removed . ( b ) the glue operation on two rings of size 3 : concatenate paths @xmath128 , @xmath129 and add an additional node @xmath80 with a label from @xmath146 . ] we now describe the procedure for adding rings to @xmath154 . in stage 1 , we consider the set @xmath160 of 3-cliques obtained from each sequence in @xmath142 by identifying each integer with a node and adding all edges between them . we take a subset @xmath161 of size at least @xmath162 such that the same advice string is provided to the algorithm for each ring in @xmath82 . the rings in @xmath82 are added to @xmath154 , which concludes stage 1 . in each stage @xmath163 , we consider the set @xmath164 of rings that were added to @xmath154 in stage @xmath165 . the elements of @xmath164 are partitioned into @xmath166 pairs of rings in an arbitrary way . for each such pair @xmath156 , we perform @xmath155 . define @xmath167 to be the set of all of the resulting rings . we take a subset @xmath168 of size at least @xmath169 such that the same advice string is provided to the algorithm for each ring in @xmath78 . the rings in @xmath78 are added to @xmath154 , which concludes stage @xmath170 . this concludes the construction of @xmath154 . it is not immediately clear that this construction can always be carried out . in particular , in order to define @xmath78 in each stage @xmath171 , there must exist @xmath169 rings in @xmath167 such that the same advice string is provided to the algorithm for each . to prove this fact , and to obtain the desired contradiction to prove the theorem , we will use the following two claims . * claim 2 * _ consider any @xmath172 and any ring @xmath173 . for every @xmath174 , there exists a ring @xmath175 such that @xmath124 contains @xmath176 as a subgraph . _ to prove the claim , we proceed by induction on @xmath50 . the case where @xmath177 is trivial . next , assume that , for some @xmath172 and any ring @xmath173 , for every @xmath174 , there exists a ring @xmath175 such that @xmath124 contains @xmath176 as a subgraph . we now prove that , for any @xmath178 , there exists a ring @xmath179 such that @xmath124 contains @xmath180 as a subgraph . by construction , @xmath181 for some disjoint rings @xmath182 . by the definition of the gluing operation , @xmath46 contains both @xmath128 and @xmath129 as disjoint subgraphs . since @xmath124 has one fewer edge than @xmath46 , at least one of @xmath128 and @xmath129 is a subgraph of @xmath124 . by induction , this proves claim 2 . * claim 3 * _ consider any @xmath183 with @xmath174 . for any ring @xmath184 , there exist disjoint @xmath185 such that @xmath186 contains @xmath187 and @xmath176 as subgraphs . _ to prove the claim , note that , by our construction , @xmath188 for some disjoint @xmath189 . it follows that @xmath128 and @xmath129 are disjoint subgraphs of @xmath186 . if @xmath190 , setting @xmath191 and @xmath192 satisfies the statement of the claim . if @xmath193 , then , by claim 2 , there exists a ring @xmath194 such that @xmath128 contains @xmath187 as a subgraph . similarly , there exists a ring @xmath175 such that @xmath129 contains @xmath176 as a subgraph . note that , since @xmath132 and @xmath137 are disjoint , it follows that @xmath187 and @xmath176 are disjoint , so @xmath195 and @xmath196 are disjoint . thus , @xmath195 and @xmath196 satisfy the statement of the claim , which concludes its proof . we now show that , in any fixed stage of the above construction of @xmath154 , the advice string that is provided for the rings added to @xmath154 in this stage is different than the advice strings provided for the rings added to @xmath154 in all previous stages . * claim 4 * _ consider any @xmath183 with @xmath174 . for any ring @xmath197 and any ring @xmath198 , we have @xmath199 . _ to prove the claim , notice that , by claim 3 , there exist disjoint @xmath185 such that @xmath200 contains @xmath187 and @xmath176 as subgraphs . recall that the algorithm is provided the same advice string for all graphs in @xmath201 , so @xmath202 . by claim 1 and the correctness of @xmath44 , it follows that @xmath203 . hence , @xmath204 , which completes the proof of claim 4 . we show that the construction of @xmath154 can always be carried out . * claim 5 * _ for all @xmath172 , in stage @xmath50 of the construction , there exist at least @xmath205 rings in @xmath206 such that the same advice string is provided to the algorithm for each of them . _ to prove the claim , first note that , for @xmath177 , there are @xmath207 different strings that could be used as advice for rings in @xmath160 . so , by the pigeonhole principle , there are at least @xmath208 rings in @xmath160 such that the same advice string is provided to the algorithm for each of them . next , for any @xmath209 , claim 4 implies that there are @xmath210 strings that are not provided to the algorithm as advice for rings in @xmath206 . namely , there are at most @xmath211 different strings used as advice for rings in @xmath206 . by the pigeonhole principle , there are at least @xmath205 rings in @xmath206 such that the same advice string is provided to the algorithm for each of them , which proves the claim . the following claim implies that @xmath212 , for all @xmath213 . later , we will use two rings from some @xmath78 to obtain the desired contradiction needed to complete the proof of the theorem . * claim 6 * _ for all @xmath171 , at the end of stage @xmath170 , @xmath214 . we prove the claim by induction on @xmath170 . when @xmath215 , we have @xmath216 . by claim 5 , there exist at least @xmath217 rings in @xmath160 such that the same advice string is provided to the algorithm for each of them . this implies that @xmath218 , as required . as induction hypothesis , assume that , at the end of some stage @xmath219 , @xmath214 . in stage @xmath220 , the set @xmath201 is partitioned into pairs and @xmath221 consists of one ( glued ) ring for each such pair . thus , @xmath222 . by claim 5 , there exist at least @xmath223 rings in @xmath221 such that the same advice string is provided to the algorithm for each of them . this implies that @xmath224 , as required . this concludes the proof of claim 6 . finally , we construct a ring @xmath225 on which algorithm @xmath118 fails . note that the rings in @xmath154 all have node labels from the sets @xmath142 and @xmath146 , and we proved that the largest integer in these sets is less than @xmath2 . thus , the rings in @xmath154 all have node labels from the range @xmath226 . to construct @xmath225 , we take any two ( disjoint ) rings @xmath227 ( which exist by claim 6 ) and form a ring @xmath225 consisting of the concatenation of paths @xmath128 , @xmath129 , and the path of @xmath2 nodes with labels @xmath228 . recall that , in the construction , all rings in @xmath78 , for any fixed @xmath213 , get the same advice string . by claim 4 , each of the @xmath207 distinct advice strings is used for rings in some @xmath78 . therefore , there exists a stage @xmath170 such that the advice provided for all graphs in @xmath78 is the string @xmath229 . by claim 2 , there exists a ring @xmath194 such that @xmath128 contains @xmath187 as a subgraph , and , there exists a ring @xmath175 such that @xmath129 contains @xmath176 . it follows that @xmath225 contains @xmath187 and @xmath176 as subgraphs , and @xmath230 . by claim 1 , when algorithm @xmath118 is executed by the nodes of ring @xmath225 , two distinct nodes output 1 , which contradicts the correctness of @xmath118 . note that the size of @xmath225 is at least @xmath2 and at most @xmath75 , which implies that the size of the label space is linear in the size of @xmath225 . the obtained contradiction was due to the assumption that @xmath141 . hence the number @xmath207 of different advice strings is at least @xmath231 , which implies that the size of advice is @xmath232 for some ring of diameter at most @xmath2 . since imposing less time can not make the selection task easier , we have the following corollary . [ cor ] for any constant @xmath31 , consider any selection algorithm @xmath118 such that , for any ring @xmath46 , algorithm @xmath118 halts within @xmath121 rounds . for every positive integer @xmath2 , there exists a ring @xmath46 of diameter at most @xmath2 for which algorithm @xmath118 requires advice of size @xmath32 . we now establish an upper bound on the size of advice that matches the lower bound from theorem [ lb1 ] . let @xmath2 be any positive integer and let @xmath13 be a power of 2 . we consider algorithms that solve selection on the class of rings with diameter at most @xmath2 and labels from @xmath233 . recall that we assume that @xmath13 is polynomial in the size of the ring , and hence also in @xmath2 . in order to prove the upper bound , we propose a family of selection algorithms such that , for each fixed @xmath234 $ ] , there is an algorithm in the family that takes @xmath34 bits of advice , and , for each ring @xmath46 , halts within @xmath121 rounds . we start with an informal description of the algorithm and the advice for any fixed @xmath234 $ ] . the algorithm consists of two stages , and the advice consists of two substrings @xmath235 and @xmath236 . for any ring @xmath46 , the substring @xmath235 of the advice is the binary representation of the integer @xmath237 . the size of this advice is @xmath34 . note that @xmath238 . in stage 1 of the algorithm the nodes perform @xmath239 communication rounds , after which each node @xmath16 has acquired knowledge @xmath17 . next , each node @xmath16 checks if its own label is the largest of the labels it has seen within @xmath18 communication rounds , i.e. , the largest in the set @xmath19 . if not , then @xmath16 outputs 0 and halts immediately . let @xmath240 be the set of _ candidate nodes _ , i.e. , nodes @xmath16 whose label is the largest in @xmath19 . clearly , the largest node in @xmath46 is in @xmath241 , and every node knows if it belongs to @xmath240 . nodes in @xmath240 proceed to the next stage of the algorithm . in stage 2 of the algorithm , each node in @xmath240 determines whether or not it is the largest node in @xmath46 , without using any further communication rounds . this is achieved using @xmath236 , the second substring of the advice , which we now describe . let @xmath242 be the family of sets of labels which contain all labels in @xmath241 and no larger labels . at a high level , we construct an integer colouring @xmath243 of the family @xmath242 such that , for any @xmath244 , when the colour @xmath245 is given as advice to candidate nodes , each of them can determine , without any communication , whether or not it is the candidate node with the largest label . call such a colouring _ discriminatory_. substring @xmath236 of the advice will be @xmath245 for some @xmath244 and some discriminatory colouring @xmath243 of @xmath242 . ( we can not simply use @xmath246 because our colouring @xmath243 will be defined on sets of fixed size , and sets of candidate nodes for different rings do not have to be of equal size . ) using @xmath236 , the candidate nodes solve selection among themselves . this concludes the high - level description of the algorithm . the main difficulty of the algorithm is finding a discriminatory colouring @xmath243 . for example , bijections are trivially discriminatory , as nodes could deduce the set @xmath247 . however , we can not use such a colouring . indeed , the colouring must use few colours , otherwise the advice would be too large . we will be able to construct a discriminatory colouring with few colours using the fact that the number of candidate nodes is bounded by a constant that depends only on @xmath248 , as given in the following lemma . [ boundcandidates ] @xmath249 . first note that @xmath250 since the largest node @xmath16 in @xmath46 also has the largest label in @xmath19 . if @xmath251 , then @xmath252 is at most the number of nodes in @xmath46 , i.e. , at most @xmath253 , which is at most @xmath254 . hence we may assume that @xmath255 . next , note that there can not be two candidate nodes @xmath256 such that the distance between them is at most @xmath18 . indeed , if @xmath257 and @xmath258 , then the node in @xmath259 with smaller label will not be a candidate . since the number of nodes in @xmath46 is @xmath260 , it follows that the number of candidate nodes is less than @xmath261 . since @xmath262 and @xmath237 , it can be shown that @xmath263 is at most @xmath264 . this completes the proof of the lemma . we define what it means for a colouring to be _ legal_. it will be shown that a legal colouring known by all nodes is discriminatory . let @xmath265 . let @xmath266 denote the set of all @xmath267-tuples of the form @xmath268 such that @xmath269 and @xmath270 . ( we identify the tuple @xmath268 with the set @xmath271 . ) a colouring @xmath243 of elements of @xmath266 by integers is _ legal _ if , for each colour @xmath272 and each integer @xmath273 , either 1 . every @xmath274 that contains @xmath275 and is coloured @xmath272 has @xmath276 , or , 2 . every @xmath274 that contains @xmath275 and is coloured @xmath272 has @xmath277 . informally , a colouring of sets of labels is legal if , for all sets of a given colour in which a label @xmath275 appears , @xmath275 is either always the largest label or never the largest label . assume that we have a legal colouring @xmath243 of @xmath266 that uses @xmath278 colours , and that each node knows @xmath243 . using @xmath243 , we provide a complete description of our algorithm with advice of size @xmath34 and prove that it is correct . we will then describe such a legal colouring @xmath243 . in order to complete the description , it remains to construct a legal colouring @xmath243 that uses @xmath279 colours . note that , since @xmath13 is polynomial in @xmath2 , that the number of colours is indeed @xmath280 . consider the following mapping @xmath281 that maps each @xmath282 to a @xmath283-tuple @xmath284 . for each @xmath285 , set @xmath286 to be the largest integer @xmath170 such that there exists an integer in the range @xmath287 $ ] that is divisible by @xmath288 . next , take any bijection @xmath289 between @xmath290 and the set @xmath291 . define the colouring @xmath243 of @xmath266 as the composition of @xmath292 and @xmath289 . note that the colouring @xmath243 uses at most @xmath293 colours . [ islegal ] @xmath243 is a legal colouring of the elements of @xmath266 . consider an arbitrary colour @xmath294 and any integer @xmath273 . to obtain a contradiction , assume that : * there exists an element of @xmath266 , say @xmath295 , that contains @xmath275 , is coloured @xmath272 , and has @xmath296 , and , * there exists an element of @xmath266 , say @xmath297 , that contains @xmath275 , is coloured @xmath272 , and has @xmath298 . let @xmath299 . since @xmath295 contains @xmath275 and @xmath296 , it follows that @xmath300 for some @xmath285 . by the definition of @xmath292 , we know that the range @xmath301 $ ] contains an integer , say @xmath302 , that is divisible by @xmath303 , and we know that no integer in this range is divisible by @xmath304 . moreover , since @xmath298 , we know that the range @xmath305 $ ] contains an integer , say @xmath306 , that is divisible by @xmath303 , and we know that no integer in this range is divisible by @xmath307 . it follows that there are two distinct integers @xmath308 in the range @xmath309 $ ] that are divisible by @xmath310 and that no integer in the range @xmath309 $ ] is divisible by @xmath307 . since @xmath311 is the smallest integer greater than @xmath302 that is divisible by @xmath310 , it follows that @xmath312 \in [ x_0 + 1,\ldots , y_i]$ ] , so @xmath311 is not divisible by @xmath307 . note that , for some positive integer @xmath313 , we can write @xmath314 and @xmath315 . since neither @xmath302 nor @xmath316 is divisible by @xmath307 , it follows that both @xmath313 and @xmath317 must be odd , a contradiction . [ ub ] consider any fixed @xmath234 $ ] and any positive integer @xmath2 . for any ring @xmath46 with diameter at most @xmath2 , algorithm select solves selection in the ring @xmath46 in time @xmath318 and with advice of size @xmath34 . notice that algorithm select halts within @xmath318 communication rounds . indeed , by line 2 of the algorithm , every node uses exactly @xmath319 communication rounds , and @xmath320 . next , note that since @xmath235 is the binary representation of @xmath321 , the length of @xmath235 is @xmath34 . further , recall that @xmath236 is the binary representation of a colour assigned by @xmath243 , which uses @xmath278 colours where @xmath322 . thus , the length of @xmath236 is @xmath34 , and hence the size of advice is @xmath34 . finally , we prove the correctness of the algorithm for an arbitrary ring @xmath46 . first , note that the construction of the advice string can indeed be carried out . in particular , at line 6 in the advice construction , the tuple @xmath323 exists since , by lemma [ boundcandidates ] , the number of candidates ( and , hence , the value of @xmath324 ) is bounded above by @xmath267 . next , recall that @xmath242 is the family of sets of labels which contain all labels in @xmath241 and no larger labels . the following claim shows that the colouring @xmath243 is discriminatory . * claim 1 * _ for any @xmath325 , let @xmath326 . when lines 6 11 of algorithm select are executed , the largest node in @xmath240 outputs 1 , and all other nodes in @xmath240 output 0 . this proves that colouring @xmath243 is discriminatory . _ to prove the claim , consider the largest node @xmath327 . we first show that @xmath328 is equal to @xmath329 s label . indeed , since @xmath329 has the largest label in @xmath330 , it follows that @xmath331 . since @xmath247 does not contain labels larger than those in @xmath241 , @xmath328 is equal to @xmath329 s label . next , note that , by line 7 , @xmath332 . since @xmath333 is equal to @xmath329 s label , the condition of the * if * statement at line 8 evaluates to true , so @xmath329 outputs 1 . next , consider any node @xmath334 that is not the largest . since @xmath333 is not equal to @xmath16 s label and @xmath243 is a legal colouring , it follows that no tuple that is coloured @xmath335 has @xmath16 s label as its first entry . in other words , no tuple in @xmath336 has @xmath16 s label as its first entry , so the * if * statement at line 8 evaluates to false , and @xmath16 outputs 0 . this completes the proof of the claim . we have shown that , if the nodes in @xmath240 are provided as advice the value of @xmath245 for any @xmath337 , then lines 6 11 of algorithm select solve selection among the nodes in @xmath240 . it remains to show that the advice substring @xmath236 created in the advice construction is indeed the binary representation of such an @xmath245 . * claim 2 * _ in advice construction , @xmath338 . _ to prove the claim , recall that , by line 4 of the advice construction , @xmath339 is the decreasing sequence of labels of nodes in @xmath240 . consider line 6 of the advice construction , and note that the largest element @xmath328 of the tuple @xmath340 is equal to @xmath341 , which proves that the labels @xmath342 are no larger than those of nodes in @xmath240 . further , since @xmath343 and @xmath344 , it follows that @xmath340 contains all of the labels of nodes in @xmath241 . therefore , @xmath338 , as claimed . we can now conclude that algorithm select solves selection in the entire ring . indeed , in lines 1 4 , each node that learns about the existence of a node with a larger label than itself outputs 0 and halts . the remaining nodes , namely those in @xmath240 ( which necessarily includes the largest node ) , proceed to lines 5 11 . from line 7 of the advice construction and by claim 2 , line 5 of algorithm select assigns to @xmath335 the value of @xmath245 for some @xmath337 . finally , claim 1 shows that the largest node outputs 1 and all other nodes output 0 . corollary [ cor ] and theorem [ ub ] imply the following tight bound on the size of advice . the optimal size of advice to complete selection in any ring in time linear in its diameter ( and not exceeding it ) is @xmath9 . we now turn attention to very fast selection for rings of diameter @xmath2 . since every such ring has size @xmath0 linear in @xmath2 , and since @xmath13 is polynomial in @xmath0 , selection ( and even election ) can be accomplished using @xmath29 bits of advice without any communication by providing the largest label as advice . it turns out that , even when the available time is @xmath345 , for any constant @xmath346 , this size of advice is necessary . compared to theorem [ ub ] , this shows that selection in time @xmath345 where @xmath346 , requires exponentially more advice than selection in time @xmath25 . for any constant @xmath347 , any selection algorithm @xmath44 for rings of diameter @xmath2 that works in time at most @xmath345 requires @xmath27 bits of advice . it is enough to consider sufficiently large values of @xmath2 . let @xmath348 , and let @xmath349 . we start by defining a special class of rings , and then proceed to show that algorithm @xmath44 requires @xmath27 bits of advice for this class . for the sake of clarity , we will use the notation @xmath350 and @xmath351 . first , we define a set of paths @xmath352 as follows . each path consists of @xmath353 nodes with labels from @xmath186 , with the middle node of the path having the largest label . more specifically , path @xmath354 is obtained by considering the path of nodes with labels @xmath355 , respectively , and reversing the order of the last @xmath356 labels . formally , for each @xmath357 , let @xmath354 be the path of nodes whose labels form the sequence @xmath358 . next , we define a set @xmath359 of rings @xmath360 . each ring @xmath361 will consist of the paths @xmath362 along with enough nodes with labels from @xmath363 to ensure that @xmath361 has size @xmath75 . in order to define the ring @xmath361 , we first construct a path @xmath364 by taking the paths @xmath365 , and , for each @xmath366 , connecting the last node of path @xmath367 with the first node of path @xmath195 . more specifically , @xmath364 is obtained by taking the union of the paths @xmath365 , and , for each @xmath366 , adding an edge between the nodes with labels @xmath368 and @xmath369 . next , we construct a path @xmath370 whose labels form the sequence @xmath371 . since the number of nodes in @xmath364 is at most @xmath372 , which , for sufficiently large @xmath2 , is strictly less than @xmath75 , it follows that path @xmath370 has at least one node . finally , the ring @xmath361 is obtained by joining the paths @xmath364 and @xmath370 . more specifically , @xmath361 is obtained by taking the union of the paths @xmath364 and @xmath370 and adding the edges @xmath373 and @xmath374 . the paths and rings constructed above are illustrated in figure [ pathsrings ] . ] the following claim asserts that , in any two rings in @xmath359 that contain path @xmath354 as a subgraph , the middle node of this path in both rings acquires the same knowledge when executing algorithm @xmath44 . * claim 1 * _ consider any @xmath357 and any @xmath375 . let @xmath376 be the node in @xmath361 with label @xmath377 and let @xmath378 be the node in @xmath379 with label @xmath377 . then , knowledge @xmath380 in @xmath361 is equal to knowledge @xmath381 in @xmath379 . _ to prove the claim , note that , in both @xmath361 and @xmath379 , the node with label @xmath377 is the middle node of path @xmath354 . since the path @xmath354 has length @xmath353 , it follows that , within @xmath80 communication rounds , node @xmath376 does not learn about any nodes in @xmath361 outside of @xmath354 . similarly , node @xmath378 does not learn about any nodes in @xmath379 outside of @xmath354 . this implies that knowledge @xmath380 in @xmath361 is equal to knowledge @xmath381 in @xmath379 , which completes the proof of the claim . we proceed to prove the theorem by way of contradiction . assume that the number of bits of advice needed by algorithm @xmath41 for the rings in @xmath154 is less than @xmath382 . it follows that the number of distinct advice strings that are provided to algorithm @xmath44 for the rings in @xmath359 is strictly less than @xmath383 . however , since the class @xmath359 consists of @xmath384 rings ( where @xmath385 for sufficiently large @xmath2 ) this means that there exist two rings in @xmath359 , say @xmath386 with @xmath387 , such that @xmath388 . first , consider the execution of @xmath44 by the nodes of ring @xmath389 . let @xmath390 be the node in @xmath389 with label @xmath391 , and note that @xmath390 is the largest node in @xmath389 . it follows that , in this execution , node @xmath390 outputs 1 . next , consider the execution of @xmath44 by the nodes of ring @xmath392 . let @xmath393 be the node in @xmath392 with the label @xmath391 . by claim 1 , @xmath394 in @xmath389 is equal to @xmath395 in @xmath392 . moreover , algorithm @xmath44 halts within @xmath80 rounds and @xmath388 . it follows that the execution of @xmath44 by @xmath390 in @xmath389 is identical to the execution of @xmath44 by @xmath393 in @xmath392 . hence , @xmath393 outputs 1 in the execution of @xmath44 in @xmath392 . however , node @xmath393 is not the largest in @xmath392 . indeed , @xmath393 s label is @xmath391 , whereas path @xmath396 ( and , hence , ring @xmath392 ) contains a node labeled @xmath397 . this contradicts the correctness of @xmath44 . for any constant @xmath346 , the optimal size of advice to complete selection in any ring of diameter at most @xmath2 in time @xmath345 is @xmath5 . we established tradeoffs between the time of choosing the largest node in a network and the amount of _ a priori _ information ( advice ) needed to accomplish two variations of this task : election and selection . for the election problem , the tradeoff is complete and tight up to multiplicative constants in the advice size . moreover , it holds for the class of arbitrary connected graphs . for selection , our results are for the class of rings and a small gap remains in the picture . we proved that in rings with diameter @xmath3 at most @xmath2 , the optimal size of advice is @xmath5 if time is @xmath6 for any @xmath7 , and that it is @xmath9 if time is at most @xmath3 . hence , the first open problem is to establish the optimal size of advice to perform selection for rings when the time is in the small remaining gap , for example , in time @xmath398 . another problem is to extend the tradeoff obtained for selection in rings to the class of arbitrary connected graphs . in particular , it would be interesting to investigate whether the optimal advice needed to perform fast selection in graphs of size much larger than the diameter depends on their size ( like in the case of election ) or on their diameter . as noted in the introduction , all known leader election algorithms in labeled networks choose as leader either the node with the largest label or that with the smallest label . it is worth noting that the situation is not always symmetric here . for example , the time slice algorithm for leader election @xcite ( which elects the leader using exactly @xmath0 messages in @xmath0-node rings , at the expense of possibly huge time ) finds the node with the smallest label , and it does not seem to be possible to convert it directly to finding the node with the largest label . ( of course it is possible to first find the node with the smallest label and then to use this leader to find the node with the largest label , but this takes additional time and communication . ) in our case , however , the situation is completely symmetric with respect to the order of labels : our results hold without change , if finding the largest node is replaced by finding the smallest . it is an open question whether they also remain valid if electing the largest node is replaced by general leader election ( i.e. , the task in which a single _ arbitrary _ node becomes the leader , and all other nodes become non - leaders and also learn the identity of the leader ) and if selecting the largest node is replaced by general leader selection ( i.e. , the task in which a single _ arbitrary _ node becomes the leader , and all other nodes become non - leaders ) . our lower bound for election in time at most the diameter and our algorithms for selection use the ordering in an essential way , and it is not clear if smaller advice would be sufficient to elect or select some arbitrary node in a given time . p. boldi , s. shammah , s. vigna , b. codenotti , p. gemmell , and j. simon , symmetry breaking in anonymous networks : characterizations . 4th israel symposium on theory of computing and systems , ( istcs 1996 _ ) , 16 - 26 . j. chalopin , local computations on closed unlabelled edges : the election problem and the naming problem _ proc . 31st conference on current trends in theory and practice of computer science ( sofsem 2005 ) _ , 82 - 91 . p. fraigniaud , d. ilcinkas , a. pelc , communication algorithms with advice , journal of computer and system sciences 76 ( 2010 ) , 222232 . p. fraigniaud , d. ilcinkas , a. pelc , tree exploration with advice , information and computation 206 ( 2008 ) , 12761287 . p. fraigniaud , a. korman , e. lebhar , local mst computation with short advice , theory of computing systems 47 ( 2010 ) , 920933 . haddar , a.h . kacem , y. mtivier , m. mosbah , and m. jmaiel , electing a leader in the local computation model using mobile agents . 6th acs / ieee international conference on computer systems and applications ( aiccsa 2008 _ ) , 473 - 480 .

we have used the radio telescope to obtain new global data for 16 giant low surface brightness ( lsb ) galaxies . our targets have optical luminosities and disk scale lengths at the high end for spiral galaxies ( and kpc for=75 mpc ) , but they have diffuse stellar disks , with mean disk surface brightnesses magnitude fainter than normal giant spirals . thirteen of the galaxies previously had been detected in by other workers , but the published observations were either confused , resolved by the telescope beam , of low signal - to - noise , or showed significant discrepancies between different authors . for the other 3 galaxies , no data were previously available . several of the galaxies were resolved by the e - w beam , so global parameters were derived from multiple - point mapping observations . typical masses for our sample are , with.3 - 1.7 ( in solar units ) . all of the observed galaxies have published optical surface photometry , and we have compiled key optical measurements for these objects from the literature . we frequently find significant variations among physical parameters of giant lsb galaxies reported by various workers .

astro-ph0010075

in 1987 , bothun et al . reported the serendipitous discovery of the extraordinarily large low surface brightness ( lsb ) galaxy now known as malin-1 . in spite of having a projected @xmath8-band central surface brightness of only 26.5 mag arcsec@xmath9 , malin-1 is the largest spiral galaxy known , with a disk scale length of 73 kpc ( assuming @xmath10=75 mpc@xmath3 ) , and an exceptionally high mass ( @xmath11 ) and optical luminosity ( @xmath1223.1 ) . subsequently , through systematic searches of photographic survey plates , other galaxies with similar ( albeit slightly less extreme ) properties to malin-1 have been uncovered ( bothun et al . 1990 ; sprayberry et al . 1993 ; sprayberry et al . we hereafter refer to these as `` lsb giants '' . a handful of lsb giants are also found in the ugc ( nilson 1973 ) , ngc ( dreyer 1953 ) , and eso ( lauberts & valentijn 1989 ) catalogues ( see gallagher & bushouse 1983 ; impey & bothun 1989 ; walsh et al . 1997 ; pickering et al . 1997 ; schombert 1998 ) . nonetheless , while recent photographic and ccd surveys have uncovered large numbers of new small and medium - sized , moderate - to - low surface brightness spiral galaxies ( e.g. , schombert et al . 1992 ; impey et al . 1996 ; oneil et al . 1997 ) , lsb giants have remained relatively rare . having faint , diffuse disks , but sizes , masses , and luminosities at the high end for disk galaxies , the lsb giants occupy a unique realm of physical parameters space and may share evolutionary histories distinct from other lsb galaxies ( e.g. , hoffman et al . 1992 ) . since a continuum of values exists for galaxy properties such as surface brightness , luminosity , and scale length , sprayberry et al . ( 1995b ) proposed to define lsb giants as those objects meeting a `` diffuseness index '' criterion : @xmath13=100 mpc@xmath3 . here we have rewritten the diffuseness criterion for @xmath2=75 mpc@xmath3 , which is assumed throughout this work . ] , where @xmath14 is the extrapolated , deprojected @xmath8-band disk central surface brightness in magnitudes arcsec@xmath9 , and @xmath15 is the disk scale length in kpc . among seven lsb giants described by sprayberry et al . ( 1995b ) , mean properties include : @xmath16 , @xmath17 mags arcsec@xmath9 , and @xmath18 kpc . the colors of these lsb giants are thus comparable to those of normal spirals ( sprayberry et al . 1995b ) , but are redder than typical colors of many small and moderate - sized lsb disks ( e.g. , mcgaugh & bothun 1994 ; matthews & gallagher 1997 ; de blok et al . 1996 ; beijersbergen et al . 1999 ) . in addition , the lsb giants are distinct from other more common lsb spirals in that they often have a significant bulge component ( e.g. , gallagher & bushouse 1983 ; knezek 1993,1998 ) , and frequently their centers harbor an active nucleus ( e.g. , schombert 1998 ) . the origin and evolutionary histories of lsb giant galaxies are still enigmatic . hoffman et al . ( 1992 ) have proposed a formation scenario whereby these systems form in very low density regions from rare , 3@xmath19 density fluctuations . they predict these galaxies should exhibit quiescent , unevolved , gas - rich disks , with rotation curves that flatten near @xmath20300 . knezek ( 1993 ) has suggested an alternative scenario , based on kormendy ( 1989 ) , whereby lsb giants may have dissipatively formed from massive , metal - poor dark matter halos . testing formation and evolution scenarios for lsb giants requires an accurate knowledge of the neutral gas properties and linewidths of these galaxies . and only by combining such measures with optical data can we begin to build a picture of the star - formation histories of these systems and their relationship to other types of lsb galaxies . other motivations also exist for improved observations . hoffman et al . ( 1992 ) have argued that for the enormous disks of lsb giants to remain quiescent over a hubble time , they must be very isolated . yet studies hint that lsb giants are in fact less isolated than other lsb spirals , although redshift surveys in the vicinities of these objects are still incomplete ( sprayberry et al . 1995b ) . pointed observations in the vicinity of lsb giants can thus reveal if these galaxies have any yet - undiscovered gas - rich neighbors . another important use of data is for exploring the tully - fisher ( tf ) relation for giant lsb spirals . sprayberry et al . ( 1995a ) have shown that at least two of the presently known lsb giant galaxies are extreme outliers from the tf relation defined by normal galaxies . this is unlike the bulk of moderate - sized , moderate luminosity lsb galaxies , which tend to follow tf ( sprayberry et al . 1995a ; zwaan et al . 1995 ; verheijen 1997 ) . it is of considerable interest therefore to assess from a larger sample whether lsb giants deviate systematically from the tf relation . while previous observations have established that lsb giants are in general very gas - rich ( @xmath21 ; e.g. , sprayberry et al . 1995b ; walsh et al . 1997 ; pickering et al . 1997,1999 ) , unfortunately existing data for many lsb giants are of dubious quality ( i.e. , the galaxy was confused or resolved by the telescope beam , the spectra are of low signal - to - noise , or measurements from different workers are highly discrepant ; see also table 5 and section [ disc ] ) . for example , based on arecibo 21-cm observations of 3 objects , sprayberry et al . ( 1995b ) suggested that peculiar , asymmetric profiles may be commonplace for lsb giants . however , since the extents of these galaxies were expected to be comparable to the size of the telescope beam , it is important to verify that these `` peculiar '' spectra do not result from some combination of source resolution and telescope mispointing . in addition , independent checks on derived parameters are valuable since it is more difficult to accurately measure integrated fluxes and linewidths when the global profiles are quite broad compared to the bandwidth used , and when sources are at large recessional velocities ( @xmath22 15000 ) , where flux calibration can become increasingly uncertain . finally , there are still a handful of known lsb giants for which no data have previously been obtained . quality single - dish spectra for lsb giants are also a useful precursor and complement to aperture synthesis studies of these galaxies ( e.g. , walsh et al . 1997 ; pickering et al . 1997,1999 ) . mapping is of course a critical part of understanding the dynamics and gas distributions for these galaxies , but since many of the known examples of lsb giants are rather distant ( @xmath2310000 ) and of relatively modest optical angular size ( @xmath24 ) , such observations are challenging and benefit from careful planning based on prior measurements . moreover , because lsb giants are generally expected to have disks with relatively low surface densities ( e.g. , pickering et al . 1997 ) , diffuse emission can be missed in aperture synthesis measurements , and total flux and maximum rotational velocity measures from pencil beam observations serve as an important check . based on the above motivations , we have used the radio telescope to obtain new global observations of a sample of 16 lsb giant galaxies for which existing global measurements were incomplete , required reconfirmation , or were nonexistent . we measure integrated line fluxes , linewidths , and recessional velocities , and attempt to clear up conundrums surrounding several of these objects in the literature . our targets for the present program were 16 lsb giant galaxies meeting the `` diffuseness '' criterion stated in section 1 . the objects were culled from knezek ( 1993 ) , sprayberry et al . ( 1995b ) , and pickering et al . ( 1997 ) , and all have available optical surface photometry . thirteen of our targets have previously been observed in by other workers , but the published parameters were uncertain or incomplete ( see below ) . for the other 3 galaxies , no data were previously available . some basic parameters for our target galaxies are summarized in table 1 . coordinates , @xmath8 magnitudes , angular sizes , axial ratios , and radial velocities were culled from the ned database . position angles and scatter in the published radial velocity values [ @xmath25 were taken from the leda database , and inclinations were obtained from the various references given in table 6 ( discussed below ) . inclinations and position angles marked with a colon ( :) were estimated by us from inspection of the digitized sky survey . lccllrrrrrlrr + ident . & r.a . & dec & & @xmath26 & @xmath27 & @xmath28 & @xmath29 & @xmath30 & @xmath31 & @xmath32 + & & leda & ned & mag & @xmath33 & & @xmath34 & @xmath34 & km / s & km / s + + + ugc 568 & 00 52 35.5 & -01 19 01 & s ? & sd & 15.08 & 1.17 & 0.70 & 40 & & 13197 & 204 + ugc 1752 & 02 13 30.0 & 24 40 00 & sc & sa(s)cd & 16.5 & 1.55 & 1.00 & 0 : & 0 : & 17847 & 60 + pgc 135657 & 02 37 38.7 & -01 59 18 & sb & sc & 15.47 & 0.78 & & 38 & 63 & 12701 & 60 + ugc 2936 & 04 00 12.6 & 01 49 36 & sbc & sb(s)d & 15.00 & 2.51 & 0.27 & 78 & 30 & 3817 & 125 + ugc 3140 & 04 40 20.1 & 00 31 35 & sbc & sa(rs)c : & 13.30 & 1.78 & 0.91 & 0 : & 175 & 4633 & 60 + ngc 2770 & 09 06 29.8 & 33 19 53 & sc & sa(s)c : & 12.77 & 3.66 & 0.30 & 79 & 148 & 1947 & 63 + pgc 135754 & 10 34 52.9 & 02 20 57 & sc & sc & 16.34 & & & 49 & & 21335 & 60 + f568 - 6 & 10 37 08.5 & 21 06 24 & scd & sd / p & 15.28 & 1.67 & 0.79 & 38 & & 13830 & + ugc 6614 & 11 36 39.1 & 17 25 14 & sa & ( r)sa(r)a : & 14.37 & 2.02 & 0.86 & 35 & 0 : & 6351 & 60 + malin-1 & 12 34 27.3 & 14 36 15 & & s & 16.20 & 0.20 & 0.93 & 45 & 45 & 24750 & + pgc 45080 & 13 00 42.6 & 01 44 12 & sc & sc & 15.61 & 0.79 & 0.22 & 72 & 84 & 12264 & 295 + ugc 9024 & 14 04 21.0 & 22 18 28 & sa & s : & 16.0 & 2.09 & 0.95 & 37 & 0 : & 2323 & + f530 - 1 & 21 05 21.5 & 26 15 00 & sc & s(r ) & 16.31 & 0.47 & 0.74 & & & 14340 & + f533 - 3 & 22 14 53.3 & 24 57 48 & sbc & sbc(r ) & 15.43 & 0.88 & 0.65 & 65 : & 165 : & 12669 & + ngc 7589 & 23 15 41.9 & -00 00 45 & sb & sab(rs)a : & 15.01 & 1.07 & 0.62 & 58 & 60 & 8938 & 219 + pgc 71626 & 23 27 58.2 & -02 44 18 & sbc & sb(r)b pec & 15.14 & 1.78 & 0.59 & 61 & 65 & 9520 & 247 + + lcccccccccl + galaxy name & rms & @xmath35 & @xmath36 & @xmath37 & @xmath38 & @xmath32 & @xmath39 & @xmath40 & s / n & notes + + & mjy & mjy & & & & & & & & + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) & ( 9 ) & ( 10 ) & ( 11 ) + + + ugc 568 & 1.15 & 12.3 & 410 : & 300 : & 13420 & 10 & 2.28 & 0.28 & 10.7 & peculiar + ugc 1752 * & 2.84 & 26.6 & 429 & 388 & 17861 & 7 & 5.13 & 0.70 & 9.4 & + pgc 135657 & 2.71 & 23.8 & 134 & 100 & 13108 & 7 & 2.17 & 0.46 & 8.8 & blended ? + ugc 2936 * & 2.32 & 38.6 & 502 & 469 & 3811 & 4 & 11.23 & 0.72 & 16.7 & + ugc 3140 * & 4.01 & 130.8 & 131 & 150 & 4623 & 2 & 13.41 & 0.74 & 32.6 & + ngc 2770 * & 4.84 & 143.6 & 350 & 322 & 1947 & 2 & 36.93 & 1.41 & 29.6 & + pgc 135754 & 1.65 & 5.5 & 222 & 197 & 21769 & 15 & 0.59 & 0.30 & 3.4 & + f568 - 6 * & 2.29 & 17.5 & 420 & 393 & 13829 & 7 & 4.20 & 0.64 & 7.6 & + ugc 6614 * & 2.86 & 81.2 & 290 & 255 & 6351 & 2 & 15.94 & 0.72 & 28.4 & + malin-1 * & 2.02 & 9.4 & 341 & 293 & 24784 & 15 & 1.80 & 0.50 & 4.7 & + pgc 45080 & 2.90 & 11.5 & 315 & 280 & 12339 & 15 & 1.57 & 0.61 & 4.0 & + ugc 9024 * & 2.69 & 115.7 & 134 & 120 & 2321 & 1 & 11.41 & 0.48 & 42.9 & + f530 - 1 & 3.0 & 10.7 & 497 & 484 & 14335 & 10 & 2.07 & 0.75 & 3.6 & + f533 - 3 * & 2.72 & 13.3 & 399 & 384 & 12670 & 8 & 2.46 & 0.66 & 4.9 & + ngc 7589 & 1.68 & 10.4 & 414 : & 357 : & 8866 & 12 & 2.96 & 0.51 & 6.2 & blended + pgc 71626 * & 3.40 & 47.0 & 126 & 88 & 10016 & 5 & 3.94 & 0.56 & 13.8 & + + + the decimentric radio telescope is a meridian transit - type instrument of the kraus / ohio state design , consisting of a fixed spherical mirror ( 300 m long and 35 m high ) , a tiltable flat mirror ( 200@xmath4140 m ) , and a focal carriage moving along a curved rail track . due to an ongoing major renovation of the focal system , the length of the focal track was reduced to @xmath4260 m during the period of our observations , thus allowing tracking of a source on the celestial equator for about 45 minutes . the effective collecting area of the telescope is roughly 7000 m@xmath43 ( equivalent to a 94-m diameter parabolic dish ) . due to the elongated geometry of the mirrors , at 21-cm wavelength the telescope has a half - power beam width of @xmath5 e - w@xmath41 22@xmath33 n - s for the range of declinations covered in this work ( e. grard , private comm . ; see also matthews & van driel 2000 ) . typical system temperatures were @xmath4240 k for our project . the observations at were made in the period june 1998 - october 1999 , using a total of about 300 hours of telescope time . we obtained our observations in total power ( position - switching ) mode using consecutive pairs of two - minute on- and two - minute off - source integrations . off - source integrations were taken at approximately 20@xmath33 e of the target position . the autocorrelator was divided into two pairs of cross - polarized receiver banks , each with 512 channels and a 6.4 mhz bandpass . this yielded a channel spacing of 2.64 , for an effective velocity resolution of @xmath423.3 at 21-cm . the center frequencies of the two banks were tuned to the expected redshifted frequency of the target based on values from the literature ( table 1 ) . depending on the signal strength , the spectra were smoothed to a channel separation of @xmath427.9 or @xmath4213.2 during the data reduction in order to increase signal - to - noise . total integration times were up to 12 hours per galaxy , depending on the strength of the source and scheduling constraints . in all cases , data were initially obtained with the telescope pointed at the published optical center of the galaxy . however , as shown by van der hulst et al . ( 1993 ) , the disks of large lsb spirals may frequently extend to up to @xmath422.5@xmath41 their optical diameter . we therefore observed several of the targets , including the 8 galaxies with @xmath44@xmath45 ( i.e. one - third the fwhp e - w beamwidth ) at three or more spatial positions : one at the target s optical center , plus additional pointings offset to the east or west by multiples of one - half beamwidth ( see table 3 , discussed below ) . because of the large n - s diameter of the beam ( @xmath4622@xmath33 ) , these mapping observations were limited to pointings along an e - w line . 14.5 cm flux calibration ( i.e.,@xmath47-to - mjy conversion ) at is determined via regular measurements of a cold load calibrator and periodic monitoring of strong continuum sources by the nancay staff . standard calibration procedures include correction for declination - dependent gain variations of the telescope ( e.g. , fouqu et al . these techniques typically yield an internal calibration accuracy of @xmath4215% at frequencies near 1420mhz . lcccc + ident . & postion & rms & @xmath39 & @xmath48 + & & mjy & & + + + ugc 1752 & @xmath49w & 3.00 & 2.42 & 17858 + & c & 2.45 & 4.72 & 17846 + & @xmath49e & 4.65 & 1.84@xmath50 & 18028 + + ugc 2936 & 4@xmath33w & 2.55 & @xmath510.67 & + & 2@xmath33w & 3.55 & 5.99 & 3797 + & c & 2.76 & 9.83 & 3812 + & 2@xmath33e & 2.56 & 4.29 & 3807 + & 4@xmath33e & 2.31 & @xmath510.29 & + + ugc 3140 & 6@xmath33w & 2.59 & 0.74 & 4596 + & 4@xmath33w & 1.98 & 0.99 & 4662 + & 2@xmath33w & 2.28 & 6.33 & 4618 + & c & 2.32 & 10.60 & 4626 + & 2@xmath33e & 3.32 & 6.09 & 4628 + & 4@xmath33e & 3.44 & 1.40 & 4632 + & 6@xmath33e & 2.91 & @xmath510.27 & + + ngc 2770 & 6@xmath33w & 3.59 & 1.39 & 1908 + & 4@xmath33w & 3.38 & 4.08 & 1966 + & 2@xmath33w & 4.00 & 16.16 & 1950 + & c & 2.56 & 28.50 & 1948 + & 2@xmath33e & 3.46 & 15.58 & 1930 + & 4@xmath33e & 2.86 & 3.82 & 1922 + & 6@xmath33e & 4.15 & 1.91 & 2022 + + f568 - 6 & 2@xmath33w & 2.88 & 2.27 & 13825 + & c & 3.36 & 4.56 & 13812 + & 2@xmath33e & 2.94 & 1.20 & 13948 + + ugc 6614 & 6@xmath33w & 2.02 & @xmath510.15 & + & 4@xmath33w & 2.95 & 1.73 & 6437 + & 2@xmath33w & 3.06 & 10.28 & 6368 + & c & 2.00 & 10.39 & 6347 + & 2@xmath33e & 3.04 & 5.99 & 6338 + & 4@xmath33e & 2.09 & 1.36 & 6329 + & 6@xmath33e & 1.79 & 0.80 & 6355 + + malin 1 & 2@xmath33w & 1.37 & 0.89 & 24750 + & c & 2.08 & 1.63 & 24774 + & 2@xmath33e & 2.08 & 0.96 & 24794 + + ugc 9024 & 4@xmath33w & 2.63 & 1.08 & 1293 + & 2@xmath33w & 2.76 & 6.04 & 2323 + & c & 3.30 & 9.98 & 2323 + & 2@xmath33e & 2.48 & 4.76 & 2322 + & 4@xmath33e & 2.40 & 0.96 & 2372 + + f533 - 3 & 2@xmath33w & 2.33 & 1.45 & 12681 + & c & 1.88 & 2.01 & 12655 + & 2@xmath33e & 2.94 & @xmath511.09 & + + pgc 71626 & 4@xmath33w & 2.56 & 0.72 & 10037 + & 2@xmath33w & 2.38 & 1.70 & 10013 + & c & 2.51 & 2.51 & 10018 + & 2@xmath33e & 2.32 & 1.78 & 10015 + & 4@xmath33e & 3.22 & @xmath510.39 & + + [ ] @xmath50three out of four scans of ugc 1752 at the 2@xmath33e position were obtained with a band - limiting filter not optimized for the redshifted frequency of the source . as a result , the observed integrated flux at this position is diminished due to filter response deterioration near the edge of the filter . in our present program several of our targets have recessional velocities @xmath5212000 and hence were observed at frequencies where calibration reliability and consistency at and other radio telescopes are less well established . to estimate the comparative accuracy of our flux density calibration at these lower frequencies as well as recheck frequency dependent changes in the noise diode temperature , we examined continuum calibration data obtained at 1400 , 1425 , and 1280 mhz , from several periods over the course of the months during which our spectral line data were acquired ( l. alsac , private comm . ; see also thuan et al . 2000 ) . over this frequency range we found the noise diode temperature to vary by less than 10% . our data were corrected for this effect based on a linear correction curve derived from the continuum data . these calibration data also confirm the expected internal calibration accuracy of our data is @xmath5315% near 1420 - 1425mhz , but only @xmath5325% near 1280mhz . an additional step was required for accurate flux calibration of our data , as it has been found that changes have occurred in the output power of the calibration diode used at since the early 1990 s ( see figure 4 of theureau et al . 1998 ; see also thuan et al . 2000 ) , resulting in an overall shift of the absolute calibration scale . this makes it necessary to appropriately renormalize the fluxes determined via the standard calibration techniques described above ( e.g. , theureau et al . 1998 ; matthews et al . 1998 ; thuan et al . 2000 ) . matthews et al . ( 1998 ) showed via a statistical comparison of integrated fluxes measured for @xmath4230 galaxies at and elsewhere that applying a scaling factor of 1.26 to the flux densities very effectively corrects for the above effect , and restores the correct normalization of the flux scale . matthews & van driel ( 2000 ) subsequently found that the application of this same factor minimized scatter between fluxes determined for a second sample of galaxies observed at both and at arecibo . theureau et al . ( 1998 and priv . comm . ) also derived similar corrections via independent observations of line calibration sources . as a final calibration step we therefore apply a renormalization factor of 1.26 to all fluxes reported in the present work . we reduced our spectra using the standard dac and sir spectral line reduction packages available at the site . with this software we subtracted baselines ( generally third order polynomials ) , averaged the two receiver polarizations , and applied a declination - dependent conversion factor to convert from units of @xmath47 to flux density in mjy . because of the broad width of the lines , careful attention was paid to baseline fitting , and scans with extremely curved baselines were discarded . in addition , the reductions were performed independently by two of us to check the consistency of the results . lccccccc + galaxy name & @xmath54 & @xmath55 & @xmath56 & @xmath57 & log@xmath58 & @xmath59 & @xmath60 + + & & & & mpc & & & @xmath33 + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) + + + ugc 568 & 385 & 286 & 13579 & 181.0 & 10.25 & 0.30 & + ugc 1752 * & 398 & 364 & 18072 & 241.0 & 10.85 & 1.29 & 2.0 + pgc 135657 & 121 & 94 & 13204 & 176.0 & 10.20 & 0.47 & + ugc 2936 * & 488 & 461 & 3860 & 51.5 & 9.85 & 1.00 & 3.3 + ugc 3140 * & 143 & 118 & 4633 & 61.8 & 10.08 & 0.45 & 3.0 + ngc 2770 * & 343 & 319 & 1936 & 25.8 & 9.76 & 0.85 & 4.1 + pgc 135754 & 200 & 182 & 21588 & 287.8 & 10.06 & 0.27 & + f568 - 6 * & 394 & 374 & 13738 & 183.2 & 10.52 & 0.41 & 1.3 : + ugc 6614 * & 277 & 248 & 6245 & 83.3 & 10.42 & 1.44 & 6.0 + malin-1 * & 308 & 269 & 24679 & 329.0 & 10.66 & 0.17 & 2.5 + pgc 45080 & 296 & 267 & 12185 & 162.5 & 9.99 & 1.72 & + ugc 9024 * & 126 & 117 & 2296 & 30.6 & 9.40 & 0.92 & 2.7 + f530 - 1 & 467 & 460 & 14596 & 194.6 & 10.27 & 1.07 & + f533 - 3 * & 375 & 367 & 12941 & 172.5 & 10.24 & 0.75 & 2.5 : + ngc 7589 & 395 : & 345 : & 9056 & 120.7 & 10.01 & 0.55 & + pgc 71626 * & 115 & 83 & 10193 & 135.9 & 10.23 & 0.34 & 2.6 + + + radial velocities , @xmath38 , peak flux densities , integrated line fluxes , velocity widths at 50% and 20% of peak maximum ( @xmath37 and @xmath36 ) , and rms noise levels of our program spectra were measured using our own idl software . velocity widths were measured interactively , by moving the cursor outward from the profile center . radial velocities were defined to be the centroid of the two 20% peak maximum points on the profile and are quoted using the optical convention . to construct the global profiles for each of the mapped galaxies , we employed the procedure of matthews et al . a gaussian model with appropriate sidelobes for the beam was assumed ( see guibert 1973 ) . we treated the beam as infinite in the n - s direction , thus reducing the analysis to a one - dimensional problem . with our model beam , the model galaxy flux distributions were then iteratively integrated numerically until the best - fit model that reproduced the observed flux distribution in each of the telescope pointings was found . in all cases an asymmetric gaussian distribution ( i.e. , a lopsided gaussian with a different @xmath19 on the e and w sides , but uniform height ) was assumed for the distribution of the galaxy . because all of our sample galaxies were only coarsely resolved by the beam in the e - w direction , use of models for the distribution more complex than a gaussian ( e.g. , containing central depressions , etc . ) was not attempted ( see also fouqu 1984 ) . moreover , we found the simple gaussian models produced a good match to the data in all but two cases ( f568 - 6 & f533 - 3 ; see sect . [ disc ] ) . our reduced global spectra for all of our target galaxies are shown in fig . 1 . for the mapped galaxies , the spectra at each individual pointing are shown in fig . 2 . parameters for the final global spectra for all of our targets , including the mapped galaxies , are given in table 2 . the columns in table 2 are defined as follows : * ( 1 ) * galaxy name . * ( 2 ) * spectrum rms , in millijanskys . * ( 3 ) * peak flux density of the line profile , in millijanskys . * ( 4 ) & ( 5 ) * raw , measured full width at 20% and 50% of the maximum profile height , respectively , in . no correction has been applied to the raw linewidths for cosmological stretching , instrumental resolution , or for the errors arising from describing equal frequency - width channels by a constant velocity width across the entire bandwidth of the spectrum ( but see table 4 ) . the latter effect is inherent in the software , but is negligible @xmath611.5 ) compared with our measurement uncertainties . * ( 6 ) * heliocentric radial velocity , in , quoted using the optical convention , @xmath62 . * ( 7 ) * uncertainty in the heliocentric radial velocity , in computed following the prescription of fouqu et al . errors in the measured linewidths may be estimated as @xmath63 and @xmath64 ( fouqu et al . 1990 ) . * ( 8) * raw , integrated line flux , in . no corrections have been applied for beam attenuation . * ( 9 ) * uncertainty in the integrated line flux , in , computed following fouqu et al . ( 1990 ) . * ( 10 ) * signal - to - noise ratio of the detected line , defined as the ratio of the peak flux density to the spectrum rms . * ( 11 ) * comments . for more detailed comments on individual spectra , see section [ disc ] . table 3 summarizes the raw , integrated line profile fluxes ( @xmath39 ) and velocity centroids ( @xmath48 ) for each pointing in our mapping observations . in cases where no flux was detected at a particular pointing , an upper limit to the integrated flux was derived by multiplying the rms noise of the spectrum by the linewidth at 50% peak maximum from the previous pointing . in table 4 we tabulate several additional parameters for our target galaxies . columns in table 4 are as follows : * ( 1 ) * galaxy name * ( 2 ) & ( 3 ) * @xmath36 and @xmath37 values corrected for cosmological stretching and spectral resolution , using the relation @xmath65/(1+z)\ ] ] ( see haynes & giovanelli 1984 ) . here @xmath66 is the raw observed linewidth at @xmath67=20% or @xmath67=50% peak maximum , @xmath68 , and @xmath69 is given by @xmath70 where @xmath71=20 and @xmath71=50 for @xmath36 and @xmath37 , respectively , and @xmath72 is the velocity resolution of the measured spectrum ( see bottinelli et al . no corrections were applied for inclination angle of the source or for turbulent motions . * ( 4 ) * radial velocity , in , corrected to the local standard of rest , following the prescription of sandage & tammann ( 1981 ) : @xmath73 * ( 5 ) * galaxy distance in mpc , computed from @xmath74 . * ( 6 ) * logarithm of the mass in solar units , computed from the integrated line flux @xmath39 in column 8 of table 2 and using the relation @xmath75 . * ( 7 ) * ratio of the mass to the optical @xmath8-band luminosity @xmath76 , in solar units . @xmath76 was derived from the mean of the absolute @xmath8 magnitudes for each galaxy given in table 6 ( discussed below ) and assuming a solar absolute magnitude of @xmath775.48 . for ngc 7589 a @xmath8-band magnitude was taken from the ned database . * ( 8) * rough estimate of the diameter of the source in arcminutes . estimates were made only for mapped galaxies where flux was detected at 2 or more positions ( see table 3 ) . following fouqu ( 1984 ) , we define the diameter as the isophote enclosing half of the mass in a flat disk model , which for a gaussian surface density , is equal to the fwhm of the model ( fouqu 1984 ) . because of the elongation of the beam and the fact that our maps were obtained along an e - w axis , a correction to the raw diameter for the position angle and inclination of the source was also applied . hence , @xmath78 where : @xmath79 here , @xmath80 is the ratio of the major to the minor axis of the distribution and @xmath30 is the position angle of the major axis . we assume @xmath81=@xmath82 and adopt the photometric position angle for @xmath30 ( fouqu 1984 ) . typical errors for this method of estimating @xmath60 are @xmath83% ( see fouqu 1984 ) . lrrrrrl|lrrrrrl + ident . & @xmath38 & @xmath55 & @xmath54 & @xmath39 & tel . & ref . & ident . & @xmath38 & @xmath55 & @xmath54 & @xmath39 & tel . & ref . + & km / s & km / s & km / s & jy km / s & & & & km / s & km / s & km / s & jy km / s & & + + + ugc 1752 & 17836 & 384 & & 3.93 & a & gh89 & malin-1 & 24750 & & 341 & 3.5 & a & bo87 + & 17861 & 364 & 398 & 4.97 & n & new & & 24745 & 315 & 340 & 4.6 & b & ibpc + & & & & & & & & 24705 & 295 & 355 & 2.5 & a & ibpc + ugc 2936 & 3809 & & 490 & 8.36 & a & he83 & & 24755 & & 322 & 2.5 & v & p97 + & 3823 & & 490 & 8.4 & a & hg84 & & 24784 & 269 & 308 & 1.68 & n & new + & 3816 & & 513 & 11.7 & g & tc88 & & & & & & & + & 3817 & 493 & 499 & 9.0 & a & s95 & pgc 45080 & 12264 & & & & a & s95 + & 3811 & 484 & & 13.29 & a & h99 & & 12339 & 267 & 296 & 1.53 & n & new + & 3813 & & 500 & 13.6 & v & p99 & & & & & & & + & 3811 & 461 & 488 & 11.20 & n & new & ugc 9024 & 2285 & & & & & bo85b + & & & & & & & & 2326 & 114 & 128 & 8.7 & a & l87 + ugc 3140 & 4650 & & & & g & s78 & & 2323 & 117 & & 9.3 & a & h97 + & 4630 & 161 & 228 & 12.08 & n & bt82 & & 2323 & 115 & 131 & 8.4 & n & t98 + & 4622 & 125 & 148 & 12.30 & j & sd88 & & 2321 & 117 & 126 & 11.38 & n & new + & 4623 & 118 & 143 & 13.35 & n & new & & & & & & & + & & & & & & & f530 - 1 & 14339 & 488 & & 1.7 & a & s92 + ngc 2770 & 1953 & 330 & 347 & 33.26 & g & ft81 & & 14335 & 460 & 467 & 2.01 & n & new + & 1950 & 334 & 357 & 37.84 & j & sd88 & & & & & & & + & 1944 & 327 & & 24.10 & a & c93 & f533 - 3 & 12660 & 406 & 430 & 2.5 & a & gh89 + & 1948 & 340 & 373 & 33.88 & w & ra96 & & 12659 & 398 & & 2.2 & a & s92 + & & 324 & & 35.32 & a & h97 & & 12670 & 367 & 375 & 2.41 & n & new + & 1947 & 319 & 343 & 36.90 & n & new & & & & & & & + & & & & & & & ngc 7589 & 8938 & 356 & 363 & & a & s95 + f568 - 6 & 13827 & 386 & & 3.48 & a & s92 & & 8925 & & 402 & 2.7 & v & p97 + & 13835 & & 392 & 4.4 & v & p97 & & 8866 & 345 : & 395 : & 2.90 & n & new + & 13829 & 374 & 394 & 4.10 & n & new & & & & & & & + & & & & & & & pgc 71626 & 10022 & 91 & 118 & 2.6 & n & t98 + ugc 6614 & 6351 & 244 & 278 & 7.08 & a & bo85a & & 10016 & 83 & 115 & 3.86 & n & new + & 6358 & 236 & 286 & 9.56 & a & hr89 & & & & & & & + & 6353 & & 287 & 15.0 & v & p97 & & & & & & & + & 6351 & 248 & 277 & 15.79 & n & new & & & & & & & + + llllll + + + bo85a & bothun et al . ( 1985a ) & bo85b & bothun et al . ( 1985b ) & bo87 & bothun et al . ( 1987 ) + bt82 & bottinelli et al . ( 1982 ) & c93 & chengalur et al . ( 1993 ) & ft81 & fisher & tully ( 1981 ) + gh89 & giovanelli & haynes ( 1989 ) & hg84 & haynes & giovanelli ( 1984 ) & h97 & haynes et al . ( 1997 ) + h99 & haynes et al . ( 1999 ) & he83 & hewitt et al . ( 1983 ) & hu83 & huchra et al . ( 1983 ) + hr89 & huchtmeier & richter ( 1989 ) & ibpc & impey & bothun ( 1989 ) , see hr89 & l87 & lewis ( 1987 ) + new & this work & p97 & pickering et al . ( 1997 ) & p99 & pickering et al . ( 1999 ) + ra96 & rhee & van albada ( 1996 ) & s92 & schombert et al . ( 1992 ) & s78 & shostak ( 1978 ) + s95 & sprayberry et al . ( 1995b ) & sd88 & staveley - smith & davies ( 1988 ) & t98 & theureau et al . ( 1998 ) + tc88 & tifft & cocke ( 1988 ) & & + + + + a & arecibo 305 m & b & green bank 43 m & e & effelsberg 100 m + g & green bank 91 m & j & jodrell bank 64 m & n & 94 m equiv + v & vla & w & westerbork & & + + not surprisingly , all of the lsb giant galaxies observed in the present program are found to be very -rich systems , having values ranging from @xmath84 7.1@xmath85 ( table 4 ) . the bulk of the sample galaxies are not only rich in an absolute sense , but are also -rich both for their hubble types and for their optical luminosities six of the galaxies in the present lsb giant sample have @xmath861 ( in solar units ; see table 4 ) . more typical values of @xmath59 for high surface brightness spirals are : @xmath420.2 ( type sab - sb ) ; @xmath420.3 ( type sbc - sc ) ; @xmath420.4 ( type scd - sd ) ( roberts & haynes 1994 ) . as expected for massive , -rich disks , broad linewidths are also hallmarks of lsb giants . because a number of our program galaxies have rather low inclinations ( @xmath87 ) , it is difficult to estimate the true rotational velocities in some cases , but even many of the uncorrected linewidths tend to be quite broad ( e.g. , @xmath88398 for ugc 1752 , which is seen nearly face - on ) . only pgc 71626 seems to have a surprisingly narrow linewidth ( @xmath54=115 ) given its inclination ( @xmath89 ) and hubble type ( sbb or sbc ) . the majority of the lsb giants in our study exhibit relatively normal , classic , double - horned rotational profiles ( fig . 1 ) . however , we do see a few unusual cases as well ; ugc 568 shows a peculiar , strongly asymmetric profile ; pgc 45080 and ngc 7589 both have rather squarish global profiles with no clear rotation horns ; and finally , f530 - 1 has a double - horned profile that is strongly `` cleft '' in the middle . pgc 135754 may also have a similar cleft morphology , but our detection of this source is relatively weak . among the 11 relatively normal , double - horned profiles , 3 cases show modest asymmetries ( ugc 2936 , ngc 2770 , and f568 - 6 ) . thus we find no evidence that strong asymmetries are a general feature of lsb giants . we comment further on our individual spectra below . [ [ comparison - of - new - data - with - previous - measurements - and - comments - on - individual - objects ] ] comparison of new data with previous measurements and comments on individual objects ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ in order to compare the parameters derived in the present work with past measurements , we have compiled global parameters from the literature for the 13 of our sample galaxies which have previously been observed in the 21-cm line . table 5 summarizes previously measured recessional velocities , linewidths , and integrated line fluxes for each case , and indicates the telescope at which the data were obtained , as well as the reference for the quoted parameters . an examination of table 5 shows that for lsb giants observed by more than one group , significant discrepancies frequently exist between the measured parameters , particularly the global integrated line fluxes . we discuss each of the 16 targets observed in the present work in more detail below , including comments on these discrepancies . as part of our investigation of our lsb giant targets , we also searched the vicinity of each of the 16 objects ( i.e. , an area about 1.5 times the hpbw ) for nearby galaxies which could possibly give rise to confusion in our observed profiles . for this we employed the ned and leda databases , as well as optical images extracted from the digitized sky survey ( dss ) . results of these searches are also described below . * ugc 568 * : no data have been previously published for this object . our global spectrum appears quite peculiar , having a strong asymmetry , with a broad , single `` horn '' on the low velocity side of the profile . published optical redshifts show a discrepancy of over 400 ( @xmath90=13600 , melnick & quintana 1981 ; @xmath90=13198@xmath91275 , impey et al . our new velocity corresponds to the mean of the optical redshifts . several other galaxies were found within our search area , but they do not seem candidates for an confusion that could explain our peculiar global profile : cgcg 384 - 029 is a @xmath92=15.0 mag sbc spiral with an optical velocity 886 higher than the lsb giant ( dale et al . 1998 ; katgert et al . 1998 ) , while pgc 3280 is an s0 with a recessional velocity 1524 lower than the lsb ; mcgg + 00 - 03 - 027 is a 15.5 mag galaxy without known velocity , but it is classified as s0/a ( leda ) or sa ( ned ) , and is therefore expected to be gas - poor . * pgc 135657 * : no data have been previously published for this object . our detection has a recessional velocity 407 larger than the optical velocity of @xmath90=12701@xmath91275 reported for pgc 135657 by impey et al . ( 1996 ) . in the present survey we detect additional marginal features at the edges of our bandpass , near @xmath9512100 and @xmath9513300 . there are a number of small angular size galaxies in this field , but none appear as obvious candidates for interlopers . * ugc 2936 * : the arecibo spectrum of ugc 2936 published by sprayberry ( 1995b ) is highly asymmetric ; however only a modest asymmetry is seen in our new data and in the vla spectrum of pickering et al . our new integrated line flux of 11.23 agrees to within errors with that of tifft & cocke ( 1988 ) ( @xmath4211.7 ) . in comparison , the arecibo spectra by hewitt et al . ( 1983 ) , haynes & giovanelli ( 1984 ) , and sprayberry et al . ( 1995b ) all yield considerable lower integrated line fluxes ( @xmath428.4 - 9.0 ) , while our new integrated flux value is somewhat lower than the arecibo measurement of haynes et al . ( 1999 13.29 ) and the vla measurement of pickering et al . ( 1999 ; 13.6 ) . co(1 - 0 ) and co(2 - 1 ) line emission were detected from ugc 2936 at sest and iram by chini et al . ( 1996 ) , making this one of the rare galaxies classified as an lsb giant that has been detected in co ( see also knezek 1993,1998 ) . however , schmelz et al . ( 1986 ) failed to detect oh line absorption from the object . * ugc 3140 (= ngc 1642 ) * : this galaxy forms a wide pair ( @xmath96 projected separation ; @xmath97=332@xmath9185 ) with 15.0 mag sb spiral ugc 3141 . their e - w separation is @xmath98 and our spectrum does not appear to be confused . our new linewidth measurements are in good agreement with the jodrell bank measurements of staveley - smith & davies ( 1988 ) and measurements we performed from new hipass data ( see barnes et al . , in prep . for a description of hipass ) , but are lower than the values published by bottinelli et al . no published observations are available for ugc 3141 . * ngc 2770 * : short westerbork radio synthesis 21-cm line observations ( rhee & van albada 1996 ; broeils & rhee 1997 ) indicate that the distribution of ngc 2770 at a surface density level of 1 pc@xmath9 is about 1.7 times as large as the optical ( @xmath27 ) diameter and that its mean surface density averaged over the entire disk @xmath99@xmath100@xmath101 is 4.5 pc@xmath9 , which is higher than typical values for an sc . our new spectra show a slightly higher peak flux density on the low - velocity side of the global line profile , hinting at a possible weak asymmetry not seen in the westerbork observations . * pgc 135754 * : no data have been previously published for this object and it was only weakly detected in the present study . our new velocity is 434 higher than the optical value reported by impey et al . ( 1996 ; 21335@xmath91275 ) . * f568 - 6 (= malin-2 = pgc 86622 ) * : flux appears to have been missed in the arecibo observations of f568 - 6 by schombert et al . our new value of @xmath39=4.20 shows good agreement with the value derived from the vla observations of pickering et al . ( 1997 ; 4.4 ) . * ugc 6614 * : our new integrated flux of 15.94 is significantly larger than previous measurements from arecibo ( bothun et al . 1985a ; haynes & giovanelli 1989 ) . however , pickering et al . ( 1997 ) measured @xmath39=15.0 from vla data , consistent with our results . our diameter estimate for this object is very poorly constrained , as we find a gaussian model gives a poor fit to the data . this is reaffirmed by the vla data of pickering et al . ( 1997 ) , which show a signficant central depression in the distribution of this galaxy . * malin-1 (= pgc 42102 ) * : one published arecibo observation of malin-1 ( see table 5 ) gives an integrated line flux ( 2.5 ; bothun et al . 1987 ) comparable to the vla value derived by pickering et al . ( 1997 ) , while a re - reduction of the same arecibo data yielded @xmath39=3.5 jy ( impey & bothun 1989 ) ; finally impey & bothun ( 1989 ) measure @xmath39=4.6 from a green bank 43 m spectrum . our new data indicate that the arecibo and green bank fluxes previously reported by impey & bothun ( 1989 ) are likely to be significantly overestimated . in contrast our new integrated flux ( 1.80 ) is somewhat lower than the vla measurement of pickering et al . ( 1997 ) , although our values marginally agree to within errors , since at the redshifted frequency of malin-1 ( @xmath421312 mhz ) , our absolute calibration uncertainties are expected to be as high as @xmath5325% ( see section [ cal ] ) . in terms of spectral morphology , our new global spectrum shows excellent agreement with that derived by pickering et al . no co(1 - 0 ) line emission was detected in the central regions of malin-1 by radford ( 1992 ) , with an estimated upper limit of 4.7@xmath102 to the h@xmath103 mass , and braine et al . ( 2000 ) recently reported an upper limit of @xmath104 . * pgc 45080 * : our new integrated line flux is smaller than the values reported by sprayberry et al . ( 1995b ) , and our new spectrum lacks the distinct rotation horns seen in the sprayberry et al . given the small optical size of this object ( @xmath27=@xmath105 ) it seems unlikely significant flux extended outside our beam . there are also no obvious interlopers in this field that would have caused confusion in the sprayberry et al . spectrum , hence the discrepencies in the two spectra are probably due to the lower signal - to - noise of our present data . * ugc 9024 * : our new integrated flux value of 11.41 is higher than past values measured at arecibo or with single - pointing measurements ( bothun et al . 1985 ; lewis 1987 ; haynes et al . 1997 ; theureau et al . 1998 ) , suggesting this galaxy has extended emission . * f530 - 1 (= pgc 87136 ) * : we detect only a slightly higher integrated flux ( 2.07 ) than the arecibo value of schombert et al . ( 1992 ; 1.7 ) . our spectrum shows an unusual `` cleft '' morphology , although our signal - to - noise is only modest due to the broad linewidth of this source . * f533 - 3 (= pgc 68495 ) * : no flux is clearly detected at 2@xmath33e in our mapping observations of this galaxy ; this may be due to the broadness and relative weakness of the line profile and/or an distribution that is very lopsided relative to the optical center of the galaxy . as a result , our diameter estimate for f533 - 3 very uncertain . * ngc 7589 * : ngc 7589 was mapped in at the vla by pickering et al . ( 1997 ) . our new profile hints at a possible small amount of additional flux on the low - velocity side of the profile compared with the vla global spectrum ; this is probably caused by confusion with f893 - 29 , which was also detected in at the vla . f893 - 29 is a 17.1 mag sb ? spiral ( ned ) or dwarf ( pickering et al . 1997 ) , @xmath98 sw of the lsb giant , with a e - w separation of @xmath106 . f893 - 29 has a systemic velocity of 8768 ( pickering et al . 1997 ) , consistent with the velocity of the `` extra '' flux we observe associated with our global profile of ngc 7589 . our new integrated line flux ( 2.96 ) agrees well with that reported by pickering et al ( 1997 ; 2.7 ) , hence the flux contamination from this second source appears to be minimal . however our line profile widths appear to be overestimated by @xmath42100 due to this second source , hence corrections have been applied to the @xmath36 and @xmath37 measurements in tables 2 & 4 . lcccccl + ident . & @xmath90 & @xmath107 & @xmath108 & @xmath15 & @xmath109 & ref . + & km / s & mag & mag @xmath110 & kpc & mag & + + + ugc 568 & 13197 & 14.76 & 23.0 & 17.1 & @xmath11121.53 & s95 + & 13198 & 14.9 & 19.9 & & @xmath11121.42 & i96 + ugc 1752 & & 15.5@xmath50 & 22.7 & 20.5 & @xmath11121.38 & k93 + pgc 135657 & 12701 & 15.47 & 22.6 & 11.1 & @xmath11120.74 & s95 + & 12701 & 15.3 & 20.9 & & @xmath11120.92 & i96 + ugc 2936 & & 13.97 & 22.23 & 9.24 & @xmath11119.58 & s95 + & 4065 & 15.1 & 22.1 & & @xmath11118.42 & i96 + & 3822 & & 20.3@xmath112 & 8.4 & @xmath11121.1@xmath112 & p99 + ugc 3140 & & 13.34 & 20.9 & 3.9 & @xmath11120.61 & dj96 + & & 13.4@xmath50 & 22.8 & 9.8 & @xmath11120.55 & k93 + ngc 2770 & & 12.94 & 20.8 & 13.0 & @xmath11119.09 & k85 + pgc 135754 & 21335 & 16.40 & 22.5 & 11.2 & @xmath11120.99 & s95 + & 21335 & 16.1 & 21.0 & & @xmath11121.22 & i96 + f568 - 6 & & 14.57 & 23.2 & 21.0 & @xmath11121.79 & mb94 + & & & 22.1@xmath112 & 18.3 & @xmath11123.6@xmath112 & p97 + ugc 6614 & & & 22.9@xmath112 & 13.6 & @xmath11122.3@xmath112 & p97 + & & 14.43 & 24.3 & 15.8 & @xmath11120.29 & mb94 + & & 14.6@xmath50 & 22.3 & 9.1 & @xmath11120.04 & k93 + malin-1 & & & 26.0@xmath113 & 82.0 & @xmath11122.9@xmath113 & p97 + & & & 26.4 & 73.2 & & s93 + & & & 25.5@xmath113 & 73.2 & @xmath11123.12 & ib89 + pgc 45080 & & 17.48 & 22.8 & 6.45 & @xmath11118.65 & s95 + & 11674 & 17.0 & 22.4 & & @xmath11119.12 & i96 + ugc 9024 & & 14.90 & 24.1 & 4.3 & @xmath11117.58 & dj96 + & & 14.74 & 24.4 & 7.4 & @xmath11118.46 & mb94 + f530 - 1 & & 16.3@xmath50 & 21.8 & 12.2 & @xmath11120.11 & k93 + f533 - 3 & & 15.7@xmath50 & 23.5 & 26.5 & @xmath11120.44 & k93 + ngc 7589 & & & 23.3@xmath112 & 12.6 & @xmath11121.9@xmath112 & p97 + pgc 71626 & 9520 & 14.48 & 22.1 & 12.6 & @xmath11121.08 & s95 + & 9521 & 14.1 & 20.0 & & @xmath11121.42 & i96 + + + + + + + + + + + & + & + & + & + & + + * pgc 71626 * : pgc 71626 was detected at a recessional velocity consistent with that recently reported by theureau et al . ( 1998 ) , but somewhat offset from the published optical velocity of 9520 reported by sprayberry et al . ( 1995b ) and impey et al . our new integrated flux value of 3.94 is higher than the value of 2.6 reported by theureau et al . ( 1998 ) based on a single - pointing observation . an additional source was detected in our spectrum at @xmath114 , with @xmath115 and @xmath116 . this may be contamination from the s ? galaxy mcg -01 - 59 - 026 ( of unknown redshift ) , at a projected distance of @xmath117 from pgc 71626 , or due to an uncatalogued lsb dwarf visible on the dss . all of the lsb giant galaxies observed in the present survey have optical surface photometry available in the literature . to offer a more complete picture of the nature of the objects in our sample , we have compiled a summary of several optical and photometric parameters for each galaxy in table 6 . radial velocities quoted in table 6 are heliocentric values derived from optical measurements . @xmath107 values are the total @xmath8-band magnitudes derived from extrapolated exponential disk fits unless otherwise noted . @xmath108 is the extrapolated central disk surface brightness in the @xmath8-band ( unless otherwise specified ) and is _ uncorrected _ for inclination and internal extinction . @xmath15 is the exponential disk scale length in kpc . @xmath109 is the absolute magnitude ( @xmath8-band , unless otherwise noted ) derived from the total magnitude in column 3 , and assuming @xmath118 if no distance was given in the original reference ; no internal extinction corrections were applied . corrections to the values in table 6 for galactic extinction , @xmath119-corrections , and cosmological surface brightness dimming were generally applied by the original authors . the values given in table 6 show that there is a great deal of scatter in the photometric parameters reported for lsb giant galaxies . surface photometry of lsb galaxies is always challenging , but it is further complicated for lsb giants by the fact that many of these galaxies have significant bulges , leading to ambiguity in the bulge - disk decompositions and hence vastly different derived scale lengths and central surface brightness values from different workers . offsets between recessional velocities derived via optical and observations are also common ( cf . table 5 ) . we include table 6 partly as a caution to readers interested in understanding the global properties of giant lsb galaxies that there still exist large uncertainties in the measured parameters for such objects , and to highlight the need for additional surface photometry ( and measurements ) of such galaxies . such data will be necessary before we can distinguish whether all lsb giants such as the ones in the present study form a distinct class of object , as suggested by hoffman et al . ( 1992 ) , or whether some or most of these objects may form a natural and continuous extension of other low and moderate surface brightness disk galaxies . well - established photometric parameters will also be key to placing the lsb giant galaxies on the tully - fisher diagram . beijersbergen m. , de blok w.j.g . , van der hulst j.m . 1999 , a&a 351 , 903 bothun g.d . , aaronson m. , schommer r.a . , et al . 1985a , apjs 57 , 423 bothun g.d . , beers t.c . , mould j.r . 1985b , aj 90 , 2487 bothun g.d . , impey c.d . , malin d.f . , mould j.r . 1987 , aj 94 , 23 bothun g.d . , schombert j.m . , impey c.d . , schneider s.e . 1990 , apj 360 , 427 bottinelli l. , gouguenheim l. , fouqu p. , paturel g. 1990 , a&as 82 , 391 bottinelli l. , gouguenheim l. , paturel g. 1982 , a&a 113 , 61 braine j. , herpin f. , radford s.j.e . 2000 , a&a 358 494 broeils a.h . , rhee m .- h . 1997 , a&a 324 , 877 chengalur j.n . , salpeter e.e . , terzian y. 1993 , apj 419 , 30 chini r. , krgel e. , lemke r. 1996 , a&as 118 , 4 dale d.a . , giovanelli r. , haynes m.p . , et al 1998 , aj 115 , 418 de blok w.j.g . , mcgaugh s.s . , van der hulst j.m . 1996 , mnras 283 , 18 de jong r.s . 1996 , a&as 118 , 557 dreyer j.l.e . 1953 , a new general catalogue of nebulae and clusters of stars , london royal astronomy society fisher j.r . , tully r.b . 1981 , apjs 47 , 139 fouqu p. 1984 , a&as 55 , 55 fouqu p. , bottinelli l. , durand n. , gouenheim l. , paturel g. 1990 , a&as 86 , 473 gallagher j.s . , bushouse h. 1983 , aj 88 , 55 giovanelli r. , haynes m.p . 1989 , aj 97 , 633 guibert j. 1973 , a&as 12 , 263 haynes m.p . , giovanelli r. 1984 , aj 89 , 758 haynes m.p . , giovanelli r. , herter t. , et al . 1997 , aj 113 , 1197 haynes m.p . , giovanelli r. , chamaraux p. , et al . 1999 , aj 117 , 2039 hewitt j.n . , haynes m.p . , giovanelli r. 1983 , aj 88 , 272 hoffman y. , silk j. , wyse r.f.g . 1992 , apj 388 , l13 huchra j. , davis m. , tonry j. , latham d. 1983 , apjs 52 , 89 huchtmeier w.k . , richter o .- g . 1989 , a general catalogue of observations of galaxies , springer - verlag impey c. , bothun , g. 1989 , apj 341 , 89 impey c.d . , sprayberry d. , irwin j.a . , bothun g.d . 1996 , apjs 105 , 209 katgert p. , mazure a. , den hartog r. , et al . 1998 , a&as 129 , 399 kent s.m . 1985 , apjs 59 , 115 knezek p.m. 1993 , ph.d . thesis , university of massachusetts knezek p.m. 1998 , iau colloquium 171 , the low surface brightness universe , asp vol . 170 , 191 kormendy j. 1989 , apj 342 , l63 lauberts a. , valentijn e.a . 1989 , the surface photometry catalogue of the eso - uppsala galaxies , european southern observatory lewis b.m . 1987 , apjs 63 , 515 matthews l.d . , gallagher j.s . 1997 , aj 114 , 1899 matthews l.d . , van driel w. 2000 , a&as 143 , 421 matthews l. d. , van driel w. , gallagher j.s . 1998 , aj 116 , 1169 mcgaugh s.s . , bothun , g.d . 1994 aj 107 , 530 melnick j. , quintana h. 1981 , aj 86,1567 nilson p. 1973 , uppsala general catalog of galaxies , uppsala astron . band 6 oneil k. , bothun g.d . , cornell m. 1997 , aj 113 , 1212 pickering t.e . , impey c.d . , van gorkom j.h . , bothun g.d . 1997 , aj 114 , 1858 pickering t.e . , van gorkom j.h . , impey c.d . , quillen a.c . 1999 , aj 118 , 765 radford s.j.e . 1992 , a&a 262 , 13 rhee m .- h . , van albada t.s . 1996 , a&as 115 , 407 roberts m.s . , haynes m.p . 1994 , araa 32 , 115 sandage a. , tammann g.a . 1981 , a revised shapeley - ames catalog of bright galaxies , carnegie institution sanders d.b . , solomon p.m. , scoville n.z . 1984 , apj 276 , 182 schmelz j.t . , baan w.a . , haschick a.d . , eder j. 1986 , aj 92 , 1291 schombert j. 1998 , aj 116 , 1650 schombert j. , bothun g.d . , schneider s.e . , mcgaugh s.s . 1992 , aj 103 , 1107 shostak g.s . 1978 , a&a 68 , 321 sprayberry d. , bernstein g.m . , impey c.d . , bothun g.d . 1995a , apj 438 , 72 sprayberry d. , impey c.d . , bothun g.d . , irwin m.j . 1995b , aj 109 , 558 sprayberry d. , impey c.d . , irwin m.j . , et al . 1993 , apj 417 , 114 staveley - smith l. , davies r.d . 1988 , mnras 231 , 833 thuan t.x . , lipovetsky v.a . , martin j .- , pustilnik s.a . 2000 , a&as , 139 , 1 theureau g. , bottinelli l. , coudreau - durand n. , et al . 1998 , a&as 130 , 333 tifft w.g . , cocke w.j . 1988 , apjs 67 , 1 van der hulst j.m . , skillman e.d . , smith t.r . , 1993 , aj 106 , 548 verheijen m.a.w . 1997 , ph.d . thesis , university of groningen walsh w. , staveley - smith l. , oosterloo , t. 1997 , aj 113 , 1591 zwaan m.a . , van der hulst j.m . , de blok w.j.g . , mcgaugh s.s . 1995 , mnras 273 , l35

in order to study partial restoration of the chiral symmetry at finite density , we investigate the density corrections of the chiral condensate up to the next - to - leading order of density expansion using the chiral ward identity and an in - medium chiral perturbation theory . in our study , we assume that all the in - vacuum quantities for the pion , the nucleon and the interaction are determined in vacuum and focus on density expansion of the in - medium physical quantities . we perform diagrammatic analysis of the correlation functions which provide the in - medium chiral condensate . this density expansion scheme shows that the medium effect to the chiral condensate beyond the linear density comes from density corrections to the sigma term as a result of the interactions between pion and nucleon in nuclear matter . we also discuss that higher density contributions beyond order of can not be fixed only by the in - vacuum dynamics and we encounter divergence in the calculation of the order corrections of the chiral condensate . to remove the divergence , we need nn two - body contact interaction , which can be fixed in vacuum .

1308.2660

dynamical breaking of chiral symmetry ( db@xmath2s ) is one of the important phenomena of qcd for low - energy hadron spectrum and dynamics of light hadrons . the light pseudoscalar mesons , @xmath3 , @xmath4 and @xmath5 , are identified as the nambu - goldstone bosons of db@xmath2s , and the quark mass generation is also explained by db@xmath2s . the quark condensate @xmath6 is one of the order parameters of db@xmath2s and its magnitude characterizes the qcd vacuum . since db@xmath2s is a phase transition phenomenon , such dynamically broken symmetry is expected to be restored in extreme environment , such as high temperature and/or high baryonic density . it is very significant to confirm phenomenologically that db@xmath2s really takes place in the qcd vacuum . one of the proofs of db@xmath2s is to make sure of partial restoration of chiral symmetry in nuclear matter . the partial restoration of chiral symmetry is an incomplete restoration of chiral symmetry with sufficient reduction of the magnitude of the quark condensate . recent observations of pionic atom spectra , especially precise measurement of the isotope dependence of deeply bound pionic atoms @xcite , and low energy pion - nucleus scattering @xcite have found out that the @xmath7 parameter appearing in the pion optical potential is substantially enhanced in nuclei . with this fact and theoretical examination @xcite , it turns out that the magnitude of the quark condensate does decrease about 30% at the saturation density . the reduction of the quark condensate in nuclear medium also leads to various phenomena , for instance , attractive enhancement of scalar - isoscalar @xmath8 correlation in nuclei @xcite and the suppression of the spectrum difference between the chiral partners , such as @xmath9-@xmath10 @xcite and @xmath11-@xmath12 @xcite . the experimental observations of these phenomena can be further confirmation of partial restoration of chiral symmetry in nuclear medium . for instance , one could observe the reduction of the @xmath11-@xmath12 mass difference from the formation spectrum of the @xmath5 mesonic nuclei @xcite . the mass difference between the @xmath5 and @xmath13 mesons is also responsible for the quark condensate through the @xmath14 anomaly effect @xcite . these phenomena are caused by substantial quark dynamics , but because we have quark - hadron duality in the description of hadron dynamics , these phenomena should be also described in terms of hadron dynamics , such as nuclear many - body theories . this means that if one could describe the suppression of the spectrum difference of the chiral partners in a nuclear many - body theory , this does _ not _ rule out partial restoration of chiral symmetry . to a greater extent , once one could understand hadronic phenomena in terms of quark - gluon dynamics , one would have more substantial and deeper insight of hadron dynamics , which will bring us its more systematic understanding in terms of qcd . theoretically the reduction of the quark condensate in the nuclear medium is naturally expected according to the model independent low density relation @xcite , in which the ratio of the in - medium and in - vacuum quark condensates is given by the @xmath0 sigma term together with the in - vacuum pion mass and pion decay constant . this relation is derived under the linear density approximation . the sign of the sigma term determines the fate of the in - medium quark condensate . since the sigma term extracted from @xmath0 scattering data has a positive sign , the quark condensate should be reduced , at least , in low density limit . however , one does not know up to which density one can apply the linear density approximation . for further detailed understanding , one needs calculation beyond the linear density based on effective theories . because the quark condensate is not a direct observable in experiments , one needs theoretical examination to conclude partial restoration of chiral symmetry phenomenologically . in ref . @xcite , an exact sum rule which relates the quark condensate and hadronic observables has been derived by using the chiral ward indentity . in the linear density approximation , the in - medium quark condensate can be written in terms of the in - medium temporal pion decay constant and the pion wavefunction renormalization constant . the importance of the wavefunction renormalization in in - medium chiral effective theories has been also discussed in refs . @xcite . with this relation , the reduction of the quark condensate has been phenomenologically confirmed by using the in - medium pion decay constant extracted from pion - nucleus dynamics @xcite and the wavefunction renormalization constant extracted from pion - nucleon scattering @xcite . since this proof of the partial restoration of chiral symmetry in the nuclear medium is based on the linear density approximation of in - medium quantities , precise determination of the density dependence of the quark condensate both in theory and experiment @xcite is strongly desired . the in - medium quark condensate has been discussed in various approaches . the in - medium correction of the condensate is given by the pion - nucleon sigma term model - independently at the first order in nucleon density @xcite and higher orders are evaluated with mean field calculations using the nambu - jona - lasinio and gell - mann - levy models@xcite and the relativistic brueckner approach@xcite . in ref . @xcite , the in - medium condensate has been obtained beyond the linear density based on the hellman - feynman theorem , in which they have calculated the energy density in nuclear matter based on a chiral effective theory and taken its derivative with respect to the quark mass to obtain the quark condensate . the pion self - energies in asymmetric nuclear matter was calculated based on chiral perturbation theory in ref . reference @xcite developed systematic framework of chiral perturbation theory in nuclear matter and calculated the equation of state of isospin - symmetric nuclear matter . in ref . @xcite , a novel formulation of chiral perturbation theory in a nuclear background was proposed and the self - energy and the nuclear optical potential of the charged pion were calculated . in ref . @xcite hadronic quantities , such as pion optical potential , have been calculated beyond the linear density . the goal of this paper is to examine higher density correction beyond linear density and to show a systematic way to calculate the in - medium quantities based on chiral effective theory . for this purpose , we use the formulation proposed in ref . @xcite and developed in ref . @xcite . in this formalism , one calculates matrix elements in the free fermi nuclear matter , which are defined by the path integral under the action of the system . all the interaction between nucleons in matter and pions are assumed to be described in the interaction lagrangian.in this formulation , one can make a double expansion in terms of fermi see insertion and chiral order counting . thus , the expansion scheme is clear . in this paper we calculate the ward identity , which connect the quark condensate and hadronic quantities , based on this in - medium chiral perturbation formulation . this paper is organized as follows . we explain the chiral ward identity which relates the chiral condensate with the correlation function of the chiral currents in section 2 . in section 3 , we introduce the in - medium chiral perturbation theory and the in - medium chiral counting scheme . in section 4 , we show the result of the calculation of the in - medium chiral condensate @xmath15 , and finally we devote sec.5 to the conclusion of the present paper . in order to calculate the density dependence of the in - medium quark condensate , we take the chiral ward identity approach proposed by ref . @xcite . in this approach , we consider the correlation function of the axial vector current @xmath16 and the pseudoscalar density @xmath17 : @xmath18 where @xmath19 is the nuclear matter ground state normalized as @xmath20 and is characterized by the proton and neutron densities , @xmath21 and @xmath22 . the axial vector current @xmath16 is associated with the su(2 ) chiral transformation whose generators are given by @xmath23 . the pseudoscalar density @xmath17 is defined in terms of the quark field by @xmath24 with the pauli matrix @xmath25 for the isospin space and transforms under the su(2 ) chiral transformation as @xmath26 = -i \delta^{ab } \bar{q}{q } ( x ) $ ] . using the operator identity @xmath27 = \delta(x_{0 } ) [ a_{0}^{a}(x),p^{b}(0 ) ] + t[\partial^{\mu } a_{\mu}^{a}(x ) p^{b}(0)]$ ] and performing the integral in the soft limit @xmath28 , we obtain the in - medium quark condensate as @xmath29 where we have written the expectation value @xmath30 as @xmath31 for operator @xmath32 , and @xmath33 and @xmath34 are defined as @xmath35 here we have used the pcac relation @xmath36 with the quark mass @xmath37 . equation implies that the in - medium quark condensate is written in terms of the green functions in the soft limit . we can evaluate the in - medium chiral condensate by calculating this correlation functions @xmath38 and @xmath39 in the soft limit @xmath40 . up to the next - to - leading order , we will confirm that @xmath41 vanishes off the chiral limit in the soft limit when there are no massless pionic modes which coupled to the axial current @xmath16 . in the chiral limit , the quark condensate can be calculated by @xmath42 as discussed in ref . we note that both methods are equivalent when one calculate @xmath43 in the chiral limit . chiral effective theories are powerful theoretical tools to describe hadron dynamics based on chiral symmetry and its spontaneous breaking@xcite . in this work , we use the in - medium extension of the chiral perturbation theory developed by refs . @xcite . in this method , first one defines the generating functional of the correlation functions by taking non - interacting fermi gas of nucleons as the asymptotic state and assumes all the interaction between nucleons and other internal fields are described by the chiral effective lagrangian . the in - medium correlation functions are calculated by taking functional derivatives of the generating functional . let us consider the non - interacting nucleon system at asymptotic times @xmath44 , @xmath45 and @xmath46 as usual scattering theory . here we assume the unpolarized nuclear matter for simplicity . the in- and out - states are described in terms of the nucleon creation operators @xmath47 with the nucleon momentum @xmath48 as @xmath49 where the nucleon fermi gas states are occupied up to the fermi momentum @xmath50 . the proton and neutron densities are given by the fermi momenta @xmath51 as @xmath52 for @xmath53 . the generating functional is given by @xmath54 = e^{i w[j , \eta , \eta^\dag ] } = \langle \omega_{\rm out } | \omega_{\rm in } \rangle_{j , \eta , \eta^\dag } \label{generating}\ ] ] under the presence of the external fields @xmath55 , @xmath5 and @xmath13 . here , @xmath56 represent the scalar , pseudo - scalar , vector and axial - vector sources respectively and @xmath57 are the nucleon external sources . we also define the generating functional for the connected green functions @xmath58 $ ] in eq.([generating ] ) . the path integral is to be performed for the fields in the lagrangian , such as the chiral field @xmath59 and the nucleon field @xmath11 : @xmath60 & = & \int du dn dn^\dagger \langle \omega_{\rm out}| n ( + \infty ) \rangle \nonumber \\ & & \hspace{-2em } \times e^{i \int dx ( { \cal l}_{\pi}+ { \cal l}_{\pi n } + \eta^\dag n + n^\dag \eta ) } \langle n(-\infty)| \omega_{\rm in } \rangle , \end{aligned}\ ] ] where @xmath61 is the pion chiral lagrangian and we take the @xmath3-@xmath11 chiral lagrangian with the nucleon bilinear interaction @xmath62 given by the @xmath0 chiral lagrangian @xmath63 . the operator @xmath62 is written by the pion fields and its derivatives together with the external fields and is subject to chiral order counting , and @xmath64 is the nucleon field with @xmath65 for proton and neutron . in appendix a , the detailed expression of @xmath62 is summarized . the parameters of the lagrangian are to be fixed in vacuum . the integral in terms of the nucleon field can be done easily by using the gauss integral formula , if the lagrangian has the bilinear form for the nucleon interaction . as shown in ref . @xcite , the generating functional is characterized by double expansion of fermi sea insertions and chiral orders . the fermi sea insertion is seen as @xmath66 & = & \int du \exp \big{\ { } i\int dx \big{[}\mathcal{l}_{\pi \pi } - \int \frac{d \bf{p}}{(2\pi)^3 2e(p ) } { \rm f.t.\ , } { \rm tr } \big { ( } i\gamma(x , y ) ( \slash{p } + m_n ) n(p ) \big { ) } \nonumber \\ & & - \frac{i}{2 } \int \frac{d \bf{p}}{(2\pi)^3 2e(p ) } \frac{d \bf{q}}{(2\pi)^3 2e(q ) } { \rm f.t.\ , } { \rm tr } \big { ( } i\gamma(x , x ' ) ( \slash{q } + m_n ) n(q ) i\gamma(y',y ) ( \slash{p } + m_n ) n(p ) \big { ) } + \cdots \big { ] } \big{\ } } \label{eq : zexp}\end{aligned}\ ] ] where f.t . denotes fourier transformation of the spacial variables except @xmath67 , @xmath68 is the relativistic nucleon energy @xmath69 and the nonlocal vertex @xmath70 is defined by @xmath71^{-1}$ ] , which is given only by the in - vacuum interactions @xmath62 and the free nucleon propagator @xmath72 . the matrix @xmath73 in the isodoublet space is defined to restrict the momentum integral up to the fermi momentum as @xmath74 fig . [ z ] shows the diagrammatic structure of the fermi sea insertion of the generating functional . in the figure , the thick line represents nucleon propagation in the fermi sea . the chiral expansion is given by the expansion of the nonlocal vacuum vertices @xmath75 together with the chiral expansion for the bilinear local vertex @xmath62 . using the generating functional , we can define the in - medium pion lagrangian @xmath76 as @xmath77= \exp [ { i \int d^{4}x \tilde{\mathcal l}_{\pi \pi}}]$ ] . is the nonlocal vertex given by the in - vacuum @xmath0 interaction.,width=325 ] the connected @xmath78-point green functions can be evaluated by taking functional derivatives of @xmath79 $ ] defined in eq.([generating ] ) with respect to the external sources @xmath80 : @xmath81 , \label{green}\ ] ] where @xmath82 is the corresponding current operator to the external source @xmath83 . the current operator @xmath84 can be represented in terms of the corresponding quark current , such as the pseudo - scalar current @xmath85 and the axial - vector current @xmath86 . the in - vacuum chiral perturbation theory has the chiral expansion scheme in which the pion energy momentum and the small quark mass are counted as small quantities . in addition to these quantities , in the in - medium chiral perturbation theory the fermi momentum is also regarded as a small quantity , since the fermi momentum at the normal nuclear density @xmath87 mev is as small as @xmath88 . according to ref . @xcite , chiral order @xmath89 for a specific diagram is given by the following : @xmath90 where @xmath91 is the number of pion loops , @xmath92 is the number of the pion propagators , @xmath93 is the chiral dimension coming from the pion chiral lagrangian @xmath94 , @xmath95 is the chiral dimension of the nonlocal in - medium vertex with @xmath78 fermi sea insertions and @xmath96 is the chiral dimension of the @xmath97 vertex . this counting rule is called standard case in ref . @xcite in which the nucleon propagator is counted as @xmath98 . we note that @xmath89 is larger than 4 , so that leading order contribution to the in - medium chiral condensate appears from @xmath99 . in this formalism , we can calculate any processes in which pions interact with fermi gas using this method . if one follows strictly the chiral expansion scheme , one has to expand also in - vacuum terms and renormalize them order by order . this kind of the expansion is useful in theoretical consideration , while it is not convenient in practical use because expanded quantities are not direct observables . here we consider that all the in - vacuum quantities for the @xmath0 dynamics are already fixed by experiments . this implies that the renormalization procedure for in - vacuum values are already done and we do not have to evaluate in - vacuum loop diagrams . this is also consistent with having taken @xmath100 to obtain eq . . the expansion scheme of the generating functional in terms of the fermi sea insertion given in eq . is equivalent to the conventional nuclear many body calculation using the pauli - blocked nucleon propagator in the fermi gas @xmath101 to see the equivalence , we examine a one - nucleon loop diagram with two interaction operator @xmath62 ( see fig . [ example ] ) . in the conventional approach , this diagram can be calculated using the pauli - blocked nucleon propagator @xmath102 by @xmath103}.\ ] ] using eq . , this can be written as @xmath104 } + i \int \frac{d^3 { \bf p}}{(2\pi)^3 2e(p ) } \frac{d^4 q}{(2\pi)^4 } { \rm tr } \big { [ } ( -ia ) id_{0}^{-1}(q ) ( -ia ) i(\slash{p } + m_{n } ) n(p ) \big { ] } \nonumber \\ & & + \frac{i}{2 } \int \frac{d^3 { \bf p}}{(2\pi)^3 2e(p ) } \frac{d^3 { \bf q}}{(2\pi)^3 2 e(q ) } { \rm tr } \big { [ } ( -ia ) i(\slash{q } + m_{n } ) n(q ) ( -ia ) i(\slash{p } + m_{n } ) n(p ) \big { ] } \label{eq:2point}\end{aligned}\ ] ] here we have integrated out in terms of @xmath105 for @xmath106 : @xmath107 given in eq . .,width=325 ] the first term of eq . is the nucleon loop in vacuum and should be renormalized into the in - vacuum lagrangian . the second term can be obtained from the one fermi sea insertion as appearing in the second term of the argument of exp in eq . after replacing the non - local vertex @xmath97 to @xmath108 which is the second term of the chiral expansion of @xmath97 in eq . . the third term can be obtained in the two fermi sea insertion by replacing @xmath97 to @xmath62 which is the first term of the chiral expansion . in the same way , one can show that eq . contains all the terms of one nucleon loop diagram with three interaction operators given by the conventional approach as @xmath109}\end{aligned}\ ] ] with the correct factor except the free nucleon loop . therefore calculation with the in - medium nucleon propagor @xmath102 is equivalent to use the expansion scheme of the generating functional given in eq . . in order to evaluate the in - medium condensate @xmath110 with eq . ( [ ward ] ) , we calculate the current - current correlation functions in the soft limit , @xmath111 and @xmath112 defined in eqs . and , by using the in - medium chiral perturbation theory . from eq . ( [ green ] ) , @xmath112 is expressed by the generating functional @xmath113 $ ] . \ \ \ \ \ \end{aligned}\ ] ] here , @xmath115 are the pseudo - scalar density and the corresponding external field . similarly , @xmath111 is expressed in terms of the generating functional @xmath113 $ ] . in sect . [ dfunction ] , we list up the feynman graphs for the calculation of @xmath116 based on the density order counting and evaluate it up to the nlo corrections . we also list up the feynman diagram for @xmath117 evaluate it in sect . [ cancellation ] . we will find that @xmath117 vanishes off the chiral limit within the nlo corrections by taking the soft limit . in sect . [ densitydependence ] , we show the density dependence of the chiral condensate within the nlo corrections . in sect . [ contactnn ] , we discuss higher order corrections beyond nlo . we will find that some diagrams are divergent and show the necessity of the @xmath118 contact terms to renormalize the higher order corrections . in the following , we consider the symmetric nuclear matter for simplicity . correlation function in vacuum . the wavy line denotes the pseudo - scalar density and the filled circle represents the tree vertex for the pions . the dashed line stands for the pion propagation . the diagram in ( a ) is for the tree level and the diagrams shown in ( b ) are examples of the radiative correction.,width=325 ] we calculate @xmath112 in the soft limit with the finite quark mass . in the following @xmath119 denotes the in - medium expectation value @xmath120 . first of all , let us evaluate @xmath112 in the vacuum using the in - vacuum chiral lagrangian @xmath121 . we draw the feynman diagrams for the @xmath122 correlation function in fig . [ p2 ] based on the chiral counting scheme . in the diagrams the wavy line denotes the pseudo - scalar density and the filled circle represents the tree vertex for the pions . feynman graph ( a ) shows the leading order graph counted as @xmath123 in the chiral counting , while the diagrams ( b ) are the radiative corrections in the higher order of the chiral expansion . calculating @xmath122 in vacuum and taking the soft limit @xmath40 in the momentum space , we obtain @xmath34 and the in - vacuum chiral condensate through eq . ( [ ward ] ) : @xmath124 where @xmath125 is the pion decay constant in the chiral limit , @xmath126 is one of the low energy constants in the chiral lagrangian @xmath127 given in appendix [ app ] , @xmath128 is the free pion propagator and @xmath129 is the pion mass . the first term is the leading order contribution of the in - vacuum chiral condensate and the dots mean higher order loop corrections , such as those given as the diagrams in fig . [ p2 ] ( b ) . using the relation between the pion and quark masses in the chiral perturbation theory , @xmath130 , we obtain the in - vacuum chiral condensate @xmath131 : @xmath132 in this work we presume that the in - vacuum quantities appearing in the lagrangian are already fixed by the experimental data . ) and the thick solid line denotes the fermi sea insertion which is the second term in eq . ( [ gprop ] ) for the pauli blocking effect . the filled circle represents the vertex in the leading order , while the unfilled circle is the vertex coming from the in - vacuum subleading @xmath0 chiral lagrangian @xmath133 . ( a ) the feynman diagrams for @xmath99 . ( b ) the feynman diagrams for the linear density order @xmath134 . ( c ) the feynman diagrams for the next to leading order in density @xmath135 . three diagrams at the bottom give the double scattering term.,width=325 ] second , we consider the density corrections . according to the chiral counting in eq . ( [ counting ] ) , the leading order contributions start from @xmath136 . in fig . [ density ] ( a ) , we show all the @xmath136 diagrams . in these graphs , the thin solid line is the free nucleon propagator shown as the first term of eq . ( [ gprop ] ) and the thick solid line is fermi sea insertion shown as the second term in eq . ( [ gprop ] ) for accounting the pauli blocking effect . the unfilled circle is the verticies coming from the in - vacuum subleading @xmath0 chiral lagrangian @xmath133 . it is notable that all the @xmath137 contributions to the correlation function @xmath138 vanishes in the soft limit . to see it , we evaluate the central diagram in fig . [ density ] ( a ) , as an example , with the external momentum @xmath139 using the in - medium propagator ( [ gprop ] ) and the vertices given in appendix [ app ] : @xmath140 } \\ % & = & - \lim_{q \to 0 } g_a^2 b_0 ^ 2 \big { ( } id_\pi ( q ) \big{)}^2 \int \frac{d^4 p}{(2 \pi ) ^4 } { \rm tr } \big { [ } \slash{q } ( \slash{p } + \slash{q } - m_n ) % \\ & & \times % \slash{q } ( \slash{p } + m_n ) \tau^a \tau^b \big { ] } ig_m ( p+q ) ig_m ( p ) \\ & = & 0 .\end{aligned}\ ] ] having taking the soft limit in the final equality , we find that the contribution of this diagram vanishes since the term in the trace becomes zero in the limit , while the pion propagator @xmath141 is finite thanks to the nonzero denominator . since the other diagrams in fig . [ density ] ( a ) have the same structure , we find that all the @xmath136 terms vanish in the soft limit and do not contribute to the in - medium chiral condensate . the reason that these terms vanish is that the pion is not a zero - mode off the chiral limit and the interaction between pion and nucleon is @xmath142-wave in the leading order . we will see that @xmath117 also have the same momentum dependence as the leading term of @xmath34 up to the nlo corrections in sec . [ cancellation ] . thus , @xmath117 does not contribute to the in - medium chiral condensate . in fig . [ density ] ( b ) , we show all the diagrams of the leading order ( lo ) contribution in the density expansion , in which there are three diagrams . we write the lo contribution for the in - medium chiral condensate as @xmath143 and evaluate the diagrams in fig . [ density ] ( b ) by the expansion of @xmath144 at the final state . in the following we first calculate the left diagram in fig . [ density ] ( b ) in the soft limit of the external momentum @xmath139 : @xmath145 } \\ & = & -\frac{16 i c_1 b_0 ^ 2}{m_\pi^2 } \delta^{ab } \left(\sigma_{p}^{1}(0 ) + \sigma_{n}^{1}(0 ) \right ) \label{eq1 } \\ & = & -\frac{16 i c_1 b_0 ^ 2 } { m_\pi^2 } \delta^{ab } \rho \left ( 1- \frac{3 k_f^2}{10 m_n^2 } \right ) .\label{eq2 } \end{aligned}\ ] ] here we have used the result of the tadpole nucleon loop @xmath146 given in eq . , in which we have expanded the result in terms of @xmath147 and taken the first two terms , and assumed the symmetric nuclear matter by taking @xmath148 . next we calculate the central and right graphs in fig . [ density ] ( b ) . these contributions denote @xmath149 and @xmath150 , respectively : @xmath151 } \nonumber \\ & = & -\frac{32 i b_0 ^ 3 c_1 m_q}{m_\pi^4 } \delta^{ab } \int \frac{d^4 p}{(2 \pi ) ^4 } { \rm tr } \big { [ } id^{-1}_m ( q ) \big { ] } \\ & = & \frac{16 i b_0 ^ 2 c_1 } { m_\pi^2 } \delta^{ab } \rho \left ( 1- \frac{3 k_f^2}{10 m_n^2 } \right ) .\end{aligned}\ ] ] here we have used again the calculation of the tadpole nucleon loop and the @xmath152 expansion has been made up to @xmath153 . we have also used the relation @xmath154 . it is important to notice that the contributions coming from the @xmath155 and @xmath156 low energy constants do vanish in the soft limit . @xmath150 can be calculated in the same way as @xmath157 @xmath158 the leading contribution in the density expansion @xmath159 is given by the sum of @xmath160 . we obtain the linear density contribution of the in - medium condensate together with the fermi motion correction up to @xmath153 : @xmath161 with this result we obtain the in - medium condensate in the normalization of the in - vacuum condensate as @xmath162 here @xmath163 is one of the low energy constants ( lecs ) in @xmath133 and can be determined by the @xmath0 sigma term @xmath164 . as we mentioned before , we presume that the quantum corrections in vacuum is already accounted into the low energy constants and the renormalization procedure is supposed to be done . thus , the loop contributions are renormalized into the low energy constants in the lagrangian which we consider and we do not have to calculate the in - vacuum loop contributions in the in - medium calculation . for instance , we show two diagrams which have ultraviolet divergences and are to be counted as higher chiral orders while with the same density counting in fig . [ p5correction ] . the left diagram is the loop correction for the nucleon mass and the right accounts the vertex correction . both loop corrections are calculated by the in - vacuum quantities . therefore , we suppose that these loop corrections should be accounted in the nucleon mass and the vertex , respectively , and we use the observed values for these quantities . if one would follow the chiral counting scheme strictly , one would expand the physical quantities in terms of the chiral order and discard higher order corrections than the order which one considers . here we do not take the strict rule for the chiral counting . we exploit the observed value , in which all order of the corrections should be included . in this way , we fix the @xmath165 parameter by the observed @xmath0 sigma term as @xmath166 admitting that the loop corrections for the @xmath0 sigma term are taken into account into the physical value , and we do not calculate further the in - vacuum loop contribution , most of which are divergent . therefore , the leading order contribution of the in - medium chiral condensate in the density expansion is given by the in - vacuum physical quantities as @xmath167 the second term is a @xmath168 correction to the linear density and it is counted as nnlo @xmath169 . to complete the @xmath170 contributions , we have to calculate further higher orders in the density expansion , as we shall see later . here it is important to emphasize again that we take the observed values . the values which we use in this work are the sigma term @xmath171 @xcite , the pion decay constant @xmath172 , the pion mass @xmath173 and the nucleon mass @xmath174 . the result coincides the well - known linear density approximation result @xcite . now we evaluate the nlo density corrections @xmath175 . the relevant diagrams for the next leading order are shown in fig . [ density ] ( c ) . these diagrams contain two loops coming from the nucleon in the fermi sea and a free pion . the diagrams in which one of the nucleon propagators is the free propagator is already accounted as higher chiral order terms into the renormalized @xmath0 vertex , because the diagram contains a loop written by only the free propagators , which is divergent and should be renormalized into the in - vacuum vertex . here we use such a parameterization of the chiral field @xmath59 in terms of the pion field that the naive perturbative expansion can be done . the details are written in appendix [ para ] . the leftmost diagram in third row in fig . [ density ] is calculated as @xmath176 } \\ & = & \delta^{ab}\int \frac{d^4 k}{(2 \pi ) ^4 } \frac{-i g_a^2 b_0 ^ 2}{f^2 m_{\pi}^2 } \left ( \frac{1}{k^2- m_{\pi}^2 + i \epsilon } \right)^2 \sigma_{n}^{2}(k ) \\ & = & -\frac{2 i g_a^2 b_0 ^ 2 } { f^2 m_{\pi}^2 } \delta^{ab } \frac{k_f^4}{6 \pi^4 } f(\frac{m_\pi^2}{4k_f^2 } ) .\end{aligned}\ ] ] where the factor @xmath177 comes from the symmetric factor for the loop . the calculation of the nucleon one - loop in the fermi sea is given in eq . . we have defined @xmath178 after we evaluate the other diagrams in the third row in fig . [ density ] , we obtain @xmath179 here , we also take the physical value of the axial coupling @xmath180 . this term is a density correction to the @xmath0 sigma term through the pion - loop and is proportional to @xmath181 . the feynman diagrams in the fourth row in fig . [ density ] also contribute to the next - to - leading order , representing the ericson - ericson double - scattering correction @xcite . for example , the middle diagram of the fourth two in fig . [ density ] can be calculated as @xmath182 } i d_{\pi}(p+k ) \label{doublescat } \\ & = & \frac{2 b_0 ^ 2 \sigma_{\pi n}^2 k_f^4}{3 f_\pi^2 \pi^4 m_\pi^4 } \sigma_{n}^{4}(0 ) \\ & = & \frac{-8i b_0 ^ 2 \sigma_{\pi n}^2 k_f^4}{3 f_\pi^2 \pi^4 m_\pi^4 } g(\frac{m_\pi^2}{4k_f^2 } ) , \end{aligned}\ ] ] where @xmath183 is given as @xmath184 the detail calculations are summarized in appendix [ loops ] . in eq . we have understood that the terms with @xmath155 and @xmath156 vanish at the soft limit because they are proportional to the pion momentum @xmath185 . including the rest of the diagram for the double scattering correction , we obtain @xmath186 in this way , we obtain nlo contributions of in - medium chiral condensate : @xmath187 from eqs . ( [ lo ] ) and ( [ nlo ] ) , we obtain the in - medium chiral condensate within nnlo corrections(@xmath188 ) @xmath189 let us comment on the @xmath190 resonance contribution up to @xmath170 . the @xmath190 resonance contributes to the in - medium amplitudes through the @xmath190-hole excitation in these orders . nevertheless , the @xmath190-hole excitation in the left diagram of fig . [ density ] ( a ) vanishes in the soft limit because of the @xmath142-wave nature of the @xmath191 coupling as we have seen in the nucleon - hole excitation . the @xmath190-hole excitation in the diagrams of fig . [ density ] ( c ) should be accounted in the in - vacuum @xmath192 vertex , since the loop contribution of the pion and @xmath190 in this diagram appears in the higher order calculation of the @xmath192 vertex in the chiral expansion . therefore , up to nlo there are no explicit @xmath190 contributions to the in - medium chiral condensate . correlation function : the leading density contribution ( a ) and the next to leading density corrections ( b ) . the left and right wavy lines express the axial and pseudoscalar currents , respectively , the dashed line denotes the pion propagation and the solid dots stand for the leading vertices given by the chiral lagrangian @xmath127 . , width=302 ] according to eq . , @xmath117 can also contribute to the in - medium condensate . nevertheless , the correlation function @xmath193 gives a null value in the soft limit off the chiral limit . in this section , we confirm the cancellation of @xmath193 in the soft limit . in vacuum the correlation function @xmath193 should vanish in the soft limit because there is no zero mode propagation off the chiral limit and the coupling of the axial current to pseudoscalar modes is proportional to the external momentum @xmath194 , which is taken to be zero in the soft limit . this can be seen in the leading order calculation of the in - vacuum contribution as @xmath195 for the in - medium contributions of the correlation function @xmath193 we show the feynman diagram @xmath196 in fig . [ vanish ] . in this figure , the left and right wavy lines express the axial and pseudoscalar currents , respectively , and the dashed line denotes the pion propagation . in the linear density order , we also find that @xmath193 vanishes in the soft limit . for example , we evaluate the upper left diagram in ( a ) : @xmath197 } \\ & = & \lim_{q \to 0 } d_\pi ( q ) ( - i ) g_{a}^{2 } b_{0 } \delta^{ab } \sigma^{2}_{n}(q ) = 0 .\end{aligned}\ ] ] where we have used the one - nucleon loop function given in eq . , and in the last equation we have used the fact that the pion propagator is finite in the soft limit owing to the finite pion mass and the nucleon one - loop vanishes . this cancellation comes from two reasons ; the one is that pion is not a zero mode and the other is that interaction between pion and the axial current is derivative interaction , in other words explicit and spontaneous symmetry breaking leads to this cancellation . in addition , the interactions between pion and nucleon is proportional to the pion momentum because of spontaneous symmetry breaking , and it causes the same result . in the same manner one can confirm that contributions coming from the other diagrams give null value in the soft limit . up to the nlo density order , we find that the @xmath117 correlation function vanishes . generally speaking , there exist zero modes which couple with the axial current such as one - particle one - hole excitation as discussed in @xcite . nevertheless , up to the nlo corrections we find that such zero modes do not contribute . we note that in the chiral limit the @xmath117 correlation function contributes the in - medium chiral condensate , while the @xmath112 correlation function vanishes . one can find easily that the momentum dependence of the pion propagator cancels to the external momentum in the soft limit and @xmath117 remains finite . the in - medium chiral condensate in the soft limit calculated by the @xmath198 correlation function reads @xmath199 for the leading order of the density expansion and @xmath200 for the next leading order . these are equivalent to the result obtained from @xmath112 off the chiral limit by taking the chiral limit afterwards . in fig . [ chiral2 ] , the density dependence of the ratio of the chiral condensates , @xmath201 , is plotted as a function of @xmath202 in symmetric nuclear matter . the solid line represents the next - leading order ( nlo ) result shown in eq . . for comparison , we also show the linear density results in and off the chiral limit as the dashed and dotted lines , respectively . for the result in the chiral limit , we have used chiral limit values of the @xmath165 parameter @xmath203 and the pion decay constant @xmath204 . we find that the linear density result in the chiral limit decreases more rapidly than the results off chiral limit . the nlo correction amounts to about as small as 5% at @xmath205 , and becomes significant in higher density , for instance , at @xmath206 the nlo correction is around 10 % . therefore the linear density approximation is good in low densities , while in higher density the nlo contribution is not ignorable . numerically , we find that at normal nuclear density up to lo @xmath207 and up to nlo @xmath208 . these values are close to the value suggested by the recent precise pionic atom determination @xmath209 @xcite . we note that this experimental value is determined by linear density extrapolation under the assumption that the pion bound in the @xmath210 orbit is in a nuclear medium with an effective density @xmath211 . we also evaluate the quark condensate at @xmath212 and find @xmath213 for lo and @xmath214 up to nlo . these values are very close to the experimentally extracted value of the ratio @xmath215 @xcite , where @xmath7 is a parameter of the optical potential for the in - medium pion presenting the in - medium isovector @xmath0 scattering length and @xmath216 is the @xmath0 isovector scattering length . under the linear density approximation and a small isoscalar @xmath0 scattering length , the ratio of @xmath217 is equivalent to the ratio of the chiral condensate . thus , this implies that the density expansion might be good in at least low density region . we note that in - medium chpt is a low energy effective theory and this theory would be applicable up to about 2 normal density because at twice the normal nuclear density fermi momentum is about @xmath218 . in further higher density region this theory will be beyond applicability . nevertheless we could estimate the density at which chiral symmetry is restored . in the fig . [ chiral2 ] , we would find that the nlo correction raises the symmetry restoration density from @xmath219 to @xmath220 , which would imply that the nlo correction is not ignorable in high density region . when one considers further higher order correction beyond nlo , one encounters divergent amplitudes even though all of the @xmath0 interactions is fixed in vacuum . for instance , diagrams of multi - pion exchange in fermi gan as shown in fig . [ contact ] give divergent amplitudes . in the figure the solid line denotes fermi see insertion and the dashed line represents pion propagation . these diagrams count as @xmath221 namely @xmath222 . the upper diagram is proportional to @xmath223 } % \nonumber \\ & & \times id_\pi ( p ' ) id_\pi ( p+p ' -k ) . \nonumber % \\ % & & \propto \int \frac{d^4 p'}{(2 \pi ) ^4 } \frac{1}{p'^2 - m_\pi^2 + i \epsilon } \frac{1}{(p+p ' -q)^2 - m_\pi^2 + i \epsilon } \\ % & & \longrightarrow \infty\end{aligned}\ ] ] in this expression the integral with respect to @xmath224 for the pion loop gives divergence . as pointed out in ref . @xcite , we need @xmath118 contact terms to control the divergence . this means that one can proceed the in - medium calculation up to nlo by using the @xmath0 dynamics , but if one considers @xmath222 and higher corrections , one needs also in - vacuum @xmath118 contact terms obtained by the @xmath118 dynamics @xcite . we emphasize that in order to evaluate the in - medium chiral condensate quantitatively with higher density corrections , we need not only the @xmath0 dynamics informations but also @xmath118 dynamics . recently , as a step in this direction , a non - perturbative chiral effective theory has been developed to improve the @xmath118 correlation by including @xmath118 contact terms using a resummation method @xcite . moreover , in ref . @xcite the @xmath190(1232 ) resonance contributions have been evaluated and it has been found that @xmath190 resonance effects , which appears from the @xmath222 contributions , are not small . therefore , we may need more sophisticated calculations for the in - medium chiral condensate including @xmath118 and more dynamics . we calculate the chiral condensate at finite nuclear density @xmath225 using the chiral ward identity and the in - medium chiral perturbation theory . we study diagrammatic structure of the current - current green s functions @xmath226 and @xmath227 and classify density corrections to the chiral condensate . in our study we fix the @xmath0 dynamics in vacuum and calculate the in - medium chiral condensate with the in - medium chiral perturbation theory . in this study , lo(@xmath228 ) reproduces the well - known linear density approximation to the chiral condensate . this leading density correction is proportional to the @xmath229 sigma term @xmath230 . the next leading correction , nlo(@xmath231 ) , represents in - medium corrections of the sigma term . as a results , we find that linear density approximation is rather good in low density region such as a normal nuclear density . we have found that for higher corrections the correlation function has divergence from the pion loop even though all the in - vacuum quantities for the @xmath0 dynamics are fixed this means that the @xmath222 corrections can not be determined only by the @xmath0 dynamics in vacuum and the information of the @xmath118 dynamics is necessary to control the divergence . it should be also emphasized that for a realistic nuclear matter one should incorporate the nn dynamics and check that the matter satisfies the nuclear matter properties . this work was partially supported by the grants - in - aid for scientific research ( no . 25400254 and no . 24540274 ) . the chiral perturbation theory successfully introduces the chiral invariant lagrangian in the spontaneous breaking of chiral symmetry . the chiral field @xmath59 transforming linearly under the chiral rotation is written nonlinearly in terms of the pion field . the parametrization of the chiral field in terms of the pion field is not unique @xcite and all of the correct parameterizations provide the same physical result . however , one should realized that the basic field for the nonlinear sigma model to maintain chiral symmetry is the chiral field @xmath59 not the pion field as one can see that the partition function of the chiral perturbation theory is defined by the path integral with respect to the chiral field @xmath59 . therefore , when one considers quantum corrections of the pion field in perturbative calculations , one should be careful with chiral invariance , naive perturbative calculations might break chiral symmetry @xcite . the perturbative expansion of the pion fields is defined by the path integral with respect to the pion field , so that one should make the integral measure to be chiral invariant @xcite . one of the popular prescription is the background field method as it was applied to the chiral perturbation theory in ref . @xcite . here , instead of the cerebrated ccwz parametrization , we take the parametrization of the @xmath59 field which can be used for the naive perturbative calculation . this was found in refs . @xcite . in this parametrization the chiral field is written @xcite as @xmath232\ ] ] where the function @xmath233 satisfies @xmath234 for the calculation we expand the chiral field in terms of the pion field @xcite as @xmath235 and take some first terms . in this section , we show the chiral lagrangian and the @xmath0 interaction which we use in this work . the chiral lagrangian for the pion sector is as follows : @xmath236 where the covariant derivative is defined with the vector external fields as @xmath237 and the external scalar fields are given by @xmath238 in the following we list up the interaction lagrangian relevant for the present calculations . these terms are obtained from the lagrangian with the expansion : * @xmath3-@xmath239 vertex : @xmath240 * @xmath241-@xmath239 vertex : @xmath242 * @xmath243 vertex : @xmath244 * @xmath241-@xmath245 vertex : @xmath246 the chiral lagrangian for the one - nucleon sector is as follows : @xmath247 where @xmath62 represents all the chiral interaction with the nucleon bilinear form and can be counted in terms of the pion momentum : @xmath248 here @xmath249 is counted as @xmath250 . the explicit form of the leading term @xmath251 is @xmath252 with the vector current @xmath253 - \frac{i}{2 } u^\dag ( v_{\mu } + a_{\mu } ) u - \frac{i}{2 } u ( v_{\mu } - a_{\mu } ) u^\dag\ ] ] and the axial current @xmath254}\ ] ] here we define @xmath255 . the explicit expression of the next leading term @xmath256 is is given as @xmath257 -c_5 \hat{\chi}_+ -\frac{ic_6}{8m_n } \gamma^{\mu } \gamma^{\nu } f_{\mu \nu}^+\nonumber \\ & & - \frac{ic_7}{8m_n } \gamma^{\mu } \gamma^{\nu } \langle f_{\mu \nu}^+ \rangle \end{aligned}\ ] ] with @xmath258 , \hspace{2em } r_\mu = v_\mu + a_\mu \\ & & f_{\mu \nu}^l = \partial_\mu l_\nu - \partial_\nu l_\mu -i [ l_\mu , l_\nu ] , \hspace{2em } l_\mu = v_\mu - a_\mu\end{aligned}\ ] ] we list up the vertices which we use in the calculation : * @xmath259 vertex : @xmath260 * @xmath261 vertex : @xmath262 = - \frac{1}{2f } \gamma^\mu \pi^i a_\mu^j \epsilon^{ijk } \tau^k \\ a_{\pi a}^{(2 ) } & = & -\frac{2c_2}{f m_n^2 } \partial_{\mu } \pi^i a_{\nu}^i \partial^{\mu } \partial^{\nu } + \frac{2c_3}{f } \partial_{\mu } \pi^i a^{\mu i } \nonumber \\ & & - \frac{ic_4}{f } \epsilon^{ijk } \tau^k \partial_\mu \pi^i a_{\nu}^j [ \gamma^\mu , \gamma^\nu ] \nonumber \\\end{aligned}\ ] ] * @xmath263 vertex : @xmath264 * @xmath265 vertex : @xmath266 * @xmath267 vertex : @xmath268 = \frac{\gamma^{\mu}}{4f^2 } \pi^i \partial_{\mu } \pi^j \epsilon^{ijk } \tau^k \\ a_{\pi \pi}^{(2 ) } & = & \frac{4b_0 c_1 m_q}{f^2}\pi^i \pi^i + \frac{c_2}{f^2 m_n^2 } \partial_{\mu } \pi^i \partial_{\nu } \pi^i \partial^{\mu } \partial^{\nu } \nonumber \\ & & -\frac{c_3}{f^2 } \partial_{\mu } \pi^i \partial^{\mu } \pi^i + i \frac{ic_4}{f^2 } \epsilon^{ijk } \tau^k \gamma^{\mu } \gamma^{\nu } \partial_{\mu } \pi^i \partial_{\nu } \nonumber \end{aligned}\ ] ] in this section we calculate the nucleon loop diagrams in the fermi sea . in these calculations we take the trace only in the spin space , which is indicated by @xmath269 , and the isospin will be considered in other places . first of all , we calculate the nucleon tadpole @xmath270 which appears in a diagram ( a ) shown in fig . [ sigma ] : @xmath271 } \nonumber \\ & = & \int \frac{d^3 p}{(2 \pi ) ^3 2 e({\bf p } ) } { \rm tr_s } \big { [ } ( \slash{p } + m_n ) \big { ] } \theta ( k_f^i - |{\bf p}| ) \nonumber \\ & = & \int^{k_f^{(i)}}_{0 } \frac{{\bf p}^2 dp}{\pi^2 } \frac{m_{n}}{e({\bf p } ) } \nonumber \\ & \approx & \int^{k_{f}}_{0 } \frac{{\bf p}^2 dp}{\pi^2 } \left ( 1- \frac{{\bf p}^2}{2 m_n^2 } \right ) \label{eq : nonrela } \nonumber \\ & = & \rho^i \left ( 1- \frac{3 k_f^{i2}}{10 m_n^2 } \right ) , \label{eq : tadpole}\end{aligned}\ ] ] in eq . , we have taken the first two terms in the @xmath147 expansion and the nucleon density is given by @xmath272 . . ( b ) the one - loop diagram of nucleons in the fermi sea , @xmath273 . ( c ) the one - loop diagram in which one nucleon is in the fermi sea while the other is in vacuum , @xmath274 . ( d ) the double scattering term represented by the 2-loop diagram with two fermi sea insertions and 1 pion propagator.,width=302 ] next we consider @xmath275 which arises in a diagram ( b ) in fig . [ sigma ] @xmath276 } \\ & = & - \int \frac{d^4 p}{(2 \pi ) ^4 } { \rm tr_s } \big { [ } ( \slash{k } \gamma_5 ) id^{-1}_m ( p+k ) ( \slash{k } \gamma_5 ) id^{-1}_m ( p ) \big { ] } \\ % & = & \frac{1}{2 } m_q \lim_{q \to 0 } \sum_{i j } ( -1 ) \int \frac{d^4 k}{(2 \pi ) ^4 } \frac{d^4 p}{(2 \pi ) ^4 } i ( 2 f b_0 ) id^{-1}_0 ( q ) i \left ( -\frac{2b_0}{5f } \big { ( } \delta^{ki } \delta^{lj } + \delta^{kj } \delta^{il } + \delta^{kl } \delta^{ij } \big { ) } \right ) \left ( id^{-1}_0 ( k ) \right)^2 \\ % & & \times \frac{- g_a^2}{2f^2 } \delta^{ij } { \rm tr_s } \big { [ } \slash{k } ( \slash{p } + \slash{k } - m_n ) \slash{k } ( \slash{p } + m_n ) \big { ] } ig_m ( p ) ig_m ( p+k ) \\ & = & \int \frac{d^3 p}{(2 \pi ) ^3 2e({\bf p})}(-2\pi ) \frac{\delta(k_{0 } + e({\bf p } ) - e({\bf k+p}))}{2 e({\bf k+p } ) } \\ & & \times{\rm tr}_s \big { [ } \slash{k } ( \slash{p } + \slash{k } - m_n ) \slash{k } ( \slash{p } + m_n ) \big { ] } 2 \delta^{ij } \\ & & \times \theta ( k_f^i - |{\bf p}+ { \bf k}| ) \theta ( k_f^i - |{\bf p}| ) , \end{aligned}\ ] ] where we have used @xmath277 with @xmath278 . since both nucleons are in the fermi sea , they are on the mass shell , @xmath279 and @xmath280 , which provides @xmath281 . using these facts , the trace can be evaluated as @xmath282 } = - 8 k^{2 } m_{n}^{2 } .\ ] ] now we consider the @xmath147 expansion in the nucleon energy and take the first term , that is , @xmath283 . then , finally we obtain @xmath284 where @xmath285 and we have used the formula @xcite @xmath286 we calculate @xmath287 which appears in a diagram ( c ) shown in fig . [ sigma ] in the soft limit : @xmath288 } \nonumber \\ & = & \lim_{k\to 0}\int \frac{d^3 p}{(2 \pi ) ^3 2e({\bf p } ) } \frac{i}{(k+p)^{2 } - m_{n}^{2 } + i \epsilon } \nonumber \\ & & \times{\rm tr_s } \big { [ } \slash{k } ( \slash{p } + \slash{k } - m_n ) \slash{k } ( \slash{p } + m_n ) \big { ] } \theta ( k_f^i - |{\bf p}| ) \nonumber \\ & = & \lim_{k\to 0}\int \frac{d^3 p}{(2 \pi ) ^3 2e({\bf p } ) } \frac{i}{k^{2 } + 2 k \cdot p+ i \epsilon } \theta ( k_f^i - |{\bf p}| ) \nonumber \\ & & \times 4 \left[(2 k\cdot p+k^{2 } ) k\cdot p - 2 k^{2 } m_{n}^{2}\right ] \nonumber \\ & = & 0.\end{aligned}\ ] ] this goes to zero in the soft limit . finally we calculate @xmath289 given as diagram ( d ) in fig . [ sigma ] in the soft limit : @xmath290 } id_\pi ( p+k ) \\ & = & \int \frac{d^3 p}{(2 \pi ) ^3 } \frac{d^3 q}{(2 \pi ) ^3 } { \rm tr}_s \big { [ } ( \slash{q } - \frac{\slash{p}}{2 } + m_n ) ( \slash{q } + \frac{\slash{p}}{2 } + m_n ) \big { ] } \\ & & \times \frac { \theta ( k_f^i - |{\bf q } - \frac{{\bf p}}{2}|)}{2e({\bf q } - \frac{\bf p}{2 } ) } \frac{\theta ( k_f^i - |{\bf q } + \frac{{\bf p}}{2 } | ) } { 2e({\bf q } + \frac{\bf p}{2 } ) } id_\pi ( p ) .\end{aligned}\ ] ] here we have performed the @xmath105 and @xmath291 integrals , which contain two delta functions for the energy conservation and give us the following relation @xmath292 in addition we have the following kinematical relations : @xmath293 using these relation , the trace can be evaluated as @xmath294 = 8 m_{n}^{2 } - 2 p^2 .\ ] ] taking the leading term of the @xmath147 expansion in which @xmath295 and @xmath296 , we obtain @xmath297 as @xmath298 here @xmath299 is defined by @xmath300}.\ ] ] t. hatsuda , t. kunihiro , and h. shimizu , phys . * 82 * , 2840 ( 1999 ) . d. jido , t. hatsuda , and t. kunihiro , phys . d * 63 * , 011901 ( 2000 ) . t. hyodo , d. jido , and t. kunihiro , nucl . a * 848 * , 341 ( 2010 ) . s. weinberg , phys . * 18 * , 507 ( 1967 ) . kapusta and e.v . shuryak , phys . rev . d*49 * , 4694 ( 1994 ) . detar and t. kunihiro , phys . rev . d*39 * , 2805 ( 1989 ) . kim , d. jido , and m. oka , nucl . a * 640 * 77 ( 1998 ) . d. jido , y. nemoto , m. oka , and a. hosaka , nucl . a * 671 * , 471 ( 2000 ) . daisuke jido , makoto oka , and atsushi hosaka , prog . theor . phys . * 106 * , 873 ( 2001 ) . d. jido , h. nagahiro , and s. hirenzaki , phys . c * 66 * , 045202 ( 2002 ) . h. nagahiro , d. jido , and s. hirenzaki , phys . c * 68 * , 035205 ( 2003 ) . d. jido , e. e. kolomeitsev , h. nagahiro , and s. hirenzaki , nucl . a * 811 * , 158 ( 2008 ) . lee and t. hatsuda , phys . d * 54 * 1871 ( 1996 ) . d. jido , h. nagahiro , and s. hirenzaki , phys . c * 85 * 032201(r ) ( 2012 ) . e. g. drukarev and e. m. levin , prog . * 27 * , 77 ( 1991 ) . v. dmitrasinovic , phys . c * 59 * , 2801 ( 1999 ) . s. weinberg , phys . rev . * 166 * , 1568 ( 1968 ) . charap , phys . d * 3 * , 1998 ( 1971 ) . gerstein , r. jackiw , s. weinberg , and b.w . lee , phys.rev . d * 3 * , 2486 ( 1971 ) . a.l . fetter and j.d . walecka , _ quntaum theory of many - particle systems _ ( dover , new york , 2003 )

v794 aql was observed with the hubble space telescope / space telescope imaging spectrograph ( _ hst_/stis ) on august 28 , 2003 , and with the far ultraviolet spectroscopic explorer ( _ fuse _ ) on may 13 , 2004 . in both observations v794 aql was found in a relatively high state with the same flux level . the _ fuse _ spectrum exhibits heavy interstellar hydrogen absorption features that implies , , and a reddening value e(b - v)=0.08 . inspection of the existing archival _ iue _ spectra also indicates that the reddening could be large . we present here a spectral analysis of the dereddened _ fuse _ and _ hst_/stis spectra separately and combined together assuming e(b - v)=0.1 & 0.2 . overall , we find that the model fits are in much better agreement with the dereddened spectra when e(b - v ) is large , as excess emission in the longer wavelengths render the slope of the observed spectra almost impossible to fit , unless e(b - v)=0.2 . for a the best fit is an accretion disk with a mass accretion rate yr with an inclination when assuming e(b - v)=0.2 , and decreases by one order of magnitude when assuming e(b - v)=0.1 . the distance to the system ispc for the e(b - v)=0.2 case and it decreases topc for e(b - v)=0.1 . the best fit accretion disk model is obtained for e(b - v)=0.2 . a single white dwarf model leads to a rather hot temperature ( depending on the assumptions ) but does not provide a fit as good as the accretion disk model . a combination of a white dwarf plus a disk does not lead to a better fit . the same best fit disk model is consistently obtained when fitting the _ fuse _ and _ hst_/stis spectra individually and when combined together , implying therefore that the disk model is the best fit not only in the least sense , but also as a consistent solution across a large wavelength span of observation . this is not the case with the single white model fitting which leads to a different ( and therefore inconsistent ) temperature for each different spectrum ( _ fuse _ , stis and _ fuse_+stis ) . subject headings : accretion , accretion disks - novae , cataclysmic variables - stars : individual ( v794 aquilae ) - ultraviolet : stars - white dwarfs

astro-ph0703083

cataclysmic variables ( cvs ) are short - period , semi - detached binary systems consisting of an accreting white dwarf ( wd ) star ( the primary ) and a low - mass main - sequence star ( the secondary ) as the roche lobe - filling mass donor @xcite . in non magnetic cv systems , the mass is accreted by means of an accretion disk reaching all the way down to the surface of the wd . on - going accretion at a low rate ( quiescence ) is interrupted every few weeks to months by intense accretion ( outburst ) of days to weeks - a dwarf nova ( dn ) accretion event . cv systems are divided in sub - classes according to the durations , occurrence and amplitude of their outburst : e.g. dwarf nova systems ( dns ) spend most of their time in the quiescent state , while nova - like systems ( nls ) are found mostly in the high outburst state . both dn and non - magnetic nl systems exhibit emission from the accretion disk during the high state . far ultraviolet ( fuv ) observations have shown that some wds in cvs can be directly viewed in the uv when the accretion disk is not dominant ( as ealy as @xcite ) . consequently , much effort has gone to observe the systems with low mass transfer rates . in those systems the wds are hotter than single field wds and their temperature increases with orbital period , as higher accretion rates are found in longer period systems @xcite . interestingly enough , while the binary period of the cv systems ranges between a fraction of an hour ( e.g. am cvn systems ) up to @xmath13 day ( e.g. gk per ) , there is a gap between about 2 and 3h where almost no system is found ( hereafter the `` period gap '' ) . however , at a given orbital period the accretion rate can vary by a large amount and unfortunately there are not many data points for long period systems above the period gap ( @xmath143hrs ) . in the accretion disk limit cycle @xcite , material accumulates in the accretion disk during quiescence and accretes onto the wd during outburst . in order for an outburst to occur , the accretion rate must be below a critical value for a given orbital period . dne at periods above the gap should have accretion rates below the limit , the rates for z cam are very close to the critical value and the novalikes have rates above the limit . thus , above the period gap one finds 3 classes of systems : u gem ( dn ) type that undergo outbursts ; z cam ( dn ) type which have standstills where they remain at about one magnitude below their outburst level for long times ; and nova - like systems that are in a permanent high state . among these high accretion nl systems , only the underlying wds of dw uma @xcite , tt ari @xcite and mv lyr @xcite have been studied for temperature , gravity , rotation and chemical abundances information . all the other nls that have been observed have been caught in a high state . the spectra of nls ( in permanent high state ) consist of h emission lines which may of may not be superimposed on broad shallow absorption features ; and the brightness fluctuates about some mean value , deviating up and down irregularly by no more than about 1 magnitude . the spectra in the near uv and optical is consistent with that of an accretion disk in outburst . while nls are characterized by an approximately steady , high rate of mass transfer ( and consequently a high luminous accretion disk - a permanent high state ) , there is a class of nls ( namely , the vy sculptoris systems ) in which the prolonged state of high accretion is interrupted unpredictably by low brightness states of little or no accretion when the disk greatly shrinks or vanishes and the underlying accretion heated hot wd is exposed , as it is the case for mv lyr @xcite . the vy scl systems are apparently all disk systems with negligible ( magnetic ) accretion at the poles . v794 aql is a nova - like system belonging to the vy sculptoris class , or at least it appears to have observational properties consistent with this class . the vy scl stars fall above the period gap , all in the range 3 - 4hr , except for v751 cyg with @xmath15h . the vy scl have all extreme ( 3-magnitude ) brightness variations in the high state and unexpectedly fall to an extreme low state ( more than 5 magnitudes ) . v794 aql has been seen to vary erratically between photographic magnitude 14 and 17 on time scales of days to years @xcite , and @xcite observed v794 aql in an unprecedented low state with magnitude 20 ! photometrically speaking , v794 aql seems to be one of the most active cataclysmic variables with brightness variations of up to 0.5 mag occurring on a timescale of minutes ( and smaller flares on timescales of tens of seconds , @xcite ) . the brightness variations in vy scl systems appear to be random in amplitude , shape and recurrence interval . however , v794 aql is an exception : photometric monitoring of v794 aql has revealed an unusual type of light curve @xcite , in which the uniform decline from high to low state is interupted by an abrupt return to the high state , giving the light curve a distinctive sawtooth appearance . this is a type of photometric variation in cvs which has never been reported . v794 aql was spectroscopically observed in high and low state in the optical @xcite , in the uv with _ iue _ @xcite and in the x - ray with einstein @xcite , and in respects other than the character of its long - term light curve , v794 aql was found to have a photometric and spectroscopic behavior very similar to other members of the nl vy sculptoris class , such as mv lyrae and tt arietis . in the x - ray , v794 aql was observed near a high state in terms of its optical range . the system had a flux of @xmath16 ergs@xmath17@xmath1s@xmath6 ( in the 0.1 - 4.5 kev band ) , corresponding to an x - ray luminosity of @xmath18ergs@xmath17s@xmath6 , for a distance d=100pc . the source was detected in both the soft ( 0.1 - 0.5 kev ) and hard ( 0.5 - 4.5 kev ) energy bands , and the hardness ratio was found to be @xmath19 ( though the apparent hardness can be affected by the absorption of soft x - rays either at the source itself or in the ism ; and ism absorption might not be negligible for v794 aql - see section 2.2 ) . simultaneous ground - based optical observations showed blue colors , strong h emission with a flat balmer decrement , plus strong he and he @xmath204686 emission . these _ iue _ , x - ray , as well as optical spectra show relatively strong emission lines of intermediate excitation , similar to those seen in other vy sculptoris stars in both the low state and the high state . in the low state reported by @xcite the accretion disk emission lines were replaced by very narrow balmer emission , in a behavior similar to ( but more extreme than ) that of mv lyr ( e.g. @xcite ) and a few other nl systems in which the accretion apparently switches off for brief periods of time . though v794 aql has many of the properties of a typical disk vy scl star , there has been no direct evidence yet that a disk exists in this system . v794 aql can not be classified as a z cam system , as the sawtooths are consistently much fainter than the occasional intervals of steady brightness characteristic of z cam systems . if v794 aql is a z cam system , then it is certainly a very unusual one . @xcite studied v794 aql from _ iue _ spectra , however the low resolution and geocoronal contamination of the _ iue _ spectra prevented accurate determination of the temperature and the low response shortward of ly@xmath21 prevented any observation of a rising continuum from the hot wd . this study estimated a white dwarf temperature near 50,000k , while the accretion rate has been estimated to be around @xmath22yr@xmath6 @xcite - based on the long term optical variations of the system . this nl system has the highest mass accretion rate , above the limit for outbursts . because of its inclination , both the hot wd and the inner disk might be observable . more recently an _ hst_/stis snapshot of v794 aql was obtained , and the system was also observed in the fuv with _ fuse _ under a cycle 4 program . in this paper , we report a spectral analysis of the _ fuse _ and _ hst_/stis spectra of v794 aql using accretion disk models , photosphere models , and models combining white dwarfs and accretion disks . our objectives are to identify the source(s ) of the fuv radiation , derive the properties of the wd ( if possible ) , the accretion disk , and characterize the hot component in the system . we report here all the ultraviolet spectra of v794 aql . this includes 5 _ iue _ archival spectra , one _ hst_/stis spectrum and one _ fuse _ spectrum . while we model mainly the _ fuse _ and _ hst_/stis spectra , we also use the _ iue _ spectra for flux comparison to estimate the relative state in which the system was found during each observation . for all the spectra we identify the emission and absorption features with a particular emphasis on the estimate of the molecular and atomic hydrogen column densities using the _ fuse _ and _ hst_/stis ( which covers the ly @xmath21 region ) . we use the _ iue _ spectra to carry out an assessment of the reddening of the system . the reason for questioning the zero color excess value of @xcite is in part the _ a posteriori _ fact that the slope of the continuum ( both in the _ fuse _ and stis spectra ) can not be easily matched with the model fits when assuming a small e(b - v ) value . we obtain a large residual in the longer wavelength indicating a net excess of flux in the red part of the spectrum . this discrepancy basically disappears when we deredden the spectra assuming @xmath23 . in addition , the hydrogen column density we obtain from the _ fuse _ and stis spectra leads to a reddening of e(b - v)=0.08 using the relation of @xcite , the _ iue _ archive contains three good observations of v794 aql with the short - wavelength prime ( swp ) camera through the large aperture at low dispersion ( see table 2 and figure 1 ) . one additional _ iue _ swp spectrum ( swp 14708 ) has an extremely low flux level and a very low signal to noise ( s / n ) ratio . this spectrum has been disregarded here as it has basically no continuum and can not be used . the _ iue _ swp spectra have a wavelength binning of 1.68 , covering the wavelength range 1150 to 1980 . there is also a data set obtained with the long - wavelength camera lwr 11782 , which we use to assess the reddening of the system . the _ iue _ lwr spectrum has a wavelength binning of 1.67 , and covers the wavelength range 1850 - 3350 . all the _ iue _ newsips spectra were flux - calibration corrected using the massa - fitzpatrick corrections @xcite . these spectra were analyzed in detail by @xcite . the swp 28501 spectrum was obtained when the system had a visual magnitude of 17 - 18 , a relatively low state , though not the lowest state of v794 aql ( magnitude 20 reported by @xcite ) . the spectrum swp 15266 has a flux only slightly larger than swp 28501 , but in addition it exhibits some broad emission lines . the most prominent ones are @xmath24 ( 1238/1243 ) , @xmath25 ( 1394/1403 ) , @xmath26 ( 1548/1550 ) , and @xmath27 ( 1640 balmer @xmath21 ) . with its broad emission lines characteristic of nova - like systems , the spectrum of v794 aql show evidence of hot gas . the swp 50754 shows the system in a relatively higher state with emission and absorption features , however , the s / n of this spectra is very low . the presently accepted value for the reddening of v794 aql is @xmath28 ( e.g. @xcite , based on the work of @xcite ) . however , the heavy interstellar hydrogen absorption in its _ fuse _ spectrum ( see next subsection ) implies that the source is either located pretty far away , or masked by an ism cloud , or some cold circumstellar matter , and consequently it is probably also affected by dust , therefore making the assumption @xmath28 very unlikely . in his pioneer work , @xcite carried out an analysis of the broad ( several hundreds ) absorption feature around 2175 of the _ iue _ spectra of 51 cv systems to determine their reddening value e(b - v ) . the reddening of the 51 systems was assessed by dereddening each spectrum assuming increasing values of e(b - v ) ( e.g. 0.05 , 0.10 , 0.15 .. ) until the broad absorption dip ( between @xmath292000 and @xmath292400 ) disappeared completely to the `` eye '' ( i.e. the acceptable range of values of e(b - v ) were found by visual inspection ) . following @xcite we first combine the _ iue _ spectra swp 28501 and lwr 11782 binned at 20 ( we have chosen the swp 28501 spectrum as it matches the flux level and shape of lwr 11782 ) . the spectra are very noisy , including a hot pixel at 2200 . this combined spectrum is then dereddened assuming values e(b - v)=0.10 , 0.20 and 0.30 . in figure 2 we plot the ab - magnitude @xcite for each dereddened spectrum for visual inspection . the broad absorption feature is more apparent in the shortest wavelengths between 2000 and 2200 in the e(b - v)=0 spectrum , an indication that the reddening is not zero . obviously for e(b - v)=0.30 there is excess emission around @xmath30 . for e(b - v)=0.10 and 0.20 the absorption feature is not that apparent as the flux seems to level off , an indication that the reddening is probably around @xmath31 but could be as large as 0.20 . since we have only one _ iue _ lwr spectrum of v794 aql , and it is extremely noisy , it is difficult to say more about the reddening . in addition , we see how difficult it can be to assess how much of the broad absorption feature is present when it is left only to a visual inspection to decide , as one can not differentiate between noise and signal . this is probably the reason why @xcite assumed an error of 0.1 for v794 aql . the difference between our reddening assessment and verbunt s might reside in the fact that verbunt co - added all the swp spectra together , even though only the swp 28501 spectrum matches the lwr 11782 spectrum . also , since the estimate is carried out by visual inspection , the subjectivity of the observer is also a decisive factor . _ fuse _ is a low - earth orbit satellite , launched in june 1999 . its optical system consists of four optical telescopes ( mirrors ) , each separately connected to a different rowland spectrograph . the four diffraction gratings of the four rowland spectrographs produce four independent spectra on two microchannel plates . two mirrors and two gratings are coated with sic to provide wavelength coverage below 1020 , while the other two mirrors and gratings are coated with al and lif overcoat . the al+lif coating provides about twice the reflectivity of sic at wavelengths @xmath141050 , and very little reflectivity below 1020 (hereafter the sic1 , sic2 , lif1 and lif2 channels ) . a time tag _ fuse _ spectrum ( d1440101 ) of v794 aql was obtained starting on 2003 , may 13 , ( at 19:50:22 ) with a total observing duration covering 6 individual spacecraft orbits through the lwrs . the system was in a relatively high state at the time of the observation . the data were processed with calfuse version 3.0 totaling 13,066s of good exposure time . the main change from previous versions of calfuse is that now the data are maintained as a photon list ( the intermediate data file - idf ) throughout the pipeline . bad photons are flagged but not discarded , so the user can examine , filter , and combine data without re - running the pipeline . a number of design changes enable the new pipeline to run faster and use less disk space than before . processing time with calfuse has decreased by a factor of up to 10 . in this version , event bursts are automatically taken care of . event bursts are short periods during an exposure when high count rates are registered on one of more detectors . the bursts exhibit a complex pattern on the detector , their cause , however , is yet unknown ( it has been confirmed that they are not detector effects ) . during the observations , fine error sensor a , which images the lif 1 aperture was used to guide the telescope . the spectral regions covered by the spectral channels overlap , and these overlap regions are then used to renormalize the spectra in the sic1 , lif2 , and sic2 channels to the flux in the lif1 channel . we then produced a final spectrum that covers almost the full _ fuse _ wavelength range @xmath32 . the low sensitivity portions of each channel were discarded . here we took particular care to discard the portion of the spectrum where the so - called _ worm _ crawls , which deteriorates lif1 longward of @xmath33 ( calfuse can not correct target fluxes for this effect ) . because of this the @xmath34 region was lost ( see the _ fuse _ instrument and data handbook or section 2.1 in @xcite ) . we combined the individual exposures and channels to create a time - averaged spectrum with a linear , @xmath35 dispersion , weighting the flux in each output datum by the exposure time and sensitivity of the input exposure and channel of origin . v794 aql , with a flux of a few @xmath36ergs@xmath17s@xmath6@xmath1@xmath6 , is actually a relatively weak source . the procedure we used to process the _ fuse _ data of v794 aql is the same as the one we used in our previous _ fuse _ analysis of cvs , such as e.g. ru peg & ss aur @xcite , vw hyi @xcite , or ww cet @xcite . + * interstellar absorption toward v794 aql * a single look at the _ fuse _ spectrum of v794 aql ( figure 3 ) immediately reveals that the continuum is moderately affected by hydrogen absorption : the continuum is basically sliced ( at almost equal intervals ) by interstellar hydrogen lines starting at wavelengths around 1110 and continuing towards shorter wavelengths all the way down to the hydrogen cut - off around 915 . in table 3 we identified all the absorption and emission lines of metals in the _ fuse _ spectrum of v794 aql . in table 4 we identified the most prominent molecular hydrogen absorption lines by their band ( werner or lyman ) , upper vibrational level ( 1 - 16 ) , and rotational transition ( r , p , or q ) with lower rotational state ( j=1,2,3 ) . next , we model the ism hydrogen absorption lines to assess the atomic and molecular column densities . we use a custom spectral fitting package to estimate the temperature and density of the interstellar absorption lines of atomic and molecular hydrogen . the ism model assumes that the temperature , bulk velocity , and turbulent velocity of the medium are the same for all atomic and molecular species , whereas the densities of atomic and molecular hydrogen , and the ratios of deuterium to hydrogen and metals ( including helium ) to hydrogen can be adjusted independently . the model uses atomic data of @xcite and molecular data of @xcite . the optical depth calculations of molecular hydrogen have been checked against those of @xcite . for v794 aql , the ratios of metals to hydrogen and deuterium to hydrogen are fixed at 0 and @xmath37 , respectively , because of the low signal - to - noise ratio data . the wings of the atomic lines are used to estimate the density of atomic hydrogen and the depth of the unsaturated molecular lines for molecular hydrogen . the temperature and turbulent velocity of the medium are primarily determined from the lines of molecular hydrogen when the ism temperatures @xmath38k . in order to model the atomic hydrogen column density correctly , we take care to include the portion of the _ hst_/stis spectrum exhibiting the ly @xmath21 feature into the modeling . in figure 4 we show our fit to the ( atomic and molecular ) hydrogen absorption lines to the _ fuse _ spectrum of v794 aql , as applied to our wd model fit ( namely , we multiplied the wd synthetic spectral model by the transmission values obtained from the ism model ) . this ism model has zero metalicity , a temperature of 150k , a turbulent velocity of 40km@xmath17s@xmath6 , a molecular hydrogen column density of @xmath39@xmath1 and an atomic hydrogen column density of @xmath40@xmath1 . the ly @xmath41 feature around 1025 is obviously far too sharp to be due to the disk , even with an inclination of only @xmath42 this feature would be much broader . this absorption feature can not be accounted by the star itself neither and it is modeled here as part of the ism model . it is interesting to note that the hydrogen column density of v794 aql that we obtain here is very similar to the hydrogen column density of v1432 aql @xcite which is located in the same region of the sky ( at about 230 pc ) . v1432 aql is much closer to v794 aql than any of the nearest neighboring sources found by the ism hydrogen column density tool from the _ euve _ web site . also in the same vicinity and at about the same distance ( @xmath43pc ) , uu aql exhibits strong hydrogen absorption features in it _ fuse _ spectrum @xcite . this seems to indicate that this region of sky in the constellation of aquila might have an ism cloud in the foreground ( say at @xmath44pc or less ) . next , we use the hydrogen column density we obtained to estimate the reddening of the system using the analytical expression given by @xcite : @xmath45 we are aware that this expression is not accurate , in that sens that it represents a mean extinction law with a large deviation , as the ratio of dust to hydrogen is not constant . for example , objects towards which the atomic column density is ( say ) @xmath46atoms @xmath1 , have a e(b - v ) value ranging from 0.03 to 0.2 , almost a factor of 10 . for hydrogen column densities higher than @xmath47atoms e(b - v ) varies only by a factor of 3 but the typical scatter is really of only about 30% . using our computed ism value for the hydrogen column density @xmath2@xmath1 and @xmath48@xmath1 in equation ( 1 ) , we obtain @xmath49 . the actual reddening toward v794 aql could be smaller or larger than 0.08 as explained in the previous paragraph . however , from the visual inspection of m(ab ) of the _ iue _ spectra of v794 aql we decide to model the spectra of v794 aql for 2 values of the reddening , namely e(b - v)=0.1 and e(b - v)=0.2 . an _ hst_/stis snapshot spectrum of v794 aql was obtained on august 28 2003 , totaling 830 sec of exposure time . the spectrum was taken with the 0.2x0.2 aperture in accum ( accumulation ) mode using the g140l grating , covering the wavelength range from 1150 to 1710 , with a wavelength binning of @xmath50 . the use of g140l provides a higher resolution spectrum than _ iue _ , but not as high as when using e140 m . the spectrum was taken in a relatively high state , similar to the state during which the _ fuse _ spectrum was obtained . the spectrum exhibits some broad emission lines , mainly @xmath51 ( 1175 ) , @xmath52 ( 1240 ) , @xmath25 ( 1400 ) , @xmath26 ( 1550 ) and @xmath27 ( 1640 ) . other possible lines are also annotated in figure 1 . also very pronounced is the ly @xmath21 absorption feature ( around 1215 ) . on the sides of the ly@xmath21 absorption feature there is some emission from @xmath53 on the left and @xmath54 on the right . the ly@xmath21 absorption feature is too sharp to be that of the wd and/or the accretion disk of v794 aql , however , it agrees very well with the hydrogen column density we found , except that it should be saturated , as the transmission values of the ism model are null there due the high column density of atomic hydrogen . the only reason the ly@xmath21 is not saturated in the observed stis spectrum of v794 aql is because the spectral resolution ( @xmath55 ) of the g140l grating combined with the short exposures required by snapshots often does not resolve the sharp geocoronal @xmath56 emission , seen in many stis g140l exposures of some systems . for example the ( mast ) archived stis snapshot of vy scl ( with e(b - v)=0.06 @xcite ) has a sharp ly@xmath21 absorption feature that is not saturated , and the stis snapshots of ss aur and v442 cen ( with e(b - v)=0.08 and 0.15 , respectively ; @xcite ) show clearly a sharp emission peak in the bottom of their ly@xmath21 profile . all these systems with a reddening value as large as that of v794 aql have most probably an atomic hydrogen column density larger than @xmath57@xmath1 and should therefore all have a saturated ly@xmath21 absorption feature ( for any line of sight with @xmath58@xmath1 the ly@xmath21 absorption feature should be saturated ) . not enough light - curve data points were available from the american association of variable star observers ( aavso ) to determine whether v794 aql was observed in a high or low state . however , a comparison of the _ fuse _ spectrum with the _ iue _ archival spectra and the _ hst_/stis snapshot ( see figure 1 ) helps us to assess the state in which the system was at the time each observations was made . from that figure it is clear that v794 aql was observed in about four different states . the lowest _ iue _ spectrum has a flux about 5 times lower than the _ fuse _ and stis spectra . this _ iue _ spectrum , however , was itself not obtained during the lowest state(@xmath59 ) . it seems therefore quite clear that the _ fuse _ and stis spectra need mainly to be modeled with a disk model and the contribution of the wd might be small . however , because of its inclination ( @xmath60 ) the contribution of the heated wd during outburst might be non - negligible . though there is a large error bar on the mass of the wd ( 0.88 @xmath61 0.4 @xmath62 ) , we deliberately decided to use only one mass in our modeling ( @xmath63 ) to limit the number of unknown parameters in the system : the distance , the inclination @xmath64 , the reddening . we deredden the spectra assuming a reddening of e(b - v)=0.1 and 0.2 . we prepared the _ fuse _ and _ hst _ spectra for fitting by masking regions containing emission lines and artifacts . these regions are emphasized with a blue color in figure 3 . these regions of the spectrum were not included in the fitting . a comparison of the _ hst_/stis spectrum with the _ fuse _ spectrum in the wavelength overlap region reveals that the flux levels match thus enabling us to carry out model fits over a substantially broader wavelength range . the high - gravity white dwarf atmosphere model spectra are generated assuming solar abundances ( unless otherwise specified ) using tlusty200 @xcite , synspec48 and rotin4 @xcite . tlusty generates numerical models of stellar atmospheres for a given surface gravity and effective temperature . synspec is then used to generate a synthetic spectrum for each particular stellar atmosphere model . the routine rotin is used last to perform a rotational and/or instrumental convolution of the synthetic spectrum obtained from synspec . for the input temperature in tlusty , we chose t ranging from 15,000k to 55,000k by increments of 5,000k at first , and when a best fit is found a refined fitting is carried out by changing the temperature in increments of 1,000k . we chose a value of @xmath65 to match the mass of the wd in v794 aql , namely @xmath66 . we also varied the stellar rotational velocity @xmath67 from @xmath68km@xmath17s@xmath6 to @xmath69km@xmath17s@xmath6 in steps of @xmath68km@xmath17s@xmath6 . we have the possibility to change the abundance of elements , however , we first ran a model with solar composition and only changed the si and c abundances in two cases , to assess how it affects the results . for the synthetic accretion disk models , we used the latest accretion disk models from the optically thick disk model grid of @xcite . the range of disk model parameters in the grid of @xcite varies as follows : wd mass ( in solar masses ) values of 0.35 , 0.55 , 0.80 , 1.03 , and 1.21 ; orbital inclination ( in degrees ) of 18 , 41 , 60 , 75 , and 81 . the accretion rate ranges from @xmath70yr@xmath6 to @xmath71yr@xmath6 , varying in increments of 0.5 in log m. for the present work we select from this grid the models with parameters consistent with table 1 , namely : wd masses 0.80 and 1.03@xmath62 , an inclination angle of 18 , 41 and 60@xmath72 and the entire range of mass accretion rates from the grid of @xcite . for combined accretion disk and stellar wd atmosphere models , the disk flux is divided by 100 to normalize it at 1000pc to match the wd flux , therefore giving explicitly the relative flux contributions of each component . then both fluxes are added for comparison with the observed flux . the best two - component models are not especially a combination of the best wd models with the best disk models . the main aim here is to find a fit that is better ( lower @xmath73 ) than both the wd model fit and the accretion disk model fit . before finding the best fit , all the synthetic spectral models ( wd , accretion disk , and wd+accretion disk ) are multiplied by the transmission values of the ism model presented in section 2.2 . in order to find the best model , we use fit - a @xmath12 minimization routine ( see e.g. @xcite ) . for each model , we computed @xmath74 and scale factor values . @xmath73 is known as the `` reduced '' chi - squared , namely @xmath12 per number of degrees of freedom . in the present case the number of degrees of freedom is the number of wavelength bins of the observed spectrum taken into account in the fitting . therefore , when we exclude ( i.e. mask ) regions of the spectrum ( such as broad emission lines and features such as air glow ) in preparation for the fitting , we reduce the number of degrees of freedom . and when we merge the _ fuse _ spectrum together with the _ hst_/stis spectrum we actually increase the number of degrees of freedom . since _ fuse _ and stis are binned differently and have different errors , one can not compare the least chi square best fit model obtained by fitting the _ fuse _ spectrum alone with the least chi square best fit model obtained when fitting the stis spectrum alone or the combined _ fuse_+stis spectrum . these fits can not be compared on the basis of their @xmath73 values , as the same synthetic model will lead to different @xmath73 values when fitting the _ fuse _ spectrum alone , the stis spectrum alone , or the combined _ fuse_+stis spectrum . so the analysis of the _ fuse _ , stis and the combined _ fuse_+stis spectra are carried out separately . for each single value of @xmath65 there exists only one value of @xmath75 and @xmath76 , since we use the mass radius relation for wds ( e.g. @xcite or @xcite for different composition and non - zero temperature wds ) . as a consequence , when fitting the theoretical flux to the observed flux , the distance to the system is obtained as an output parameter when scaling the fluxes . when the wd mass and distance of a system are known , the fitting technique leads to only one precise value of the wd temperature . in the present case , since neither the mass nor the distance are known , the best fit model consists actually of a whole domain in the @xmath65-@xmath77 plane . because of that we fix the wd mass to @xmath78 ( for the wd models ) and use the @xmath79 models from the grid of accretion disk models . we then chose the lowest @xmath12 model ( best fit ) and obtain as output the wd temperature ( or the mass accretion rate ) and the distance to the system . since the reddening is also not known accurately , the choice @xmath80 helps us limit the number of solutions to a single table . we first tried a single white dwarf model , then an accretion disk alone , and then a combination of both for the _ fuse _ spectrum alone , then for the _ hst_/stis spectrum alone and then for the combined _ fuse _ + _ hst/_stis spectrum . . we note that if we do not deredden the observed spectra , they are basically impossible to fit with our synthetic spectral models . the best solar composition wd model fit we found to the dereddened _ fuse _ spectrum of v794 aql assuming e(b - v)=0.1 has an effective temperature of 44,000k , a rotation rate of 200km@xmath17sec@xmath6 and a distance of 301pc . this model is shown in figure 4 and listed in table 5 . the wd model has some absorption features ( around 1120 - 1130 ) not present at all in the observed spectrum . in order to try and improve the fit , we decrease the abundances of c and si ( responsible for these absorption feature ) to 0.01 their solar value and kept all the other abundances solar . this provides a 10 percent reduction in the @xmath73 value ( table 5 ) . next we ran the fitting subroutine on the grid of accretion disk models and found that the fitting is slightly better for the disk than for the wd . the best fit accretion disk models are also listed in table 5 for the @xmath81 and @xmath82 cases . the mass accretion rate we obtained is rather large ( @xmath83yr@xmath6 ) and the distance is much larger than for a wd model , namely @xmath84pc . in figure 5 we show the best fit accretion disk model with @xmath82 and @xmath85yr@xmath6 . the inclination angle we obtained is 60 degrees . in fact for all the disk models we found in the present work , the best fit is obtained for the higher inclination . this comes from the fact that the slope of the continuum is better matched by a rather flat continuum , which is obtained naturally with a higher inclination . a flatter continuum can also be obtained by dereddening the spectrum assuming a higher e(b - v ) value , which is what we did next . we ran our fitting subroutine for the wd and accretion disk models assuming now e(b - v)=0.2 , and dereddening the _ fuse _ spectrum of v794 aql accordingly . we found that the @xmath73 value decreases by about 20 percent for all the models . however , now the best wd model has a temperature of 51,000k ( since the spectrum is now `` bluer '' ) and a distance of 200pc . the best wd model fit assuming low c and si abundances is presented in figure 6 ( see also table 5 ) . for the best accretion disk model we found a slightly higher mass accretion rate ( @xmath86yr@xmath6 ) than for the e(b - v)=0.1 case , and a distance @xmath87pc . here again the best fit is for the accretion disk models . the @xmath88 model ( figure 7 ) only marginally better than the @xmath89 model ( all the models are listed in table 5 ) . next we ran two - component ( disk+wd ) model fits to find whether the fitting can be improved , but we found that the least @xmath73 obtained for the two - component is never smaller than the least @xmath73 obtained for the disk alone . from the _ fuse _ spectrum alone , we find that the best fit model is that of an accretion disk with a high accretion rate , in agreement with the fact that v794 aql was caught in a rather high state . the distance inferred from the modeling is pretty large @xmath90pc , consistent with the assumption of a large reddening value ( 0.2 ) and with the rather low _ fuse _ flux for a high state . next we carried out exactly the same analysis but for the _ hst_/stis spectrum of v794 aql , assuming both e(b - v)=0.1 and e(b - v)=0.2 . for easy comparison all these results are also listed in table 5 . the main difference with the _ fuse _ results is that the temperature of the best wd model is 6,000k higher for the e(b - v)=0.1 case and 4,000k higher for the e(b - v)=0.2 case . the distance for the wd models is about the same , namely d=200 - 300pc . for the e(b - v)=0.1 case , the best accretion disk model ( figure 8) has a slightly smaller mass accretion rate ( @xmath91yr@xmath6 ) than for the _ fuse _ best fit which leads to about half the distance obtained for the _ fuse _ spectrum with the same reddening value . for the e(b - v)=0.2 case we obtained exactly the same best accretion disk solution as for the _ fuse _ spectrum . the best fit wd model ( for e(b - v)=0.2 ) is presented in figure 9 and the best fit accretion disk model ( for e(b - v)=0.2 ) is presented in figure 10 . note that for all the stis spectral fits the @xmath73 values for the disk models are significantly smaller than for the best wd models , namely : @xmath73=13.25 versus 21.65 for the e(b - v)=0.1 case , and @xmath73=9.63 versus 13.65 for the e(b - v)=0.2 case . this is an indication that from the stis spectrum alone , the accretion disk is unambiguously the best model fit , and that e(b - v)=0.2 might be the correct value for the reddening . here too the two - component ( disk+wd ) model did not lead to a better fit . in order to combine the _ fuse _ spectrum with the stis spectrum we check how well their fluxes match in the overlap region between about 1150and 1180 . at very short wavelengths stis is very noisy , and the longer wavelengths of _ fuse _ are represented only by one channel ( because of the worm ) and are , therefore , less reliable too . consequently , we are left only with the @xmath51 emission region and its immediate vicinity . we find that the spectra have the same flux level there and can therefore be combined together . again we ran our fitting subroutine , but now for the combined _ fuse_+stis spectrum of v794 aql . this time we found that the temperature for the best wd model fit is much lower than for the individual spectra , namely 30,000k for the e(b - v)=0.1 case and t=45,000k for the e(b - v)=0.2 case ( shown in figure 12 ) with a distance of only d=145 - 176pc . for the e(b - v)=0.1 case ( figure 11 ) , we found the same accretion disk solution as for the stis spectrum , while for the e(b - v)=0.2 case ( figure 13 ) we found the same accretion disk solution as for the _ fuse _ and stis spectra alone . again the lowest @xmath73 values were obtained for the accretion disk models . again the e(b - v)=0.2 solutions were better than the e(b - v)=0.1 solutions , and here too the two - component ( disk+wd ) model did not lead to a better fit . since the _ fuse _ and _ hst_/stis spectra were obtained in a relatively high state , we could not get much information on the wd itself . in order to try and gain some knowledge of the wd directly we decided to model the _ iue _ spectrum swp 15266 of v794 aql in a relatively low state . in this state we expect to see mainly the wd and therefore we model the spectrum with a @xmath89 wd to agree with the grid of accretion disk models . since the spectrum is pretty noisy , more than one temperature is obtained and to limit the number of solution we fixed the distance to d@xmath92200pc and d@xmath92600pc and dereddened the spectrum assuming e(b - v)=0.20 . the best fit model for the shortest distance consists of a wd alone with t=30,000k , with @xmath73= 5.17 and a distance of 244pc . for the larger distance the best fist consists of a wd plus and accretion disk model ( figure 14 ) . the wd has t=47,000k and the disk has @xmath93yr@xmath6 , i=41deg . this best model has a distance of 643pc and @xmath73=4.52 . in this model the wd contributes 56 percent of the flux and the accretion disk contributes the remaining 44 percent . if this temperature is correct , then we have to reject the wd models from table 5 with @xmath94 for e(b - v)=0.20 ; this would be the wd model for the combined _ fuse_+stis spectrum for ( e(b - v)=0.20 ) with a temperature of only 45,000k , as this would be inconsistent , namely the system during the higher state can not have a temperature lower than during quiescence ( implying @xmath95 ) . we have analyzed the fuv spectra of v794 aql observed in a relatively high state with _ fuse _ and _ hst_/stis and found evidence for the presence of a hot accretion disk accreting at a high rate . the spectra exhibit some broad emission lines ( @xmath96 , @xmath51 , @xmath24 , @xmath25 , @xmath26 and @xmath27 ) , a clear sign of a hot gas where a high rotational velocity in a disk might be responsible for the broadening of these lines . we were limited in our spectral analysis by the large number of unknown parameters ( distance , wd mass , inclination , reddening , wd temperature , .. ) and decided to limit the search of solution assuming @xmath97 . for the _ fuse _ , stis and _ fuse_+stis spectra of v794 aql , we found that the accretion disk model provides a better fit to the observed spectra , which is not surprising since the system was observed in a relatively high state when the contribution of the disk is dominant . the best fit model had a mass accretion rate of @xmath98yr@xmath6 and an inclination of 60 degrees , assuming a mass @xmath80 . the distance we obtained was @xmath99pc and in all the cases the least @xmath73 was obtained for e(b - v)=0.2 . we also found that v794 aql is moderately affected by the ism with a molecular hydrogen column density of @xmath39@xmath1 and an atomic hydrogen column density of @xmath40@xmath1 . pg wishes to thank the space telescope science institute for its kind hospitality . this research was partly based on observations made with the nasa - cnes - csa far ultraviolet spectroscopic explorer . _ fuse _ is operated for nasa by the johns hopkins university under nasa contract nas5 - 32985 . funding was provided by nasa _ fuse _ ( cycle 4 ) grant nng04gl45 g to villanova university ( p.godon ) , nasa _ fuse grant _ nng04gc97 g to the university of washington ( p.szkody ) , and nasa _ fuse _ grant nng04gl18 g ( p.e.barrett ) . additional support for this work was provided by nasa through grant number hst - ar-10657.01-a ( hst cycle 14 archival ) to villanova university ( p.godon ) and grant number hst - go-09724 to the university of washington ( p.szkody ) , from the space telescope science institute , which is operated by the association of universitie for research in astronomy , incorporated , under nasa contract nas5 - 26555 . ( a)@xcite ( b ) the visual magnitude of v794 aql usually ranges between @xmath110 and @xmath111 @xcite but it has also been observed in an extremely low state where the magnitude drops to @xmath112 @xcite . clcccl + + + telescope & dataset & date & t@xmath113(s ) & aperture & additional characteristics + + + iue & swp 15266 & 10/14/81 & 7200 & large & disp.=low , contin= 54 cts / s , bckgr=28 cts / s + iue & swp 28501 & 06/16/86 & 23400 & large & disp.=low , contin=118 cts / s , bckgr=73 cts / s + iue & swp 50754 & 05/13/94 & 7198 & large & disp.=low , contin=166 cts / s , bckgr=96 cts / s + iue & swp 14708@xmath114 & 08/10/81 & 3600 & large & disp.=low , contin=0 cts / s , bckgr=23 cts / s + iue & lwr 11782 & 15/10/81 & 3600 & large & disp.=low , contin=72 cts / s , bckgr34 cts / s + fuse & d1440101 & 05/13/04 & 13,333 & lwrs & mode = ttag , cent.wave.=959.998 + hst / stis & o8mz66010 & 08/28/03 & 830 & 0.2x0.2 & mode = accum , filter / grating = g140l + + + lccl + + line & wavelength & absorption@xmath116 & origin@xmath117 + identification & & emission & + @xmath118 & 944.50 & e & s + @xmath118 & 933.50 & e & s + @xmath119 & 964.63 & a & ism + @xmath51 & 977.02 & e & c , s + @xmath54 & 1020.70 & a & ism , s + @xmath96 & 1031.93 & e & c , s + @xmath96 & 1037.62 & e & c , s + @xmath120 & 1048.20 & a & ism + @xmath120 & 1066.66 & a & ism + @xmath121 & 1073.52 & a & s + @xmath122 & 1083.99 & a , e & c , ism + & 1084.56 & a , e & c , ism + & 1084.58 & a , e & c , ism + & 1085.53 & a , e & c , ism + & 1085.55 & a , e & c , ism + & 1085.70 & a , e & c , ism + @xmath123 & 1096.60 & a & s + @xmath124 & 1096.88 & a & ism + @xmath124 & 1125.45 & a & ism + @xmath119 & 1134.16 & a & ism + & 1134.42 & a & ism + & 1134.98 & a & ism + @xmath124 & 1144.94 & a & ism + @xmath125 & 1152.82 & a & ism + @xmath126 & 1168.61 & e & c + @xmath123 & 1172.55 & a & s + + ( 1 ) : e = emission , a = absorption . + ( 2 ) : c = contaminated by airglow , geocoronal or heliocoronal emission , s = system , ism = interstellar medium + the contamination is from geocoronal lines mainly : @xmath56 , @xmath127 , @xmath119 and @xmath122 , and scattered solar light , such as the sharp @xmath51 ( 977 ) and @xmath96 ( 1032 ) which are mainly seen here in the sic channels . llll + + line & wavelength & line & wavelength + identification & ( ) & identification & ( ) + @xmath128 blend & 946 - 948@xmath129 & l6r1 & 1024.99 + l13r1 & 955.06 & l5r0 & 1036.55 + l13p1 & 955.71 & l5r1 & 1037.15 + l12r0 & 962.98 & l5p1 & 1038.16 + l12r1 & 963.61 & l5r2 & 1038.69 + @xmath128 blend & 965.00 & l5p2 & 1040.37 + w2q1 & 966.09 & l5r3 & 1041.16 + w2q2 & 967.28 & l5p3 & 1043.50 + l10r0 & 981.44 & l4r0 & 1049.37 + l10r1 & 982.07 & l4r1 & 1049.96 + l10p1 & 982.84 & l4p1 & 1051.03 + l10r2 & 983.59 & l4r2 & 1051.50 + l10p2 & 984.86 & l4p2 & 1053.28 + w1r0+w1r1 & 985.60 & l4r3 & 1053.98 + w1r1 & 986.80 & l4p3 & 1056.47 + w1r2 & 987.97 & l3r0 & 1062.88 + l9r0 & 991.38 & l3r1 & 1063.46 + l9r1 & 992.01 & l3p1 & 1064.61 + l8r0 & 1001.82 & l3r2 & 1065.00 + l8r1 & 1002.45 & l3p2 & 1066.90 + l8p1 & 1003.29 & l3r3 & 1067.48 + l8r2 & 1003.98 & l3p3 & 1070.14 + w0r0+w0r1 & 1008.50 & l2r0 & 1077.14 + w0r2 & 1009.02 & l2r1 & 1077.70 + w0r1 & 1009.77 & l2p1 & 1078.93 + w0r2 & 1010.94 & l2r2 & 1079.23 + w0p2 & 1012.17 & l1r0 & 1092.20 + l7r0 & 1012.70 & l1r1 & 1092.73 + l7r1 & 1013.44 & l1p1 & 1094.05 + l7p1+w0p3 & 1014.50 & l1p3 & 1099.79 + l7r2 & 1014.97 & l0r0 & 1108.13 + l7p2 & 1016.46 & l0r1 & 1108.63 + l7r3 & 1017.42 & l0p1 & 1110.06 + l6r0 & 1024.37 & l0r2 & 1110.12 + + ccccccccc + @xmath76 & @xmath130 & log(@xmath8 ) & @xmath103 & @xmath73 & @xmath131&wd / disk&e(b - v ) & spectrum + ( @xmath62)&(1000k ) & ( @xmath62yr@xmath6)&(deg ) & & ( pc ) & & & + 0.90 & 44 & - & - & 0.02959 & 301 & wd & 0.10 & fuse + 0.90 & 44 & - & - & 0.02767 & 304 & wd@xmath114 & 0.10 & fuse + 0.80 & - & -8.5 & 60 & 0.02651 & 834 & disk & 0.10 & fuse + 1.03 & - & -9.0 & 60 & 0.02584 & 616 & disk & 0.10 & fuse + 0.90 & 51 & - & - & 0.02355 & 199 & wd & 0.20 & fuse + 0.90 & 51 & - & - & 0.02257 & 201 & wd@xmath114 & 0.20 & fuse + 0.80 & - & -8.0 & 60 & 0.02206 & 791 & disk & 0.20 & fuse + 1.03 & - & -8.5 & 60 & 0.02177 & 585 & disk & 0.20 & fuse + 0.90 & 50 & - & - & 21.65 & 281 & wd & 0.10 & stis + 0.80 & - & -9.0 & 60 & 13.70 & 420 & disk & 0.10 & stis + 1.03 & - & -9.5 & 60 & 13.25 & 311 & disk & 0.10 & stis + 0.90 & 55 & - & - & 13.65 & 201 & wd & 0.20 & stis + 0.80 & - & -8.0 & 60 & 9.63 & 794 & disk & 0.20 & stis + 1.03 & - & -8.5 & 60 & 9.82 & 589 & disk & 0.20 & stis + 0.90 & 30 & - & & 10.9 & 145 & wd & 0.10 & fuse+stis + 0.80 & - & -9.0 & 60 & 5.22 & 451 & disk & 0.10 & fuse+stis + 1.03 & - & -9.5 & 60 & 4.97 & 333 & disk & 0.10 & fuse+stis + 0.90 & 45 & - & & 4.47 & 176 & wd & 0.20 & fuse+stis + 0.80 & - & -8.0 & 60 & 3.10 & 791 & disk & 0.20 & fuse+stis + 1.03 & - & -8.5 & 60 & 3.22 & 585 & disk & 0.20 & fuse+stis + 0.80 & 47 & -9.5 & 41 & 4.52 & 643 & 56/44 & 0.20 & iue + 0.80 & 30 & - & - & 5.17 & 244 & wd & 0.20 & iue + fig.1 - the _ fuse _ , _ hst_/stis and _ iue _ spectra of v794 aql with line identification in the stis and _ iue _ range . _ fuse _ is in orange , stis is in black , the _ iue _ swp 50754 is in blue , swp 28501 is in red and swp 15266 is in green . the _ fuse _ spectrum has been binned to 0.5here . the flux ( y - axis ) is given ergs@xmath17s@xmath6@xmath1@xmath6 and the wavelength ( x - axis ) is given in . the sharp emission lines in the _ fuse _ spectrum and the strong ly@xmath21 emission in the _ iue _ spectra are all due to air glow . line identification for _ fuse _ is provided in figure 3 and tables 3 & 4 . fig.2 - the _ iue _ swp 28501 and lwr 11782 spectra of v794 aql have been combined together and binned at 20 following the procedure of @xcite . the ab magnitude is shown as a function of wavelengths ( ) for the dereddened spectra assuming different value of e(b - v ) . the broad absorption feature between 2000 - 2400 is clearly seen in the lower graph , a sign that e(b - v)@xmath140 . fig.3 - the _ fuse _ spectrum of v794 aql with line identification . the flux ( y - axis ) is in erg@xmath17s@xmath6@xmath1@xmath6 and the wavelength ( x - axis ) is in . the airglow ( sharp ) emission lines have been annotated with a circle with a cross ; on top of the o and c ( @xmath20977 ) broad emission from the source , there is a contribution from helio - coronal emission ; the helium and hydrogen emission from the airglow have been annotated under the x - axis for clarity . the short vertical lines without annotation are the molecular hydrogen absorption lines due to the ism . the nitrogen i and ii are contaminated with airglow and ism absorption lines . the spectrum here has not yet been dereddened . fig.4 - the best synthetic spectral fit ( solid black ) using a wd stellar atmosphere model ( with solar composition ) to the dereddened _ fuse _ spectrum ( in red and blue ) of v794 aql assuming e(b - v)=0.1 . the wd has a temperature @xmath132k , a projected rotation rate of @xmath133km@xmath17s@xmath6 , a mass @xmath63 , a distance of 301pc and a @xmath134=0.02959 . the synthetic spectrum has been multiplied by the transmission values of the ism model to match the ism absorption feature . the strong emission lines ( such as @xmath96 and @xmath135 ) , the ism absorption features and the airglow lines have all been masked and are shown here in blue . the dotted line shows the synthetic spectrum without the ism model . the fit is carried between the solid black line and the red portions of the observed spectrum . note in the synthetic spectrum the strong absorption features between 1120 and 1130 , which are not present in the observed spectrum . fig.5 - the best synthetic spectral fit using a disk model to the dereddened _ fuse _ spectrum of v794 aql assuming e(b - v)=0.1 . using the grid of models from wade and hubeny , we find the best fit is for a wd with a mass @xmath82 , a mass accretion rate is @xmath136yr@xmath6 an inclination of @xmath137deg and a distance of 616pc . the resulting reduced chi - square is @xmath134=0.02584 . fig.6 - the best synthetic spectral fit ( solid black ) using a wd stellar atmosphere model ( with low c and si abundances ) to the dereddened _ fuse _ spectrum of v794 aql assuming e(b - v)=0.2 . the wd has a temperature @xmath138k , a projected rotation rate of @xmath133km@xmath17s@xmath6 , a mass @xmath63 , a distance of 201pc and a @xmath134=0.02257 . note in the synthetic spectrum that the absorption features between 1120 and 1130have been reduced by decreasing the abundances of c and si to better match the observed spectrum . fig.7 - the best synthetic spectral fit using a disk model to the dereddened _ fuse _ spectrum of v794 aql assuming e(b - v)=0.2 . the best fit is for a wd with a mass @xmath82 , a mass accretion rate is @xmath139yr@xmath6 an inclination of @xmath137deg and a distance of 585pc . the resulting reduced chi - square is @xmath134=0.02177 . fig.8 - the best synthetic spectral fit ( solid black ) to the dereddened stis spectrum ( in red and blue ) of v794 aql assuming e(b - v)=0.1 . here too the masked portions are shown in blue and the synthetic spectrum without the ism model is shown with the dashed line . the best fit consist of an accretion disk model with @xmath140yr@xmath6 , around a wd with a mass @xmath82 , an inclination of @xmath137deg and a distance of 311pc . the resulting reduced chi - square is @xmath134=13.25 . fig.9 - the best synthetic spectral fit ( solid black ) using a wd stellar atmosphere model ( solar abundances ) to the dereddened stis spectrum of v794 aql assuming e(b - v)=0.2 . the wd has a temperature @xmath141k , a projected rotation rate of @xmath133km@xmath17s@xmath6 , a mass @xmath63 , a distance of 201pc and a @xmath134=13.65 . fig.10 - the best synthetic spectral fit to the dereddened stis spectrum of v794 aql assuming e(b - v)=0.2 . the best fit consist of an accretion disk model with @xmath142yr@xmath6 , around a wd with a mass @xmath143 , an inclination of @xmath137deg and a distance of 794pc . the resulting reduced chi - square is @xmath134=9.63 . fig.11 - the best synthetic spectral fit ( solid black ) to the dereddened combined ( _ fuse_+stis ) spectrum ( in red and blue ) of v794 aql assuming e(b - v)=0.1 . here too the masked portions are shown in blue and the synthetic spectrum without the ism model is shown with the dashed line . the best fit consist of an accretion disk model with @xmath140yr@xmath6 , around a wd with a mass @xmath82 , an inclination of @xmath137deg and a distance of 333pc . the resulting reduced chi - square is @xmath134=4.97 . fig.12 - the best synthetic spectral fit using a wd stellar atmosphere model ( solar abundances ) to the dereddened combined ( _ fuse_+stis ) spectrum of v794 aql assuming e(b - v)=0.2 . the wd has a temperature @xmath144k , a projected rotation rate of @xmath133km@xmath17s@xmath6 , a mass @xmath63 , a distance of 176pc and a @xmath134=4.47 . fig.13 - the best synthetic spectral fit ( solid black ) to the dereddened combined ( _ fuse_+stis ) spectrum of v794 aql assuming e(b - v)=0.2 . the best fit consist of an accretion disk model with @xmath142yr@xmath6 , around a wd with a mass @xmath143 , an inclination of @xmath137deg and a distance of 791pc . the resulting reduced chi - square is @xmath134=3.10 . fig.14 the best fit model ( in red ) to the dereddened _ iue _ spectrum swp 15266 of v794 aql ( assuming e(b - v)=0.20 ) in a relatively low state . the best fit consist of a wd plus an accretion disk model . the wd has a mass @xmath81 wd to agree with the grid of accretion disk models of wade & hubeny . the best fit wd temperature for this model is 47,000k , the mass accretion rate is @xmath140yr@xmath6 , the inclination @xmath145deg and the distance 643pc . the resulting reduced chi - square is @xmath134=4.52 . in this model the wd ( in blue ) contributes to 56% of the flux and the disk ( in green ) contributes the remaining 44% .

the control design problem is considered for nonlinear systems with unknown internal system model . it is known that the nonlinear control problem can be transformed into solving the so - called hamilton - jacobi - isaacs ( hji ) equation , which is a nonlinear partial differential equation that is generally impossible to be solved analytically . even worse , model - based approaches can not be used for approximately solving hji equation , when the accurate system model is unavailable or costly to obtain in practice . to overcome these difficulties , an off - policy reinforcement leaning ( rl ) method is introduced to learn the solution of hji equation from real system data instead of mathematical system model , and its convergence is proved . in the off - policy rl method , the system data can be generated with arbitrary policies rather than the evaluating policy , which is extremely important and promising for practical systems . for implementation purpose , a neural network ( nn ) based actor - critic structure is employed and a least - square nn weight update algorithm is derived based on the method of weighted residuals . finally , the developed nn - based off - policy rl method is tested on a linear f16 aircraft plant , and further applied to a rotational / translational actuator system . shell : bare demo of ieeetran.cls for journals control design ; reinforcement learning ; off - policy learning ; neural network ; hamilton - jacobi - isaacs equation .

1311.6107

learning ( rl ) is a machine learning technique that has been widely studied from the computational intelligence and machine learning scope in the artificial intelligence community @xcite . rl technique refers to an actor or agent that interacts with its environment and aims to learn the optimal actions , or control policies , by observing their responses from the environment . in @xcite , sutton and barto suggested a definition of rl method , i.e. , any method that is well suited for solving rl problem can be regarded as a rl method , where the rl problem is defined in terms of optimal control of markov decision processes . this obviously established the relationship between the rl method and control community . moreover , rl methods have the ability to find an optimal control policy from unknown environment , which makes rl a promising method for control design of real systems . in the past few years , many rl approaches @xcite have been introduced for solving the optimal control problems . especially , some extremely important results were reported by using rl for solving the optimal control problem of discrete - time systems @xcite . such as , liu and wei suggested a finite - approximation - error based iterative adaptive dynamic programming approach @xcite , and a novel policy iteration ( pi ) method @xcite for discrete - time nonlinear systems . for continuous - time systems , murray et al . @xcite presented two pi algorithms that avoid the necessity of knowing the internal system dynamics . vrabie et al . @xcite introduced the thought of pi and proposed an important framework of integral reinforcement learning ( irl ) . modares et al . @xcite developed an experience replay based irl algorithm for nonlinear partially unknown constrained - input systems . in @xcite , an online neural network ( nn ) based decentralized control strategy was developed for stabilizing a class of continuous - time nonlinear interconnected large - scale systems . in addition , it worth mentioning that the thought of rl methods have also been introduced to solve the optimal control problem of partial differential equation systems @xcite . however , for most of practical real systems , the existence of external disturbances is usually unavoidable . to reduce the effects of disturbance , robust controller is required for disturbance rejection . one effective solution is the @xmath1 control method , which achieves disturbance attenuation in the @xmath2-gain setting @xcite , that is , to design a controller such that the ratio of the objective output energy to the disturbance energy is less than a prescribed level . over the past few decades , a large number of theoretical results on nonlinear @xmath1 control have been reported @xcite , where the @xmath1 control problem can be transformed into how to solve the so - called hamilton - jacobi - isaacs ( hji ) equation . however , the hji equation is a nonlinear partial differential equation ( pde ) , which is difficult or impossible to solve , and may not have global analytic solutions even in simple cases . thus , some works have been reported to solve the hji equation approximately @xcite . in @xcite , it was shown that there exists a sequence of policy iterations on the control input such that the hji equation is successively approximated with a sequence of hamilton - jacobi - bellman ( hjb)-like equations . then , the methods for solving hjb equation can be extended for the hji equation . in @xcite , the hjb equation was successively approximated by a sequence of linear pdes , which were solved with galerkin approximation in @xcite . in @xcite , the successive approximation method was extended to solve the discrete - time hji equation . similar to @xcite , a policy iteration scheme was developed in @xcite for the constrained input system . for implementation purpose of this scheme , a neuro - dynamic programming approach was introduced in @xcite and an online adaptive method was given in @xcite . this approach suits for the case that the saddle point exists , thus a situation that the smooth saddle point does not exist was considered in @xcite . in @xcite , a synchronous policy iteration method was developed , which is the extension of the work @xcite . to improve the efficiency for computing the solution of hji equation , luo and wu @xcite proposed a computationally efficient simultaneous policy update algorithm ( spua ) . in addition , in @xcite the solution of the hji equation was approximated by the taylor series expansion , and an efficient algorithm was furnished to generate the coefficients of the taylor series . it is observed that most of these methods @xcite are model - based , where the full system model is required . however , the accurate system model is usually unavailable or costly to obtain for many practical systems . thus , some rl approaches have been proposed for @xmath1 control design of linear systems @xcite and nonlinear systems @xcite with unknown internal system model . but these methods are on - policy learning approaches @xcite , where the cost function should be evaluated by using system data generated with policies being evaluated . it is found that there are several drawbacks ( to be discussed in section [ sec_3 ] ) to apply the on - policy learning to solve real @xmath1 control problem . to overcome this problem , this paper introduces an off - policy rl method to solve the nonlinear continuous - time @xmath1 control problem with unknown internal system model . the rest of the paper is rearranged as follows . sections [ sec_2 ] and [ sec_3 ] present the problem description and the motivation . the off - policy learning methods for nonlinear systems and linear systems are developed in [ sec_4 ] and [ sec_5 ] respectively . the simulation studies are conducted in section [ sec_6 ] and a brief conclusion is given in section [ sec_7 ] . _ notations _ : @xmath3 and @xmath4 are the set of real numbers , the @xmath5-dimensional euclidean space and the set of all real matrices , respectively . @xmath6 denotes the vector norm or matrix norm in @xmath7 or @xmath4 , respectively . the superscript @xmath8 is used for the transpose and @xmath9 denotes the identify matrix of appropriate dimension . @xmath10 denotes a gradient operator notation . for a symmetric matrix @xmath11 means that it is a positive ( semi - positive ) definite matrix . @xmath12 for some real vector @xmath13 and symmetric matrix @xmath14 with appropriate dimensions . @xmath15 is function space on @xmath16 with first derivatives are continuous . @xmath17 is a banach space , for @xmath18 . let @xmath19 and @xmath20 be compact sets , denote @xmath21 . for column vector functions @xmath22 and @xmath23 , where @xmath24 define inner product @xmath25 and norm @xmath26 @xmath27 . @xmath28 is a sobolev space that consists of functions in space @xmath29 such that their derivatives of order at least @xmath30 are also in @xmath29 . let us consider the following affine nonlinear continuous - time dynamical system : \(t ) & = f(x(t ) ) + g(x(t))u(t ) + k(x(t))w(t ) [ eq_sys ] + z(t ) & = h(x ) [ eq_h(x ) ] where @xmath31^t \in \mathcal{x } \subset \mathbb{r}^n$ ] is the state , @xmath32^t \in \mathcal{u } \subset \mathbb{r}^m$ ] is the control input and @xmath33 , @xmath34^t \in \mathcal{w } \subset \mathbb{r}^q$ ] is the external disturbance and @xmath35 , @xmath36^t \in \mathbb{r}^p$ ] is the objective output . @xmath37 is lipschitz continuous on a set @xmath38 that contains the origin , @xmath39 . @xmath37 represents the internal system model which is assumed to be _ unknown _ in this paper . @xmath40 and @xmath41 are known continuous vector or matrix functions of appropriate dimensions . the @xmath1 control problem under consideration is to find a state feedback control law @xmath42 such that the system is closed - loop asymptotically stable , and has @xmath2-gain less than or equal to @xmath43 , that is , @xmath44 for all @xmath45 and @xmath46 is some prescribed level of disturbance attenuation . from @xcite , this problem can be transformed to solve the so - called hji equation , which is summarized in lemma [ lemma_1 ] . [ lemma_1 ] assume the system and is zero - state observable . for @xmath46 , suppose there exists a solution @xmath47 to the hji equation [ hji ] & [ v^*(x)]^t f(x ) + h^t(x)h(x ) + & - [ v^*(x)]^t g(x ) r^-1 g^t(x ) v^*(x ) + & + [ v^*(x)]^t k(x ) k^t(x ) v^*(x ) = 0 . where @xmath48 and @xmath49 . then , the closed - loop system with the state feedback control @xmath50 has @xmath2-gain less than or equal to @xmath43 , and the closed - loop system , and ( when @xmath51 ) is locally asymptotically stable . from lemma [ lemma_1 ] , it is noted that the @xmath0 control relies on the solution of the hji equation . therefore , a model - based iterative method was proposed in @xcite , where the hji equation is successively approximated by a sequence of linear pdes : @xmath53^t ( f + gu^{(i ) } + kw^{(i , j ) } ) + h^th + \vert u^{(i ) } \vert _ r^2 \nonumber \\ - \gamma^2 \vert w^{(i , j ) } \vert ^2=0\end{aligned}\ ] ] and then update control and disturbance policies with w^(i , j+1 ) & ^-2 k^t v^(i , j+1)[eq_3.2 ] + u^(i+1 ) & - r^-1 g^t v^(i+1)[eq_3.3 ] with @xmath54 . from @xcite , it was indicated that the @xmath55 can converge to the solution of the hji equation , i.e. , @xmath56 . [ remark1 ] _ note that the key point of the iterative scheme - depends on the solution of the linear pde . thus , several related methods were developed , such as , galerkin approximation @xcite , synchronous policy iteration @xcite , neuro - dynamic programming approach @xcite and online adaptive control method @xcite for constrained input systems , and galerkin approximation method for discrete - time systems @xcite . obviously , the iteration - will generate two iterative loops since the control and disturbance policies are updated at the different iterative steps , i.e. , the inner loop for updating disturbance policy @xmath57 with index @xmath58 , and the outer iterative loop for updating control policy @xmath59 with index @xmath60 . the outer loop will not be activated until the inner loop is convergent , which results in low efficiency . therefore , luo and wu @xcite proposed a simultaneous policy update algorithm ( spua ) , where the control and disturbance policies are updated at the same iterative step , and thus only one iterative loop is required . it worth noting that the word simultaneous " in @xcite and the word synchronous / simultaneous " in @xcite represent different meanings . the former emphasizes the same iterative step, while the latter emphasizes the same time instant " . in other words , the spua in @xcite updates control and disturbance policies at the same " iterative step , while the algorithms in @xcite update control and disturbance policies at the different " iterative steps . @xmath61 _ the procedure of model - based spua is given in algorithm [ algorithm_1 ] . ' '' '' [ algorithm_1 ] ' '' '' it worth noting that algorithm [ algorithm_1 ] is an infinite iterative procedure , which is used for theoretical analysis rather than for implementation purpose . that is to say , algorithm [ algorithm_1 ] will converge to the solution of the hji equation when the iteration goes to infinity . by constructing a fixed point equation , the convergence of algorithm [ algorithm_1 ] is established in @xcite by proving that it is essentially a newtons iteration method for finding the fixed point . with the increase of index @xmath60 , the sequence @xmath62 obtained by the spua with equations - can converge to the solution of hji equation , i.e. , @xmath63 . [ remark2a ] _ it is necessary to explain the rationale of using equations and for control and disturbance policies update . the @xmath1 control problem - can be viewed as a two - players zero - sum differential game problem @xcite . the game problem is a minimax problem , where the control policy @xmath64 acts as the minimizing player and the disturbance policy @xmath65 is the maximizing player . the game problem aims at finding the saddle point @xmath66 , where @xmath67 is given by expression and @xmath68 is given by @xmath69 . correspondingly , for the @xmath1 control problem - , @xmath67 and @xmath68 are the associated @xmath1 control policy and the worst disturbance signal @xcite , respectively . thus , it is reasonable using expressions and ( that are consistent with @xmath67 and @xmath68 in form ) for control and disturbance policies update . similar control and disturbance policy update method could be found in references @xcite . @xmath61 _ observe that both iterative equations and require the full system model . for the @xmath0 control problem that the internal system dynamic @xmath37 is unknown , data based methods @xcite were suggested to solve the hji equation online . however , most of related existing online methods are on - policy learning approaches @xcite . from the definition of on - policy learning @xcite , the cost function should be evaluated with the data generated from the evaluating policies . for example , @xmath70 in equation is the cost function of the policies @xmath71 and @xmath72 , which means that @xmath70 should be evaluated with system data by using evaluating policies @xmath71 and @xmath72 . it is observed that these on - policy learning approaches for solving the @xmath0 control problem have several drawbacks : * \1 ) for real implementation of on - policy learning methods @xcite , the approximate evaluating control and disturbance policies ( rather than the actual policies ) are used to generate data for learning their cost function . in other words , the on - policy learning methods using the inaccurate " data to learn their cost function , which will increase the accumulated error . for example , to learn the cost function @xmath70 in equation , some approximate policies @xmath73 and @xmath74 ( rather than its actual policies @xmath71 and @xmath72 , which are usually unknown because of estimate error ) are employed to generate data ; * \2 ) the evaluating control and disturbance policies are required to generate data for on - policy learning , thus disturbance signal should be adjustable , which is usually impractical for most of real systems ; * \3 ) it is known @xcite that the issue of exploration " is extremely important in rl for learning the optimal control policy , and the lack of exploration during the learning process may lead to divergency . nevertheless , for on - policy learning , exploration is restricted because only the evaluating policies can be used to generate data . from the literature investigation , it is found that the exploration " issue is rarely discussed in existing work that using rl techniques for control design ; * \4 ) the implementation structure is complicated , such as in @xcite , three nns are required for approximating cost function , control and disturbance policies , respectively ; * \5 ) most of existing approaches @xcite are implemented online , thus they are difficult for real - time control because the learning process is often time - consuming . furthermore , online control design approaches just use current data while discard past data , which implies that the measured system data is used only once and thus results in low utilization efficiency . to overcome the drawbacks mentioned above , we propose an off - policy rl approach to solve the @xmath0 control problem with unknown internal system dynamic @xmath37 . in this section , an off - policy rl method for @xmath0 control design is derived and its convergence is proved . then , a nn - based critic - actor structure is developed for implementation purpose . to derive the off - policy rl method , we rewrite the system as : @xmath75 + k [ w - w^{(i ) } ] .\ ] ] for @xmath76 . let @xmath77 be the solution of the linear pde , then taking derivative along the state of system yields , [ eq_4_2 ] & = [ v^(i+1)]^t ( f + gu^(i ) + kw^(i ) ) + & + [ v^(i+1)]^t g[u - u^(i ) ] + [ v^(i+1)]^t k [ w - w^(i ) ] . with the linear pde , conducting integral on both sides of equation in time interval @xmath78 $ ] and rearranging terms yield , & _ t^t+t [ v^(i+1)(x())]^t g(x())[u ( ) - u^(i)(x ( ) ) ] d + & + _ t^t+t [ v^(i+1)(x())]^t k(x ( ) ) [ w()-w^(i)(x ( ) ) ] d + & + v^(i+1)(x(t ) ) - v^(i+1)(x(t+t ) ) [ eq_4_3 ] & = _ t^t+t ( h^t(x())h(x ( ) ) + u^(i ) ( x ( ) ) _ r^2 . - ^2 w^(i ) ( x ( ) ) ^2 ) d it is observed from the equation that the cost function @xmath79 can be learned by using arbitrary input signals @xmath64 and @xmath65 , rather than the evaluating policies @xmath72 and @xmath80 . then , replacing linear pde in algorithm [ algorithm_1 ] with results in the off - policy rl method . to show its convergence , theorem [ theorem_4.1 ] establishes the equivalence between iterative equations and . [ theorem_4.1 ] let @xmath81 and @xmath82 . @xmath77 is the solution of equation iff @xmath83if and only if @xmath84 it is the solution of the linear pde , i.e. , equation is equivalent to the linear pde . * * from the derivation of equation , it is concluded that if @xmath77 is the solution of the linear pde , then @xmath77 also satisfies equation . to complete the proof , we have to show that @xmath77 is the unique solution of equation . the proof is by contradiction . before starting the contradiction proof , we derive a simple fact . consider [ eq_4_4 ] & _ t 0 _ t^t+t ( ) d + & = _ t 0 ( _ 0^t+t ( ) d - _ 0^t ( ) d ) + & = _ 0^t ( ) d + & = ( t ) . from , we have [ eq_4_5 ] & = _ t 0 + & = _ t 0 _ t^t+t [ v^(i+1)(x())]^t g(x ( ) ) + & d + & + _ t 0 _ t^t+t [ v^(i+1)(x())]^t k(x ( ) ) + & d + & - _ t 0 _ t^t+t d. by using the fact , the equation is rewritten as [ eq_4_6 ] & = [ v^(i+1)(x(t))]^t g(x(t))[u(t ) - u^(i)(x(t ) ) ] + & + [ v^(i+1)(x(t))]^t k(x(t ) ) [ w(t)-w^(i)(x(t ) ) ] + & - . suppose that @xmath85 is another solution of equation with boundary condition @xmath86 . thus , @xmath87 also satisfies equation , i.e. , [ eq_4_7 ] & = [ w(x(t))]^t g(x(t))[u(t ) - u^(i)(x(t ) ) ] + & + [ w(x(t))]^t k(x(t ) ) [ w(t)-w^(i)(x(t ) ) ] + & - . substituting equation from yields , [ eq_4_7a ] & ( v^(i+1)(x ) -w(x ) ) + & = ^t g(x)[u - u^(i)(x ) ] + & + ^t k(x ) [ w -w^(i)(x ) ] . this means that equation holds for @xmath76 . if letting @xmath88 , we have @xmath89 = 0.\ ] ] then , @xmath90 for @xmath91 , where @xmath92 is a real constant , and @xmath93 . thus,@xmath94 , i.e. , @xmath95 for @xmath91 . this completes the proof . @xmath61 [ remark3 ] _ it follows from theorem [ theorem_4.1 ] that the solution of equation is equivalent to equation , and thus the convergence of the off - policy rl is guaranteed , i.e. , the solution of the iterative equation will converge to the solution of hji equation as iteration step @xmath60 increases . different from the equation in algorithm [ algorithm_1 ] , the off - policy rl with equation uses system data instead of the internal system dynamic @xmath37 . hence , the off - policy rl can be regarded as a direct learning method for @xmath0 control design , which avoids the identification of @xmath37 . in fact , the information of @xmath37 is embedded in the measurement of system data . that is to say , the lack of knowledge about @xmath37 does not have any impact on the off - policy rl to obtain the solution of hji equation and the @xmath0 control policy . it worth pointing out that the equation is similar with the form of the irl @xcite , which is an important framework for control design of continuous - time systems . the irl in @xcite is an online optimal control learning algorithm for partially unknown systems . @xmath61 _ to solve equation for the unknown function @xmath77 based on system data , we develop a nn based actor - critic structure . from the well known high - order weierstrass approximation theorem @xcite , a continuous function can be represented by an infinite - dimensional linearly independent basis function set . for real practical application , it is usually required to approximate the function in a compact set with a finite - dimensional function set . we consider the critic nn for approximating the cost function on a compact set @xmath38 . let @xmath96^t$ ] be the vector of linearly independent activation functions for critic nn , where @xmath97 is the number of critic nn hidden layer neurons . then , the output of critic nn is given by @xmath98 for @xmath99 , where @xmath100^t$ ] is the critic nn weight vector . it follows from , and that the disturbance and control policies are given by : ^(i ) ( x ) & = - r^-1 g^t ( x ) ^t(x ) ^(i)[eq_4_10 ] + ^(i ) ( x ) & = ^-2 k^t ( x ) ^t(x ) ^(i ) [ eq_4_9 ] for @xmath99 , and @xmath101^t$ ] is the jacobian of @xmath102 . expressions and can be viewed as actor nns for the disturbance and control policies respectively , where @xmath103 and @xmath104 are the activation function vectors and @xmath105 is the actor nn weight vector . due to estimation errors of the critic and actor nns - , the replacement of @xmath106 and @xmath107 in the iterative equation with @xmath108 and @xmath109 respectively , yields the following residual error : [ eq_4_11 ] & ^(i ) ( x(t ) , u(t ) , w(t ) ) + & _ t^t+t [ u ( ) - ^(i)(x())]^t g^t(x ( ) ) ^t(x ( ) ) ^(i+1 ) d + & + _ t^t+t [ w ( ) -^(i)(x ( ) ) ] ^t k^t(x ( ) ) ^t(x ( ) ) ^(i+1 ) d + & + [ ( x(t ) ) - ( x(t+t))]^t ^(i+1 ) + & - _ t^t+t d + & = _ t^t+t u^t ( ) g^t(x ( ) ) ^t(x ( ) ) ^(i+1 ) d + & + _ t^t+t ( ^(i ) ) ^t ( x ( ) ) g(x ( ) ) r^-1 g^t(x ( ) ) + & ^t(x ( ) ) ^(i+1 ) d + & + _ t^t+t w^t ( ) k^t(x ( ) ) ^t(x ( ) ) ^(i+1 ) d + & - ^-2 _ t^t+t ( ^(i ) ) ^t ( x ( ) ) k(x ( ) ) k^t(x ( ) ) + & ^t(x ( ) ) ^(i+1 ) d + [ ( x(t ) ) - ( x(t+t))]^t ^(i+1 ) + & - _ t^t+t ( ^(i ) ) ^t ( x ( ) ) g(x ( ) ) r^-1 g^t(x ( ) ) + & ^t(x ( ) ) ^(i ) d + & + ^-2 _ t^t+t ( ^(i ) ) ^t ( x ( ) ) k(x ( ) ) k^t(x ( ) ) + & ^t(x ( ) ) ^(i ) d - _ t^t+t h^t(x())h(x ( ) ) d for notation simplicity , define _ ( x(t ) ) & ^t + _ g ( x(t ) ) & _ t^t+t ( x ( ) ) g(x ( ) ) r^-1 g^t(x ( ) ) + & ^t(x ( ) ) d + _ k ( x(t ) ) & _ t^t+t ( x ( ) ) k(x ( ) ) k^t(x ( ) ) + & ^t(x ( ) ) d + _ u ( x(t),u(t ) ) & _ t^t+t u^t ( ) g^t(x ( ) ) ^t(x ( ) ) d + _ w ( x(t),w(t ) ) & _ t^t+t w^t ( ) k^t(x ( ) ) ^t(x ( ) ) d + _ h ( x(t ) ) & _ t^t+t h^t(x())h(x ( ) ) d then , expression is rewritten as [ eq_4_12 ] & ^(i ) ( x(t ) , u(t ) , w(t ) ) + & = _ u ( x(t),u(t ) ) ^(i+1 ) + ( ^(i ) ) ^t _ g ( x(t))^(i+1 ) + & + _ w ( x(t),w(t ) ) ^(i+1 ) - ^-2 ( ^(i ) ) ^t _ k ( x(t ) ) ^(i+1 ) + & + _ ^(i+1 ) - ( ^(i ) ) ^t _ g ( x(t))^(i ) + & + ^-2 ( ^(i ) ) ^t _ k ( x(t ) ) ^(i ) - _ h ( x(t ) ) . for description convenience , expression is represented as a compact form @xmath110 where & ^(i)(x(t ) , u(t ) , w(t ) ) _ u ( x(t),u(t ) ) + ( ^(i ) ) ^t _ g ( x(t ) ) + & + _ w ( x(t),w(t ) ) - ^-2 ( ^(i ) ) ^t _ k ( x(t ) ) + _ + & ^(i)(x(t ) ) ( ^(i ) ) ^t _ g ( x(t))^(i ) + & - ^-2 ( ^(i ) ) ^t _ k ( x(t ) ) ^(i ) + _ h ( x(t ) ) . for description simplicity , denote @xmath111^t$ ] . based on the method of weighted residuals @xcite , the unknown critic nn weight vector @xmath112 can be computed in such a way that the residual error @xmath113 ( for @xmath114 ) of is forced to be zero in some average sense . thus , projecting the residual error @xmath113 onto @xmath115 and setting the result to zero on domain @xmath116 using the inner product , @xmath117 , i.e. , @xmath118 then , the substitution of into yields , @xmath119 where the notations @xmath120 and @xmath121 are given by + @xmath122 $ ] and @xmath123^t $ ] . + thus , @xmath124 can be obtained with [ eq_4_15 ] ^(i+1 ) = & ^-1 _ + & _ . the computation of inner products @xmath125 and @xmath126 involve many numerical integrals on domain @xmath116 , which are computationally expensive . thus , the monte - carlo integration method @xcite is introduced , which is especially competitive on multi - dimensional domain . we now illustrate the monte - carlo integration for computing @xmath125 . let @xmath127 , and @xmath128 be the set that sampled on domain @xmath116 , where @xmath129 is size of sample set @xmath130 . then , @xmath125 is approximately computed with [ eq_4_16 ] & _ + & = _ ( ^(i)(x , u , w ) ) ^t ^(i)(x , u , w ) d(x , u , w ) + & = _ m=1^m ( ^(i)(x_m , u_m , w_m ) ) ^t ^(i)(x_m , u_m , w_m ) + & = ( z^(i ) ) ^t z^(i ) where @xmath131^t $ ] . similarly , [ eq_4_17 ] & _ + & = _ m=1^m ( ^(i)(x_m , u_m , w_m ) ) ^t ^(i)(x_m ) + & = ( z^(i ) ) ^t ^(i ) where @xmath132^t $ ] . then , the substitution of and into yields , @xmath133^{-1 } \left ( z^{(i ) } \right)^t \eta^{(i)}.\ ] ] it is noted that the critic nn weight update rule is a least - square scheme . based on the update rule , the procedure for @xmath0 control design with nn - based off - policy rl is presented in algorithm [ algorithm_2 ] . [ remark4a ] _ in the least - square scheme , it is required to compute the inverse of matrix @xmath134 . this means that the matrix @xmath135 should be full column rank , which depends on the richness of the sampling data set @xmath136 and its size @xmath137 . to attain this goal in real implementation , it would be useful by increasing the size @xmath137 , starting from different initial states , and using rich input signals , such as random noises , sinusoidal function noises with enough frequencies . of course , it would be nice , if possible but is not a necessity , to use the persistent exciting input signals , while it is still a difficult issue @xcite that requires further investigation . in a word , the choices of rich input signals and the size @xmath137 are generally experience - based . @xmath61 _ ' '' '' [ algorithm_2 ] ' '' '' note that algorithm [ algorithm_2 ] has two parts : the first part is step 1 for data processing , i.e. , measure system data @xmath138 for computing @xmath139 and @xmath140 ; the second part is steps 2 - 4 for offline iterative learning the solution of the hji equation . [ remark4 ] _ from theorem [ theorem_4.1 ] , the proposed off - policy rl is mathematically equivalent to the model - based spua ( i.e. , algorithm [ algorithm_2 ] ) , which is proved to be a newton s method @xcite . hence , the off - policy rl have the same advantages and disadvantages as the newton s method . that is to say , the off - policy rl is a local optimization method , and thus there exists a problem that an initial critic nn weight vecotr @xmath141 should be given such that the initial solution @xmath142 locates in a neighbourhood @xmath143 of the hji equation . in fact , this problem also widely arises in many existing works for solving optimal and @xmath0 control problems of either linear or nonlinear systems through the observations from computer simulation , such as @xcite . till present , it is still a difficult issue for finding proper initializations or developing global approaches . there is no exception for the proposed off - policy rl algorithm , where the selection of initial weight vector @xmath144 is still experience - based and requires further investigation . @xmath61 _ algorithm [ algorithm_2 ] can be viewed as an off - policy learning method according to references @xcite , which overcomes the drawbacks mentioned in section [ sec_3 ] , i.e. , * \1 ) in the off - policy rl algorithm ( i.e. , algorithm [ algorithm_2 ] ) , the control @xmath64 and disturbance @xmath65 can be arbitrarily on @xmath145 and @xmath20 , where no error occurs during the process of generating data , and thus the accumulated error ( exists in the on - policy learning methods mentioned in section [ sec_3 ] ) can be reduced ; * \2 ) in the algorithm [ algorithm_2 ] , the control @xmath64 and disturbance @xmath65 can be arbitrarily on @xmath145 and @xmath20 , and thus disturbance @xmath65 does not required to be adjustable ; * \3 ) in the algorithm [ algorithm_2 ] , the cost function @xmath79 of control and disturbance policies @xmath146 can be evaluated by using system data generated with other different control and disturbance signals @xmath147 . thus , the obvious advantage of the developed off - policy rl method is that it can learn the cost function and control policy from system data that are generated according to a more exploratory or even random policies ; * \4 ) the implementation of algorithm [ algorithm_2 ] is very simple , in fact _ only one _ nn is required , i.e. , critic nn . this means that once the critic nn weight vector @xmath124 is computed via , the action nns for control and disturbance policies can be obtained based on and accordingly ; * \5 ) the developed off - policy rl method learns the @xmath1 control policy offline , which is then used for real - time control . thus , it is much more practical than online control design methods since less computational load will generate during real - time application . meanwhile , note that in algorithm [ algorithm_2 ] , once the terms @xmath139 and @xmath140 are computed with sample set @xmath130 ( i.e. , step 1 is finished ) , no extra data is required for learning the @xmath1 control policy ( in steps 2 - 4 ) . this means that the collected data set can be utilized repeatedly , and thus the utilization efficiency is improved compared to the online control design methods . [ remark5a ] _ observe that the experience replay based irl method @xcite can be viewed as an off - policy method based on its definition @xcite . there are three obvious differences between the method and the work of this paper . firstly , the method in @xcite is for solving the optimal control problem without external disturbance , while the off - policy rl algorithm in this paper is for solving the @xmath1 control problem with external disturbance . secondly , the method in @xcite is online adaptive control approach . the off - policy rl algorithm in this paper uses real system information , and learns the @xmath1 control policy by using an offline process . after the learning process is finished , the convergent control policy is employed for real system control . thirdly , the method in @xcite involves two nns ( i.e. , one critic nn and one actor nn ) for adaptive optimal control realization , while only one nn ( i.e. , critic nn ) is required in the algorithm of this paper.@xmath61 _ it is necessary to analyze the convergence of the nn - based off - policy rl algorithm . from theorem [ theorem_4.1 ] , the equation in the off - policy rl is equivalent to the linear pde , which means that the derived least - square scheme is essentially for solving the linear pde . in @xcite , a similar least - square method was suggested to solve the first order linear pde directly , wherein some theoretical results are useful for analyzing the convergence of the proposed nn - based off - policy rl algorithm . the following theorem [ theorem_4.2 ] is given to show the convergence of critic nn and actor nns . [ theorem_4.2 ] for @xmath148 , assume that @xmath149 is the solution of , the critic nn activation functions @xmath150 , @xmath151 are selected such that they are complete when @xmath152 , @xmath79 and @xmath153 can be uniformly approximated , and the set @xmath154 is linearly independent and complete for @xmath155 . then , & _ x |^(i+1)(x ) - v^(i+1)(x)| 0 [ eq_4_3.1 ] + & _ x |^(i+1)(x ) - v^(i+1)(x)| 0 [ eq_4_3.2 ] + & _ x |^(i+1)(x ) - u^(i+1)(x)| 0 [ eq_4_3.3 ] + & _ x |^(i+1)(x ) - w^(i+1)(x)| 0 [ eq_4_3.4 ] . * proof . * the proof procedure of the above results is very similar with that in reference @xcite , and thus some similar proof steps will be omitted for avoidance of repetition . to use the theoretical results in @xcite , we firstly prove the @xmath156 is linear independent by contradiction . assume this is not true , then there exists a vector @xmath157^t \neq 0 $ ] such that @xmath158 which means that for @xmath91 , & _ t^t+t _ l=1^l _ l _ l ( f+gu^(i)+kw^(i ) ) d + & = _ l d + & = _ l=1^l _ l [ _ l(x(t+t ) ) - _ l(x(t ) ) ] + & = _ l=1^l _ l ( x(t+t),x(t ) ) + & = 0 . this contradicts the fact that the set @xmath159 is linearly independent , which implies that the set @xmath156 is linear independent . + from theorem [ theorem_4.1 ] , @xmath79 is the solution of the linear pde . then , with the same procedure used in theorem 2 and corollary 2 of the reference @xcite , the results - can be proven . and the result can be proven in a similar way for . @xmath61 + the results - in theorem [ theorem_4.2 ] imply that the critic nn and actor nns are convergent . in the following theorem [ theorem_4.3 ] , we prove that the nn - based off - policy rl algorithm converges uniformly to the solution of the hji equation and the @xmath0 control policy . [ theorem_4.3 ] if the conditions in theorem [ theorem_4.2 ] hold , then , for @xmath160 , @xmath161 , when @xmath162 and @xmath163 , we have & _ x |^(i)(x ) - v^*(x)| < [ eq_4_3.5 ] + & _ x |^(i)(x ) - u^*(x)| < [ eq_4_3.6 ] + & _ x |^(i)(x ) - w^*(x)| < [ eq_4_3.7 ] . * proof . * by following the same proof procedures in theorems 3 and 4 in @xcite , the results - can be proven directly . similar to , the result can also be proven . [ remark7 ] _ the proposed off - policy rl method is to learn the solution of the hji equation and the @xmath1 control policy . it follows from theorem [ theorem_4.3 ] that the control policy @xmath74 designed by the off - policy rl will uniformly converge to the @xmath1 control policy @xmath67 . with the @xmath1 control policy , it is noted from lemma [ lemma_1 ] that the closed - loop system with @xmath51 is locally asymptotically stable . furthermore , it is observed from that for the closed - loop system with disturbance @xmath35 , the output @xmath164 is in @xmath17 @xcite , i.e. , the closed - loop system is ( bounded - input bounded - output ) stable.@xmath61 _ in this section , the developed nn - based off - policy rl method ( i.e. , algorithm [ algorithm_2 ] ) is simplified for linear @xmath1 control design . consider the linear system : \(t ) = & ax(t ) + b_2u(t ) + b_1w(t ) [ eq_5.1 ] + z(t ) = & cx [ eq_5.2 ] where @xmath165 , @xmath166 , @xmath167 and @xmath168 . then , the hji equation of the linear system and results in an algebraic riccati equation ( are ) @xcite : @xmath169 where @xmath170 . if are has a stabilizing solution @xmath171 , the solution of the hji equation of the linear system and is @xmath172 , and then the linear @xmath1 control policy is accordingly given by @xmath173 consequently , @xmath174 , then the iterative equations - in algorithm [ algorithm_1 ] are respectively represented with u^(i ) = & -r^-1 b_2^t p^(i ) x [ eq_5.3 ] + w^(i ) = & ^-2 b_1^t p^(i ) x [ eq_5.4 ] @xmath175 where @xmath176 and @xmath177 @xmath178 @xmath179 . similar to the derivation of the off - policy rl method for nonlinear @xmath1 control design in section [ sec_4 ] , rewrite the linear system as @xmath180 + b_1 [ w - w^{(i ) } ] .\ ] ] based on equations - , the equation is given by [ eq_5.7 ] & _ t^t+t x^t ( ) p^(i+1 ) b_2 d + & + _ t^t+t x^t ( ) p^(i+1 ) b_1 d + & + [ x(t ) - x(t+t)]^t p^(i+1 ) [ x(t ) - x(t+t ) ] + & = _ t^t+t x^t ( ) ^(i ) x ( ) d where @xmath181 is a @xmath182 unknown matrix to be learned . for notation simplicity , define _ x(x(t ) ) & x(t ) - x(t+t ) + _ xx ( x(t ) ) & _ t^t+t x ( ) x()d + _ ux ( x(t),u(t ) ) & _ t^t+t u ( ) x()d + _ wx ( x(t),w(t ) ) & _ t^t+t w ( ) x()d where @xmath183 denotes kronecker product . each term of equation can be written as : & _ t^t+t x^t ( ) p^(i+1 ) b_2 u ( ) d + & = ^t_ux ( x(t),u(t ) ) ( b^t_2 i ) vec ( p^(i+1 ) ) + & _ t^t+t x^t ( ) p^(i+1 ) b_2 r^-1 b_2^t p^(i ) x ( ) d + & = ^t_xx ( x(t ) ) ( p^(i ) b_2 r^-1b^t_2 i ) vec ( p^(i+1 ) ) + & _ t^t+t x^t ( ) p^(i+1 ) b_1 w ( ) d + & = ^t_wx ( x(t),w(t ) ) ( b^t_1 i ) vec ( p^(i+1 ) ) + & ^-2 _ t^t+t x^t ( ) p^(i+1 ) b_1 b_1^t p^(i ) x ( ) d + & = ^-2 ^t_xx ( x(t ) ) ( p^(i ) b_1 b^t_1 i ) vec ( p^(i+1 ) ) + & [ x(t ) - x(t+t)]^t p^(i+1 ) [ x(t ) - x(t+t ) ] + & = ^t_x(x(t ) ) vec ( p^(i+1 ) ) + & _ t^t+t x^t ( ) ^(i ) x ( ) d= ^t_xx ( x(t ) ) vec ( ^(i ) ) where @xmath184 denotes the vectorization of the matrix @xmath185 formed by stacking the columns of @xmath185 into a single column vector . then , equation can be rewritten as @xmath186 with & ^(i)(x(t ) , u(t ) , w(t ) ) = ^t_ux ( x(t),u(t ) ) ( b^t_2 i ) + & + ^t_wx ( x(t),w(t ) ) ( b^t_1 i ) + ^t_x(x(t ) ) + & - ^-2 ( p^(i ) b_1 b^t_1 i ) ] + & ^(i)(x(t ) ) = ^t_xx ( x(t ) ) vec ( ^(i ) ) . it is noted that equation is equivalent to the equation with residual error @xmath187 . this is because no cost function approximation is required for linear systems . then , by collecting sample set @xmath130 for computing @xmath188 and @xmath189 , a more simpler least - square scheme can be derived to obtain the unknown parameter vector @xmath190 accordingly . at each iteration.,width=163 ] at each iteration.,width=163 ] in this section , the efficiency of the developed nn - based off - policy rl method is tested on a f16 aircraft plant . then , it is applied to the rotational / translational actuator ( rtac ) nonlinear benchmark problem . consider a f16 aircraft plant that used in @xcite , where the system dynamics is described by a linear continuous - time model : = & x + & + u+ w [ eq_6.1 ] + z = & x.[eq_6.2 ] where the system state vector is @xmath191^t $ ] , @xmath192 denotes the angle of attack , @xmath193 is the pitch rate and @xmath194 is the elevator deflection angle . the control input @xmath64 is the elevator actuator voltage and the disturbance @xmath65 is wind gusts on angle of attack . select @xmath195 and @xmath196 for the @xmath2-gain performance . then , solve the associated are with the matlab command care , we obtain @xmath197 . \nonumber\ ] ] weights at each iteration.,width=163 ] weights at each iteration.,width=163 ] for linear systems , the solution of the hji equation is @xmath172 , thus the complete activation function vector for critic nn is @xmath198^t$ ] of size @xmath199 . then , the idea critic nn weight vector is @xmath200^t $ ] @xmath201 @xmath202^t$ ] . letting initial critic nn weight @xmath203 , iterative stop criterion @xmath204 and integral time interval @xmath205 , algorithm [ algorithm_2 ] is applied to learn the solution of the are . to generate sample set @xmath130 , let sample size @xmath206 and generate random noise in interval [ 0,0.1 ] as input signals . figures [ fig1]-[fig2 ] give the critic nn weight @xmath144 at each iteration , in which the dash lines represent idea values of @xmath207 . it is observed from the figures that the critic nn weight vector converges to the idea values of @xmath207 at @xmath208 iteration . then , the efficiency of the developed off - policy rl method is verified . in addition , to test the influence of the parameter @xmath209 to algorithm [ algorithm_2 ] , we re - conduct simulation with different parameter cases : @xmath210 , and the results show that the critic nn weight vector @xmath144 still converges to the idea values of @xmath207 at @xmath208 iteration for all cases . this implies that the developed off - policy rl algorithm [ algorithm_2 ] is insensitive to the parameter @xmath209 . the rtac nonlinear benchmark problem @xcite has been widely used to test the abilities of control methods . the dynamics of this nonlinear plant poses challenges because the rotational and translation motions are coupled . the rtac system is given as follows : = & + u + & + w [ eq_simulation_11 ] + z = & i x[eq_simulation_12 ] where @xmath211 . for the @xmath2-gain performance , let @xmath195 and @xmath212 . of the closed - loop rtac system.,width=163 ] of the closed - loop rtac system.,width=163 ] then , the developed off - policy rl method is used to solve the nonlinear @xmath1 control problem of system and . select the critic nn activation function vector as @xmath213^t \nonumber\end{aligned}\ ] ] of size @xmath214 . with the initial critic nn weight @xmath215 , iterative stop criterion @xmath204 and integral time interval @xmath216 , algorithm [ algorithm_2 ] is applied to learn the solution of the hji equation . to generate sample set @xmath130 , let sample size @xmath217 and generate random noise in interval [ 0,0.5 ] as input signals . it is found that the critic nn weight vector converges fast to @xmath218^t \nonumber\end{aligned}\ ] ] at @xmath219 iteration . figures [ fig3]-[fig4 ] show first 10 critic nn weights ( i.e. , @xmath220 ) at each iteration . with the convergent critic nn weight vector @xmath221 , the @xmath1 control policy can be computed with . under the disturbance signal @xmath222 $ ] is a random number ) , closed - loop simulation is conducted with the @xmath1 control policy . figures [ fig5]-[fig7 ] give the trajectories of state and control policy . to show the relationship between @xmath2-gain and time , define the following ratio of disturbance attenuation as @xmath223 figure [ fig8 ] shows the curve of @xmath224 , where it converges to 3.7024@xmath225 as time increases , which implies that the designed @xmath1 control law can achieve a prescribed @xmath2-gain performance level @xmath43 for the closed - loop system . of the closed - loop rtac system.,width=163 ] of the closed - loop rtac system.,width=163 ] a nn - based off - policy rl method has been developed to solve the @xmath0 control problem of continuous - time systems with unknown internal system model . based on the model - based spua , an off - policy rl method is derived , which can learn the solution of hji equation from the system data generated by arbitrary control and disturbance signals . the implementation of the off - policy rl method is based on an actor - critic structure , where only one nn is required for approximating the cost function , and then a least - square scheme is derived for nn weights update . the effectiveness of the proposed nn - based off - policy rl method is tested on a linear f16 aircraft plant and a nonlinear rtac problem . j. j. murray , c. j. cox , g. g. lendaris , and r. saeks , `` adaptive dynamic programming , '' _ ieee transactions on systems , man , and cybernetics , part c : applications and reviews _ , vol . 32 , no . 2 , pp . 140153 , 2002 . v. yadav , r. padhi , and s. balakrishnan , `` robust / optimal temperature profile control of a high - speed aerospace vehicle using neural networks , '' _ ieee transactions on neural networks _ , vol . 18 , no . 4 , pp . 11151128 , 2007 . h. zhang , q. wei , and y. luo , `` a novel infinite - time optimal tracking control scheme for a class of discrete - time nonlinear systems via the greedy hdp iteration algorithm , '' _ ieee transactions on systems , man , and cybernetics , part b : cybernetics _ , vol . 38 , no . 4 , pp . 937942 , 2008 . h. zhang , y. luo , and d. liu , `` neural - network - based near - optimal control for a class of discrete - time affine nonlinear systems with control constraints , '' _ ieee transactions on neural networks _ 20 , no . 9 , pp . 14901503 , 2009 . h. zhang , l. cui , x. zhang , and y. luo , `` data - driven robust approximate optimal tracking control for unknown general nonlinear systems using adaptive dynamic programming method , '' _ ieee transactions on neural networks _ , vol . 22 , no . 12 , pp . 22262236 , 2011 . d. liu , d. wang , d. zhao , q. wei , and n. jin , `` neural - network - based optimal control for a class of unknown discrete - time nonlinear systems using globalized dual heuristic programming , '' _ ieee transactions on automation science and engineering _ , vol . 9 , no . 3 , pp . 628634 , 2012 . b. luo and h .- n . wu , `` approximate optimal control design for nonlinear one - dimensional parabolic pde systems using empirical eigenfunctions and neural network , '' _ ieee transactions on systems , man , and cybernetics , part b : cybernetics _ , vol . 42 , no . 6 , pp . 15381549 , 2012 . h .- n . wu and b. luo , `` heuristic dynamic programming algorithm for optimal control design of linear continuous - time hyperbolic pde systems , '' _ industrial & engineering chemistry research _ 51 , no . 27 , pp . 93109319 , 2012 . q. wei and d. liu , `` a novel iterative @xmath226-adaptive dynamic programming for discrete - time nonlinear systems , '' _ ieee transactions on automation science and engineering _ , doi : 10.1109/tase.2013.2280974 , online available , 2013 . d. liu , d. wang , and h. li , `` decentralized stabilization for a class of continuous - time nonlinear interconnected systems using online learning optimal control approach , '' _ ieee transactions on neural networks and learning systems _ , vol . 25 , no . 2 , pp . 418428 , 2014 . h. modares , f. l. lewis , and m .- b . naghibi - sistani , `` integral reinforcement learning and experience replay for adaptive optimal control of partially - unknown constrained - input continuous - time systems , '' _ automatica _ , vol . 50 , no . 1 , 193202 , 2014 . d. liu and q. wei , `` policy iteration adaptive dynamic programming algorithm for discrete - time nonlinear systems , '' _ ieee transactions on neural networks and learning systems _ , vol . 25 , no . 3 , pp . 621634 , 2014 . b. luo , h .- wu , and h .- x . li , `` data - based suboptimal neuro - control design with reinforcement learning for dissipative spatially distributed processes , '' _ industrial & engineering chemistry research _ , doi : 10.1021/ie4031743 , online available , 2014 . m. abu - khalaf , f. l. lewis , and j. huang , `` policy iterations on the hamilton isaacs equation for @xmath0 state feedback control with input saturation , '' _ ieee transactions on automatic control _ , vol . 51 , no . 12 , pp . 19891995 , 2006 . k. g. vamvoudakis and f. l. lewis , `` online solution of nonlinear two - player zero - sum games using synchronous policy iteration , '' _ international journal of robust and nonlinear control _ , vol . 22 , no . 13 , pp . 14601483 , 2012 . y. feng , b. anderson , and m. rotkowitz , `` a game theoretic algorithm to compute local stabilizing solutions to hjbi equations in nonlinear @xmath0 control , '' _ automatica _ , vol . 45 , no . 4 , pp . 881888 , 2009 . d. liu , h. li , and d. wang , `` neural - network - based zero - sum game for discrete - time nonlinear systems via iterative adaptive dynamic programming algorithm , '' _ neurocomputing _ , vol . 110 , no . 13 , pp . 92100 , 2013 . n. sakamoto and a. v. d. schaft , `` analytical approximation methods for the stabilizing solution of the hamilton jacobi equation , '' _ ieee transactions on automatic control _ , vol . 53 , no . 10 , pp . 23352350 , 2008 . r. beard , g. saridis , and j. wen , `` approximate solutions to the time - invariant hamilton bellman equation , '' _ journal of optimization theory and applications _ , vol . 96 , no . 3 , pp . 589626 , 1998 . s. mehraeen , t. dierks , s. jagannathan , and m. l. crow , `` zero - sum two - player game theoretic formulation of affine nonlinear discrete - time systems using neural networks , '' _ ieee transactions on cybernetics _ , vol . 43 , no . 6 , pp . 16411655 , 2013 . h. modares , f. l. lewis , and m .- b . n. sistani , `` online solution of nonquadratic two - player zero - sum games arising in the @xmath0 control of constrained input systems , '' _ international journal of adaptive control and signal processing _ , p. in press , 2014 . b. luo and h .- n . wu , `` computationally efficient simultaneous policy update algorithm for nonlinear @xmath1 state feedback control with galerkin s method , '' _ international journal of robust and nonlinear control _ 23 , no . 9 , pp . 9911012 , 2013 . h .- n . wu and b. luo , `` neural network based online simultaneous policy update algorithm for solving the hji equation in nonlinear @xmath1 control , '' _ ieee transactions on neural networks and learning systems _ , vol . 23 , no . 12 , pp . 18841895 , 2012 . h. zhang , l. cui , and y. luo , `` near - optimal control for nonzero - sum differential games of continuous - time nonlinear systems using single - network adp , '' _ ieee transactions on cybernetics _ , vol . 43 , no . 1 , pp . 206216 , 2013 . z. chen and s. jagannathan , `` generalized hamilton bellman formulation - based neural network control of affine nonlinear discrete - time systems , '' _ ieee transactions on neural networks _ , 19 , no . 1 , pp . 90106 , 2008 . g. escobar , r. ortega , and h. sira - ramirez , `` an @xmath227 disturbance attenuation solution to the nonlinear benchmark problem , '' _ international journal of robust and nonlinear control _ , vol . 8 , no . 4 - 5 , pp . 311330 , 1999 . biao luo received the b.e . degree in measuring and control technology and instrumentations and the m. e. degree in control theory and control engineering from xiangtan university , xiangtan , china , in 2006 and 2009 , respectively . he is now working for the ph . d. degree in control science and engineering with beihang university ( beijing university of aeronautics and astronautics ) , beijing , china . from february 2013 to august 2013 , he was a research assistant with the department of system engineering and engineering management ( seem ) , city university of hong kong , kowloon , hong kong . from september 2013 to december 2013 , he was a research assistant with department of mathematics and science , texas a&m university at qatar , doha , qatar . his current research interests include distributed parameter systems , optimal control , data - based control , fuzzy / neural modeling and control , hypersonic entry / reentry guidance , learning and control from big data , reinforcement learning , approximate dynamic programming , and evolutionary computation . huai - ning wu was born in anhui , china , on november 15 , 1972 . he received the b.e . degree in automation from shandong institute of building materials industry , jinan , china and the ph.d . degree in control theory and control engineering from xian jiaotong university , xian , china , in 1992 and 1997 , respectively . from august 1997 to july 1999 , he was a postdoctoral researcher in the department of electronic engineering at beijing institute of technology , beijing , china . in august 1999 , he joined the school of automation science and electrical engineering , beihang university ( formerly beijing university of aeronautics and astronautics ) , beijing . from december 2005 to may 2006 , he was a senior research associate with the department of manufacturing engineering and engineering management ( meem ) , city university of hong kong , kowloon , hong kong . from october to december during 2006 - 2008 and from july to august in 2010 , he was a research fellow with the department of meem , city university of hong kong . from july to august in 2011 and 2013 , he was a research fellow with the department of systems engineering and engineering management , city university of hong kong . he is currently a professor with beihang university . his current research interests include robust control , fault - tolerant control , distributed parameter systems , and fuzzy / neural modeling and control . wu serves as associate editor of the ieee transactions on systems , man & cybernetics : systems . he is a member of the committee of technical process failure diagnosis and safety , chinese association of automation . tingwen huang is a professor at texas a & m university at qatar . he received his b.s . degree from southwest normal university ( now southwest university ) , china , 1990 , his m.s . degree from sichuan university , china , 1993 , and his ph.d . degree from texas a & m university , college station , texas , 2002 . after graduated from texas a&m university , he worked as a visiting assistant professor there . then he joined texas a & m university at qatar ( tamuq ) as an assistant professor in august 2003 , then he was promoted to professor in 2013 . huang s focus areas for research interests include neural networks , chaotic dynamical systems , complex networks , optimization and control . he has authored and co - authored more than 100 refereed journal papers .

in this paper we adopt a chemical evolution model , which is an improved version of the chiappini , matteucci , & gratton ( 1997 ) model , assuming two main accretion episodes for the formation of the galaxy : the first one forming the halo and bulge in a short timescale followed by a second one that forms the thin - disk , with a timescale which is an increasing function of the galactocentric distance ( being of the order of 7 gyrs at the solar neighborhood ) . the present model takes into account in more detail than previously the halo density distribution and explores the effects of a threshold density in the star formation process , during both the halo and disk phases . the model also includes the most recent nucleosynthesis prescriptions concerning supernovae of all types , novae and single stars dying as white dwarfs . in the comparison between model predictions and available data , we have focused our attention on abundance gradients as well as gas , stellar and star formation rate distributions along the disk , since this kind of model has already proven to be quite successful in reproducing the solar neighborhood characteristics . we suggest that the mechanism for the formation of the halo leaves detectable imprints on the chemical properties of the outer regions of the disk , whereas the evolution of the halo and the inner disk are almost completely disentangled . this is due to the fact that the halo and disk densities are comparable at large galactocentric distances and therefore the gas lost from the halo can substantially contribute to building up the outer disk . we also show that the existence of a threshold density for the star formation rate , both in the halo and disk phase , is necessary to reproduce the majority of observational data in the solar vicinity and in the whole disk . in particular , a threshold in the star formation implies the occurrence of a gap in the star formation at the halo - disk transition phase , in agreement with recent data . we conclude that a relatively short halo formation timescale ( 0.8 gyr ) , in agreement with recent estimates for the age differences among galactic globular clusters , coupled with an `` inside - out '' formation of the galactic disk , where the innermost regions are assumed to have formed much faster than the outermost ones , represents , at the moment , the most likely explanation for the formation of the milky way . this scenario allows us to predict abundance gradients and other radial properties of the galactic disk in very good agreement with observations . moreover , as a consequence of the adopted `` inside - out '' scenario for the disk , we predict that the abundance gradients along the galactic disk must have increased with time and that the averagefe>$ ] ratios in stars ( halo plus disk ) slightly decrease going from 4 to 10 kpcs from the galactic centre . we also show that the same ratios increase substantially towards the outermost disk regions and the expected scatter in the stellar ages decreases , because the outermost regions are dominated by halo stars . more observations at large galactocentric distances are needed to test these predictions .

astro-ph0102134

most of the progress in understanding the formation and evolution of the milky way galaxy comes , on one side , from observations concerning chemical abundances in stars and gas and from the continuous improvement of stellar nucleosynthesis calculations on the other side . other important quantities relevant to the chemical evolution of the milky way , such as the star formation rate ( sfr ) , the initial mass function ( imf ) and gas flows , are still poorly constrained . however , a good model of chemical evolution can allow us to impose constraints on such quantities . in particular , a good chemical evolution model should be able to reproduce observables larger in number than the number of adopted free parameters . among the observables there are some which can help us to constrain the model better than others ( see matteucci 2000 ) . the best observables from the point of view of galactic chemical evolution are : the abundances in stars , in hii regions and planetary nebulae ( pne ) from which we derive the abundance gradients along the galactic disk , and the g - dwarf metallicity distribution in the solar neighborhood . in past years a great deal of theoretical work has appeared concerning the chemical evolution of the milky way ( matteucci & franois 1989 ; carigi 1994 , 1996 ; timmes et al . 1995 ; tsujimoto et al . 1995 ; chiappini et al . 1997 ( hereafter cmg97 ) , 1999 ; chang et al . 1999 ; portinari & chiosi 1999 , 2000 ; boissier & prantzos 1999 ; goswami & prantzos 2000 ; romano et al . 2000 ) and most of them dealt with the evolution of the galactic disk only . only a few models have attempted to model the galactic halo ( cmg97 ; samland et al . 1997 ; goswami & prantzos 2000 ) . in particular , cmg97 have suggested a scenario where the galaxy forms as a result of two main infall episodes . during the first episode the halo forms and the gas lost by the halo rapidly ( roughly 1 gyr ) accumulates in the center with the consequent formation of the bulge . during the second episode , a much slower infall of primordial gas gives rise to the disk with the gas accumulating faster in the inner than in the outer regions . in this scenario the formation of the halo and disk are almost completely dissociated although some halo gas falls into the disk . this mechanism for disk formation is known as `` inside - out '' scenario which is quite successful in reproducing the main features of the milky way ( cmg97 ) as well as of external galaxies especially concerning abundance gradients ( see prantzos & boissier , 2000 ) . in this paper we present a model similar to that of cmg97 and romano et al . ( 2000 ) but with an improved treatment of the halo evolution . the model is always based on the `` two - infall '' scenario but more attention is paid to the evolution of the halo and to the possibility of a density threshold for the sfr in the halo . in the cmg97 paper we did not investigate in detail the effects of a density threshold for the star formation in the halo as our main goal was to study the solar neighborhood . moreover , as will be shown in the next sections , the evolution of the solar neighborhood is almost independent of the enrichment history of the halo . it is worth recalling that a surface density threshold for star formation has been observed in a variety of objects , including normal spirals , starburst galaxies and low surface brightness galaxies ( e.g kennicutt 1989 , 1998 ; van der hulst et al . the existence of a threshold in spheroidal systems ( halos , bulges and ellipticals ) is not yet clear although there are theoretical arguments by elmegreen ( 1999 ) suggesting the existence of a threshold also in such systems . the threshold adopted by cmg97 produced results in very good agreement with the majority of the observational constraints in the solar neighborhood and in the whole disk . these include the most recent findings which suggest a discontinuity in the chemical properties between the halo ( and part of the thick - disk ) and the thin - disk ( e.g. pagel & tautvaisiene 1995 ; beers & sommer - larsen 1995 ; gratton et al . in particular , one of the main effects of the threshold was that it naturally produced a hiatus in the sfr between the end of the halo / thick - disk phase and the beginning of the thin - disk phase . such a hiatus seems to be real since it is observed both in the plot of [ fe / o ] versus [ o / h ] ( cmg97 ; gratton et al . 1996 , 2000 ; see also pagel 2000 for a discussion ) and in the plot of [ fe / mg ] versus [ mg / h ] ( fuhrmann 1998 ; gratton et al . 2000 ) . the evidence for this is shown by the steep increase of [ fe / o ] and [ fe / mg ] at a particular value of [ o / h ] and [ mg / h ] , respectively , indicating that at a certain epoch ( coinciding with the halo - disk transition ) sne ii , responsible for the production of o and mg , stopped exploding while fe , produced by the long living sne ia , continued to be produced . the star formation must have ceased for a period which can not be longer than @xmath3 1 gyr , as suggested by theoretical models ( cmg97 ; gratton et al . 2000 and the present paper ) . the `` two - infall '' model also allowed us to fit the observed metallicity distribution of the g - dwarfs by assuming a long timescale for the thin - disk formation . this long timescale for the thin - disk formation at the solar vicinity was then suggested also by more recent chemical evolution models ( e.g. prantzos & silk 1998 ; portinari , chiosi , & bressan 1998 ; chang et al . 1999 ; and more recently boissier & prantzos 1999 ; hou et al . the same result is also suggested by chemodynamical models ( samland et al . 1997 ; hensler 1999 ) . we not that in cmg97 we focused mainly on the solar vicinity modelling , and adopted very simplified assumptions to compute the radial profiles . we predicted non - linear gradients along the galactic disk and a tendency for them to be flatter in the outermost regions and almost constant with time , whereas in the innermost parts they were steeper and steepened with time . in particular , the gradients we obtained were flatter than the standard values presently adopted in the literature . for example , the gradient of oxygen was flatter than @xmath40.07 dex / kpc , which is the value suggested by different data sources ( b stars , hii regions and pne , see maciel & quireza 1999 ) . in this paper , we study in detail the formation of abundance gradients in the galactic disk by trying to fit at the same time the abundance , sfr , star and gas profiles in the milky way ( mw ) disk . clearly the evolution of gradients with time is strictly related to the mechanism of galaxy formation , including the formation of the halo and this is an important point to assess . for this purpose , we test several assumptions about the halo and disk formation and the existence and variation of a density threshold for the sfr both in the halo and disk . our results concern mostly the elements for which more data are available , such as n , o , s and fe . we believe that by improving the mw model we will provide a solid basis for a more detailed understanding of the history of chemical enrichment of our galaxy and other spirals . an important aspect of this work will be to provide specific predictions for galaxy sizes of mw - like galaxies as a function of redshift type ( in a forthcoming paper ) , which could in principle be tested against the growing body of observational data on distant galaxies . in particular , the slow timescale for the formation of the mw disk implied by our chemical evolution model implies that at high redshifts we should see smaller disks . some observational tests , using the predictions of cmg97 with the growth of the thin - disk timescale as a function of galactocentric distance , have already been done ( roche et al . 1998 ) . those authors estimate that galaxies at redshift 2 @xmath5 @xmath6 @xmath5 3.5 show 2.79 @xmath7 0.31 mag of surface brightness evolution relative to those at @xmath6 @xmath5 0.35 , which is significantly larger than the luminosity evolution over this redshift range . they suggest that this can be explained by a size and luminosity evolution model , in which the outer regions of spiral galaxies form later and with a longer timescale than the inner regions , causing the half - light radius to increase with time . although the interpretation of the observational data is still controversial ( see simard et al . 1999 ) , they clearly represent a powerful test for disk formation models . the paper is organized as follows : in section 2 the observational constraints are discussed ; section 3 presents the model assumptions and in section 4 the results are shown . some conclusions are drawn in section 5 . radial abundance profiles constitute one of the most important observational constraints for models of the evolution of the mw disk . however , over the past decade , different authors using different observational tools often came to contradictory views on both the shape , the magnitude and the evolution of the abundance gradients along the disk ( see tosi 1996 , 2000 for a discussion on this point ) . the controversial results originate from both theoretical and observational considerations . in particular , observations of b - stars seemed to rule out the existence of a gradient ( or to allow for only a very mild one ; gehren et al . 1985 ; fitzsimmons et al . 1992 ; kaufer et al . 1994 ; kilian - montenbruck et al . 1994 ) , in contrast to observations of hii regions ( shaver et al . 1983 ; fich & silkey 1991 ; simpson et al . 1995 ; afflerbach et al . 1997 ; rudolph et al . 1997 ) and pne of type ii ( maciel & chiappini 1994 ; maciel & kppen 1994 ; maciel & quireza 1999 ) . this controversy was solved by smartt & rolleston ( 1997 ) and gummersbach et al . ( 1998 ) who found , by means of the reanalysis of a sample of b - stars , a gradient similar to that derived from hii regions and pne . although the b - star discrepancy seems now to be solved , a new one appeared recently in the literature concerning the hii regions abundance gradients . deharveng et al . ( 2000 ) analysed a new sample of 34 hii regions located between 6.6 and 17.7 kpcs from the galactic center and after a careful estimate of the electron temperatures in those objects they obtained , using their best observations , an oxygen abundance gradient which is flatter ( by a factor of 2 ) than the one obtained in previous works based on hii regions . their result seems to be in good agreement with results by esteban et al . ( 1998 , 1999a , b ) . proposals for flatter gradients or bimodal ones have also been made by works based on open clusters ( friel 1999 and twarog et al . 1997 ) although the situation is still very unclear ( see carraro et al . observations of open clusters seem to support the existence of an iron gradient along the disk , but large uncertainties related to both the metallicity calibration and the ages of these objects are still present . another open question related to the observed abundance gradients concerns their variation with time : do the gradients steepen or flatten with time ? this question can not be answered properly by the presently available data ( see maciel 2000 for a discussion ) . however , pne are the most promising objects to solve this problem . as has been extensively discussed in the past few years , pne galactic distribution , kinematics , chemical composition and morphology clearly indicate that pne comprise objects of different populations ( peimbert 1978 ; maciel 1997 ) . previous work has shown that disk objects of type ii are particularly useful in tracing the chemical enrichment of the interstellar medium at the time of the formation of the pn progenitors ( maciel & kppen 1994 ; maciel & chiappini 1994 ) . in a recent work , maciel & quireza ( 1999 ) obtained the gradients for a sample that includes the objects studied by maciel & kppen ( 1994 ) , maciel & chiappini ( 1994 ) and costa et al . ( 1997 ) ( the latter consists of a sample of pne near the anticentre direction intended to derive a better estimate of the gradients at galactocentric distances larger than the solar position ) . their main conclusions were : _ i ) _ there is an average gradient of @xmath40.04 to @xmath40.07 dex / kpc for what they call `` inner '' galaxy ( between 4 and 10 kpc those authors assume @xmath8 = 7.6 kpc ) ; _ ii ) _ the gradients show a small variation for the different element ratios ( see their table 2 ) ; _ iii ) _ the pn gradients are generally slightly flatter than those derived from younger objects ; _ iv ) _ for larger galactocentric distances the pn gradients show some flattening . however the authors pointed out that _ a ) _ the precise region where the gradient flattens out is not well defined and larger samples are needed , particularly at @xmath9 @xmath10 12 kpc ; and _ b ) _ although the tendency of the gradients to steepen with time suggested by maciel & kppen ( 1994 ) seems to be confirmed by the more recent analysis , this is only an indicative result because of the difficulty of assigning precise ages to disk pne and of estimating the importance of dynamical effects which could have flattened the gradients of the older objects . from the theoretical point of view , several mechanisms have been proposed in order to explain the existence of abundance gradients in disk galaxies ( see matteucci & chiappini 1999 ) . among the alternatives , the so called `` biased - infall '' is the most popular one . in this approach the galactic disk formed by infall of gas occurring at a faster rate in the innermost regions than in the outermost ones ( `` inside - out '' formation picture ) . the physical reason for a `` biased - infall '' can be found in the fact that the gas tends to collapse more quickly in the center of the spheroid so that a gradient in the gas density , lower in the more external parts , is established . in this situation the gas which continues to infall towards the disk falls more efficiently towards the center than towards the external regions , owing to the stronger tidal force acting at the center ( see larson 1976 ) . several authors have shown that the `` biased - infall '' can well reproduce the steep abundance gradients along the disk , especially if coupled with a sfr proportional to some power @xmath11 of the gas density ( see section 3 ) with @xmath11 @xmath10 1 ( tosi 1988 ; matteucci & franois 1989 ) . however , the situation is still not clear also from a theoretical point of view . in fact , as pointed out in many recent papers ( maciel & quireza 1999 ; tosi 2000 ; portinari & chiosi 2000 ; deharveng et al . 2000 ; hou et al . 2000 and others ) , different authors can fit the solar vicinity constraints and even the present time abundance gradients , but they seem to be divided into two groups concerning the evolution of the abundance gradients : in some of them the gradients steepen with time ( tosi 1988 ; cmg97 ; samland et al . 1997 ) while in others the abundance gradients flatten with time ( allen et al . 1998 ; portinari & chiosi 2000 ; hou et al . 2000 ) . the lack of good data for the _ outer _ mw disk abundances still prevents us from testing this prediction and represents one of the main reasons for the non - uniqueness of the various chemical evolution models . table 1 summarizes the abundance data available to trace the gradients in our galaxy . in this work we present results for the galactic abundance gradients of n , o , s and fe as those are the elements for which more data exist . the observed ( h i + h@xmath12 ) distribution is taken from dame ( 1993 ) . the surface density distribution of the total gas @xmath13 is obtained from the sum of the h i and h@xmath12 distributions , @xmath14 , accounting for the helium and heavy elements fractions ( thick line in figure 4 ) . the stellar profile is exponentially decreasing outwards , with characteristic scale length @xmath15 @xmath3 2.53 kpc ( sackett 1997 and references therein ; freudenreich 1998 ) . moreover , cobe observations suggest that the stellar disk has an outer edge 4 kpc from the sun ( freudenreich 1998 ) . to compare our model predictions on the stellar density profile along the galactic disk to the observed one we consider @xmath16 = 35 @xmath7 5 @xmath17 pc@xmath18 ( gilmore et al . 1989 ) and @xmath15 = 2.5 kpc ( see figure 5 ) . the distributions of supernova remnants ( guibert , lequeux , & viallefond 1978 ) , of pulsars ( lyne , manchester , & taylor 1985 ) , of lyman - continuum photons ( gsten & mezger 1982 ) and of molecular gas ( rana 1991 and references therein ) all can be used to derive an estimate for the sfr along the galactic disk . since these observables can not directly yield the absolute sfr without further assumptions e.g. on the imf and mass ranges for producing pulsars , supernovae , etc . ( see lacey & fall 1985 ) , it is common practice to normalize them to their values at the solar radius , and then to trace the radial profile sfr(@xmath9)/sfr(@xmath8 ) ( the thick lines plotted in figure 6 refer to the upper and lower limits obtained from the observational data listed above ) . as we stated in the introduction , a good model for the chemical evolution of our galaxy should honour a minimal number of observational constraints both in the solar neighborhood and in the whole disk . the model we will adopt here is an updated version of the so called `` two - infall '' model of cmg97 . this model assumes that the galaxy forms out of two main accretion episodes almost completely dissociated . during the first episode , the primordial gas collapses very quickly and forms the spheroidal components ( halo and bulge ) ; during the second episode , the thin - disk forms , mainly by accretion of a significant fraction of matter of primordial chemical composition plus traces of halo gas . the disk is built - up in the framework of the `` inside - out '' scenario of galaxy formation ( larson 1976 ) which ensures the formation of abundance gradients along the disk ( matteucci & franois 1989 ) . as in matteucci & franois ( 1989 ) , the galactic disk is approximated by several independent rings , 2 kpc wide , without exchange of matter between them . the rate of accretion of matter in each shell is : @xmath19 where @xmath20 is the surface mass density of the infalling material , which is assumed to have primordial chemical composition ; @xmath21 is the time of maximum gas accretion onto the disk , coincident with the end of the halo / thick - disk phase and set here equal to 1 gyr ; @xmath22 and @xmath23 are the timescales for mass accretion onto the halo / thick - disk and thin - disk components , respectively . in particular , @xmath22 = 0.8 gyr and , according to the `` inside - out '' scenario , @xmath24 = 1.033@xmath25/kpc)@xmath41.267 gyr ( see romano et al . we adopt a linear approximation for the variation of @xmath24 . however , this will be best constrained by the observed radial profiles of gas , sfr and abundances . in some cases the observed radial profiles suggest that this timescale can approach a constant value for galactocentric distances larger than @xmath0 12 kpc ( see next section ) . the quantities @xmath26 and @xmath27 are derived from the condition of reproducing the current total mass surface density distribution in the halo and along the disk respectively at the present time . the sfr adopted here has the same formulation as in cmg97 : @xmath28 where @xmath29 is the efficiency of the star formation process , @xmath30 is the total mass surface density at a given radius @xmath9 and given time @xmath31 , @xmath32 is the total mass surface density at the solar position and @xmath33 is the gas surface density . note that the gas surface density exponent , @xmath11 , equal to 1.5 , was obtained from the best model of cmg97 in order to ensure a good fit to the observational constraints at the solar vicinity . this value is also in very good agreement with the recent observational results of kennicutt ( 1998 ) and with n - body simulation results by gerritsen & icke ( 1997 ) . the efficiency of star formation is set to @xmath34=1 gyr@xmath35 to ensure the best fit to the observational features in the solar vicinity , and becomes zero when the gas surface density drops below a certain critical threshold ( kennicutt 2000 ) . we adopt a threshold density @xmath36 @xmath3 7 @xmath17 pc@xmath18 in the disk ( cmg97 ) . as far as the halo / thick - disk phase is concerned , a similar value for the threshold is expected ( see elmegreen 1999 and the model a description below for details ) . the imf is that of scalo ( 1986 ) , assumed to be constant during the evolution of the galaxy ( see chiappini , matteucci & padoan 2000 ) . our present model differs from that of cmg97 in _ ( i ) _ the adopted yields for the low and intermediate mass range stars which are now taken from van den hoek & groenewegen ( 1997 ) instead of renzini & voli ( 1981 ) ( see romano et al . 2000 ) ; _ ( ii ) _ the fact that now we are including the explosive nucleosynthesis from nova outbursts ( see romano & matteucci 2000 ) and _ ( iii ) _ the adopted solar galactocentric distance ( @xmath8 = 8 @xmath7 0.5 kpc , reid 1993 ) and age of the galaxy ( @xmath37 = 14 gyr ) . moreover , in the present model we adopt a primordial @xmath38he abundance of 0.241 ( by mass ) instead of 0.23 as recently suggested by viegas , gruenwald , & steigman ( 2000 ) . the primordial abundances by mass of d and @xmath39he were taken to be 4.5 @xmath40 10@xmath41 and 2.0 @xmath40 10@xmath41 respectively ( see chiappini & matteucci 2000 ) . we run several models by varying the sfr threshold in the halo phase , the surface mass density distribution of the halo and the timescale for the halo formation . the present surface mass density distribution of the disk has been always kept the same and is exponential with scale length @xmath42 = 3.5 kpc normalized to @xmath43 = 54 @xmath17 pc@xmath18 ( see romano et al . 2000 for a discussion of the choice of these parameters ) . the models are labelled a , b , c , d and the model parameters are summarized in table 5 . in particular , model a assumes : @xmath44 the total halo mass density profile is constant and equal to 17 @xmath17 pc@xmath18 for @xmath9 @xmath45 8 kpc and decreases as @xmath46 outwards . we adopt the halo mass profile given by binney & tremaine ( 1987 ) where the density is proportional to @xmath47 and this implies a surface mass density profile proportional to @xmath46 ; the value of the total surface mass density of the halo at the solar position is quite uncertain , and we assumed it to be 17 @xmath17 pc@xmath18 from the fact that the total surface mass density ( halo plus disk ) in the solar vicinity is @xmath0 71 @xmath17 pc@xmath18 ( kuijken and gilmore 1991 ) , where @xmath0 54 @xmath17 pc@xmath18 corresponds to the disk total surface mass density ( see also a discussion in romano et al . 2000 ) . @xmath44 @xmath22 is constant and equal to 0.8 gyr along the overall galactic disk , and @xmath24 is an increasing function of @xmath9 in the range 414 kpc , as described above , and constant and equal to @xmath23(14 kpc ) for @xmath9 @xmath10 14 kpc . @xmath44 the thresholds in the gas density are 4 and 7 @xmath17 pc@xmath18 during the halo and disk phase , respectively . these choices are not arbitrary since the threshold for the disk agrees with the observational values found by kennicutt ( 1989 , 1998 ) . not much is known about the threshold in the halo and so this particular choice is a result of several tests and ensures a good fit to the main observational constraints . in particular , these choices for the density thresholds lead to a very good fit of the gap in the sfr during the halo - disk transition phase , as suggested by both the [ fe / o ] vs. [ o / h ] and [ fe / mg ] vs. [ mg / h ] relations , as well as of the radial profiles in the disk . a possible physical motivation for different star formation thresholds in the halo and the disk might be related to a metallicity effect . since the opacities are lower in the halo due to a lower metal content , this would make it easier for a protostellar cloud to cool and collapse and in this case we would expect a lower threshold density . moreover , if the threshold density is close to the tidal limit ( which is the average total density inside a radius r ) , then it would be higher in the inner regions than in the outer ones . consequently one would expect a higher threshold density in the bulge than in the disk outside the bulge and higher in the disk than in the remote halo that is further out from the center ( elmegreen 1999 ) . however , it is worth nothing that in the present model we adopt a simplification of the problem as both the disk and halo threshold density are taken as constant with the galactocentric distance . a more physical treatment of the threshold ( varying with position ) will be pursued in a forthcoming paper . model b assumes that : @xmath44 same as model a , but considering a constant halo surface mass density of @xmath48 = 17 @xmath17 pc@xmath18 at all galactocentric distances . model c assumes that : @xmath44 same as model a , but no threshold in the gas density during the halo / thick - disk phase . finally , model d assumes that : @xmath44 same as model a , but also @xmath22 varies as a function of @xmath9 : @xmath49 = 0.8 gyr for @xmath9 @xmath45 10 gyr , @xmath49 = 2 gyr for @xmath9 @xmath50 12 gyr . this slower formation of the outer halo might represent a merging scenario for the halo formation as originally suggested by searle & zinn ( 1978 ) . in other words this simulates an `` inside - out '' formation also for the halo . the situation , in this case , is similar to that predicted by models which do not take the halo into account in studying the formation of the disk ( e.g. boissier & prantzos 1999 ; portinari & chiosi 1999 , 2000 ) . these models predict steep abundance gradients in the outermost disk regions owing to the fact that they do not take into account the pre - enrichment of the disk due to the halo , which is not negligible at large galactocentric distances . all the above models can fit the available constraints in the solar vicinity ( see table 2 ) , however they differ substantially in reproducing the properties of the whole disk . in particular , concerning the solar neighborhood , our models are in good agreement with what is called the minimum set of observational constraints , among which the most important is the g - dwarf metallicity distribution ( see tosi 2000 for a recent review ; see figure 1 for model a ) . in table 3 we present the values for the observables in the solar vicinity predicted by models a and c ( the results from models b and d are the same as model a for the solar vicinity and they differ only for @xmath9 @xmath10 8 kpc ) . in table 4 we show the solar abundances predicted by model a compared to the observed ones ( anders & grevesse 1989 ; grevesse & sauval 1998 ) and the agreement is quite good . model c gives almost the same predictions as model a concerning solar abundances , so we do not list them here . this is because _ at the solar radius _ the chemical evolution of the thin - disk is not influenced by that of the halo / thick - disk . on the contrary , the abundances observed at the outermost radii strongly reflect the chemical enrichment occurring during the halo phase , as we will see in the next section . it is worth noting that all of these models provide a good fit of the [ @xmath51/fe ] versus [ fe / h ] relation in the solar vicinity , as shown in chiappini et al . ( 1999 ) . here we only show the plot of [ fe / o ] versus [ o / h ] which indicates the existence of a gap in the sfr occurring at [ o / h ] @xmath3 @xmath40.2 dex ( see figure 2a ) . in fact , if there is a gap in the sfr we should expect both a steep increase of [ fe / o ] at a fixed [ o / h ] and a lack of stars corresponding to the gap period . this is what both our model ( model a ) and the data show . this gap , suggested also by the [ fe / mg ] data of fuhrmann ( 1998 ) , in our models is due to the adoption of the threshold in the star formation process coupled with the assumption of a slow infall for the formation of the disk . this is evident from figure 3 which shows clearly a halt in the sfr at the halo - disk transition . a similar explanation for the gap in the star formation rate is that it might result from the fact that the star formation in the halo begins while the protogalaxy is still in an overall expansion phase . consequently the disk can only form later as the protogalaxy collapses and virializes ( see mathews and schramm 1993 ) . it is worth noting here , that the data in figure 2a belong to a large compilation presented in gratton et al . these data contain abundances obtained both from permitted and forbidden lines and do not show the linear decrease in [ fe / o ] ( or the growth in [ o / fe ] ) for low metallicity exhibited by recent data from israelian et al . ( 1998 ) and boesgaard et al . ( 1999 ) obtained from uv oh lines . this oxygen abundance controversy is particularly embarassing from a theoretical point of view , since a steep increase of [ o / fe ] at low metallicities would imply the same behaviour also for the other @xmath51-elements ( but see ramaty et al . 2000 for a solution involving only oxygen ) , and this , it seems , is not observed . if such behaviour is confirmed also for the other @xmath51- elements ( sharing a common origin with oxygen ) we should revise the yields of fe and/or those of oxygen . in fact , a linear increase of the [ o / fe ] could be due either to a lower contribution of fe from type ii sne or to a metallicity dependent yield of oxygen or , as a last possibility , to a strong variation of the imf during the halo phase . an earlier appearance of type ia sne during the halo phase ( at around [ fe / h ] = @xmath43.0 ) is unrealistic since it would imply that stars more massive than 8 @xmath52 can be progenitors of such supernovae . in fact , most of the models in the literature ( including the present one ) do not produce the observed linear behaviour of the [ fe / o ] found by israelian et al . ( 1998 ) , although those models assume an 8 @xmath52 star as the maximum mass allowed to form a co white dwarf , corresponding to a timescale of only about 300 myr for the first explosions . however , it is worth noting that although the first type ia sne may occur quite early , the bulk of them occurs only after @xmath53 1 gyr ( matteucci & greggio 1986 ) and produce a visible change in the abundance ratios for metallicities larger than [ fe / h ] @xmath0 @xmath41 . recent sn ia models with maximum allowed progenitors smaller than 8 @xmath52 have been suggested , but the chemical evolution results seem to be unaffected ( kobayashi et al . 1999 ) , thus confirming that only the sn ia originating from low mass stars have an impact on the abundances of the ism . the small increase of [ o / fe ] for [ fe / h ] @xmath5 @xmath54 is far from being the observed trend by israelian et al . ( 1998 ) , contrary to what is stated by goswami & prantzos ( 2000 ) . as explained in chiappini et al . ( 1999 ) the reason why the so - called `` [ o / fe]-plateau '' is not exactly flat is due to a ) the fact that some contribution by sn ia can happen already after about 300 myr as explained above and b ) the fact that the stellar evolution calculations for massive stars ( woosley & weaver 1995 ; thielemann et al . 1996 ) predict that the ejected mass of ( o / fe)@xmath55 by a massive star is an increasing function of the stellar mass ( see figure 2b ) . our results for the radial profiles of gas , stars and sfr are shown in figures 4 , 5 and 6 , respectively . it can be seen that model a is the model in best agreement with the observations . model b ( constant halo density profile ) overestimates the star density in the outer regions while model d underestimates this quantity . moreover , model b gives a constant gas profile for galactocentric distances between 820 kpc , at variance with observations . in figures 7 , 8 and 9 we compare the model predictions with the abundance data for n , o and s , respectively . in the upper panels the present time gradient is shown : our model predictions are compared with abundance data obtained from hii regions and b - stars ( see section 2 ) . in the lower panels a comparison is made with data from pne of type ii ( roughly 2 gyrs old objects ) . in figure 10 the abundance data on open clusters are compared with our model predictions for the present day abundance gradient . we did not attempt to separate the open cluster sample by age . among all the gradient figures , model a seems again to be the one which best agrees with the observations . model b predicts gradients that are flatter than the observed ones ( as in cmg97 ) , while model d predicts too steep gradients . model c is instead an alternative model which could be supported by the observations . in fact , given the scatter of the observations for a given galactocentric distance , it is very difficult to constrain our models , especially in the outer parts of the disk . model c assumes no density threshold for the sfr in the halo phase . as a consequence , a higher initial level of enrichment due to the halo evolution is reached in the galactic disk , which is almost constant for galactocentric distances larger than 12 kpc ( see also the almost constant stellar profile predicted by this model in figure 5 ) . this is due to the predominance of the halo gas in the outer regions of the disk and it creates a pre - enriched disk gas dominating the subsequently accreted primordial one . figure 11 shows the behaviour of the abundance gradients , for each element and each model , at four different epochs of the evolution of the galaxy . in figure 12 we show the abundance gradient behaviour as a function of time ( between 4 and 14 kpcs ) . as is clearly shown by this figure the inner gradients tend to steepen with time while the outer gradients remain essentially constant . this conclusion is in agreement with cmg97 and with the chemodynamical model results ( samland et al . this behaviour is expected in a model taking into account the `` inside - out '' formation of the disk , where the external disk regions are still forming now and the abundance gradient is still building up . in fact , at early epochs , the efficiency in the chemical enrichment of the inner galactic regions is low ( due to the large amount of primordial infalling gas ) leading to a flat initial abundance gradient . then , at late epochs , while the sfr is still much higher in the central than in the external regions the infall of metal poor gas is stronger in the outer than in the inner regions , thus steepening the gradients . however , other authors find a flattening of the gradients ( see tosi 2000 for a clear discussion on the possible scenarios for the evolution of the abundance gradients ) . a flattening of the gradients with time can clearly be achieved if one assumes that the disk formed on the same timescale at any galactocentric distance , but with this scenario it is very difficult to reproduce the correct gradients at the present time . however , models adopting the `` inside - out '' picture for the disk formation also seem to create gradients flattening with time . this can be explained as follows : in the inner parts of the disk those models assume a very high efficiency in the chemical enrichment process already in the earliest phases of the galaxy evolution , thus soon reaching a maximum metallicity in the gas which then remains constant or decreases due to the gas recycled by dying stars ( portinari & chiosi 1999 ; hou et al . 2000 ; boissier & prantzos 1999 ) . at the same time , in the outermost disk regions the lack of any pre - enrichment from the halo phase ( a similar situation is shown by our model d discussed above ) and the fact that those models do not include a threshold in the star formation process in the disk , produces a growth of metallicity which we do not find . this also explains an important difference between the results shown in our figure 11 and those of figure 4 of hou et al . ( 2000 ) . as in their model the halo phase is completely decoupled from the disk evolution ( even in the outer parts ) , their initial metallicities at larger galactocentric distances are very small . moreover , since they do not consider a threshold on the star formation process in the disk , their predicted abundances in the outermost parts of the disk keep increasing with time leading to a flattening of the abundance gradients . in summary , from figure 12 we can conclude that : _ a ) _ model b gives the flattest gradients . this model is in good agreement with the recent flatter gradient suggested by deharveng et al . the predicted gas and stellar density profiles are at variance with the observed ones in the outer parts of the disk . however , we stress that large uncertainties are still present in the observed profiles especially at larger galactocentric distances . _ b ) _ model c predicts a present time gradient of o in agreement with the `` standard '' adopted value of @xmath40.07 to @xmath40.06 dex / kpc as suggested by b - stars , pne and most of the data based on hii regions . morever , it predicts a steeper gradient for fe and n than for o and s , which is in agreement for instance with the results by shaver et al . ( 1983 ) ( see table 1 ) where a gradient of @xmath40.07 dex / kpc is measured for oxygen while a value of @xmath40.09 dex / kpc is obtained for n. the fact that the n gradient is steeper than the o gradient is not surprising since n is mainly a secondary element ( i.e. its abundance is metallicity dependent ) and is mainly produced on long timescales as is iron . _ c ) _ model a also predicts gradients that steepen with time but only in the first 5 gyrs of the galactic disk evolution , remaining essentially constant with time after that . again the gradients obtained for o and s are flatter than the ones obtained for n and fe . the abundance gradients obtained in model a for the present time are slightly flatter than the ones of model c ( for a galactocentric range of 414 kpc ) . finally , in figure 13 we show the predicted @xmath5[o / fe]@xmath10 ratios in the stars born and still alive at any galactocentric distance ( including halo , thick- and thin - disk stars ) as a function of such distance . in particular , this ratio refers to the stellar population dominating in mass at any radius . it is clear from the figure that , while the ratios are slightly decreasing in the galactocentric distance range 410 kpcs , as expected from the `` inside - out '' disk formation , they substantially increase at large radii . this is due to the fact that for @xmath56 kpcs the halo / thick - disk stars are dominating and they have large [ o / fe ] ratios . in the same figure we show also the plot of the same ratios versus the stellar age of the dominating stellar population and this confirms that the stars dominating in the outermost disk regions are the oldest ones . _ this also means that the spread in stellar ages is a decreasing function of the galactocentric distance_. edvardsson ( 1998 ) showed similar plots but only for disk stars older than 10 gyrs and located between 4 and 14 kpcs and , although a precise comparison can not be made , they seem to be in agreement with our predictions . @xmath44 the outer gradients are sensitive to the halo evolution , in particular to the amount of halo gas which ends up in the disk . this result is not surprising since the halo density is comparable to that of the outer disk , whereas it is negligible when compared to that of the inner disk . therefore , the inner parts of the disk ( @xmath9 @xmath5 @xmath8 ) evolve independently from the halo evolution in agreement with cmg97 result . @xmath44 our best - model predicts gradients in good agreement with the observed ones in pne , h ii regions and open clusters . in cmg97 we found flatter gradients than in the present paper and in matteucci & franois ( 1989 ) , although a very similar `` inside - out '' scenario was assumed in all these papers . the reason for being flatter than in matteucci & franois ( 1989 ) results from the fact that in cmg97 we adopted a threshold density for the sfr . this means that every time the gas density goes below the threshold then the star formation stops and starts again only after the gas has reached a density above the threshold , because of new infalling gas and/or the gas restored by dying stars ( see figure 3 ) . in particular , the effect of such a threshold was to keep the gas density always close to the threshold value especially at late times and large radii , thus contributing to maintaining an almost constant metallicity irrespective of the galactocentric distance . the reason for the fact that the present model predicts steeper gradients than cmg97 , although a threshold is also assumed , is that we considered here a more realistic halo density distribution , namely , a decreasing function of the galactocentric distance for @xmath57 8kpc , leading to a smaller contribution of the halo to the disk component metallicity enrichment at large radii . @xmath44 we predict that the abundance gradients along the galactic disk must have increased with time . this is a direct consequence of the assumed `` inside - out '' scenario for the formation of the galactic disk . moreover , the gradients of different elements are predicted to be slightly different , owing to their different nucleosynthesis histories . in particular , fe and n , which are produced on longer timescales than the @xmath51-elements , show steeper gradients . unfortunately , the available observations can not yet confirm or disprove this , because the predicted differences are below the limit of detectability . @xmath44 our model demonstrates a satisfactory fit to the elemental abundance gradients and it is also in good agreement with the observed radial profiles of the sfr , gas density and the number of stars in the disk . @xmath44 our best model suggests that the average @xmath1fe]@xmath10 ratio in stars slightly decreases from 4 to 10 kpcs . this is due to the predominance of disk over halo stars in this distance range and to the fact that the `` inside - out '' scenario for the disk predicts a decrease of such ratios ( matteucci 1991 ) . on the other hand we predict a substantial increase ( @xmath58 dex ) of these ratios in the range 1018 kpcs , owing to the predominance , in this region , of the halo over the disk stars . in this paper we have presented an improved version of the model of cmg97 for the chemical evolution of the milky way . this model assumes that the halo and disk form almost independently out of two different episodes of infall of gas of primordial chemical composition . this scenario is suggested by many observational facts and seems the most likely at the present time , even though our understanding of the formation and evolution of the milky way is still far from clear . we explored several cases where we varied in turn the timescale for the formation of the halo , which is always much smaller than for the disk , the halo density distribution as a function of galactocentric distance and the density threshold for the sfr in the halo phase . the existence of such a threshold is suggested both by observational and theoretical arguments ( kennicutt 1989 ; elmegreen 1999 ) . concerning the disk we have assumed an `` inside - out '' formation scenario , where the timescale for disk formation is a linear function of the galactocentric distance . this assumption has proven to be successful in explaining the main features of the galactic disk as well as those of other spirals . the novelty of our approach consists in the fact that we have found that the halo formation strongly affects the evolution of the outermost regions of the disk whereas it leaves untouched the innermost ones . we have considered mostly the radial properties of the disk , such as abundance gradients , star , gas and sfr distributions . the evolution of the solar neighborhood is unchanged relative to our previous papers ( cmg97 and chiappini et al . our model did not consider radial flows ( a thorough discussion of the effects of radial flows can be found in portinari & chiosi , 2000 ) . in any case , radial flows by themselves can not be the main cause of gradients along the disk although their presence can help in building up gradients , but always under very specific conditions . * a decoupling between halo and disk phases is needed in order to best fit all the observational constraints . however , the halo evolution can affect the outer regions of the disk evolution given the fact that at larger galactocentric distances the disk is still in process of forming and has very low gas densities and metallicities . * the threshold in the star formation process is important not only as a factor shaping the abundance gradients but also because it naturally produces a star formation gap between the halo and the disk phases ( in agreement with some recent abundance results , e.g. fuhrmann 1998 ; gratton et al . moreover our results indicate that the threshold gas density in the ( outer ) halo is lower than the one in the disk . * long timescale for the disk formation : @xmath59 = 8 kpc ) = 7 gyrs and longer times at larger radii are needed ( in agreement with cmg97 ) . this has been confirmed also by recent papers ( portinari et al . 1998 ; boissier & prantzos 1999 ) . * better agreeement with the observational constraints is obtained for a constant rather than variable imf ( chiappini et al . 2000 ) and this conclusion is mostly based on the abundance gradients and radial profiles of gas and sfr . * we predict abundance gradients along the galactic disk in very good agreement with observations ( although there is still some controversy about the absolute value of such gradients ) and gradients of different elements are sligthly different according to their nucleosynthetic origin . in particular , we predict that although there is a slight flattening of the abundance gradients at intermediate galactocentric distances , they are very steep in the outermost regions of the disk . unfortunately , there are no available data for abundance gradients in the outermost disk regions , but a study of the metallicity in the gas in the extreme outer regions of disks of spirals ( ferguson et al . 1998 ) , has revealed that there is no evidence for a flattening of the gradients . we also predict that abundance gradients have increased with time . * we predict a slight decrease with distance in the average @xmath60/fe]@xmath10 ratios in stars born in the galactocentric distance range 410 kpcs and an increase with distance of this ratio in the range 1018 kpc . in addition , a smaller spread of stellar ages with increasing galactocentric distance is found . * we conclude that a scenario where the halo ( mostly the inner one ) formed relatively quickly ( @xmath3 0.8 gyr ) , in agreement with recent estimates for the age differences among galactic globular clusters ( rosenberg et al . 1999 ) , and the disk grew differentially ( `` inside - out '' ) , represents the most likely explanation for the formation and evolution of the milky way . * however , in order to draw more secure conclusions it is necessary to have , in the future , more data in the outermost regions of the disk ! one of us ( f.m . ) thanks the institute of astronomy in cambridge ( u.k . ) for its kind hospitality . support from the italian ministry for university and scientific and technological research ( murst ) is gratefully acknowledged . wish to thank thomas m. dame for having kindly sent his data on the gas density distribution along the disk and e. friel for having sent us her data on iron abundances in open clusters in advance of publication . c.c . acknowledges financial support from columbia university . the authors also thank dr . john danziger and the referee dr . grant mathews for their comments and suggestions . afflerbach , a. , churchwell , e. , & werner , m.w . 1997 , apj , 478 , 190 allen , c. , carigi , l. & peimbert , m. 1998 , apj , 494 , 247 anders , e. , & grevesse , n. 1989 , geochim . acta , 53 , 197 bahcall , j.n . , flynn , c. , & gould , a. 1992 , apj , 389 , 234 beers , t.c . , & sommer - larsen , j. 1995 , apjs , 96 , 175 binney , j. , tremaine , s. 1987 , galactic dynamics ( princeton : princeton univ . press ) boesgaard , a. n. , king , j. r. , deliyannis , c. p. , & vogt , s. s. 1999 , aj , 117 , 492 boissier , s. , & prantzos , n. 1999 , mnras , 307 , 857 cappellaro , e. , & turatto , m. 1996 , in thermonuclear supernovae , ed . ruiz - lapuente , r. canal , & j. isern ( kluwer academic publishers ) , 77 carigi , l. 1996 , rev . mex . & astrofis . , 4 , 123 carigi , l. 1994 , apj , 424 , 181 carraro , g. , ng , y.k . , & portinari , l. 1998 , mnras , 296 , 1045 chang , r.x . , hou , j.l . , shu , c.g . , & fu , c.q . 1999 , a&a , 350 , 38 chiappini , c. , matteucci , f. , & padoan , p. 2000 , apj , 528 , 711 chiappini , c. , & matteucci , f. 2000 , in asp conf . , the light elements and their evolution , ed . l. da silva , r. de medeiros & spite ( san francisco : asp ) , in press chiappini , c. , matteucci , f. , beers , t. & nomoto , k. 1999 , apj , 515 , 226 chiappini , c. , matteucci , f. , & gratton , r. 1997 , apj , 477 , 765 ( cmg97 ) costa , r.d.d . , chiappini , c. , maciel , w.j . , & de freitas pacheco , j.a . 1997 , in advances in stellar evolution ( cambridge : cambridge univ . press ) , 159 crz , m. , chereul , e. , bienaym , o. , & pichon , c. 1998 , a&a , 329 , 920 dame , t.m . 1993 , in back to the galaxy , ed . s. holt & f. verter , 267 deharveng , l. , pea , m. , caplan , j. , & costero , r. 2000 , mnras , 311 , 329 dickey , j.m . 1993 , in asp conf . 39 , the minnesota lectures on the structure and dynamics of the milky way , ed . r.m . humphreys ( san francisco : asp ) , 93 edvardsson , b. 1998 , in abundance profiles : diagnostic tools for galaxy history , asp conference series , vol . 147 , p.45 elmegreen , b.g . 1999 , apj , 517 , 103 esteban , c. , peimbert , m. , torres - peimbert , s. , & escalante , v. 1998 , mnras , 295 , 401 esteban , c. , peimbert , m. , torres - peimbert , s. , & garc ' ia - rojas , j. 1999a , rev . 35 , 65 esteban , c. , peimbert , m. , torres - peimbert , s. , garc ' ia - rojas , j. , & rodr ' iguez , m. 1999b , apjs , 120 , 113 ferguson , a. , wyse , r. , gallagher , j.s . 1998 in `` abundance profiles : diagnostic tools for galaxy history '' a.s.p . series vol . 147 , p.103 fich , m. , & silkey , m. 1991 , apj , 366 , 107 fitzsimmons , a. , dufton , p.l . , & rolleston , w.r.j . 1992 , mnras , 259 , 489 flynn , c. , & fuchs , b. 1994 , mnras , 270 , 471 freudenreich , h. 1998 , apj , 492 , 495 friel , e.d . , & janes , k.a . 1993 , a&a , 267 , 75 friel , e.d . 1999 , ap&ss , 265 , 271 fuhrmann , k. 1998 , a&a , 338 , 161 gehren , t. , nissen , p.e . , kudritzki , r.p . , & butler , k. 1985 , in production and distribution of cno elements , ed . i.j . danziger , f. matteucci , & k. kjaer ( garching : eso ) , 171 gerritsen , j.p.e . , & icke , v. 1997 , a&a , 325 , 972 gilmore , g. , wyse , r.f.g . , & kuijken , k. 1989 , ara&a , 27 , 555 gilmore , g. , wyse , r.f.g . , & jones , j.b . 1995 , aj , 109 , 1095 goswami , a. & prantzos , n. 2000 , a&a , 359 , 191 gratton , r. , carretta , e. , matteucci , f. , & sneden , c. 1996 , in asp conf . 92 , formation of the galactic halo inside and out , ed . h. morrison , & a. sarajedini ( san francisco : asp ) , 307 gratton , r.g . , carretta , e. , matteucci , f. , & sneden , c. 2000 , a&a , 358 , 671 grevesse , n. , & sauval , a.j . 1998 , space sci . , 85 , 161 guibert , j. , lequeux , j. , viallefond , f. 1978 , a&a , 68 , 1 gummersbach , c.a . , kaufer , a. , schfer , d.r . , szeifert , t. , & wolf , b. 1998 , a&a , 338 , 881 gsten , r. , & mezger , p.g . 1982 , vistas astron . , 26 , 159 hensler , g. 1999 , ap&ss , 265 , 397 holmberg , j. , & flynn , c. 1998 , mnras , submitted ( astro - ph/9812404 ) hou , j.l . , prantzos , n. , & boissier , s. 2000 , a&a , in press hou , j.l . , chang , r. & fu , c. 1998 in pacific rim conference on stellar astrophysics , asp conf . ser , vol . kiwing lam chan , k.s . cheng and h.p . singh , p. 143 israelian , g. , garcia - lopez , r. & rebolo , r. 1998 , apj , 507 , 805 kaufer , a. , szeifert , t. , krenzin , r. , baschek , b. , & wolf , b. 1994 , a&a , 289 , 740 kennicutt , r.c . , jr . 1989 , apj , 344 , 685 kennicutt , r.c . , jr . 1998 , apj , 498 , 541 kennicutt , r.c . 2000 , in asp conf . galaxy disks and disk galaxies , ( san francisco : asp ) , in press kilian - montenbruck , j. , gehren , t. , & nissen , p.e . 1994 , a&a , 291 , 757 kobayashi , c. , tsujimoto , t. , nomoto , k. , hachisu , i. , kato , m. 1998 , apj , 503 , 155 kuijken , k. , & gilmore , g. 1991 , apj , 367 , l9 kulkarni , s.r . , & heiles , c. 1987 , in interstellar processes , ed . d. hollenbach , & h. thronson ( kluwer : dordrecht ) , 87 lacey , c.g . , fall , m. 1985 , apj , 290 , 154 larson , r.b . 1976 , mnras , 176 , 31 lyne , a.g . , manchester , r.n . , taylor , j.h . 1985 , mnras , 213 , 613 maciel , w.j . , & chiappini , c. 1994 , ap&ss , 219 , 231 maciel , w.j . , & kppen , j. 1994 , a&a , 282 , 436 maciel , w.j . 1997 , iau symp . habing and j. lamers , dordrecht , p.397 maciel , w.j . , & quireza , c. 1999 , a&a , 345 , 629 maciel , w.j . 2000 , in the chemical evolution of the milky way : stars versus clusters , ed . f. giovannelli , & f. matteucci ( kluwer : dordrecht ) , in press matteucci , f. greggio , l. 1986 , a&a , 154 , 279 matteucci , f. , & franois , p. 1989 , mnras , 239 , 885 matteucci , f. , ferrini , f. , pardi , c. , & penco , u. 1990 , in chemical and dynamical evolution of galaxies , ed . f. ferrini , j. franco & f. matteucci ( pisa : ets editrice ) , 586 matteucci , f. 1991 in supernova 1987a and other supernovae , eso workshop , eds . i.j . danziger and k. kjr , eso publ . p.703 matteucci , f. & chiappini , c. 1999 , in chemical evolution from zeroto high redshift , eso workshop , eds j.r . walsh and m.r . rosa , springer , p.83 matteucci , f. 2000 , in euroconference:the evolution of galaxies i- observational clues , apss , in press mathews , g. j. & schramm , d. n 1993 , apj 404 , 468 mra , d. , chabrier , g. , & schaeffer , r. 1998 , a&a , 330 , 937 pagel , b.e.j . , & patchett , b.e . 1975 , mnras , 172 , 13 pagel , b.e.j . , & tautvaisiene , g. 1995 , mnras , 276 , 505 pagel , b.e.j . 2000 , in the chemical evolution of the milky way : stars versus clusters , ed . f. giovannelli , & f. matteucci ( kluwer : dordrecht ) , in press peimbert , m. 1978 , in planetary nebulae ( kluwer : dordrecht ) , 215 portinari , l. , chiosi , c. , & bressan , a. 1998 , a&a 334 , 505 portinari , l. , & chiosi , c. 1999 , a&a , 350 , 827 portinari , l. , & chiosi , c. 2000 , a&a , 355 , 929 prantzos , n. , & silk , j. 1998 , apj , 507 , 229 prantzos , n. , boissier , s. 2000 , mnras , 313 , 338 ramaty , r. , scully , s.t . , lingenfelter , r.e . , lozlovsky , b. 2000 , apj , 534 , 747 rana , n.c . 1991 , ara&a , 29 , 129 reid , m. 1993 , ara&a , 31 , 345 renzini , a. , & voli , m. 1981 , a&a , 94 , 175 roche , n. , ratnatunga , k. , griffiths , r.e . , i m , m. , & naim , a. 1998 , mnras , 293 , 157 romano , d. & matteucci , f. 2000 , in the chemical evolution of the milky way : stars versus clusters , ed . f. matteucci & f. giovannelli ( kluwer : dordrecht ) , p.547 romano , d. , matteucci , f. , salucci , p. , & chiappini , c. 2000 , apj , 539 , 235 rosenberg , a. , saviane , i. , piotto , g. , & aparicio , a. 1999 , aj , 118 , 2306 rudolph , a.l . , simpson , j.p . , haas , m.r . , erickson , e.f . , & fich , m. 1997 , apj , 489 , 94 sackett , p.d . 1997 , apj , 483 , 103 samland , m. , hensler , g. , & theis , c. 1997 , apj , 476 , 544 scalo , j.m . 1986 , fundam . cosmic phys . , 11 , 1 searle , l. , zinn , r. 1978 apj , 225 , 357 shafter , a.w . 1997 , apj , 487 , 226 shaver , p.a . , mcgee , r.x . , newton , l.m . , danks , a.c . , & pottasch , s.r . 1983 , mnras , 204 , 53 simard , l. , koo , d.c . , faber , s.m . , sarajedini , v.l . , vogt , n.p . , phillips , a.c . , gebhardt , k. , illingworth , g.d . , & wu , k.l . 1999 , apj , 519 , 563 simpson , j.p . , colgan , s.w.j . , rubin , r.h . , erickson , e.f . , & haas , m.r . 1995 , apj , 444 , 721 smartt , s.j . , & rolleston , w.r.j . 1997 , apj , 481 , l47 thielemann , f. k. , nomoto , k. , hashimoto , m. 1996 , apj 460 , 408 timmes , f. x. , woosley , s.e . , & weaver , t. a. 1995 , apjs , 98 , 617 tosi , m. 1988 , a & a , 197 , 47 tosi , m. 1996 , in asp conf . 98 , from stars to galaxies : the impact of stellar physics on galaxy evolution , ed . c. leitherer , u. fritze - von - alvensleben , & j. huchra ( san francisco : asp ) , 299 tosi , m. 2000 , in the chemical evolution of the milky way : stars versus clusters , ed . f. giovannelli , & f. matteucci ( kluwer : dordrecht ) , in press tsujimoto , t. , yoshii , y. , nomoto , k. , shigeyama , t. 1995 a&a , 302 , 704 twarog , b.a . , ashman , k.m . , & anthony - twarog , b.j . 1997 , apj , 114 , 2556 van den hoek , l.b . , & groenewegen , m.a.t . 1997 , a&as , 123 , 305 van der hulst , j.m . , skillman , e.d . , smith , t.r . , bothum , g.d . , mcgaugh , s.s . & de blok 1993 , aj , 106 , 548 viegas , s.m . , gruenwald , r. , & steigman , g. 2000 , apj , 531 , 813 v ' ilchez , j.m . , & esteban , c. 1996 , mnras , 280 , 720

higher - dimensional field theory has been applied to explore various issues in recent particle physics such as the gauge hierarchy problem . in order for such approaches to be viable , a crucial ingredient is to fix the sizes of extra dimensions at some finite values , which sizes are generically free parameters in the theory . in this paper , we present several schemes to determine the radius of extra dimension in warped five - dimensional theory . in every case , a non - vanishing fayet - iliopoulos term for abelian gauge factor plays a crucial role for the radius stabilization . it is radiatively generated in the presence of charged matter fields and the compactification is therefore spontaneous , not forced by selected operators . the low - energy supersymmetry is broken or unbroken , and the radius can be fixed to give a small or large scale hierarchy without any fine tuning of parameters . we also discuss a model of the radius stabilization correlated with yukawa hierarchy and supersymmetry breaking . kuns-1935 , kyushu - het-74 + hep - ph/0409355 * radiative stabilization of warped space * tatsuo kobayashi and koichi yoshioka department of physics , kyoto university , kyoto 606 - 8502 , japan + department of physics , kyushu university , fukuoka 812 - 8581 , japan ( september , 2004 )

hep-ph0409355

field theory in higher dimensions has been providing novel approaches to theoretical and phenomenological problems in recent particle physics . the existence of extra spatial dimensions beyond our fours is applied to various issues such as the generation of large scale hierarchies @xcite . in these approaches , a key ingredient is to determine the size of extra space so that they are viable approaches and do not conflict with current observations . for example , the planck / weak mass hierarchy is attained by assuming that the radii of compactified dimensions are huge @xcite or small but a bit larger than the planck length @xcite . the compactification radius is also conjectured to have anticipated values in other phenomenological discussions such as small neutrino masses @xcite , yukawa hierarchies of quarks and leptons @xcite , and supersymmetry breaking @xcite . therefore adjusting the sizes of extra dimensions to desired values is one of the most important issues in higher - dimensional framework . there have been in the literature various resolutions to this stabilization problem in large- and small - sized extra dimensions @xcite . in this paper , we present three different schemes to stabilize the radius modulus in five - dimensional supersymmetric theory with or without charged matter fields . one is based on the model with only boundary charged fields and another with only bulk fields . in every scheme , the fayet - iliopoulos ( fi ) term @xcite in five - dimensional theory is found to play a crucial role for the radius stabilization . the resultant metric factor can be significant or nearly flat , depending on model parameters . it is stressed that the fi term is not introduced by hand in order to stabilize the radius . a non - vanishing fi term is radiatively generated even if it is set to zero in the classical lagrangian @xcite . this is unlike the four - dimensional theory where a fi term does not receive any renormalization if theory has a vanishing gravitational anomaly @xcite . the induced fi term depends on how charged matter multiplets are distributed in the extra dimensions and is therefore controllable . further it is known in four - dimensional models that the fi term is connected to yukawa hierarchy and supersymmetry breaking @xcite . we construct a five - dimensional model for generating fermion mass hierarchy which is correlated with the radius stabilization . the model also predicts characteristic spectrum of sfermions and gauginos . this paper is organized as follows . in section 2 , we give a generic form of globally supersymmetric lagrangian for five - dimensional @xmath2 theory with the fi term . the simplest stabilization scheme is found in section 3 without introducing any matter fields . in section 4 , a different scheme is presented to fix the size of extra dimension . the model contains bulk hypermultiplets with non - trivial wavefunctions whose forms are determined by fi - term coefficients . the bulk multiplets fix the distance between the two boundaries in terms of boundary couplings . in the vacuum of this model , supersymmetry is unbroken . with the generic lagrangian at hand , we show in section 5 that the radius can be stabilized by only boundary matter fields . that is established by analyzing the vacuum energy in effective four - dimensional theory with broken supersymmetry . we also construct a toy model for yukawa hierarchy and supersymmetry breaking , deeply correlated with the radius stabilization phenomenon . such a model predicts a new type of sparticle spectrum in low - energy effective theory . section 6 is devoted to summarizing our results . we consider the globally supersymmetric abelian gauge theory in five dimensions . the fifth dimension is compactified on a line segment @xmath3 , where the radius of the circle @xmath4 is @xmath5 . the radius @xmath5 is a free parameter of the theory and corresponds to a massless moduli field @xmath6 in four - dimensional effective theory ( @xmath7 ) . the fifth dimension @xmath8 has two boundaries at @xmath9 and @xmath10 . they are fixed points under the @xmath11 orbifolding of physical spacetime . we are now interested in the case that the extra dimension has curved geometry . a particularly interesting example is the warped ( ads ) geometry @xcite whose line element is given by @xmath12 where @xmath13 is the ads curvature and @xmath14 the minkowski metric in four dimensions . this background metric has been intensively studied for realistic model construction with the planck / weak scale difference , quarks and leptons mass hierarchy , the cosmological constant problem , etc . in these approaches the radius @xmath5 of the compact extra dimension was often assumed to have a desired value , and the radius stabilization is therefore one of the most important problems in constructing realistic ` brane - world ' models . in this paper we present the schemes to stabilize @xmath5 at a finite value due to the existence of @xmath2 gauge factor in supersymmetric warped dimensions . we adopt the superspace formalism of higher - dimensional supersymmetry @xcite . there are two types of supermultiplets generally introduced in five - dimensional theory ; vector and hyper multiplets . a vector multiplet contains an @xmath15 vector multiplet @xmath16 and a chiral multiplet @xmath17 , whose auxiliary components are denoted by @xmath18 and @xmath19 , respectively . a hypermultiplet consists of oppositely - charged two chiral multiplets @xmath20 and @xmath21 . in the superspace language , the most generic lagrangian for five - dimensional @xmath2 gauge theory is given by @xmath22 @xmath23 ^ 2 , \label{lv } \\[1 mm ] l_h & \!=\ ! & \int\ ! d^4\theta\ , e^{-2k|y| } \big(\phi^\dagger e^{qv}\phi + \phi^c e^{-qv } { \phi^c}^\dagger \big ) \nonumber \\ & & \hspace{2 cm } + \int\ ! d^2\theta\ , e^{-3k|y| } \phi^c\bigg[\partial_y+\frac{q}{\sqrt{2}}\chi -\big(\frac{3}{2}-c\big)k\epsilon(y)\bigg]\phi + { \rm h.c . } , \label{lh}\end{aligned}\ ] ] where @xmath24 denotes the gauge coupling constant , and @xmath25 and @xmath26 are the @xmath2 charge and the bulk mass parameter of the chiral multiplet @xmath20 , respectively . the sign function @xmath27 is inserted in order for the orbifold @xmath11 invariance . we have also included the lagrangian for chiral multiplets confined on the uv ( @xmath9 ) and ir ( @xmath28 ) boundaries ; @xmath29 the orbifold boundary conditions are imposed on each supermultiplet . the vector multiplet @xmath16 has the neumann boundary conditions at both uv and ir branes and its superpartner multiplet @xmath17 has the dirichlet ones because it contains the fifth component of the bulk gauge field . the boundary conditions of @xmath20 must be opposite to those of superpartner @xmath21 for respecting the @xmath11 symmetry . the @xmath11 boundary conditions break a half of bulk supersymmetry and thus the boundary lagrangians @xmath30 and @xmath31 preserve only the four - dimensional @xmath15 supersymmetry . for example , yukawa couplings of quarks and leptons are expected to come from these boundary interactions in the present framework . the boundary chiral multiplets @xmath32 and @xmath33 couple only to bulk multiplets with neumann boundary conditions . we have assumed , for simplicity , that there are no @xmath8-derivative couplings of @xmath11-odd chiral multiplets and no four - dimensional gauge fields on the boundaries , while these assumptions are irrelevant to the following discussion . the exponential warp factors are explicitly included in the above lagrangian . these warp factors describe the metric dependences , such as from @xmath34 , of the lowest component of each supermultiplet in the warped background . for other component fields , the proper metric factors in the warped five - dimensional action are obtained after some rescaling , for example , @xmath35 , @xmath36 for the auxiliary fields . since we now consider the abelian gauge theory , a fi term of vector multiplet @xmath16 is gauge invariant in globally supersymmetric theory and can also be added to the lagrangian as @xmath37 we have defined the coefficient @xmath38 into which the metric warp factor is absorbed . even if there is no fi term in classical lagrangian , it is radiatively generated via tadpole graphs of the @xmath18 component where charged matter fields circulate in the loop . in the case of flat extra dimensions , its form was investigated @xcite and found to reside only on the orbifold fixed points . moreover the fi term vanishes in anomaly - free low - energy effective theory when one integrates out the fifth - dimensional physics . this is consistent with the fact that , in four - dimensional theory , a coefficient of radiatively - generated fi term is proportional to the sum of matter @xmath2 charges which also gives the mixed @xmath2-gravitational anomaly . in five - dimensional theory on curved backgrounds including the warped geometry , the situation is rather different . since the fifth direction is curved , the fundamental length depends on the position @xmath8 . the implication of this fact appears through the metric - factor dependences in the fi - term calculation . in the warped geometry ( [ warp ] ) , the brane - localized fi terms are written as @xmath39 with the constant coefficients @xmath40 and @xmath41 . if the fi term is set to vanish at classical level , radiative corrections give rise to @xmath40 and @xmath41 which are given by specific combinations of @xmath2 charges of bulk and boundary fields @xcite . for a @xmath2 factor free from gravitational anomaly in low - energy theory , the two coefficients are equal to each other ; @xmath42 . the exponential factor in the second term of ( [ fi ] ) indicates that the fundamental length is redshifted at the @xmath28 boundary . this factor may be described by proper regularization , for example , la pauli - villars , and more simply implemented by a position - dependent cutoff for the four - dimensional momentum in the one - loop calculations , which dependence is suggested by the ads / cft correspondence @xcite . that has been recently confirmed by detailed analysis of five - dimensional supergravity @xcite . in four - dimensional effective theory , the fi term does not vanish as the zero modes of vector multiplet have flat wavefunctions . as a result , either @xmath2 gauge symmetry or four - dimensional supersymmetry is broken at the scale of @xmath38 . this reflects the known fact in four - dimensional theory that a fi term for anomaly - free @xmath2 gauge theory does not coexist with unbroken supersymmetry . that is , in four - dimensional supergravity theory , a fi term can be introduced only when @xmath2 is @xmath5 symmetry or non - linearly realized , i.e. the @xmath2 gauge boson becomes massive . an important point here is that even if low - energy theory is totally free of anomalies like qed , the effective fi term is non - vanishing due to the curved extra dimension ( @xmath43 ) and has important phenomenological implications @xcite . in this paper , we show that the presence of fi term also provides the stabilization mechanisms of the radius modulus field . let us first see the simplest case where we have no charged matter fields in the theory . integrating out the fifth - dimensional physics , we find that the presence of the fi term ( [ fi ] ) leads to the potential @xmath44 this vacuum energy depends on the radius @xmath5 . it is therefore determined so that the vacuum energy is minimized . we find from ( [ vr - uvir ] ) a possibility that the radius is fixed to a finite value @xmath45 the vacuum energy vanishes at this point which is the potential minimum in globally supersymmetric theory . thus four - dimensional supersymmetry is unbroken while the radius is stabilized . note that the limit @xmath46 also gives a vanishing vacuum energy , where the low - energy gauge theory becomes a free theory . however the potential barrier between the two minima can be as high as @xmath47 whose natural size is around the planck scale . therefore the vacuum ( [ kr ] ) might be made stable within the present age of the universe . moreover the parameter region far away from the origin could be lifted by supersymmetry breaking which we have not included here . it is noticed that the existence of the vacuum ( [ kr ] ) calls a restriction on the fi - term coefficients ; @xmath48 . for example , in case that low - energy theory is anomaly free , we have @xmath42 and hence the radius is not settled at a finite value . for radiatively - generated fi terms , the inequality @xmath49 is realized with ` anomalous ' matter content . then one should assume some anomaly cancellation mechanism that does not affect the fi terms . in this paper , we do not pursue such a possibility further . instead we will discuss the radius stabilization with bulk / boundary matter fields in the presence of ` non - anomalous ' fi term : @xmath50 . in this section we present a scheme for stabilizing the size of warped extra dimension which involves only bulk hypermultiplets . a hypermultiplet which has non - trivial profile of bulk wavefunction connects two localized fi terms , and determines the distance between the boundaries . in this way the radius is fixed by the equations of motion of bulk fields together with their boundary conditions on the branes . we consider the five - dimensional @xmath2 gauge theory with non - vanishing boundary fi terms . its lagrangian is given by ( [ lv ] ) and ( [ lh ] ) as well as boundary superpotentials @xmath51 and @xmath52 for bulk chiral multiplets with even @xmath11 parity . we do not include any boundary supermultiplets on the branes . the five - dimensional scalar potential is generally given by @xmath53 the auxiliary fields @xmath18 and @xmath54 s are expressed in terms of bulk scalars through their equations of motion ( see below ) . here we have simply assumed that the fifth component of the @xmath2 vector field does not have a nonzero expectation value . the model we present in this section contains two bulk hypermultiplets with the following @xmath2 charges and bulk masses : @xmath55 notice that two hypermultiplets have opposite @xmath2 charges so that they can form a mixing mass term . this is however just a simplifying assumption . we will mention other choices of @xmath2 charges in the end of this section and show that the charge assignment of hypermultiplets is irrelevant to the radius stabilization mechanism . the boundary conditions on the branes , namely , the orbifold parities are taken to be positive for @xmath20 and @xmath56 ( therefore , negative for @xmath21 and @xmath57 ) , which lead to the zero modes of @xmath20 and @xmath56 in low - energy effective theory . there exists the supersymmetric vacuum in the presence of fi term ( [ fi ] ) when the bulk scalars take appropriate expectation values so that the flatness conditions are satisfied . as mentioned in the previous section , we consider a conceivable case that @xmath50 . without loss of generality , we take @xmath58 and then find that among the bulk matter scalars only @xmath11-even @xmath56 develops a vacuum expectation value . the @xmath18 and @xmath54 flatness conditions now reduce to @xmath59 , \label{dflat } \\ 0 & = & \big[\partial_y-\frac{q}{2}\sigma -\big(\frac{3}{2}-c_\varphi\big)k\epsilon(y)\big]\varphi.\end{aligned}\ ] ] the field @xmath60 is the real part of the neutral scalar in the chiral multiplet @xmath17 . it seems difficult to write down the generic solutions of these vacuum equations but we can analytically solve them for a specific value @xmath61 . in this case , the solutions @xmath62 and @xmath63 are given by @xmath64 where @xmath65 and @xmath66 are the integration constants . these constants are determined by the boundary fi terms through the @xmath18-term equation ( [ dflat ] ) as @xmath67 the existence of non - vanishing fi terms therefore fixes the unique supersymmetric vacuum away from the origin of the field space . when the warp factor is significant ( @xmath68 ) and a fi term is small ( @xmath69 ) , the explicit forms of the integration constants are approximately given by @xmath70 the wavefunctions @xmath62 and @xmath63 do not have singularities between the two boundaries . in the following analysis , we adopt this value @xmath61 as an example . it is however stressed that one may expect similar effects of radius stabilization for other generic values of the bulk mass parameters . as we will show below , the only required is the existence of non - trivial unperturbed solutions . we introduce the following gauge - invariant superpotential terms onto the boundaries ; @xmath71 it is noted that a supersymmetric mass term between @xmath20 and @xmath56 is forbidden by bulk ads@xmath72 supersymmetry and is forced to be confined on the boundaries , at which bulk supersymmetry is broken to the four - dimensional one . on the other hand , the mass terms @xmath73 and @xmath74 are allowed to exist by the bulk supersymmetry , but in the present model , the @xmath2 gauge invariance makes it vanish . therefore ( [ wmass ] ) is the most generic superpotential for the matter fields involved . as we will see , the above potential terms play important roles of acting as the sources of @xmath20 which , in turn , stabilizes the radius and of lifting the @xmath56 direction from the equation of motion of @xmath20 . let us first see whether supersymmetry is broken in this model by examining the four - dimensional scalar potential . the @xmath18-term contribution to the scalar potential is written in the usual form with a non - vanishing fi term . the @xmath54-term contribution comes from the superpotential obtained by reducing ( [ wmass ] ) to the four - dimensional zero - mode part . expanding @xmath75 and @xmath76 , it is given by @xmath77 \phi_4(x)\varphi_4(x ) \>\equiv\ , m_{\phi\varphi}\,\phi_4(x)\varphi_4(x),\ ] ] where @xmath78 and @xmath79 are the solutions of their equations of motion . if the effective mass parameter @xmath80 is nonzero , @xmath81 and @xmath82 are lifted and supersymmetry is broken because the @xmath18-term equation enforces specific values on these fields . however @xmath80 now depends on the radius @xmath5 , which is generically not a frozen parameter . the radius @xmath5 thus fixes itself so as to give a vanishing effective mass @xmath80 , for which the vacuum energy is minimized and supersymmetry is restored . in other words , if the radius modulus @xmath6 is included as a dynamical variable , the minimization of scalar potential with respect to @xmath81 , @xmath82 and @xmath6 leads to a vanishing effective mass @xmath80 ( at least local , supersymmetric vacuum ) . since supersymmetry is unbroken , all the equations of motion in the five - dimensional theory are @xmath83 & & \hspace*{7 cm } -g^2\xi_{\rm fi}\big[\delta(y ) -e^{-2k\pi r}\delta(y-\pi r)\big ] , \label{d } \\ 0 \!&=&\ ! f_\chi^\dagger \,=\ , -\frac{q g^2}{\sqrt{2}}e^{-k|y| } ( \phi^c\phi-\varphi^c\varphi ) , \label{fchi } \\ 0 \!&=&\ ! f_\phi^\dagger \,=\ , e^{-k|y|}\big[\partial_y -\frac{q}{2}\sigma -\big(\frac{3}{2}+c_\phi\big)k\epsilon(y)\big ] \phi^c -\big[m_0\delta(y)-e^{-k\pi r}m_\pi\delta(y-\pi r)\big]\varphi , \quad \label{fphi } \\ 0 \!&=&\ ! f_\varphi^\dagger \,=\ , e^{-k|y|}\big[\partial_y + \frac{q}{2}\sigma -k\epsilon(y)\big ] \varphi^c -\big[m_0\delta(y)-e^{-k\pi r}m_\pi\delta(y-\pi r)\big]\phi , \label{fvarphi } \\ 0 \!&=&\ ! f_{\phi^c}^\dagger \!\,=\ , -e^{-k|y| } \big[\partial_y + \frac{q}{2}\sigma -\big(\frac{3}{2}-c_\phi\big ) k\epsilon(y)\big]\phi , \\ 0 \!&=&\ ! f_{\varphi^c}^\dagger \!\,=\ , -e^{-k|y| } \big[\partial_y -\frac{q}{2}\sigma -2k\epsilon(y)\big]\varphi . \label{fvarphic}\end{aligned}\ ] ] the localized operators enforce the specific boundary conditions on the parity - odd functions @xmath60 , @xmath21 and @xmath57 such that @xmath84 with the even functions @xmath85 , @xmath86 and @xmath87 which satisfy the conditions @xmath88 the @xmath54-term equations ( [ fphi ] ) and ( [ fvarphi ] ) are simplified in the bulk as @xmath89f_\phi , \label{fp } \\ 0 \!&=&\ ! \big[\partial_y + \frac{q}{2}\sigma -k\epsilon(y)\big ] f_\varphi.\end{aligned}\ ] ] since we have @xmath58 without loss of generality , the scalar field @xmath20 do not develop vacuum expectation values . this is also understood from the view of four - dimensional effective theory . in turn , the equations ( [ fchi ] ) and ( [ fvarphi ] ) together with ( [ bcfvp ] ) mean @xmath90 . the independent equations of motion now reduce to ( [ d ] ) , ( [ fvarphic ] ) , and ( [ fp ] ) with @xmath91 . the vacuum solutions are explicitly derived by solving these equations in perturbation of @xmath92 . the leading - order solutions are given by the unperturbed ones ( [ sol0s ] ) , ( [ sol0p ] ) , and @xmath93 . it is interesting to notice that the radius determination does not require a precise form of @xmath60 , which follows from the @xmath18-term equation ( [ d ] ) . the formal solutions to ( [ fvarphic ] ) and ( [ fp ] ) are @xmath94 with @xmath95 and @xmath96 being the integration constants . inserting the solutions into ( [ bcfp ] ) , we find that the boundary conditions of @xmath21 determine the value of @xmath5 : @xmath97 the boundary conditions also constrain the ratio of integration constants as @xmath98 . their individual values are fixed by the @xmath18-term equation which is satisfied by @xmath99 fluctuation of @xmath60 around the unperturbed solution ( [ sol0s ] ) . the fact that @xmath100 ensures the relevance of the perturbative analysis . the above derivation makes it clear that explicit solutions to the equations of motion are not needed to find a stabilized value of the radius . in fact , the radius in the minimum can easily be evaluated for a generic value of @xmath101 . that is , a similar analysis shows that @xmath5 is determined so that @xmath102 . several comments are in order . at least at this order of perturbation , the stabilized value of the radius does not seem to depend on the fi term . it is however noticed that the wavefunction factors @xmath103 s depend on the fi term via @xmath65 . if one turns off the fi term ( @xmath104 ) , the vacuum goes to the origin of field space , that is , @xmath105 [ see ( [ afix ] ) ] . consequently , the expectation value of @xmath21 also vanishes which can not lead to the radius determination ( [ rstab ] ) . in this way the existence of non - vanishing fi term is crucial for the stabilization of radius modulus field . secondly , as for the parameters @xmath106 and @xmath107 , realizing a significant warp factor does not need any fine tuning . for example , the values @xmath108 and @xmath109 give @xmath110 in ( [ rstab ] ) . what is needed is a parameter choice of order @xmath111 , which is similar to that in the original randall - sundrum model where @xmath112 is assumed for solving the gauge hierarchy problem . of course , a radius stabilization with no significant warp factor @xmath113 is easier to be achieved . finally , we comment on the possibility for other choices of @xmath2 charges . when the bulk mass @xmath114 is exactly one half , the equation ( [ rstab ] ) implies that the radius is not stabilized . this is however simply because of our @xmath2 charge assignment . if one supposes the @xmath2 charge of @xmath20 is @xmath115 for example , the gauge - invariant boundary superpotentials take the form @xmath116 . in this case ( also with a general unfixed @xmath101 ) , the stabilized value of @xmath5 is replaced with @xmath117 thus the radius is still stabilized as long as there is no principle relating bulk mass parameters and @xmath2 charges in a specific way . in this section we examine whether the radius modulus can be stabilized only with boundary field dynamics unlike the model presented in the previous section . in this case , the stabilization procedure is to look for the minimum of four - dimensional effective potential of the radius modulus field . let us first derive low - energy effective theory for generic boundary superpotential terms . we assume that supergravity effects except for the radius modulus are irrelevant to stabilization . integrating out the fifth dimension , we obtain the low - energy effective theory of the @xmath2 multiplet zero modes , boundary matter multiplets , and the radius modulus @xmath6 . the effective lagrangian is derived from ( [ lv ] ) , ( [ luv ] ) , ( [ lir ] ) , and ( [ ld ] ) with ( [ fi ] ) : @xmath118 \nonumber \\ & & \hspace*{-1 mm } + \int\!d^2\theta\,\big [ w_{\textrm{uv}}(\phi_{\rm uv } ) + e^{-3k\pi t } w_{\rm ir}(\phi_{\rm ir } ) \big ] + { \rm h.c . } + \!\int\!d^4\theta\big(2\xi_{\rm fi}v-\frac{6m^3}{k}\big ) \big[1-e^{-k\pi(t+t^\dagger)}\big ] , \nonumber\\[-1 mm ] \label{l4d}\end{aligned}\ ] ] where @xmath119 is the fundamental scale in five - dimensional theory . we have included the proper khler term of the radius modulus field in the warped background @xcite . it is assumed that extra bulk dynamics to lead to potential terms of @xmath6 is not introduced . as mentioned in the previous section , however , the fi term is automatically generated in the presence of charged matter fields and provides @xmath6-dependent terms . the scalar potential is obtained by integrating out all the auxiliary components @xmath120 ^ 2 \nonumber \\ & & + \left|\frac{\partial w_{\rm uv}}{\partial \phi_{\rm uv } } \right|^2 + e^{-4k\pi r } \left|\frac{\partial w_{\rm ir}}{\partial \phi_{\rm ir } } \right|^2 + \frac{ke^{-4k\pi r}}{6m^3 } \left|3w_{\rm ir}-\phi_{\rm ir } \frac{\partial w_{\rm ir}}{\partial \phi_{\rm ir}}\right|^2 . \label{potential}\end{aligned}\ ] ] for later discussion , we present the equations of motion for the auxiliary fields ; @xmath121 f_{\phi_{{}_{\rm uv}}}^\dagger \!\!&= & -\frac{\partial w_{\rm uv}}{\partial \phi_{\rm uv } } , \\ f_{\phi_{\rm ir}}^\dagger \!&= & -e^{-k\pi r } \frac{\partial w_{\rm ir}}{\partial \phi_{\rm ir } } + \frac{ke^{-k\pi r}}{6m^3}\ , \phi_{\rm ir}^\dagger \big(3w_{\rm ir}-\phi_{\rm ir } \frac{\partial w_{\rm ir}}{\partial \phi_{\rm ir } } \big ) , \\ f_t^\dagger \,&= & \frac{e^{-k\pi r}}{6\pi m^3 } \big(3w_{\rm ir}-\phi_{\rm ir } \frac{\partial w_{\rm ir}}{\partial \phi_{\rm ir } } \big),\end{aligned}\ ] ] where we have simply assumed that the graviphoton field does not have a nonzero expectation value . the scalar potential is a function of boundary scalar fields and the modulus @xmath5 . assuming an appropriate form of boundary superpotentials , we first minimize the potential ( [ potential ] ) with respect to matter scalars and find the minimum value @xmath122 of the potential , which generally depends on @xmath5 . then doing the minimization of @xmath122 , we obtain the vacuum with a stabilized value of @xmath5 . from the generic form of the scalar potential ( [ potential ] ) , we have several observations for the radius stabilization : ( i ) first , unless the fi term is present , the radius is not stabilized . for a vanishing value of @xmath123 , there is a @xmath18-flat direction in the potential . along this direction , the potential has the @xmath5 dependences only in the form of @xmath124 and the vacuum goes to infinity . ( ii ) another observation is that , unless supersymmetry is broken , the radius is not stabilized at a finite value . generally speaking , the scalar potential vanishes for unbroken supersymmetry and hence can not fix the radius . in the present case , the fi term itself gives rise to a non - vanishing potential . however it is a monotonous function of @xmath5 and consequently , the radius @xmath5 is fated to have a runaway behavior . we thus find that , for stabilizing the radius modulus , the fi term must be present and also four - dimensional supersymmetry must be broken ( leading to non - vanishing vacuum energy ) . an interesting point is that , in curved five - dimensional theory , a fi term is automatically induced and can cause required supersymmetry breaking . note that the above arguments are applied to the models with boundary multiplets only , and therefore including hypermultiplets and/or bulk dynamics may change the conclusion . for example , as in the model of section 4 , non - trivial @xmath5-dependences of bulk - field wavefunctions generate @xmath5-dependent ( super)potential terms , which can lead to ( supersymmetric ) radius stabilization . we have shown in the above that this can not be obtained by boundary matter only and supersymmetry needs to be broken . as we mentioned , four - dimensional supersymmetry must be broken to stabilize the size of the compact fifth dimension . in the absence of charged matter fields , the fi term leads to a non - vanishing potential for the radius modulus , as discussed in section 3 . however it does not have any minimum with respect to @xmath5 for anomaly - free theory . moreover it might be unfavorable with such a vacuum energy that supersymmetry is broken at a high - energy scale . in this section , we are thus interested in including the charged matter contribution . with a non - vanishing fi term , the @xmath18-flatness condition points to the unique vacuum with nonzero expectation values of charged matter scalars . therefore adding appropriate perturbation to superpotential terms , the charged field directions are lifted and hence supersymmetry is broken , as needed . the most simple case is to introduce on the @xmath9 boundary a vector - like chiral multiplets @xmath32 and @xmath125 ( with @xmath2 charges @xmath126 and @xmath127 respectively ) and their mass term : boundary , to introduce only a constant superpotential , and so on . we find that , in either of these cases , the radius modulus is not stabilized at a finite value . ] @xmath128 together with the @xmath18-term potential , one can see that supersymmetry is broken by small perturbation ( @xmath129 tev ) . it is however found from ( [ potential ] ) that the modulus potential in this case has the maximum only . we thus incorporate a constant superpotential on the @xmath28 boundary : @xmath130 . such a constant superpotential can be obtained in various ways and here we do not consider any details of its origin . without loss of generality , @xmath131 is taken to be positive and the vacuum is given by @xmath132 the vacuum energy , which depends on the radius @xmath5 , becomes @xmath133 minimizing the vacuum energy determines the value of @xmath5 . we find the solutions to @xmath134 ; @xmath135 in this vacuum with a stabilized @xmath5 , supersymmetry is broken by nonzero @xmath18 and @xmath54 terms ; @xmath136 there are two typical scales of the constant term @xmath137 such that @xmath138 is on the order of supersymmetry - breaking scale @xmath139 . the first case is given by a suppressed value of @xmath137 compared to the fundamental scale @xmath140 . in this case , it is easily found by analyzing the second derivative of the potential that the minimum of @xmath122 is given by @xmath141 . therefore the radius is stabilized around the value @xmath142 and a large metric warp factor does not arise . the supersymmetry - breaking contribution from the radius modulus is found to be @xmath143 . [ vrfig ] shows an explicit form of the vacuum energy ( the potential of the radius modulus ) for a suppressed value of @xmath137 . ( 8,81.5)@xmath144 ( -237,167)@xmath122 the stabilized modulus obtains the mass squared @xmath145 which is always positive definite at the minimum . in the parameter region we now consider , it is approximately given by @xmath146 with the canonical kinetic term of the radius modulus . another typical scale of @xmath137 is a natural scale in the theory , namely , @xmath147 . in this case , the solution ( [ sol ] ) means that the minimum is around @xmath148 and therefore , @xmath149 as a result , the metric warp factor gives significant effects , and also we have a suppressed value of radius modulus @xmath54 term ; @xmath150 . in fig . [ vrfig2 ] , we show a typical behavior of the vacuum energy @xmath122 . ( 8,84)@xmath144 the mass of the radius modulus with the proper normalization in this region is @xmath151 again we have a tev - scale massive modulus field . for a large metric factor , kk - excited modes have suppressed masses above @xmath152 in four - dimensional theory . it might require a careful treatment to examine whether the effective lagrangian of zero modes is valid in this parameter region . in addition , the minimum might not be so steep that it is not meta - stable within the cosmological evolution . however a higher potential barrier can be achieved by a smaller value of the constant superpotential , and a unstable vacuum is easily avoided . we have shown that a nonzero fi term can stabilize the radius modulus to a realistic value . as well known in four - dimensional theory , the fi term is capable of explaining yukawa hierarchy of quarks and leptons and also of providing interesting sparticle spectrum . in this subsection , we present a toy model towards constructing realistic theory in higher dimensions where a single existence of fi term has various important implications to phenomenology . let us consider one - generation ` lepton ' multiplets @xmath153 , @xmath154 , and ` higgs ' @xmath155 as well as a vector - like multiplets @xmath20 and @xmath156 . the latters play the radius stabilizer discussed in the previous section . for simplicity , we focus only on the @xmath2 factor and ignore the standard model gauge groups . incorporating these gauge factors is rather straightforward . now suppose that @xmath153 comes from a five - dimensional hypermultiplet @xmath157 with orbifold parities @xmath158 . so @xmath153 contains a massless mode in four - dimensional effective theory . all other multiplets @xmath154 , @xmath155 , @xmath20 and @xmath156 are assumed to be confined on the uv boundary . the @xmath2 charges of these multiplets are listed in the table 1 . .the @xmath2 charge assignment . we take the charge of the higgs field zero , for simplicity , and the matter charges @xmath159 and @xmath160 are positive . [ cols="^,^,^,^,^,^",options="header " , ] as in usual four - dimensional case , the @xmath2 charges of matter fields are taken as positive , which will be important to have non - vanishing yukawa couplings , positive sfermion masses squared , and also the potential analysis in the previous section to be valid . the charge @xmath25 has been set to + 1 without loss of any generalities . note that , with only these multiplets at hand , the zero - mode effective @xmath2 theory is anomalous . but some anomaly cancellation may easily be assumed , for example , introducing additional charged multiplets . an important point here is that , even if effective four - dimensional theory is anomaly free , an induced fi term can be nonzero due to the curved extra dimension . note also that there are no gauge anomalies for the standard gauge groups , if included , provided that anomalies are cancelled within massless modes @xcite . therefore in case that light - mode spectrum is that of the standard model , we do not worry about gauge ( and gravitational ) anomalies for any field configurations in the extra dimension . the fi term may be radiatively generated even for anomaly - free particle contents . in the @xmath2 theory above , the one - loop contribution is given by ( [ fi ] ) with @xmath161 and @xmath162 where @xmath163 is near the fundamental scale @xmath119 . as seen in the previous section , the successful radius stabilization needs a positive value of @xmath41 , which is not satisfied in the present form . ( @xmath40 must also be positive to have yukawa couplings and supersymmetry breaking . ) a simple way to cure this problem is to introduce charged hypermultiplets which have even ( odd ) orbifold parity at the ur ( ir ) boundary , or vice verse . these additional multiplets contribute to the fi coefficients @xmath164 where @xmath165 is the sum of @xmath2 charges of added multiplets . moreover they contain no zero modes and do not change the standard model spectrum . with this implementation , both fi - term coefficients can safely be positive . in the previous analysis , @xmath123 is replaced with @xmath41 in the solution ( [ sol ] ) and the expressions of the radion mass , but in the expectation values of @xmath125 and @xmath166 , @xmath123 is approximately given by @xmath40 , which is suitably positive if @xmath41 is made positive . the yukawa couplings for matter multiplets are described by gauge - invariant higher - dimensional operators @xcite on the @xmath9 boundary : @xmath167 where @xmath168 is the @xmath169 coupling constant . if one included other generations originated from bulk hypermultiplets , their yukawa couplings are allowed by bulk supersymmetry only on the boundaries as ( [ wy ] ) . the potential analysis shows that only @xmath156 develops a non - vanishing expectation value , i.e.([vev ] ) . also has a negative @xmath2 charge and might be worried to obtain a nonzero expectation value . however the equation of motion for the @xmath153 scalar implies that this is not the case as long as there is no source term of @xmath153 on the uv boundary . ] as a result , the above operator gives an effective yukawa coupling @xmath170 . the expectation value of @xmath171 is proportional to the fi term and thus an one - loop order quantity . therefore we obtain @xmath172 that is just suitable for describing realistic yukawa hierarchies . furthermore , in this model , there is an additional possibility to have yukawa suppression unlike in pure four - dimensional theory . that is a wavefunction factor of bulk hypermultiplet zero modes . the zero - mode wavefunction @xmath173 depends on its bulk mass @xmath26 and @xmath2 charge , and is given by @xcite @xmath174,\ ] ] with @xmath175 being the normalization constant determined by @xmath176 and roughly given by @xmath177 . the present setup predicts @xmath178 derived from the background expectation value of @xmath60 , which is fixed by the five - dimensional @xmath18-term equation . the second term in the bracket is thus tiny in all region of the fifth dimension and can be dropped . we then find that @xmath175 provides additional suppression of the effective yukawa coupling in the present model . for @xmath179 , the corresponding zero mode is localized at @xmath9 and yields no suppression of yukawa couplings which come from the operator on the @xmath9 boundary . on the other hand , the @xmath180 case gives a yukawa suppression by the factor @xmath181 . this reflects the fact that the zero mode is peaked at away from the @xmath9 boundary . in the conformal limit @xmath182 , the normalization constant @xmath175 becomes a volume - suppression factor @xmath183 as in the case of flat extra dimension . supersymmetry - breaking spectrum is related to the radius stabilization and yukawa coupling structure . according to ( [ dff ] ) , there are three types of contributions to supersymmetry - breaking parameters . since we now introduced only boundary multiplets , the equation of motion of @xmath60 in five dimensions implies that the auxiliary field @xmath18 has a flat wavefunction in the fifth direction . therefore the @xmath18-term contribution is universal to all charged scalars in the theory . the @xmath54 component of @xmath20 provides soft masses and trilinear couplings of scalar fields from superpotential and/or khler terms . it is found that they are higher - dimensional operators and suppressed by powers of @xmath184 compared to the leading @xmath18-term contribution . ignoring these higher - dimensional corrections , charges , like scalar top quark , might receive the dominant soft masses from higher - dimensional khler terms involving @xmath20 . ] non - holomorphic scalar masses are given by @xmath185}\,\frac{f_t}{2r}\bigg|^2 \qquad\;\ ; \textrm{(bulk scalars ) } , \\ m^2_r & = & 2q_rm^2 \hspace*{4 cm } \textrm{($y=0 $ boundary scalars)},\end{aligned}\ ] ] where @xmath186 is the appropriate superspace extension of the normalization constant @xmath175 . the first terms are the @xmath18-term contributions which are positive definite ( @xmath187 ) . the second term in the bulk scalar mass @xmath188 comes from the radius modulus @xmath54 term ( [ dff ] ) . we have dropped the @xmath60 contribution in @xmath189 since it is suppressed in the wavefunction factor @xmath173 as discussed above . the boundary scalars at @xmath9 receive no @xmath190 contribution as seen from the lagrangian ( [ l4d ] ) and have rather different spectrum than those of bulk scalar fields . in case that quarks and leptons originate from bulk hypermultiplets , the standard model gauge multiplets must also reside in the five - dimensional bulk . then the gauginos obtain supersymmetry - breaking masses from two contributions ; the radius modulus @xmath54 term and higher - dimensional operators @xmath191 . the masses of zero - mode gauginos @xmath192 are given by @xmath193 let us concentrate on the vacuum with @xmath194 . in this vacuum , the radius modulus @xmath54 term is @xmath195 for bulk scalar masses , the dominant part therefore comes from the @xmath18-term contribution and the spectrum is similar to four - dimensional anomalous @xmath2 models . the @xmath18-term contributed scalar masses have rich phenomenological implications such as flavor violation @xcite . on the other hand , the two contributions to gaugino masses in ( [ gauginomass ] ) are comparable in size or the @xmath190 contribution can be dominant if low - energy effective theory has a weak gauge coupling constant @xmath196 . the gauginos are found to have rather non - universal ( non - unified ) mass spectrum in this scenario . as seen in this toy model , the existence of fi term provides various schemes to discuss phenomenological issues in higher - dimensional theory . it can stabilize the sizes of extra dimensions , create yukawa hierarchies , and predict characteristic sparticle spectrum testified in future particle experiments . therefore more realistic model construction along this line may deserve to be investigated . in this work we have discussed the fayet - iliopoulos @xmath18 term as a possible origin of radius stabilization in brane world models . we have presented three different schemes for the stabilization . the simplest case has no matter multiplets to stabilize the radius , but the theory needs some cancellation mechanism of gravitational anomaly . the one of the others contains bulk hypermultiplets , whose non - trivial wavefunctions connect the two boundaries and then fix the size of the extra dimension in terms of boundary couplings . on the other hand , the third model involves only boundary dynamics and does not need the presence of bulk matter fields . the radius is , in this case , determined so that the vacuum energy is minimized after supersymmetry breaking . every scheme can lead to a significant warp factor or near - flat metric , depending on the model parameters . it should be noted that , in any model , the fi term is not a device introduced just in order to stabilize the radius modulus . a non - vanishing fi term is radiatively generated even if it is set to be zero at classical level . an induced fi term depends on how charged matter multiplets are distributed in the extra dimensions and is therefore controllable . moreover it is known in four - dimensional models that the fi term is deeply connected with yukawa hierarchy and supersymmetry breaking . we have presented a toy model for fermion yukawa hierarchy correlated to the radius stabilization . the model also predicts characteristic spectrum of sfermions and gauginos . in the model we have drawn in section 4 , the bulk scalar fields develop non - trivial wavefunction profiles in the extra dimension and four - dimensional supersymmetry is unbroken . when included supersymmetry breaking , it might give impacts on sparticle spectroscopy and also deserve cosmological considerations . we leave these phenomenological analysis to future investigations . the authors wish to thank h. abe , k. choi , t. hirayama and h. nakano for valuable discussions . t. k. is supported in part by the grant - in - aid for scientific research ( # 16028211 ) and the grant - in - aid for the 21st century coe `` the center for diversity and universality in physics '' from the ministry of education , culture , sports , science and technology of japan . 99 i. antoniadis , phys . b * 246 * ( 1990 ) 377 ; i. antoniadis , c. munoz and m. quiros , nucl . b * 397 * ( 1993 ) 515 ; n. arkani - hamed , s. dimopoulos and g.r . dvali , phys . b * 429 * ( 1998 ) 263 ; i. antoniadis , n. arkani - hamed , s. dimopoulos and g.r . dvali , phys . b * 436 * ( 1998 ) 257 . l. randall and r. sundrum , phys . rev . * 83 * ( 1999 ) 3370 ; _ ibid . _ * 83 * ( 1999 ) 4690 . dienes , e. dudas and t. gherghetta , nucl . b * 557 * ( 1999 ) 25 ; n. arkani - hamed , s. dimopoulos , g.r . dvali and j. march - russell , phys . d * 65 * ( 2002 ) 024032 ; a.e . faraggi and m. pospelov , phys . b * 458 * ( 1999 ) 237 ; g.r . dvali and a.y . smirnov , nucl . b * 563 * ( 1999 ) 63 ; r.n . mohapatra , s. nandi and a. perez - lorenzana , phys . b * 466 * ( 1999 ) 115 ; y. grossman and m. neubert , phys . lett . b * 474 * ( 2000 ) 361 ; r. barbieri , p. creminelli and a. strumia , nucl . b * 585 * ( 2000 ) 28 ; a. lukas , p. ramond , a. romanino and g.g . ross , jhep * 0104 * ( 2001 ) 010 ; g. barenboim , g.c . branco , a. de gouvea and m.n . rebelo , phys . d * 64 * ( 2001 ) 073005 ; a. de gouvea , g.f . giudice , a. strumia and k. tobe , nucl . b * 623 * ( 2002 ) 395 ; h. davoudiasl , p. langacker and m. perelstein , phys . rev . d * 65 * ( 2002 ) 105015 ; t. appelquist , b.a . dobrescu , e. ponton and h.u . yee , phys . d * 65 * ( 2002 ) 105019 ; q.h . cao , s. gopalakrishna and c.p . yuan , phys . d * 69 * ( 2004 ) 115003 ; t. gherghetta , phys . * 92 * ( 2004 ) 161601 . dienes , e. dudas and t. gherghetta , nucl . b * 537 * ( 1999 ) 47 ; s.a . abel and s.f . king , phys . d * 59 * ( 1999 ) 095010 ; n. arkani - hamed and m. schmaltz , phys . d * 61 * ( 2000 ) 033005 ; h.c . cheng , phys . d * 60 * ( 1999 ) 075015 ; k. yoshioka , mod . lett . a * 15 * ( 2000 ) 29 ; dvali and m.a . shifman , phys . b * 475 * ( 2000 ) 295 ; m. bando , t. kobayashi , t. noguchi and k. yoshioka , phys . b * 480 * ( 2000 ) 187 ; phys . d * 63 * ( 2001 ) 113017 ; t. gherghetta and a. pomarol , nucl . b * 586 * ( 2000 ) 141 ; s.j . huber and q. shafi , phys . b * 498 * ( 2001 ) 256 ; s. nussinov and r. shrock , phys . b * 526 * ( 2002 ) 137 ; n. haba , y. shimizu , t. suzuki and k. ukai , prog . . phys . * 107 * ( 2002 ) 151 ; l. hall , j. march - russell , t. okui and d.r . smith , arxiv : hep - ph/0108161 ; d.e . kaplan and t.m.p . tait , jhep * 0111 * ( 2001 ) 051 ; a. hebecker and j. march - russell , phys . b * 541 * ( 2002 ) 338 ; r. kitano and t.j . li , phys . d * 67 * ( 2003 ) 116004 ; c.a . scrucca , m. serone and l. silvestrini , nucl . b * 669 * ( 2003 ) 128 ; c. biggio , f. feruglio , i. masina and m. perez - victoria , nucl . b * 677 * ( 2004 ) 451 ; z.y . han and w. skiba , phys . d * 70 * ( 2004 ) 035013 ; y. grossman , r. harnik , g. perez , m.d . schwartz and z. surujon , arxiv : hep - ph/0407260 . i. antoniadis , c. munoz and m. quiros , nucl . b * 397 * ( 1993 ) 515 ; e.a . mirabelli and m.e . peskin , phys . d * 58 * ( 1998 ) 065002 ; i. antoniadis , s. dimopoulos , a. pomarol and m. quiros , nucl . b * 544 * ( 1999 ) 503 ; l. randall and r. sundrum , nucl . b * 557 * ( 1999 ) 79 ; a. delgado , a. pomarol and m. quiros , phys . d * 60 * ( 1999 ) 095008 ; d.e . kaplan , g.d . kribs and m. schmaltz , phys . d * 62 * ( 2000 ) 035010 ; z. chacko , m.a . luty , a.e . nelson and e. ponton , jhep * 0001 * ( 2000 ) 003 ; t. kobayashi and k. yoshioka , phys . * 85 * ( 2000 ) 5527 ; r. barbieri , l. j. hall and y. nomura , phys . d * 63 * ( 2001 ) 105007 ; t. gherghetta and a. pomarol , nucl . b * 602 * ( 2001 ) 3 ; j. bagger , f. feruglio and f. zwirner , jhep * 0202 * ( 2002 ) 010 ; t. gherghetta and a. riotto , nucl . phys . b * 623 * ( 2002 ) 97 ; g. von gersdorff and m. quiros , phys . rev . d * 65 * ( 2002 ) 064016 ; a. delgado , g. von gersdorff and m. quiros , jhep * 0212 * ( 2002 ) 002 ; k.w . choi , d.y . kim , i.w . kim and t. kobayashi , eur . j. c * 35 * ( 2004 ) 267 ; j. bagger and m. redi , phys . b * 582 * ( 2004 ) 117 . n. arkani - hamed , s. dimopoulos and j. march - russell , phys . d * 63 * ( 2001 ) 064020 ; k.r . dienes , e. dudas , t. gherghetta and a. riotto , nucl . phys . b * 543 * ( 1999 ) 387 ; w.d . goldberger and m.b . wise , phys . * 83 * ( 1999 ) 4922 ; m.a . luty and r. sundrum , phys . d * 62 * ( 2000 ) 035008 ; j. garriga , o. pujolas and t. tanaka , nucl . b * 605 * ( 2001 ) 192 ; i. brevik , k.a . milton , s. nojiri and s.d . odintsov , nucl . b * 599 * ( 2001 ) 305 ; r. hofmann , p. kanti and m. pospelov , phys . d * 63 * ( 2001 ) 124020 ; b. grinstein , d.r . nolte and w. skiba , phys . d * 63 * ( 2001 ) 105016 ; a. mazumdar and a. perez - lorenzana , phys . b * 508 * ( 2001 ) 340 ; e. ponton and e. poppitz , jhep * 0106 * ( 2001 ) 019 ; s. nasri , p. j. silva , g.d . starkman and m. trodden , phys . d * 66 * ( 2002 ) 045029 ; j. garriga and a. pomarol , phys . b * 560 * ( 2003 ) 91 ; g. von gersdorff , m. quiros and a. riotto , nucl . phys . b * 689 * ( 2004 ) 76 ; b. grzadkowski and j.f . gunion , phys . d * 68 * ( 2003 ) 055002 ; n. maru and n. okada , phys . d * 70 * ( 2004 ) 025002 . p. fayet and j. iliopoulos , phys . b * 51 * ( 1974 ) 461 . d.m . ghilencea , s. groot nibbelink and h.p . nilles , nucl . phys . b * 619 * ( 2001 ) 385 . w. fischler , h.p . nilles , j. polchinski , s. raby and l. susskind , phys . lett . * 47 * ( 1981 ) 757 . l.e . ibanez and g.g . ross , phys . b * 332 * ( 1994 ) 100 ; e. dudas , s. pokorski and c.a . savoy , phys . lett . b * 369 * ( 1996 ) 255 ; g.r . dvali and a. pomarol , phys . * 77 * ( 1996 ) 3728 ; p. binetruy and e. dudas , phys . b * 389 * ( 1996 ) 503 . n. arkani - hamed , t. gregoire and j. wacker , jhep * 0203 * ( 2002 ) 055 ; d. marti and a. pomarol , phys . d * 64 * ( 2001 ) 105025 ; a. hebecker , nucl . b * 632 * ( 2002 ) 101 . scrucca , m. serone , l. silvestrini and f. zwirner , phys . b * 525 * ( 2002 ) 169 ; r. barbieri , r. contino , p. creminelli , r. rattazzi and c.a . scrucca , phys . d * 66 * ( 2002 ) 024025 . s. groot nibbelink , h.p . nilles and m. olechowski , phys . b * 536 * ( 2002 ) 270 ; nucl . b * 640 * ( 2002 ) 171 ; d. marti and a. pomarol , phys . d * 66 * ( 2002 ) 125005 ; h. abe , t. higaki and t. kobayashi , prog . . phys . * 109 * ( 2003 ) 809 ; s. groot nibbelink , jhep * 0307 * ( 2003 ) 011 ; h.m . lee , h.p . nilles and m. zucker , nucl . b * 680 * ( 2004 ) 177 . t. hirayama and k. yoshioka , arxiv : hep - th/0311233 . maldacena , adv . * 2 * ( 1998 ) 231 ; s.s . gubser , i.r . klebanov and a.m. polyakov , phys . b * 428 * ( 1998 ) 105 ; e. witten , adv . * 2 * ( 1998 ) 253 ; h. verlinde , nucl . b * 580 * ( 2000 ) 264 ; j. de boer , e. verlinde and h. verlinde , jhep * 0008 * ( 2000 ) 003 ; n. arkani - hamed , m. porrati and l. randall , jhep * 0108 * ( 2001 ) 017 ; r. rattazzi and a. zaffaroni , jhep * 0104 * ( 2001 ) 021 ; l. randall and m.d . schwartz , jhep * 0111 * ( 2001 ) 003 ; phys . * 88 * ( 2002 ) 081801 . k. sfetsos , nucl . b * 612 * ( 2001 ) 191 ; h. abe , t. kobayashi , n. maru and k. yoshioka , phys . d * 67 * ( 2003 ) 045019 ; a. falkowski and h.d . kim , jhep * 0208 * ( 2002 ) 052 ; l. randall , y. shadmi and n. weiner , jhep * 0301 * ( 2003 ) 055 . h. abe , k. choi and i.w . kim , arxiv : hep - th/0405100 ; f.p . correia , m.g . schmidt and z. tavartkiladze , arxiv : hep - th/0408138 . j. bagger , d. nemeschansky and r.j . zhang , jhep * 0108 * ( 2001 ) 057 ; m.a . luty and r. sundrum , phys . rev . * d64 * ( 2001 ) 065012 ; a. falkowski , z. lalak and s. pokorski , nucl . phys . * b613 * ( 2001 ) 189 . n. arkani - hamed , a.g . cohen and h. georgi , phys . b * 516 * ( 2001 ) 395 . l. pilo and a. riotto , phys . b * 546 * ( 2002 ) 135 . scrucca , m. serone and m. trapletti , nucl . b * 635 * ( 2002 ) 33 ; h.d . kim , j.e . kim and h.m . lee , jhep * 0206 * ( 2002 ) 048 ; f. gmeiner , s. groot nibbelink , h.p . nilles , m. olechowski and m.g . walter , nucl . b * 648 * ( 2003 ) 35 ; t. asaka , w. buchmuller and l. covi , nucl . b * 648 * ( 2003 ) 231 ; t. gherghetta and a. kehagias , phys . d * 68 * ( 2003 ) 065019 ; g.von gersdorff and m. quiros , phys . d * 68 * ( 2003 ) 105002 ; s. groot nibbelink , m. hillenbach , t. kobayashi and m.g . walter , phys . d * 69 * ( 2004 ) 046001 ; c.a . scrucca and m. serone , int . j. mod . phys . a * 19 * ( 2004 ) 2579 ; e. dudas , t. gherghetta and s. g. nibbelink , arxiv : hep - th/0404094 . froggatt and h.b . nielsen , nucl . b * 147 * ( 1979 ) 277 . b. murakami , k. tobe and j. d. wells , phys . b * 526 * ( 2002 ) 157 ; i. jack , d.r.t . jones and r. wild , phys . b * 535 * ( 2002 ) 193 ; t. kobayashi , h. nakano , h. terao and k. yoshioka , prog . * 110 * ( 2003 ) 247 ; k.s . babu , t. enkhbat and i. gogoladze , nucl . b * 678 * ( 2004 ) 233 .

in this paper we describe the asymptotic behavior , in the exponential time scale , of solutions to quasi - linear parabolic equations with a small parameter at the second order term and the long time behavior of corresponding diffusion processes . in particular , we discuss the exit problem and metastability for the processes corresponding to quasi - linear initial - boundary value problems . 2000 mathematics subject classification numbers : 60f10 , 35k55 . keywords : nonlinear perturbations , exit problem , metastability .

0903.0428

consider a dynamical system @xmath0 together with its stochastic perturbations @xmath1 here @xmath2 is a small parameter , @xmath3 is a wiener process in @xmath4 , and the coefficients @xmath5 and @xmath6 are assumed to be lipschitz continuous . the diffusion matrix @xmath7 is assumed to be non - degenerate for all @xmath8 . let @xmath9 be a bounded domain in @xmath4 with infinitely smooth boundary @xmath10 . in this paper , with the exception of the last section , we assume that there is a point @xmath11 such that for each @xmath12 the trajectory of the dynamical system ( [ dsyst ] ) starting at @xmath8 is attracted to @xmath13 . we assume that @xmath14 for @xmath15 , where @xmath16 is the exterior normal to the boundary of @xmath9 . let @xmath17 be the first time when @xmath18 reaches the boundary of @xmath9 . if @xmath19 is small , then on any finite time interval the trajectories of the process @xmath18 defined by ( [ perturb ] ) are close to the corresponding non - perturbed trajectory with probability close to one . therefore , with high probability @xmath20 enters a small neighborhood of the equilibrium point @xmath13 before leaving @xmath9 . the process eventually exits @xmath9 as a result of large deviations of @xmath21 from @xmath22 ( @xcite , see also @xcite ) . the large deviations are governed by the normalized action functional @xmath23 , \overline{d}),\ ] ] and the quasi - potential @xmath24 , \overline{d } ) , \varphi(0 ) = x_0 , \varphi(t ) = x \},~~x \in \overline{d}.\ ] ] here @xmath25 be the elements of the inverse matrix , that is @xmath26 , and @xmath27 if @xmath28 is not absolutely continuous . it is proved in @xmath29 that @xmath30 converges in probability , as @xmath31 , to @xmath32 . moreover , if the minimum @xmath33 of @xmath34 on @xmath10 is achieved at a unique point @xmath35 ( which is true in the generic case ) , then @xmath36 converges to @xmath35 in probability as @xmath37 . these statements imply various results for pde s with a small parameter at the second order derivatives . in particular , consider the following initial - boundary value problem : @xmath38 @xmath39 where @xmath40 , for the sake of brevity , is assumed to be continuous on @xmath41 . the case when @xmath42 with @xmath43 can be considered in a similar way . assume that the minimum @xmath33 of @xmath34 on @xmath10 is achieved at a unique point @xmath35 . let @xmath44 be a function such that @xmath45 as @xmath31 with @xmath46 , that is @xmath47 as @xmath31 . then @xmath48 @xmath49 for @xmath12 . note that the solution to ( [ y1])-([y2 ] ) can be expressed in terms of the transition semigroup associated with the family of processes @xmath20 , @xmath50 . namely , let @xmath51 , @xmath52 . then the function @xmath53 is the solution to ( [ y1])-([y2 ] ) . the semigroup @xmath54 can be viewed as a small perturbation of the semigroup of shifts @xmath55 associated with the dynamical system ( [ dsyst ] ) . more general perturbations of @xmath56 may lead to nonlinear semigroups . namely , consider the following problem : @xmath57 @xmath58 when the coefficients are sufficiently smooth and the matrix @xmath59 is positive - definite , the solution @xmath60 exists and is unique in the appropriate function space ( see section [ qle ] below ) . we can now define the semigroup @xmath54 on @xmath61 via @xmath62 , where @xmath60 is the solution of ( [ eq1])-([bc ] ) with initial - boundary data @xmath40 . for @xmath63 and @xmath50 , we can define @xmath64 , @xmath65 $ ] , as the process which starts at @xmath8 and solves @xmath66 @xmath67 where @xmath68 , @xmath69 , are lipschitz continuous and such that @xmath70 . the process @xmath64 will be called the nonlinear stochastic perturbation of ( [ dsyst ] ) . more precisely , @xmath64 corresponds to the nonlinear semigroup defined by ( [ eq1 ] ) . as in the linear case , we have the following relation between @xmath60 and the process @xmath64 : @xmath71 one of the important questions in the study of parabolic linear and nonlinear equations is the one concerning the behavior of solutions ( or , in probabilistic terms , behavior of the corresponding diffusion process ) as @xmath72 . in our case , when the small parameter @xmath73 is present in front of the second order term , the limit of @xmath74 as @xmath75 , @xmath72 , depends on the manner in which @xmath76 approaches @xmath77 . in the linear case this problem has been studied in @xcite ( see also @xcite , @xcite ) . in section [ asc ] we study the asymptotic behavior of solutions to ( [ eq1])-([bc ] ) when @xmath31 and @xmath78 . as a first step , we shall introduce a family of linear problems which can be obtained from ( [ eq1])-([bc ] ) by replacing the second variable in the coefficients @xmath79 in the right hand side of ( [ eq1 ] ) by a constant @xmath80 . the asymptotics of @xmath60 can be then expressed in terms of the functions @xmath81 and @xmath82 , where @xmath81 is the minimum of the quasi - potential of the linear problem and @xmath83 is the point where this minimum is achieved . in section [ exitprob ] we study the exit problem for the process @xmath64 . we shall see that new effects appear when nonlinear stochastic perturbations are considered . in particular , even in the generic case , the distribution of the exit location @xmath84 need not be concentrated in one point . some related problems concern the notion of metastability for nonlinear perturbations of dynamical systems with several equilibrium points . let us consider the dynamical system ( [ dsyst ] ) in @xmath4 and its perturbations ( [ perturb ] ) . as before , @xmath5 and @xmath6 are assumed to be lipschitz continuous . now we shall assume that the system has a finite number of asymptotically stable equilibrium points @xmath85 such that for almost every @xmath86 , with respect to the lebesgue measure , the trajectory of ( [ dsyst ] ) starting at @xmath8 is attracted to one of the points @xmath87 . we shall also assume that the vector field @xmath6 satisfies @xmath88 for some positive constants @xmath89 and @xmath90 . the case of more general asymptotically stable attractors ( for instance , limit cycles ) can be considered similarly , however for the sake of brevity we restrict ourselves to the case of equilibriums . the general theory of metastability was developed in @xcite in the framework of large deviations ( see also @xcite , @xcite , @xcite ) . it was shown , in particular , that for a generic vector field @xmath6 satisfying the assumptions above , for almost every @xmath86 and @xmath46 , with probability which tends to one when @xmath31 , the trajectory @xmath20 of ( [ perturb ] ) spends most of the time in the time interval @xmath91 $ ] near a point @xmath92 . this point is called the metastable state for the trajectory starting at @xmath8 in the time scale @xmath93 . the metastable state can be determined by examining the values of the quasi - potential . namely , let @xmath94 , \mathbb{r}^d ) , \varphi(0 ) = x_i , \varphi(t ) = x_j \},~~1 \leq i , j \leq m.\ ] ] these numbers determine a hierarchy of cycles along which the system switches from one metastable state to another with the growth of @xmath95 ( @xcite ) . we can also study metastability for nonlinear perturbations of dynamical systems . it turns out that now the transition between the equilibrium points does not occur `` immediately in the exponential time scale '' . this implies that now metastable states should be replaced by metastable distributions between the equilibriums . the description of metastable distributions is based on the study of the asymptotic behavior of solutions to ( [ eq1])-([bc ] ) when @xmath31 and @xmath78 . note that metastable distributions also arise in @xcite , @xcite , but for reasons which are different from what is discussed in this paper . such a modification to the notion of metastability leads to a modified notion of stochastic resonance . we briefly address the problems of metastability in section [ genex ] , where we also consider other generalizations and some examples . the issue of metastability in the case of an arbitrary number of equilibrium points and cycles will be addressed in a forthcoming paper . in this paper we considered nonlinear perturbations of a system with an asymptotically stable equilibrium . in this case , the exit from a domain containing this equilibrium occurs due to large deviations , and the exit time and exit distribution essentially depend on the perturbation . a related singular perturbation problem arises in the case when the equilbrium is stable but not asymptotically stable : for instance when the unperturbed system is hamiltonian . nonlinear stochastic perturbations in this case lead to a nonlinear version of the averaging principle . say , in the case of one degree of freedom , the limiting slow motion is a diffusion process corresponding to a nonlinear operator on the graph ( compare with @xcite , chapter 8) related to the hamiltonian . we will consider these problems in one of the forthcoming papers . let @xmath96 be a bounded domain with infinitely smooth boundary @xmath10 . we shall say that @xmath97 belongs to @xmath98 if @xmath99 and all of its partial derivatives up to the second order are bounded and continuous in @xmath9 . we shall say that a function @xmath100 belongs to @xmath101 if @xmath99 , its partial derivative in @xmath102 , and all of its partial derivatives up to the second order in @xmath8 are bounded and continuous in @xmath103 . note that a function @xmath104 can be extended to a continuous function on @xmath41 and @xmath105 can be extended to a continuous function on @xmath106 . let @xmath107 , @xmath108 , and @xmath109 , @xmath110 . we also assume that there is a positive constant @xmath111 such that @xmath112 , @xmath113 , @xmath114 . let @xmath40 be an infinitely smooth function defined in a neighborhood of @xmath41 . if @xmath2 and the coefficients @xmath59 and @xmath6 and the function @xmath40 satisfy the assumptions listed above , then the equation ( [ eq1])-([bc ] ) has a unique solution in the class of functions @xmath115 ( see theorem 5 , chapter 6.2 of @xcite ) . if @xmath40 were to be only continuous on @xmath41 , the existence and uniqueness of solutions to ( [ eq1])-([bc ] ) would hold in the class of functions which are locally @xmath116-smooth inside @xmath103 and continuous up to the boundary . however , to simplify notations in later sections , we impose the smoothness condition on @xmath40 . let @xmath117 be a symmetric @xmath118 matrix whose elements @xmath119 are bounded and lipschitz continuous on @xmath120 and satisfy @xmath121 , @xmath86 , @xmath122 . let @xmath123 be the elements of the inverse matrix , that is @xmath124 , and @xmath5 be a square matrix such that @xmath125 . we choose @xmath5 in such a way that @xmath68 are also bounded and lipschitz continuous . let @xmath126 be the normalized action functional for the family of processes @xmath20 satisfying @xmath127 where @xmath6 is a bounded lipschitz continuous vector field on @xmath4 . thus @xmath128 for absolutely continuous @xmath28 defined on @xmath129 $ ] , @xmath130 , and @xmath131 if @xmath28 is not absolutely continuous or if @xmath132 ( see @xcite ) . let @xmath133 be the quasi - potential for the family @xmath20 in @xmath41 , that is @xmath134 , \overline{d } ) , \varphi(0 ) = x , \varphi(t ) = y \},~~x , y \in \overline{d}.\ ] ] recall that @xmath135 for @xmath136 , where @xmath16 is the exterior normal to the boundary of @xmath9 . we shall assume that there is an equilibrium point @xmath137 for the vector field @xmath6 , and that all the trajectories of the dynamical system @xmath138 starting in @xmath9 are attracted to @xmath13 . we also assume that there is @xmath139 such that @xmath140 for some positive constant @xmath80 and all @xmath8 in the @xmath141-neighborhood of @xmath13 . let @xmath142 , @xmath143 , and @xmath60 be the solution of ( [ eq1])-([bc ] ) . we shall be interested in the asymptotic behavior of @xmath144 , where @xmath95 is fixed , @xmath145 , and @xmath146 . let @xmath147 thus @xmath148 \subseteq [ g_{\rm min } , g_{\rm max}]$ ] . let @xmath149 \rightarrow \mathbb{r}$ ] be defined by @xmath150 where @xmath151 is extended to an arbitrary bounded lipschitz continuous function satisfying @xmath152 , @xmath114 , @xmath153 . we next make some assumptions about the quasi - potential . it is not difficult to see that these assumptions are satisfied by a quasi - potential corresponding to generic @xmath59 and @xmath6 . we shall assume that for all but finitely many points @xmath154 $ ] the minimum in ( [ eqt ] ) is attained at a single point which will be denoted by @xmath83 . we assume that in the remaining points @xmath155 the minimum is attained at two points of the boundary . in this case the function @xmath156 \rightarrow \partial d$ ] is piece - wise continuous and has left and right limits at the points of discontinuity , as follows from the formula for the quasi - potential . let @xmath157 if @xmath158 and @xmath159 if @xmath160 , @xmath161 . if @xmath162 , we define @xmath163 as the point distinct from @xmath164 where the minimum of the quasi - potential is attained , and similarly we define @xmath164 if @xmath165 as the point distinct from @xmath163 where the minimum of the quasi - potential is attained . if @xmath177 , then @xmath178 . if @xmath179 , then @xmath180 . + note that @xmath181 $ ] since @xmath182 ) \subseteq [ g_1,g_2]$ ] . we shall require that the graph of @xmath183 pass from the left of the diagonal to the right of the diagonal at @xmath176 . more precisely , we shall assume that if @xmath184 , then for every @xmath185 there exists @xmath186 $ ] such that @xmath187 and if @xmath188 , then for every @xmath185 there exists @xmath186 $ ] such that @xmath189 we also require that @xmath190 not coincide with any of the points of discontinuity @xmath169 for which @xmath191 . let @xmath192 and define function @xmath193 as follows : for @xmath194 , let @xmath195 . for @xmath196 and @xmath197 , let @xmath195 . for @xmath196 and @xmath198 , let @xmath199 , m(c ) = \lambda \ } \}$ ] . for @xmath196 and @xmath200 , let @xmath201 , m(c ) = \lambda \ } \}$ ] . here we use the convention that the minimum of an empty set is @xmath202 and the maximum of an empty set is @xmath203 . ( see picture 1 , where the thick line represents the graph of the function @xmath193 and @xmath204 is a point of discontinuity for the function @xmath193 . ) we also define @xmath205 } m(c)$ ] if @xmath206 and @xmath207 } m(c)$ ] if @xmath208 . [ pic1 ] the thick line represents the graph of @xmath193 [ mte ] let the above assumptions concerning the differential operator @xmath209 and the function @xmath183 be satisfied . suppose that the function @xmath193 is continuous at a point @xmath210 . then for every @xmath142 the following limit @xmath211 is uniform in @xmath145 , where @xmath60 is the solution to ( [ eq1])-([bc ] ) . * remark . * from theorem [ mte ] and the definition of the function @xmath193 it follows that @xmath212 uniformly in @xmath145 if @xmath213 . moreover , from the proof of theorem [ mte ] provided below it easily follows that the limit is uniform in @xmath214 for each @xmath215 . therefore , for each @xmath142 and @xmath215 there is @xmath216 such that @xmath217 whenever @xmath218 , @xmath145 and @xmath219 . it is important to note that with the boundary values of @xmath40 fixed , the limit @xmath176 may still depend on the initial function through its value @xmath190 at the equilibrium point . in the generic case , when the interval @xmath220 $ ] can be represented as a finite union of intervals @xmath221 , such that on the interior of each of the intervals the function @xmath222 is either strictly positive or strictly negative , the values of @xmath190 belonging to the interior of the same interval will correspond to the same value of @xmath176 . + the proof of theorem [ mte ] will use some properties of diffusion processes stated in the following section . theorem [ mte ] implies various results concerning the exit problem and metastability for the process @xmath64 defined above . these questions will be considered in sections [ exitprob ] and [ genex ] . in this section we shall consider diffusion processes which are somewhat more general than those introduced in section [ af ] . namely , we shall allow the diffusion matrix to be time dependent . the results stated in this section easily follow from the arguments of @xcite , chapter 4 . let @xmath117 be a symmetric @xmath118 matrix whose elements @xmath119 are lipschitz continuous on @xmath223 and satisfy @xmath224 where @xmath111 and @xmath225 are positive constants . let @xmath5 be a square matrix such that @xmath125 . we choose @xmath5 in such a way that @xmath68 are also bounded and lipschitz continuous . let @xmath20 satisfy @xmath226 and @xmath227 where @xmath6 is a bounded lipschitz continuous vector field on @xmath4 . clearly , the law of this process depends on @xmath5 only through @xmath125 . for @xmath12 , let @xmath228 be the first time when the process reaches the boundary of @xmath9 . thus @xmath229 is the location of the first exit of the process @xmath230 from the domain @xmath9 . if @xmath117 is close to a function which does not depend on time , then the asymptotics , as @xmath31 , of @xmath229 and @xmath228 can be described in terms of the quasi - potential . more precisely , let @xmath231 be a bounded lipschitz continuous matrix valued function on @xmath4 such that @xmath232 where @xmath233 . let @xmath234 be the set of points in @xmath10 at which @xmath235 is attained . this minimum will be denoted by @xmath236 . [ au1 ] suppose @xmath237 is as above , and positive constants @xmath111 and @xmath225 are fixed . for every @xmath142 there is positive @xmath238 and a function @xmath239 with @xmath240 , such that for every @xmath117 that is lipschitz continuous , satisfies ( [ elliptic1 ] ) and @xmath241 and every @xmath145 we have : ( a ) @xmath242 , ( b ) @xmath243 ( c ) @xmath244 . the next lemma only requires the boundedness of the quadratic form @xmath117 from above and below . [ sec2 ] suppose that positive constants @xmath111 and @xmath225 are fixed . there exists @xmath245 such that for every @xmath246 there is a function @xmath239 with @xmath240 , such that for every @xmath117 that is lipschitz continuous and satisfies ( [ elliptic1 ] ) and every @xmath145 we have : \(a ) @xmath247 ( b ) @xmath248 for all @xmath249 ) \geq 1 - \rho(\varepsilon),$ ] ( c ) @xmath250 for all @xmath251 ) \geq 1 - \rho(\varepsilon)$ ] . an easy corollary of this lemma is that at an exponential time the process either can be found in a small neighborhood of @xmath13 or has earlier crossed the boundary of the domain . [ co1 ] suppose that positive constants @xmath111 and @xmath225 are fixed . for every @xmath142 there is a function @xmath252 with @xmath253 , such that for every @xmath117 that is lipschitz continuous and satisfies ( [ elliptic1 ] ) , every @xmath12 and @xmath254 we have : @xmath255 let @xmath256 be sufficiently small so that there is a domain @xmath257 with smooth boundary such that @xmath258 . if the process does not reach @xmath10 by the time @xmath259 , then we can apply part ( b ) of lemma [ sec2 ] to the domain @xmath257 and the process starting at @xmath260 , and the result follows from the markov property . the next step in the proof of theorem [ mte ] is to establish that @xmath261 is nearly constant on @xmath262 if @xmath63 is fixed and @xmath19 is sufficiently small . this is accomplished in lemma [ letwo ] below . [ small ] for every positive @xmath263 and @xmath264 there are positive @xmath265 and @xmath266 such that @xmath267 whenever @xmath268 , @xmath269 and @xmath270 . let @xmath271 , @xmath272 , @xmath273 . let @xmath274 and @xmath275 , where @xmath276 is the ball of radius @xmath141 centered at the origin . then @xmath277 satisfies the following partial differential equation : @xmath278 here @xmath279 are uniformly bounded in @xmath19 and satisfy @xmath280 , @xmath272 , @xmath281 , @xmath114 . moreover , @xmath282 is bounded uniformly in @xmath19 and @xmath283 can be estimated from above by a constant times @xmath284 for @xmath285 , uniformly in @xmath19 . since the distance between @xmath286 and the boundary of @xmath287 is positive and @xmath288 is uniformly bounded in @xmath287 by @xmath289 , we can apply the a priori estimate ( see theorem 4 , chapter 5.2 of @xcite or theorem 6 , chapter 6.2 of @xcite ) to bound latexmath:[$\sup_{(t , x ) \in q_0 } @xmath19 . this immediately implies ( [ epsi ] ) . we shall need the following simple lemma about diffusion processes with the drift directed towards the origin . [ dipro ] let @xmath291 be a @xmath292 smooth vector field on @xmath120 such that @xmath293 for some positive @xmath294 and all @xmath86 . let @xmath295 be a lipschitz continuous function such that @xmath296 , @xmath297 , for all @xmath298 and @xmath86 . let @xmath299 be the process starting at @xmath8 that satisfies @xmath300 then for every @xmath301 there are @xmath264 , @xmath302 and @xmath266 , which depend on @xmath291 and @xmath303 only through @xmath294 , @xmath304 , such that @xmath305 holds for @xmath306 and @xmath307 . let @xmath308 $ ] be a smooth even function with negative derivative on @xmath309 , such that @xmath310 for @xmath311 and @xmath312 for @xmath313 . let @xmath314 if @xmath264 and @xmath302 are sufficiently large and @xmath266 is sufficiently small , then @xmath315 for @xmath306 , @xmath316 . by the ito formula , @xmath317 @xmath318 where @xmath319 is the generator of the process @xmath299 . in order to estimate the integral in the right hand side , we note that @xmath320}(h''(x ) ) \exp(-2k_1 t)/r^2,\ ] ] where the constant @xmath265 depends on @xmath294 and @xmath304 . by taking @xmath264 sufficiently large , we can bound the expectation of the integral from below by @xmath321 , thus proving ( [ inside2 ] ) . [ letwo ] for every positive @xmath322 and @xmath323 there is positive @xmath266 such that @xmath324 whenever @xmath145 , @xmath325 and @xmath326 . choose @xmath327 small enough so that @xmath328 for some positive constant @xmath80 and all @xmath329 . first , let us prove a slightly weaker version of the lemma , namely that under the same assumptions ( [ kap ] ) holds for all @xmath330 , @xmath325 and @xmath326 . let @xmath331 , @xmath332 $ ] , be the process that starts at @xmath8 and satisfies @xmath333 @xmath334 where @xmath335 , @xmath108 , are lipschitz continuous and such that @xmath336 . ( if the minimum in the definition of @xmath228 is taken over an empty set , then it is considered to be equal to @xmath337 . ) in order to avoid confusion , let us note that we switched the notation , using @xmath331 with superscript @xmath95 for the process on the interval @xmath338 $ ] with @xmath339 inside the coefficient in the right hand side of ( [ pro ] ) . the notation @xmath64 with superscript @xmath102 still applies to the process defined in ( [ smttt ] ) . given @xmath340 , we can define @xmath341 from lemma [ sec2 ] applied to the domain @xmath342 and lemma [ dipro ] it follows that for every @xmath343 there are positive @xmath302 , @xmath266 and @xmath264 such that @xmath344 for all @xmath326 , @xmath330 and @xmath325 . using the feynman - kac formula , we can write @xmath345 as follows : @xmath346 therefore , @xmath347 the first term in the right hand side can be made smaller than @xmath348 by ( [ inside3x ] ) , while the second term can be made smaller than @xmath348 by lemma [ small ] . thus ( [ kap ] ) is true if we require that @xmath330 . next , let us prove the original statement . for @xmath245 , define @xmath349 from lemma [ sec2 ] it follows that there is @xmath350 such that @xmath351 where the limit is uniform in @xmath326 , @xmath352 . using the feynman - kac formula , we can write @xmath345 as follows : @xmath353 therefore , @xmath354 @xmath355 the first term in the right hand side tends to zero by ( [ inside3 ] ) , while the second term tends to zero since we proved that the lemma is true if we require that @xmath330 . in this section we prove theorem [ mte ] . first , we examine the behavior of @xmath60 for times which are small in the logarithmic scale . [ simp ] there is a positive @xmath356 such that for every @xmath246 there is @xmath216 such that @xmath357 whenever @xmath145 , @xmath358 and @xmath359 . this lemma immediately follows from lemma [ sec2 ] . + the next three lemmas , central to the proof of theorem [ mte ] , rule out certain types of behavior for the function @xmath60 . [ fo ] suppose that @xmath360 for some constants @xmath361 , @xmath362 as @xmath363 , and @xmath364 with @xmath365 . then there is @xmath142 such that @xmath366 for all large enough @xmath367 . consider the process @xmath368 given by ( [ pro ] ) with @xmath369 being the first time when this process reaches the boundary of @xmath9 . define @xmath370 then @xmath371 the left hand side in this formula is equal to @xmath372 . if ( [ gap ] ) does not hold , then by part ( c ) of lemma [ sec2 ] and lemma [ letwo ] the right hand side can be made arbitrarily close to @xmath373 along a subsequence . [ tet ] suppose that @xmath360 for some constants @xmath361 , @xmath362 as @xmath363 , and @xmath374 if @xmath375 , then neither of the following is possible : \(a ) there is @xmath142 such that @xmath376 , \(b ) there is @xmath142 such that @xmath377 for @xmath378 $ ] . + if @xmath379 , then neither of the following is possible : ( a@xmath380 ) there is @xmath142 such that @xmath376 , ( b@xmath380 ) there is @xmath142 such that @xmath381 for @xmath378 $ ] . \(a ) let us assume that @xmath376 . we can find @xmath382 $ ] such that @xmath383 and @xmath384 for @xmath385 $ ] . let @xmath386 . let @xmath387 be the first time when the process defined in ( [ diffpt ] ) reaches the boundary of @xmath9 . by part ( b ) of lemma [ au1 ] , we can choose @xmath388 such that whenever @xmath117 is lipschitz continuous and satisfies ( [ elliptic1 ] ) and ( [ close ] ) , we have @xmath389 for @xmath145 , where @xmath390 does not depend on @xmath117 and satisfies @xmath240 . choose @xmath391 such that @xmath392 whenever @xmath393 $ ] , @xmath394 . choose a sequence @xmath395 $ ] such that @xmath396 and @xmath397 $ ] for @xmath398 $ ] . by lemma [ letwo ] , we have @xmath399 , @xmath400 $ ] , if @xmath401 is sufficiently small . consider the process @xmath402 given by ( [ pro ] ) , with @xmath369 now being the first time when this process reaches the boundary of @xmath9 , and @xmath403 then @xmath404 the left hand side in this formula is equal to @xmath372 , while the right hand side can be made arbitrarily close to @xmath405 by considering sufficiently small @xmath401 due to ( [ er1 ] ) , corollary [ co1 ] ( which applies due to lemma [ fo ] ) and lemma [ letwo ] . this leads to a contradiction . \(b ) assume that @xmath377 for @xmath378 $ ] . let @xmath234 be the set ( consisting of either one or two points ) where the minimum of @xmath406 is attained . from the definition of @xmath183 it follows that @xmath407 . by part ( c ) of lemma [ au1 ] , we can choose @xmath388 such that whenever @xmath117 is lipschitz continuous and satisfies ( [ elliptic1 ] ) and ( [ close ] ) , we have @xmath408 for @xmath145 and all sufficiently small @xmath19 . as in case ( a ) , we can find @xmath409 and @xmath410 such that @xmath411 and ( [ as ] ) holds for @xmath412 , @xmath413 $ ] , if @xmath401 is sufficiently small . we can again employ formula ( [ interv ] ) in which the left hand side is equal to @xmath372 . the right hand side can be written as @xmath414 @xmath415 the first term in the right hand side here can be made arbitrarily close to @xmath416 by corollary [ co1 ] and lemma [ letwo ] . the second term on the right hand side can be estimated from above for large @xmath367 by @xmath417 due to ( [ pp ] ) . this leads to a contradiction . the proof of ( a@xmath380 ) and ( b@xmath380 ) is completely similar to the proof of ( a ) and ( b ) . [ rone ] suppose that @xmath418 . if @xmath198 , then @xmath419 if @xmath200 , then @xmath420 we shall only consider the first statement since the second one is completely similar . note that @xmath421 indeed , otherwise by lemma [ simp ] there are @xmath422 and sequences @xmath423 and @xmath362 such that ( [ ml ] ) holds . note that the graph of @xmath183 goes above the diagonal in a neighborhood of @xmath190 , while @xmath373 and @xmath372 can be taken arbitrarily close to @xmath190 . therefore , part ( b@xmath380 ) of lemma [ tet ] leads to a contradiction . thus , if ( [ fsta ] ) does not hold , then is there are @xmath424 and a sequence @xmath362 such that @xmath425 we choose @xmath323 sufficiently small so that the graph of @xmath183 goes above the diagonal on the interval @xmath426 $ ] . take @xmath343 which will be specified later . for each @xmath427 $ ] , by part ( a ) of lemma [ au1 ] , we can choose @xmath428 such that whenever @xmath117 is lipschitz continuous and satisfies ( [ elliptic1 ] ) and ( [ close ] ) with @xmath429 we have @xmath430 @xmath431}g(\overline{c } ) - \delta ' ) \geq 1 - \rho(\varepsilon)\ ] ] for each @xmath427 $ ] , find @xmath432 such that @xmath433 whenever @xmath434 $ ] , @xmath435 . choose a finite subcovering of the interval @xmath436 $ ] by the intervals @xmath437 and take @xmath438 . let @xmath439 be such that @xmath440 , @xmath167 . we claim that if @xmath441 , @xmath442 and @xmath443 $ ] along a subsequence , then @xmath444 for large enough @xmath367 along the same subsequence . if this were not the case , then we would have @xmath445~~{\rm and}~~u^{\varepsilon_n}(\exp(\lambda'/\varepsilon_n^2 ) , x_0 ) \geq \beta_{i-1}\ ] ] along a further subsequence . note that the function @xmath446 must take values in the interval @xmath447 $ ] for @xmath448 $ ] , otherwise part ( b@xmath380 ) of lemma [ tet ] leads to a contradiction . by the construction above and lemma [ letwo ] , there is @xmath449 such that @xmath450 and @xmath451 for @xmath452 , @xmath453 $ ] , if @xmath401 is sufficiently small . consider the process @xmath454 given by ( [ pro ] ) , with @xmath369 being the first time when this process reaches the boundary of @xmath9 , and @xmath455 then @xmath456 the left hand side does not exceed @xmath457 . if @xmath458 ( which is true if @xmath459 is sufficiently small and @xmath367 is sufficiently large ) , then from ( [ hn ] ) and ( [ hn2 ] ) it follows that the right hand side can be made larger than @xmath460}g(\overline{c } ) - 2\delta'$ ] . this leads to a contradiction if @xmath459 is small enough since @xmath183 is a piece - wise continuous function which stays above the diagonal on @xmath461 $ ] . we have thus established that @xmath444 . we can then extract a further subsequence such that @xmath462 belongs to one of the intervals @xmath463 $ ] with @xmath464 . we can then take @xmath465 and repeat the argument above to show that @xmath466 . after at most @xmath111 such steps , we obtain @xmath467 such that @xmath468 along a subsequence , and @xmath469 can be chosen to be arbitrarily close to @xmath95 . this , however , is a contradiction with ( [ spll ] ) . + _ proof of theorem [ mte]_. by lemma [ letwo ] , it is sufficient to prove that @xmath470 _ case 1 _ : @xmath471 . assume that @xmath472 take @xmath473 such that @xmath474 . by lemma [ simp ] , there are sequences @xmath362 and @xmath475 such that ( [ ml ] ) holds . thus part ( a ) of lemma [ tet ] leads to a contradiction with ( [ fs1x ] ) . the inequality @xmath476 can be ruled out in the same way by referring to part ( a@xmath380 ) of lemma [ tet ] . case 2 : @xmath196 , @xmath197 . assume that @xmath477 then , since @xmath183 is piece - wise continuous and passes from the left of the diagonal to the right of the diagonal at @xmath176 , we can find @xmath142 and @xmath373,@xmath372 such that @xmath478 and @xmath377 for @xmath378 $ ] . by lemma [ simp ] , there are sequences @xmath362 and @xmath475 such that ( [ ml ] ) holds . thus part ( b ) of lemma [ tet ] leads to a contradiction with ( [ fs1x3 ] ) . the inequality @xmath476 can be ruled out in the same way by referring to part ( b@xmath380 ) of lemma [ tet ] . case 3 : @xmath196 , @xmath479 . first assume that @xmath480 we can repeat the arguments of case 2 to show that ( [ fs1x3z ] ) implies that @xmath481 . then , since @xmath95 is a point of continuity of @xmath193 , we can find @xmath482 such that @xmath483 and @xmath474 . by lemma [ simp ] , there are sequences @xmath362 and @xmath484 such that ( [ ml ] ) holds . thus part ( a ) of lemma [ tet ] leads to a contradiction with ( [ fs1x3z ] ) . finally , from lemma [ rone ] it follows that @xmath485 case 4 : @xmath196 , @xmath200 . this is completely similar to case 3 . + * remark . * if instead of the constant @xmath95 in the argument of the function @xmath60 in theorem [ mte ] , we have a positive function @xmath486 such that @xmath487 , then @xmath488 the proof of this statement requires only simple modifications to the proof of theorem [ mte ] . let the differential operator @xmath209 and the function @xmath183 satisfy the assumptions of section [ asc11 ] . let @xmath12 and @xmath46 . recall that @xmath331 , @xmath332 $ ] , is the process defined in ( [ pro ] ) , with @xmath228 being the first time when this process reaches the boundary of @xmath9 . we put @xmath489 on the event that the process does not reach the boundary by the time @xmath339 . let @xmath490 . thus , if @xmath491 , then @xmath492 is the location where the process first exits the domain . let @xmath493 be the measure on @xmath41 induced by @xmath492 : @xmath494 let @xmath495 be the restriction of @xmath493 to @xmath10 : @xmath496 note that @xmath495 is not a probability measure , since @xmath497 . in this section we shall examine the asymptotics of @xmath493 and @xmath495 when @xmath31 . we shall distinguish several cases corresponding to different values of @xmath95 . first consider the case when @xmath498 . [ fox ] if @xmath12 and @xmath194 , then @xmath499 . from lemma [ sec2 ] it follows that for @xmath142 there is @xmath500 such that @xmath501 uniformly in @xmath145 . we claim that for each @xmath502 , @xmath503 \times d^\varkappa.\ ] ] indeed , otherwise by lemma [ letwo ] we could find sequences @xmath362 and @xmath504 such that either @xmath505 or @xmath506 . suppose that the former is the case and that @xmath197 ( the argument in the cases when @xmath198 and @xmath200 is similar ) . then , from the conditions imposed on the function @xmath183 in section [ asc11 ] it follows that there are @xmath507 such that the graph of @xmath183 goes below the diagonal in a neighborhood of the interval @xmath508 $ ] . moreover , there are @xmath509 such that @xmath510 and @xmath511 . this contradicts part ( b ) of lemma [ tet ] , thus establishing ( [ ucno ] ) . from ( [ ucno ] ) and part ( a ) of lemma [ au1 ] it follows that @xmath512 tends to zero as @xmath31 . from corollary [ co1 ] it then follows that @xmath513 tends to one as @xmath31 . by the markov property of the process and due to ( [ smmm ] ) , this implies the statement of the lemma . next , let us examine the case when @xmath213 . ( recall that @xmath205 } m(c)$ ] if @xmath206 and @xmath207 } m(c)$ ] if @xmath208 . ) [ fttw ] suppose that @xmath12 and @xmath514 . if the minimum of the quasi - potential @xmath515 is achieved at a single point @xmath516 , then @xmath495 weakly converges to a probability measure @xmath517 concentrated at @xmath516 . if the minimum is achieved at two points @xmath518 and @xmath519 and @xmath520 , then @xmath495 weakly converges to a probability measure @xmath517 concentrated at those two points . in this case @xmath521 . let @xmath522 . similarly to the proof of lemma [ fox ] , and using the fact that @xmath523 converges to @xmath176 for each @xmath213 , we can show that @xmath524 \times d^\varkappa,\ ] ] for each @xmath525 . let @xmath526 if the minimum @xmath515 is achieved at a single point , and @xmath527 if the minimum is achieved at two points . recall that @xmath228 is the first time when the process @xmath331 reaches the boundary of @xmath9 . from ( [ ucnov ] ) and lemma [ au1 ] it follows that @xmath528 and for every @xmath142 we have @xmath529 this immediately implies the desired result for the case of a single minimum point . let @xmath530 denote the @xmath323 neighborhood of a point @xmath531 on the boundary . in the case when the minimum is achieved at two points , we note that @xmath532 where the left hand side tends to @xmath176 , while the right hand side is equal to @xmath533 where @xmath534 , as follows from ( [ jjj1 ] ) and ( [ jjj2 ] ) . it also follows from ( [ jjj1 ] ) and ( [ jjj2 ] ) that @xmath535 , which , together with ( [ ji1 ] ) , implies the desired result . finally , we consider the case when @xmath536 and the function @xmath193 is continuous at a point @xmath537 . the cases @xmath200 and @xmath198 are completely similar to each other , so we shall only deal with the latter one . let us introduce the needed notations . fix @xmath538 . by theorem [ mte ] , @xmath539 and @xmath540 . for each @xmath541 $ ] we define @xmath542 let @xmath543 , @xmath541 $ ] , be the probability that the process @xmath331 reaches the boundary of @xmath9 by the time @xmath544 , that is @xmath545 since @xmath546 is left - continuous , it defines a measure @xmath547 on @xmath548)$ ] via @xmath549 ) = \alpha^\varepsilon(c)$ ] . it will be important to identify the limit of @xmath546 as @xmath31 . we define the function @xmath550 \rightarrow [ 0,1]$ ] by : @xmath551 since @xmath183 is piece - wise continuous and its graph is above the diagonal in a neighborhood of @xmath552 $ ] , the function @xmath117 is a unique continuous function which satisfies the differential equation @xmath553 in the points of continuity of @xmath183 and the terminal condition @xmath554 . notice that @xmath555 for @xmath541 $ ] . the function @xmath117 defines a measure @xmath556 on @xmath557)$ ] via @xmath558 ) = \alpha(c)$ ] . if @xmath198 , @xmath193 is continuous at a point @xmath559 , then @xmath560 for @xmath541 $ ] . assume first that the minimum of the quasi - potential is achieved at a unique point @xmath83 for each @xmath80 in a neighborhood of @xmath561 $ ] , and therefore @xmath183 is a continuous function there . take @xmath562 which will be specified later and @xmath563 such that @xmath564 . let @xmath565 be such that @xmath566 , @xmath567 and @xmath568 , @xmath167 . consider the processes @xmath569 , @xmath570 , and @xmath571 . let @xmath572 be the first time when the process @xmath573 reaches the boundary of @xmath9 . let @xmath574 , @xmath575 , be the event that @xmath576 , and @xmath577 be the event that @xmath578 . using lemma [ au1 ] , it is not difficult to show that for each @xmath579 there is @xmath142 such that @xmath580 uniformly in @xmath581 . since @xmath183 is continuous , we can also make sure that @xmath323 is small enough so that @xmath582 uniformly in @xmath581 . we can write @xmath583 in two different ways @xmath584 @xmath585 and @xmath584 @xmath586 upon subtracting the right hand sides of these two equalities , using the markov property , corollary [ co1 ] , lemma [ letwo ] and ( [ miop ] ) , we obtain @xmath587 where @xmath588 and @xmath589 . this implies the desired result once we recall that @xmath459 and @xmath323 can be taken arbitrarily small , since ( [ nmmn ] ) shows that @xmath546 is a type of euler s method approximation to the solution of ( [ inequ ] ) . the condition of continuity of @xmath183 can be easily removed once we recall that @xmath183 may have at most finitely many points of discontinuity . + * remark . * using similar arguments it is not difficult to show that @xmath590 moreover , using ( [ nmmn ] ) and ( [ exe ] ) it is possible to show that in order to find the limit of @xmath495 , one can take @xmath556 , which is the limit of @xmath547 , and then take its push - forward by the function @xmath35 ( since @xmath556 is an absolutely continuous measure , it is not essential that @xmath35 may be undefined in a finite number of points ) . the push - forward of @xmath556 will be denoted by @xmath517 . thus @xmath591 : x^*(c ) \in a)$ ] , @xmath592 . + combining this with lemmas [ fox ] and [ fttw ] and corollary [ co1 ] , we can can formulate the following theorem . [ mte2 ] let @xmath493 and @xmath495 be defined by ( [ fox1 ] ) and ( [ fox2 ] ) , respectively . if @xmath12 and @xmath593 , then @xmath594 , where @xmath595 is the probability measure concentrated at @xmath13 and @xmath596 , where @xmath517 is the trivial measure , that is @xmath597 . if @xmath213 and the minimum of the quasi - potential @xmath598 is achieved at a single point @xmath516 , then @xmath493 and @xmath495 weakly converge to a probability measure @xmath517 concentrated at @xmath516 . if the minimum is achieved at two points @xmath518 and @xmath519 and @xmath520 , then @xmath493 and @xmath495 weakly converge to a probability measure @xmath517 concentrated at those two points . moreover , in this case @xmath599 . if @xmath198 and the function @xmath193 is continuous at a point @xmath600 , then take the measure @xmath556 on @xmath548)$ ] defined via @xmath558 ) = \alpha(c)$ ] , where @xmath117 is the solution of ( [ inequ ] ) . the measures @xmath495 weakly converge to the measure @xmath517 which is the push - forward of @xmath556 by the function @xmath35 . the measures @xmath493 weakly converge to the measure @xmath601 , where @xmath602 . * this theorem still holds if instead of @xmath95 in the definition of @xmath495 we have a positive function @xmath486 such that @xmath603 . if @xmath198 and the function @xmath193 is continuous at @xmath600 , then for every @xmath142 and @xmath145 we have @xmath604),\ ] ] where @xmath556 is the measure on @xmath557)$ ] defined via @xmath558 ) = \alpha(c)$ ] . the corollary immediately follows from theorems [ mte ] and [ mte2 ] and the probabilistic representation of the solution to the initial - boundary value problem . we could allow the coefficient at the first order term to depend on @xmath60 in ( [ eq1])-([bc ] ) : @xmath605 @xmath606 @xmath607 all the assumptions made in sections [ qle ] and [ asc11 ] remain in force , other than the following : instead of assuming that @xmath6 is a vector valued function on @xmath9 , we assume that @xmath608 , and there is a positive constant @xmath609 such that @xmath610 for @xmath136 , @xmath611 , where @xmath16 is the exterior normal to the boundary of @xmath9 . moreover , we assume that for each @xmath612 the vector field @xmath613 has a unique equilibrium point @xmath13 which does not depend on @xmath612 and that all the trajectories of the dynamical system @xmath614 starting in @xmath9 are attracted to @xmath13 . we now assume that there is a smooth function @xmath236 defined on @xmath41 , such that @xmath615 , @xmath616 for @xmath617 , and @xmath618 for some positive constant @xmath80 , all @xmath612 and all @xmath8 . the definition of the function @xmath619 from section [ asc11 ] needs to be modified to allow for the dependence of the drift term on a parameter . namely , now @xmath620 where @xmath621 is the quasi - potential for the process whose generator is equal to @xmath622 with this definition of @xmath619 , theorems [ mte ] and [ mte2 ] remain valid , and the proofs do not require serious modifications . in this section we again consider the solutions @xmath60 to ( [ eq1])-([bc ] ) . let all the assumptions about the domain @xmath9 and the operator @xmath209 made in sections [ qle ] and [ fot ] remain in force , except the following : instead of assuming the existence of a singe equilibrium point , we assume that there are two asymptotically stable equilibrium points @xmath623 such that for almost every @xmath12 , with respect to the lebesgue measure , the trajectory of ( [ dsyst ] ) starting at @xmath8 is attracted to either @xmath624 or @xmath625 . let @xmath626 be the set of points in @xmath9 which are attracted to @xmath624 and @xmath627 the set of points attracted to @xmath625 . as before , we need to study the quasi - potential in order to determine the asymptotic behavior of @xmath60 . while in the case of a single equilibrium , the function @xmath60 was nearly constant in @xmath262 at times of order @xmath93 ( lemma [ letwo ] ) , now @xmath628 will be close to @xmath629 for @xmath630 and close to @xmath631 for @xmath632 . this explains why instead of freezing the second variable in the coefficients @xmath79 in the right hand side of ( [ eq1 ] ) , the way it was done is section [ qle ] , now we put the variable equal to @xmath176 in @xmath633 and @xmath634 in @xmath635 . more precisely , for @xmath636 $ ] , let @xmath637 where @xmath638 is the indicator function of a set @xmath639 . for a measurable positive - definite matrix - valued function @xmath117 on @xmath41 , we define @xmath134 , \overline{d } ) , \varphi(0 ) = x , \varphi(t ) = y \},~~x , y \in \overline{d},\ ] ] where the normalized action functional @xmath640 was defined in section [ af ] . instead of function @xmath641 used in section [ qle ] , we now have functions @xmath642 , @xmath643 , @xmath644 and @xmath645 . these are defined by @xmath646 @xmath647 @xmath648 @xmath649 it is not difficult to check that the right hand side of ( [ eqtxx1 ] ) does not depend on @xmath634 and the right hand side of ( [ eqtxx2 ] ) does not depend on @xmath176 . for the process governed by equation ( [ eq1 ] ) , with @xmath650 instead of the nonlinear coefficients @xmath651 , the transition from @xmath624 to a small neighborhood of @xmath625 occurs in time of order @xmath652 ( provided that the process does not exit the domain @xmath9 earlier ) . similarly , the transition from @xmath625 to a neighborhood of @xmath624 occurs in time of order @xmath653 , while the transition from @xmath624 and @xmath625 to the boundary occurs in time of order @xmath654 and @xmath655 , respectively . in this section we would like to study the equation at a time scale which is sufficiently large for the process to make excursions between the neighborhoods of @xmath624 and @xmath625 and back , yet not too large so that the process starting at @xmath624 or @xmath625 does not exit the domain . therefore we assume that @xmath656 } m_{x_1,x_2}(c_1 ) , \sup_{c_2 \in [ g_{\rm min } , g_{\rm max } ] } m_{x_2,x_1}(c_2 ) ) < m^\partial,\ ] ] where @xmath657 } m_{x_1 , \partial d}(c_1,c_2 ) , \inf_{c_1 , c_2 \in [ g_{\rm min } , g_{\rm max } ] } m_{x_2 , \partial d}(c_1,c_2)).\ ] ] for example , if @xmath59 and @xmath6 are defined in the entire space @xmath120 , @xmath59 is bounded and @xmath6 satisfies @xmath658 for some positive constants @xmath89 and @xmath90 , then this condition will be satisfied for any domain @xmath9 which contains a sufficiently large ball centered at the origin . let @xmath659 and @xmath660 . without loss of generality we may assume that @xmath661 . in order to formulate the results on the asymptotics of @xmath60 , we need to introduce functions @xmath662 and @xmath663 which are similar to the function @xmath193 from in section [ asc11 ] . let @xmath673 . assume that at least one of the functions @xmath674 and @xmath675 is continuous at @xmath676 . let @xmath677 if @xmath674 is continuous at @xmath676 and @xmath678 otherwise . let @xmath679 and @xmath680 . [ mtex1 ] let the above assumptions be satisfied . suppose that the function @xmath681 is continuous at a point @xmath682 . then for every @xmath142 the following limit @xmath683 is uniform in @xmath684 . suppose that the function @xmath685 is continuous at a point @xmath682 . then for every @xmath142 the following limit @xmath686 is uniform in @xmath687 . * remark . * if @xmath688 , then @xmath689 . it is not difficult to see that the limit @xmath690 is uniform in @xmath691 for each @xmath692 . therefore , for each @xmath142 and @xmath693 there is @xmath694 such that @xmath695 whenever @xmath218 , @xmath145 and @xmath219 . recall that @xmath331 , @xmath696 $ ] , is the process defined in ( [ pro ] ) , @xmath228 is the first time when this process reaches the boundary of @xmath9 and @xmath697 . since we assume that @xmath698 , the probability that @xmath699 now tends to zero as @xmath31 . the distribution of the random variable @xmath492 will be concentrated near the points @xmath624 and @xmath625 . [ tenn ] suppose that @xmath700 . if the function @xmath681 is continuous at a point @xmath682 and @xmath701 , then the distribution of the random variable @xmath492 converges to the measure @xmath702 , where the coefficients @xmath703 and @xmath704 can be found from the equations @xmath705 , @xmath706 . if the function @xmath685 is continuous at a point @xmath707 and @xmath708 , then the distribution of the random variable @xmath492 converges to the measure @xmath709 , where the coefficients @xmath703 and @xmath704 can be found from the equations @xmath710 , @xmath706 . if @xmath711 and @xmath12 , then the distribution of the random variable @xmath492 converges to the measure @xmath712 , where the coefficients @xmath703 and @xmath704 can be found from the equations @xmath713 , @xmath706 . the proofs of theorems [ mtex1 ] and [ tenn ] rely on the same techniques as those used in the proofs of theorems [ mte ] and [ mte2 ] , and therefore will not be presented here . if @xmath59 is bounded and @xmath6 satisfies @xmath714 for some positive constants @xmath89 and @xmath90 , then similar results can be formulated for the cauchy problem in @xmath4 and the corresponding nonlinear perturbations of the dynamical system in @xmath120 . in this case we do not need the condition @xmath715 , but can instead consider all @xmath716 . in the first example , we assume that the domain @xmath9 is one dimensional : @xmath717 . we assume that @xmath718\times \mathbb{r})$ ] is positive , @xmath719\times \mathbb{r})$ ] , @xmath720 and there is a point @xmath721 such that @xmath722 for all @xmath612 . in this case the operator @xmath723 satisfies the assumptions of section [ fot ] . the quasi - potential , which will now be denoted by @xmath724 , is given by @xmath725 as is easily seen from the definition of the action functional ( see section [ af ] ) . therefore , @xmath726 the function @xmath727 may take at most two values : @xmath728 and @xmath729 . in particular , the value @xmath728 is taken on the set @xmath730 , while the value @xmath729 is taken on the set @xmath731 . picture 2 shows an example of the graphs of functions @xmath619 and @xmath727 in the case when @xmath732 } g(x ) < \sup_{x \in [ a , b ] } g(x ) = g(b)$ ] . from theorem [ mte2 ] and the discussion in section [ fot ] it follows that for @xmath733 , the distribution of the random variable @xmath492 converges to the probability measure @xmath517 concentrated at the end points of the segment . this measure can be found from the relations @xmath734 in the second example we assume that @xmath96 contains the origin @xmath735 . let the operator @xmath209 be as follows @xmath736 where @xmath59 is a positive - definite matrix which depends smoothly on @xmath612 , and @xmath89 is a matrix with negative eigenvalues which depends smoothly on @xmath612 . as has been demonstrated in @xcite , the quasi - potential , which we shall denote by @xmath724 , is given by the quadratic form @xmath737 where the matrix @xmath90 is given by @xmath738 therefore , @xmath739 . + + * acknowledgements * : while working on this article , m. freidlin was supported by nsf grant dms-0803287 and l. koralov was supported by nsf grant dms-0706974 .

progress on string theory in curved spacetimes since 1992 are reviewed . after a short introduction on strings in minkowski and curved spacetimes , we focus on strings in cosmological spacetimes . the classical behaviour of strings in frw and inflationary spacetimes is now understood in a large extent from various types of explicit string solutions . three different types of behaviour appear in cosmological spacetimes : * unstable , dual * to unstable and * stable*. for the unstable strings , the energy and size grow for large scale factors , proportional to . for the dual to unstable strings , the energy and size blow up for as . for stable strings , the energy and proper size are bounded . ( in minkowski spacetime , all string solutions are of the stable type ) . recent progress on self - consistent solutions to the einstein equations for string dominated universes is reviewed . the energy - momentum tensor for a gas of strings is then considered as source of the spacetime geometry and from the above string behaviours the string equation of state is determined . the self - consistent string solution exhibits the realistic matter dominated behaviour for large times and the radiation dominated behaviour for early times . finally , we report on the exact integrability of the string equations plus the constraints in de sitter spacetime that allows to systematically find * exact * string solutions by soliton methods and the multistring solutions . * multistring solutions * are a new feature in curved spacetimes . that is , a single world - sheet simultaneously describes many different and independent strings . this phenomenon has no analogue in flat spacetime and appears as a consequence of the coupling of the strings with the spacetime geometry . lectures delivered at the erice school `` current topics in astrofundamental physics '' 4 - 16 september 1994 , to appear in the proceedings edited by n. snchez .

hep-th9504098

since the previous erice school on ` string quantum gravity ' a host of impressive developments has taken place on strings in curved spacetime . for the state of the art in 1992 we refer to the 1992 proceedings @xcite where the programme on string quantization in curved spacetimes initiated by us in 1987 @xcite is reviewed . a consistent quantum theory of gravity is the strongest motivation for string theory and hence to study strings in curved spacetime@xcite . as stressed in the sec . 1 of the 1992 lectures , a quantum theory of gravity must be a theory able to describe all physics below the planck scale @xmath6tev . that means that a sensible theory of quantum gravity is necessarily part of a unified theory of all interactions . * pure gravity * ( a model containing only gravitons ) can not be a physical and realistic quantum theory . to give an example , a theoretical prediction for graviton - graviton scattering at energies of the order of @xmath7 must include all particles produced in a real experiment . that is , in practice , all existing particles in nature , since gravity couples to all matter . string theory is therefore a serious candidate for a quantum description of gravity since it provides a unified model of all interactions overcoming at the same time the nonrenormalizable character of quantum fields theories of gravity . the present lectures deal mainly with strings in cosmological spacetimes ; substantial results were achieved in this field since 1992 . the classical behaviour of strings in frw and inflationary spacetimes is now understood in a large extent@xcite . this understanding followed the finding of various types of exact and numerical string solutions in frw and inflationary spacetimes@xcite -@xcite . for inflationary spacetimes , the exact integrability of the string propagation equations plus the string constraints in de sitter spacetime @xcite is indeed an important help . this allowed to systematically find * exact * string solutions by soliton methods using the linear system associated to the problem ( the so - called dressing method in soliton theory ) and the multistring solutions @xcite -@xcite . in summary , three different types of behaviour are exhibited by the string solutions in cosmological spacetimes : * unstable , dual * to unstable and * stable*. for the unstable strings , the energy and size grow for large scale factors @xmath0 , proportional to @xmath1 . for the dual to unstable strings , the energy and size blow up for @xmath2 as @xmath3 . for stable strings , the energy and proper size are bounded . ( in minkowski spacetime , all string solutions are of the stable type ) . the equation of state for these string behaviours take the form * \(i ) * unstable * for @xmath8 * \(ii ) * dual to unstable * for @xmath9 . * \(iii ) * stable * for @xmath10 . here @xmath11 and @xmath12 stand for the corresponding string energies and @xmath13 for the number of spatial dimensions where the string solutions lives . for example , @xmath14 for a straight string , @xmath15 for a ring string , etc . this number @xmath16 is obviously less or equal than the number of spacetime dimensions @xmath17 . as we see above , the dual to unstable string behavior leads to the same equation of state than radiation ( massless particles or hot matter ) . the stable string behavior leads to the equation of state of massive particles ( cold matter ) . the unstable string behavior is a purely ` stringy ' phenomenon . the fact that in entails a negative pressure is however physically acceptable . for a gas of strings , the unstable string behaviour dominates in inflationary universes when @xmath18 and the dual to unstable string behavior dominates for @xmath19 . the unstable strings correspond to the critical case of the so called _ coasting universe _ @xcite . that is , classical strings provide a _ concrete _ realization of such cosmological models . the ` unstable ' behaviour is called ` string stretching ' in the cosmic string literature @xcite . it must be stressed that while time evolves , a * given * string solution may exhibit two and even three of the above regimes one after the other ( see sec . intermediate behaviours are also observed in ring solutions @xcite . that is , @xmath20 another new feature appeared in curved spacetimes : * multistring solutions*. that is a single world - sheet simultaneously describes many different and independent strings . this phenomenon has no analogue in flat spacetime . this is a new feature appearing as a consequence of the interaction of the strings with the spacetime geometry . the world - sheet time @xmath21 turns out to be an multi - valued function of the target string time @xmath22 ( which can be the cosmic time @xmath23 , the conformal time @xmath24 or for de sitter universes it can be the hyperboloid time @xmath25 ) . each branch of @xmath21 as a function of @xmath22 corresponds to a different string . in flat spacetime , multiple string solutions are necessarily described by multiple world - sheets . here , a single world - sheet describes one string , several strings or even an infinite number of different and independent strings as a consequence of the coupling with the spacetime geometry . these strings do not interact among themselves ; all the interaction is with the curved spacetime . one can decide to study separately each of them ( they are all different ) or consider all the infinite strings together . of course , from our multistring solution , one _ could _ just choose only one interval in @xmath26 ( or a subset of intervals in @xmath27 ) and describe just one string ( or several ) . this will be just a * truncation * of the solution . the really remarkably fact is that all these infinitely many strings come * naturally together * when solving the string equations in de sitter spacetime as we did in refs . @xcite -@xcite . here , interaction among the strings ( like splitting and merging ) is neglected , the only interaction is with the curved background . the study of string propagation in curved spacetimes provide essential clues about the physics in this context but is clearly not the end of the story . the next step beyond the investigation of * test * strings , consist in finding * self - consistently * the geometry from the strings as matter sources for the einstein equations or better the string effective equations ( beta functions ) . this goal is achieved in ref.@xcite for cosmological spacetimes at the classical level . namely , we used the energy - momentum tensor for a gas of strings as source for the einstein equations and we solved them self - consistently . to write the string equation of state we used the behaviour of the string solutions in cosmological spacetimes . strings continuously evolve from one type of behaviour to another , as is explicitly shown by our solutions @xcite -@xcite . for intermediate values of @xmath28 , the equation of state for gas of free strings is clearly complicated but a formula of the type : @xmath29 where @xmath30 this equation of state is qualitatively correct for all @xmath28 and becomes exact for @xmath19 and @xmath31 . the parameters @xmath32 and @xmath33 are positive constants and the @xmath34 varies smoothly with @xmath1 . the pressure associated to the energy density ( [ rogeni ] ) takes then the form @xmath35 inserting this source into the einstein - friedman equations leads to a self - consistent solution for string dominated universes ( see sec . iv ) @xcite . this solution exhibits the realistic matter dominated behaviour @xmath36 for large times and the radiation dominated behaviour @xmath37 for early times . for the sake of completeness we analyze in sec . iv the effective string equations @xcite . these equations have been extensively treated in the literature @xcite and they are not our central aim . it must be noticed that there is no satisfactory derivation of inflation in the context of the effective string equations . this does not mean that string theory is not compatible with inflation , but that the effective string action approach _ is not enough _ to describe inflation . the effective string equations are a low energy field theory approximation to string theory containing only the _ massless _ string modes . the vacuum energy scales to start inflation are typically of the order of the planck mass where the effective string action approximation breaks down . one must also consider the _ massive _ string modes ( which are absent from the effective string action ) in order to properly get the cosmological condensate yielding inflation . de sitter inflation does not emerge as a solution of the the effective string equations . in conclusion , the effective string action ( whatever be the dilaton , its potential and the central charge term ) is not the appropriate framework in which to address the question of string driven inflation . early cosmology ( at the planck time ) is probably the best place to test string theory . in one hand the quantum treatment of gravity is unavoidable at such scales and in the other hand , observable cosmological consequences are derivable from the inflationary stage . the natural gravitational background is an inflationary universe as de sitter spacetime . such geometries are not string vacua . this means that conformal and weyl symmetries are broken at the quantum level . in order to quantize consistently strings in such case , one must enlarge the physical phase space including , in particular , the factor @xmath38 in the world - sheet metric [ see eq.([liou ] ) ] . this is a very interesting and open problem . physically , the origin of such difficulties in quantum string cosmology comes from the fact that one is not dealing with an * empty * universe since a cosmological spacetime necessarily contains matter . in the other hand , conformal field theory techniques are only adapted to backgrounds for which the beta functions are identically zero , i. e. sourceless geometries . the outline of these lectures is as follows . section ii presents an introduction to strings in curved spacetimes including basic notions on classical and quantum strings in minkowski spacetime and introducing the main physical string magnitudes : energy - momentum and invariant string size . section iii deals with the string propagation and the string energy - momentum tensor in cosmological spacetimes . ( in sections iii.a , iii.b and iii.c we treat the straight strings , ring strings and generic strings respectively and derive the corresponding string equations of state ) . in section iv we treat self - consistent string cosmology including the string equations of state . ( section iv.a deals with general relativity , iv.b with the string thermodynamics ) . section v discuss the effective ( beta functions ) string equations in the cosmological perspective and the search of inflationary solutions . finally , in sec . vi , we briefly review the systematic construction of string solutions in de sitter universe _ via _ soliton methods and the new feature of multistring solutions . let us consider bosonic strings ( open or closed ) propagating in a curved d - dimensional spacetime defined by a metric @xmath39 . the action can be written as @xmath40 here @xmath41 ( @xmath42 ) is the metric in the worldsheet , @xmath43 stands for the string tension . as in flat spacetime , @xmath44 fixes the scale in the theory . there are no other free parameters like coupling constants in string theory . we will start considering given gravitational backgrounds @xmath45 . that is , we start to investigate _ test _ strings propagating on a given spacetime . in section iv , the back reaction problem will be studied . that is , how the strings may act as source of the geometry . string propagation in massless backgrounds other than gravitational ( dilaton , antisymmetric tensor ) can be investigated analogously . the string action ( [ accion ] ) classically enjoys weyl invariance on the world sheet @xmath46 plus the reparametrization invariance @xmath47 here @xmath48 and @xmath49 are arbitrary functions . the dynamical variables being here the string coordinates @xmath50 , ( @xmath51 ) and the world - sheet metric @xmath41 . extremizing @xmath52 with respect to them yields the classical equations of motion : @xmath53 & = & \frac{1}{2}\ , \sqrt g\ ; \partial_{a}g_{cd}(x ) \ , \partial_{\a}x^{c}(\s,\tau ) \ , \partial^{\a}x^{d}(\s,\tau ) \\ & & 0 \le a \le d-1 \cr \label{vincu } \cr t_{\a\b } ~ & \equiv & ~ g_{ab}(x ) [ \ , \partial_{\a}x^{a}(\s,\tau ) \ , \partial_{\b}x^{b}(\s,\tau ) \cr & - & \frac{1}{2 } \ , g_{\a\b}(\s,\tau ) \ , \partial_{\g}x^{a}(\s,\tau ) \ , \partial^{\g}x^{b}(\s,\tau)\ , ] = 0 ~~ , \quad 0 \le \a , \beta \le 1.\end{aligned}\ ] ] eqs . ( [ vincu ] ) contain only first derivatives and are therefore a set of constraints . classically , we can always use the reparametrization freedom ( [ reparam ] ) to recast the world - sheet metric on diagonal form @xmath54\ , \ , { \rm diag } ( -1 , + 1)\ ] ] in this conformal gauge , eqs . ( [ movi ] ) - ( [ vincu ] ) take the simpler form : @xmath55 @xmath56 where we introduce light - cone variables @xmath57 on the world - sheet and where @xmath58 stand for the connections ( christoffel symbols ) associated to the metric @xmath59 . notice that these equations in the conformal gauge are still invariant under the conformal reparametrizations : @xmath60 here @xmath61 and @xmath62 are arbitrary functions . the string boundary conditions in curved spacetimes are identical to those in minkowski spacetime . that is , @xmath63 in flat spacetime eqs.([conouno ] ) become linear @xmath64 and one can solve them explicitly as well as the quadratic constraint ( [ conodos ] ) [ see below ] : @xmath65 ^ 2 - \sum_{j=1}^{d-1 } \left [ \partial_{\pm}x^{j}(\s,\tau ) \right]^2 = 0\ ] ] the solution of eqs.([dalam ] ) is usually written for closed strings as @xmath66 + \tilde \a^{a}_{n } \exp[-in(\s + \tau ) ] \ } \label{solm}\ ] ] where @xmath67 and @xmath68 stand for the string center of mass position and momentum and @xmath69 and @xmath70 describe the right and left oscillator modes of the string , respectively . since the string coordinates are real , @xmath71 this resolution is no more possible in general for curved spacetime where the equations of motion ( [ conouno ] ) are non - linear . in that case , right and left movers interact with themselves and with each other . in minkowski spacetime we can also write the solution of the string equations of motion ( [ dalam ] ) in the form @xmath72 where @xmath73 and @xmath74 are arbitrary functions . now , making an appropriate conformal transformation ( [ conftr ] ) we can turn any of the string coordinates @xmath75 ( but only one of them ) into a constant times @xmath21 . the most convenient choice is the light - cone gauge where @xmath76 that is , there are no string oscillations along the @xmath77 direction in the light - cone gauge . we have still to impose the constraints ( [ vinm ] ) . in this gauge they take the form @xmath78 ^ 2\ ] ] where @xmath79 . this shows that @xmath80 is not an independent dynamical variable since it expresses in terms of the transverse coordinates @xmath81 . only @xmath82 is an independent quantity . the physical picture of a string propagating in minkowski spacetime clearly emerges in the light - cone gauge . the gauge condition ( [ gauge ] ) tells us that the string ` time ' @xmath21 is just proportional to the physical null time @xmath77 . eqs.([solm ] ) shows that the string moves as a whole with constant speed while it oscillates around its center of mass . the oscillation frequencies are all integers multiples of the basic one . the string thus possess an infinite number of normal modes ; @xmath83 classically describe their oscillation amplitudes . only the modes in the direction of the transverse coordinates @xmath81 are physical . this is intuitively right , since a longitudinal or a temporal oscillation of a string is meaningless . in summary , the string in minkowski spacetime behaves as an extended and composite relativistic object formed by a @xmath84-infinite set of harmonic oscillators . integrating eq.([v ] ) on @xmath85 from @xmath86 to @xmath87 and inserting eq . ( [ solm ] ) yields the classical string mass formula : @xmath88\ ] ] we explicitly see how the mass of a string depends on its excitation state . the classical string spectrum is continuous as we read from eq.([masac ] ) . it starts at @xmath89 for an unexcited string : @xmath90 for all @xmath91 and @xmath92 . the independent string variables are : the transverse amplitudes @xmath93 , the transverse center of mass variables @xmath94 , @xmath82 and @xmath95 . up to now we have considered a classical string . at the quantum level one imposes the canonical commutation relations ( ccr ) @xmath96 & = & n \ ; \delta_{n ,- m}\ , \delta^{i , j } \ ; , \cr\cr [ { \tilde \a^{i}_{n } , \a^{j}_{m } } ] & = & n \ ; \delta_{n ,- m}\ , \delta^{i , j } \ ; , \cr\cr [ { \tilde \a^{i}_{n } } , \a^{j}_{m } ] & = & 0 \ ; , \cr\cr [ q^i , p^j ] = i \ ; \delta^{i , j } \quad & , & \quad [ q^v , p^u ] = i\end{aligned}\ ] ] all other commutators being zero . an order prescription is needed to unambiguously express the different physical operators in terms of those obeying the ccr . the symmetric ordering is the simplest and more convenient . the space of string physical states is the the tensor product of the hilbert space of the @xmath97 center of mass variables @xmath98 , times the fock space of the harmonic transverse modes . the string wave function is then the product of a center of mass part times a harmonic oscillator part . the center of mass can be taken , for example , as a plane wave . the harmonic oscillator part can be written as the creation operators @xmath99 acting on the oscillator ground state @xmath100 . this state is defined as usual by @xmath101 notice that a string describes * one particle*. the kind of particle described depends on the oscillator wave function . the mass and spin can take an infinite number of different values . that is , there is an infinite number of different possibilities for the particle described by a string . let us consider the quantum mass spectrum . upon symmetric ordering the mass operator becomes , @xmath102 \ ; .\ ] ] using the commutation rules ( [ ccr ] ) yields @xmath103\ ] ] the divergent sum in the first term can be defined through analytic continuation of the zeta function @xmath104 one finds @xmath105 @xcite . thus , @xmath106\ ] ] hence , the string ground state @xmath100 has a negative mass squared @xmath107 such particles are called tachyons and exhibit unphysical behaviours . when fermionic degrees of freedom are associated to the string the ground state becomes massless ( superstrings ) @xcite . notice that the appearance of a negative mass square yields a dispersion relation @xmath108 similar to waves with jeans unstabilities @xcite . let us consider now excited states . the constraints ( [ vinm ] ) integrated on @xmath85 from @xmath86 to @xmath87 impose @xmath109 on the physical states . this means that the number of left and right modes coincide in all physical states . the first excited state is then described by @xmath110 . it describes therefore a spin two particle plus a spin zero particle ( the trace part ) . from eqs.([masaq]-[grav ] ) follows that @xmath111 this state is then a massless particle only for @xmath112 . in such critical dimension we have then a graviton ( massless spin 2 particle ) and a dilaton ( massless spin 0 particle ) as string modes of excitation . for superstrings the critical dimension turns to be @xmath113 . we shall consider , as usual , that only four space - time dimensions are uncompactified . that is , we shall consider the strings as living on the tensor product of a curved four dimensional space - time with lorentzian signature and a compact space which is there to cancel the anomalies . therefore , from now on strings will propagate in the curved ( physical ) four dimensional space - time . however , we will find instructive to study the case where this curved space - time has dimensionality @xmath17 , where @xmath17 may be 2 , 3 or arbitrary . the spacetime string energy - momentum tensor follows ( as usual ) by taking the functional derivative of the action ( [ accion ] ) with respect to the metric @xmath114 at the spacetime point @xmath115 . this yields , @xmath116 where dot and prime stands for @xmath117 and @xmath118 , respectively . notice that @xmath115 in eq.([tens ] ) is just a spacetime point whereas @xmath119 stands for the string dynamical variables . one sees from the dirac delta in eq.([tens ] ) that @xmath120 vanishes unless @xmath115 is exactly on the string world - sheet . we shall not be interested in the detailed structure of the classical strings . it is the more useful to integrate the energy - momentum tensor ( [ tens ] ) on a volume that completely encloses the string . it takes then the form @xcite @xmath121 when @xmath122 depends only on @xmath21 , we can easily integrate over @xmath21 with the result , @xmath123 _ { \tau = \tau(x^0)}\ ] ] another relevant physical magnitude for strings is the invariant size . we define the invariant string size @xmath124 using the metric induced on the string world - sheet : @xmath125 inserting @xmath126 , into eq.([intinv ] ) and taking into account the constraints ( [ conodos ] ) yields @xmath127 thus , we define the string size as the integral of @xmath128 over @xmath85 at fixed @xmath21 . let us now consider a circular string as a simple example of a string solution in minkowski spacetime . @xmath129 \\ & & \cr x^2(\s,\tau ) & = & \a ' m \cos\tau \sin\s = { { \a ' m } \over 2 } \left [ \sin(\tau + \s ) - \sin(\tau-\s ) \right ] \nonumber\end{aligned}\ ] ] this is obviously a solution of eqs.([dalam ] ) where only the modes @xmath130 are excited . the constraints ( [ vinm ] ) yields @xmath131 eqs.([anilm ] ) describe a circular string in the @xmath132 plane , centered in the origin and with an oscillating radius @xmath133 . in addition the string moves uniformly in the @xmath134-direction with speed @xmath135 . ( that is , @xmath136 is its momentum in the @xmath134-direction ) . the oscillation amplitude @xmath137 can be identified with the string mass and @xmath138 with the string energy . notice that the string time @xmath21 is here proportional to the physical time @xmath22 ( this solution is not in the light - cone gauge ) . it is instructive to compute the integrated energy - momentum tensor ( [ inten ] ) for this string solution . we find in the rest frame ( @xmath139 ) that it takes the fluid form @xmath140 where @xmath141 we see that the total energy coincides with @xmath137 as one could expect and that the ( space averaged ) pressure oscillates around zero . that is the string pressure goes through positive and negative values . the time average of @xmath136 vanishes . the string behaves then as cold matter ( massive particles ) . the upper value of @xmath136 equals @xmath142 . this is precisely the relation between @xmath138 and @xmath136 for radiation ( massless particles ) . ( notice that this circular strings lives on a two - dimensional plane ) . the lower value of @xmath136 correspond to the limiting value allowed by the strong energy condition in general relativity @xcite . we shall see below that these two extreme values of @xmath136 appear in the general context of cosmological spacetimes . the invariant size of the string solution ( [ anilm ] ) follows by inserting eq.([anilm ] ) into eq.([tam ] ) . we find @xmath143 therefore , this string solution has a constant size @xmath144 . one obtains the invariant size for a generic solution in minkowski spacetime inserting the general solution ( [ solm ] ) into @xmath145 . this gives constant plus oscillatory terms . in any case the invariant string size is always * bounded * in minkowski spacetimes . we shall see how differently behave strings in curved spacetimes . we obtain in this section physical string properties from the exact string solutions in cosmological spacetimes . we consider strings in spatially homogeneous and isotropic universes with metric @xmath146 where @xmath23 is the cosmic time and the function @xmath147 is called the scale factor . in terms of the conformal time @xmath148 the metric ( [ met ] ) takes the form @xmath149 \label{metc}\ ] ] the classical string equations of motion can be written here as @xmath150=0\ , , \qquad 1\leq i\leq d-1\,,\nonumber\end{aligned}\ ] ] and the constraints are @xmath151 the most relevants universes corresponds to power type scale factors . that is , @xmath152 where @xmath153 . for different values of the exponents we have either frw or inflationary universes . @xmath154 @xmath155 as we will see below , @xmath156 takes the fluid form for string solutions in cosmological spacetimes , allowing us to define the string pressure @xmath136 and energy density @xmath157 : @xmath158 notice that the continuity equation @xmath159 takes here the form @xmath160 where @xmath161 . for an equation of state of the type of a perfect fluid , that is @xmath162 eqs.([cont ] ) and ( [ eflu ] ) can be easily integrated with the result @xmath163 for @xmath164 this corresponds to cold matter @xmath165 and for @xmath166 this describes radiation with @xmath167 . let us start by considering strings in this simpler case . for a @xmath168 dimensional universe or for a straight string parallel to the @xmath115-axis in a @xmath17-dimensional universe , the metric ( [ met ] ) takes the form @xmath169 it is convenient to start by solving the constraints ( [ vinc ] ) @xmath170 they reduce to @xmath171 where @xmath172 . using the conformal time ( [ tco ] ) , eq.([raiz ] ) yields @xmath173 we find a first family of solutions choosing @xmath174 . then @xmath175 where @xmath176 and @xmath177 are arbitrary functions of one variable . it is easy now to check that eq.([sold ] ) fulfills the string equations of motion ( [ eqmov ] ) . the solution ( [ sold ] ) is analyzed in detail for de sitter spacetime @xmath178 in ref.@xcite where the global topology of the space is taken into account . since one can always perform conformal transformations @xmath179 with arbitrary functions @xmath180 and @xmath181 , the solution ( [ sold ] ) has no degrees of freedom other than topological ones . let us compute the energy momentum tensor for the string solution ( [ sold ] ) . we find from eqs.([tens ] ) and ( [ sold ] ) , @xmath182 where @xmath183 is the jacobian . from eq.([sold ] ) we find @xmath184 and @xmath185 . then , @xmath186 we analogously find @xmath187 . then @xmath188 that implies , @xmath189 we find a constant energy density and a constant * negative * pressure . they exactly fulfill the continuity equation ( [ cont ] ) . these results hold for * arbitrary * cosmological spacetimes in @xmath190 dimensions . that is , for arbitrary factors @xmath147 . in particular they are valid for strings winded @xmath191-times around the de sitter universe @xcite . a second family of string solutions follows from eq.([vincud ] ) by choosing @xmath192 where @xmath193 is a constant . then , the string equations of motion ( [ eqmov ] ) become @xmath194=0 \label{buf}\ ] ] using eq.([etx ] ) , we find that the energy - momentum tensor ( [ tens ] ) is traceless for this string solution : @xmath195 let us call @xmath196 then , eq.([buf ] ) implies that @xmath197 the general solution @xmath198 is implicitly defined by : @xmath199 where @xmath200 and @xmath201 are arbitrary functions . upon a conformal transformation , without loss of generality we can set @xmath202 hence , @xmath24 is in general a function solely of @xmath21 with @xmath203 the energy - momentum tensor ( [ tens ] ) takes for this solution the form : @xmath204 the @xmath205 characterizes a localized object propagating on the characteristics at the speed of light . this solution describes a massless point particle since it has been possible to gauge out the @xmath206 dependence . in summary , the two - dimensional string solutions in cosmological spacetimes generically obey perfect fluid equations of state with either @xmath207 the respectives energy densities being @xmath208 these behaviors fulfill the continuity equation ( [ cont ] ) for @xmath209 . a large class of exact solutions describing one string and multistrings has been found in @xmath210 dimensional de sitter universe @xcite -@xcite . for power - like expansion factors @xmath211 only ring solutions are known@xcite . ( @xmath212 corresponds to de sitter spacetime ) . these string solutions are on a two - dimensional plane that can be considered embedded on a d - dimensional universe . ring solutions correspond to the ansatz @xcite : @xmath213 the total energy of one string is then given by ( recall @xmath214 ) @xmath215 more generally , the energy - momentum tensor integrated on a volume that completely encloses the string , takes the form @xcite @xmath216 _ { \tau = \tau(t ) } \nonumber\end{aligned}\ ] ] for multistring solutions , one must sum over the different roots @xmath217 of the equation @xmath218 , for a given @xmath23 . we find for the ring ansatz eq.([ani ] ) : @xmath219 \nonumber \\\nonumber \theta^{01}(x ) & = & \theta^{02}(x ) = \theta^{12}(x ) = 0 ~. \label{tent}\\ \nonumber\end{aligned}\ ] ] that is , @xmath220 where , @xmath221 \label{epa}\ ] ] let us first consider ring strings in de sitter universe @xcite - @xcite . we find there three different asymptotic behaviors : stable , unstable and its dual . the unstable regime appears for @xmath222 . that is , when the scale factor tends to infinity . from eqs.([epa ] ) and ref.@xcite , we find @xmath223 \to + \infty \nonumber \\ p(\tau ) & \buildrel{\tau \to 0 ^ -}\over= & -{1\over { 2 \alpha ' h |\tau| } } \simeq -{{r(\tau)}\over { 2 \alpha'h } } \to -\infty \label{desu } \nonumber\end{aligned}\ ] ] ( here @xmath224 stands for the hubble constant ) . the invariant string size grows as @xmath225 in this unstable regime . notice that the pressure is * negative * for the unstable strings and proportional to the expansion factor @xmath1 . in this regime we also see that @xmath226 with @xmath227 $ ] . the regime dual to the unstable regime appears when the conformal time @xmath24 tends to infinity . for the solution @xmath228 of ref.@xcite in de sitter universe , @xmath24 diverges for finite @xmath229 as @xmath230 here @xmath231 then , @xmath232 we call dual to this regime since it appears related to the unstable regime ( [ desu ] ) through the exchange @xmath233 . the invariant string size tends to @xmath234 in this regime . in the stable regime , @xmath235 , ( and the cosmic time @xmath236 ) , from eq.([epa ] ) and ref.@xcite , we find @xmath237 for stable strings in de sitter universe , the pressure is positive and vanishes asymptotically , and the invariant string size tends to @xmath238 . the solution @xmath228 in ref.@xcite describes two ring strings : a stable string for @xmath239 ( @xmath240 being the hyperboloid time - like coordinate ) and a unstable one for @xmath241 . the pressure @xmath242 depends on @xmath21 ; it is negative for @xmath243 and positive for @xmath244 , where @xmath245 . it can be noticed that the behaviour eq.([desu ] ) for the energy can be interpreted as an unstable piece @xmath246 _ plus _ a stable one @xmath247 . the constant term @xmath247 is precisely the energy for the stable solution eq.([esta ] ) . let us now study the energy and pressure for the ring string solutions in power - type inflationary universes considered in @xcite . in terms of the conformal time @xmath24 , we have as expansion factor @xmath248 with @xmath249 . near @xmath250 , two types of behavior were found @xcite . the first one is a linear behavior @xmath251 eqs.([etan ] ) describe a expanding string for @xmath252 with invariant size @xmath253 . that is , in inflationary universes @xmath254 , the string size grows indefinitely for @xmath255 as the universe radius @xmath256 . the growing of the proper string size for @xmath257 is a typical feature of string unstability . using eqs.([epa ] ) and ( [ etan ] ) , the energy and pressure take the form , @xmath258 that yields @xmath259 eq.([eqes ] ) is also valid in de sitter universe for unstable strings [ eq.([peun ] ) ] . in all these cases strings exhibit * negative * pressure with an equation of state @xmath260 . this equation of state exactly saturates the strong energy condition in general relativity . this * unstable * string behavior dominates in all inflationary universes for @xmath0 . the second behavior present near @xmath250 is @xcite @xmath261 where @xmath262 must be set equal to zero for @xmath263 . the invariant string size behaves for @xmath264 as , @xmath265 this solution describes a string that collapses for @xmath266 ( power inflation ) and blows up for @xmath267 ( super inflation ) . in both cases the string size is proportional to the horizon size which is of order @xmath23 ( cosmic time ) . here , the expansion factor tends to zero as @xmath268 when @xmath269 and blows up for @xmath270 . from eq.([epa ] ) we find that @xmath271 that is , @xmath272 notice that the pressure is here * positive*. this second behaviour is related by duality ( @xmath233 ) to the first behaviour described by eqs.([epun])-([eqes ] ) . this is the * dual to unstable * regime . in such regime the strings behave as massless particles ( radiation ) . let us now consider ring string solutions in frw universes considered in @xcite . these solutions start ex - nihilo at the big bang ( @xmath273 ) . in terms of the conformal time @xmath24 , we have as expansion factor @xmath248 with @xmath274 . near @xmath250 , two types of behavior were found @xcite . the first one is a linear behavior @xmath275 eqs.([etanf ] ) describe a collapsing string for @xmath276 with invariant size @xmath253 . that is , in frw universes @xmath277 , the string size goes to zero for @xmath255 as the universe radius @xmath278 . using eqs.([epa ] ) and ( [ etanf ] ) , the energy and pressure take the form , @xmath279 that yields @xmath280 as we shall see below , this behaviour is subdominant in frw universes . [ notice that eq.([eqesf ] ) is identical eq.([eqes ] ) that holds for @xmath281 in power - like inflationary universes ( @xmath249 ) ] . the second behavior present near @xmath250 is @xcite @xmath282 the invariant string size behaves for @xmath264 as , @xmath283 this solution describes a string that collapses for @xmath284 . here , the expansion factor tends to zero as @xmath268 since @xmath284 . from eq.([epa ] ) we find that @xmath285 that is , @xmath286 notice that the pressure is here * positive*. this second behaviour is related by duality ( @xmath233 ) to the first behaviour described by eqs.([epunf])-([eqesf ] ) . hence , we see from eqs.([epunf]-[epdf ] ) that near the big bang ( @xmath2 ) the * dual * behavior dominates over the unstable behavior . it should also be noticed that for this dual behavior , the energy redshifts exactly as @xmath3 for @xmath2 . for large @xmath21 , the ring strings exhibit a stable behaviour@xcite , @xmath287 where @xmath288 is a constant phase and the oscillation amplitude has been normalized . for large @xmath289 , the energy and pressure of the solution ( [ solasim ] ) behave as @xmath290 this is the analog of the stable behaviour ( [ esta ] ) in de sitter universe for @xmath291 . notice that eqs.([epes ] ) hold both for frw @xmath292 and inflationary spacetimes @xmath254 . the behaviour described by eqs.([epes ] ) is similar to minkowski spacetime . notice that the factor @xmath293 in the comoving radius [ eq.([epes ] ) ] is just the scale factor . that is , the physical string radius oscillates with constant amplitude as in minkowski spacetime . the fact that the spacetime curvature tends to zero ( except for de sitter universe ) when @xmath294 explains here the presence of a minkowski - type behaviour . for power - like inflationary universes with @xmath295 a special exact ring solution exist @xcite with @xmath296 where @xmath297 is an arbitrary constant . for @xmath298 ( de sitter universe ) this is the solution @xmath299 in ref.@xcite which has constant string size . for this solution we find @xmath300 = k \ ; r^{1 + 2/k } \nonumber \\ p & = & -({1 \over 2}+{1 \over k})~e \label{eexo } \nonumber\end{aligned}\ ] ] where @xmath301 is a constant . this is a fluid - like equation of state with @xmath302 . notice that @xmath303 . for this solution , the energy grows with @xmath1 as @xmath1 to a power @xmath304 where , since @xmath305 , @xmath306 ( @xmath138 and @xmath242 of this solution are constants in de sitter spacetime @xmath307 ) . this means that these special strings are subdominant both for @xmath0 and for @xmath19 where the unstable strings ( @xmath308 ) and their dual ( @xmath309 ) dominate respectively . in ref.@xcite the exact general evolution of circular strings in @xmath210 dimensional de sitter spacetime is described closely and completely in terms of elliptic functions . such solutions follow the form of eq.([ansatz ] ) . the evolution depends on a constant parameter @xmath310 , related to the string energy , and falls into three classes depending on whether @xmath311 ( oscillatory motion ) , @xmath312 ( degenerated , hyperbolic motion ) or @xmath313 ( unbounded motion ) . the novel feature here is that one single world - sheet generically describes _ infinitely many _ ( different and independent ) strings . the world - sheet time @xmath21 is an infinite - valued function of the string physical time , each branch yields a different string . this has no analogue in flat spacetime . we computed in ref.@xcite the string energy @xmath138 as a function of the string proper size @xmath52 , and analyze it for the expanding and oscillating strings . for expanding strings @xmath314 : @xmath315 even at @xmath316 , @xmath138 decreases for small @xmath52 and increases @xmath317 for large @xmath52 . for an oscillating string @xmath318 , the average energy @xmath319 over one oscillation period is expressed as a function of @xmath320 as a complete elliptic integral of the third kind @xcite . let us briefly describe here the periodic solutions found in ref.@xcite . the scale factor takes the form @xmath321 } \ ; \frac{\omega\vartheta'_1(0)}{\pi\vartheta_1(\omega y)}\ ; \mid \frac{\vartheta_4(\omega(\tau - y))}{\vartheta_4(\omega\tau)}\mid\end{aligned}\ ] ] where @xmath322 stands for the elliptic modulus , @xmath323 and @xmath324 stands for a 1st . class elliptic integral . notice that @xmath325 for @xmath322 . @xmath310 relates with @xmath326 through @xmath327 . we see that @xmath147 is an oscillatory function of @xmath21 . the cosmic time can be explicitly written as a fourier series : @xmath328 + \log\mid\frac{\omega\vartheta'_1(0)}{\pi\vartheta_1(\omega y)}\mid \nonumber \cr -4\ , \sum_{n=1}^{\infty}\frac{q^n}{n(1-q^{2n})}\sin ( n y \omega ) \sin\left[n\pi\omega(2\tau - y)\right].\end{aligned}\ ] ] the comoving string radius @xmath329 [ see eq.([ansatz ] ) ] takes here the form @xmath330.\end{aligned}\ ] ] the cosmic time and the comoving string radius are therefore completely regular functions of @xmath21 , and it follows that this string solution which is _ oscillating regularly _ as a function of world - sheet time @xmath21 , is also oscillating regularly when expressed in terms of hyperboloid time or cosmic time . this solution represents one _ stable _ string . the invariant string size results : @xmath331 we see that the string oscillates * inside * the horizon : @xmath332 we see from eqs . ( [ tper]-([fper ] ) that the string oscillations do not follow a pure harmonic motion as in flat minkowski spacetime , but they are precise superpositions of all frequencies @xmath333 ( @xmath334 ) with uniquely defined coefficients . the non - linearity of the string equations in de sitter spacetime fixes the relation between the mode coefficients . in the present case the basic frequency @xmath335 depends on the string energy , while in minkowski spacetime the frequencies are merely @xmath191 . the energy ( [ enani ] ) of this solution is given by@xcite : @xmath336\;+\ ; { \mbox oscillating\;terms},\end{aligned}\ ] ] where zn@xmath337 stands for the jacobi zeta - function @xcite . averaging over a period @xmath338 , the average energy @xmath319 is just the square bracket term . the pressure ( [ enani ] ) averaged over a period can be expressed here as a combination of complete elliptic integrals of the third kind . numerical evaluation gives zero with high accuracy . we have checked in addition that the first orders in the expansion in the @xmath326 identically vanish . we conclude that the average pressure is zero as in minkowski spacetime . thus , in average the _ oscillating strings _ describe _ cold matter_. in the @xmath339 limit the amplitude of the string size oscillation and the energy goes to zero [ see eq.([sper]-[eper ] ) ] . actually the solution disappears in such limit . the other interesting limit is @xmath340 where the elliptic solutions turn into hyperbolic solutions . then @xmath341 . more generally a numerical analysis shows that @xmath319 is a monotonically increasing function of @xmath310 for @xmath342 $ ] . in summary , three asymptotic behaviors are exhibited by ring solutions . \i ) unstable for @xmath0 and @xmath343 : @xmath344 . \ii ) dual to unstable for @xmath343 and @xmath0 : @xmath345 ( except for de sitter where @xmath346 ) . \iii ) stable for @xmath347 @xmath348 . recall that the unstable behaviour dominates for @xmath0 in inflationary universes whereas the dual behaviour dominates in frw universes for @xmath343 . in addition we have the special behavior ( [ eexo ] ) for @xmath295 . notice that the three behaviors i)-iii ) appear for all expansion factors @xmath349 . the behaviours i ) and ii ) are related by the duality transformation @xmath233 , the case iii ) being invariant under duality . in the three cases we find perfect fluid equations [ see eq.([eflu ] ) ] with different @xmath350 : @xmath351 where the indices @xmath352 and @xmath353 stand for ` unstable ' , ` dual ' and ` stable ' , respectively . assuming a perfect gas of strings on a volume @xmath354 , the energy density @xmath157 will be proportional to @xmath355 . this yields the following scaling with the expansion factor using the energies from i)-iii ) : @xmath356 all three densities and pressures obey the continuity equation ( [ cont ] ) , as it must be . the factor @xmath357 in the relation between @xmath242 and @xmath138 for the cases i ) and ii ) is purely geometric . notice that this factor was one in @xmath358 dimensions [ eq.([prou ] ) ] . the solutions investigated in secs . ii.a and ii.b can exist for any dimensionality of the spacetime . embedded in d - dimensional universes , the @xmath359 solutions of section ii.a describe straight strings , the @xmath360 solutions of section ii.b are circular rings . in d - dimensional spacetime , one expects string solutions spread in @xmath97 spatial dimensions . their treatment has been done asymptotically in ref.@xcite . one finds for @xmath361 , @xmath18 , @xmath362 this relation coincides with eq.([rop ] ) for @xmath363 and with eqs.([peun ] ) and ( [ eqes ] ) for @xmath364 . the energy density scales with @xmath1 as @xmath365 ( where @xmath366 is a constant ) in accordance with eq.([rodu ] ) for @xmath367 and with eq.([rodd ] ) for @xmath368 . for the dual regime , @xmath369 , we have : @xmath370 ( where @xmath16 is a constant ) . eq.([estdu ] ) reduces to eq.([duud ] ) for @xmath367 and to eqs.([dual ] ) and ( [ duat ] ) for @xmath368 . in this dual regime strings have the same equation of state as massless radiation . finally , for the stable regime we have @xmath371 ( where @xmath353 is a constant ) . this regime is absent in @xmath367 and appears for @xmath368 solutions in eqs.([esta ] ) and ( [ epes ] ) . the lack of string transverse modes in @xmath367 explains the absence of the stable regime there . the equation of state for stable strings coincides with the one for cold matter . in conclusion , an ideal gas of classical strings in cosmological universes exhibit three different thermodynamical behaviours , all of perfect fluid type : \1 ) unstable strings : negative pressure gas with @xmath372 \2 ) dual behaviour : positive pressure gas similar to radiation , @xmath373 \3 ) stable strings : positive pressure gas similar to cold matter , @xmath374 . tables i and ii summarize the main string properties for any scale factor @xmath375 . string energy and pressure as obtained from exact string solutions for various expansion factors @xmath376 . * string properties for arbitrary @xmath376 * [ cols= " < , < , < , < " , ] among the cosmological backgrounds , de sitter spacetime occupies a special place . this is , in one hand relevant for inflation and on the other hand string propagation turns to be specially interesting there @xcite - @xcite . string unstability , in the sense that the string proper length grows indefinitely is particularly present in de sitter . the string dynamics in de sitter universe is described by a generalized sinh - gordon model with a potential unbounded from below @xcite . the sinh - gordon function @xmath377 having a clear physical meaning : @xmath378 determines the string proper length . moreover the classical string equations of motion ( plus the string constraints ) turn to be integrable in de sitter universe @xcite . more precisely , they are equivalent to a non - linear sigma model on the grassmannian @xmath379 with periodic boundary conditions ( for closed strings ) . this sigma model has an associated linear system @xcite and using it , one can show the presence of an infinite number of conserved quantities @xcite . in addition , the string constraints imply a zero energy - momentum tensor and these constraints are compatible with the integrability . the so - called dressing method @xcite in soliton theory allows to construct solutions of non - linear classically integrable models using the associated linear system . in ref.@xcite we systematically construct string solutions in three dimensional de sitter spacetime . we start from a given exactly known solution of the string equations of motion and constraints in de sitter @xcite and then we `` dress '' it . the string solutions reported there indeed apply to cosmic strings in de sitter spacetime as well . the invariant interval in @xmath17-dimensional de sitter space - time is given by @xmath380~ \sum_{i=1}^{d-1 } ( dx^{i})^2 .\ ] ] here @xmath23 is the so called cosmic time . in terms of the conformal time @xmath24 , @xmath381 } \o h } \quad , -\infty < \eta \leq 0 ~ , \ ] ] the line element becomes @xmath382 ~.\ ] ] the de sitter spacetime can be considered as a @xmath17-dimensional hyperboloid embedded in a d+1 dimensional flat minkowski spacetime with coordinates @xmath383 : @xmath384\ ] ] where @xmath385~ \sum_{i=1}^{d-1 } ( x^i)^2 ~~ , \cr \cr q^1 & = & ~\cosh{h t}~ - { { h^2 } \o 2 } \exp[h t]~ \sum_{i=1}^{d-1 } ( x^i)^2 ~~ , \cr \cr q^{i+1 } & = & ~h \exp[h t]~ x^{i } \quad , \quad 1 \leq i \leq d-1 , ~ -\infty < t , ~ x^{i } < + \infty .\end{aligned}\ ] ] the complete de sitter manifold is the hyperboloid @xmath386 the coordinates @xmath387 and @xmath388 cover only the half of the de sitter manifold @xmath389 . we will consider a string propagating in this d - dimensional space-time.the string equations of motion ( [ conouno ] ) in the metric ( [ mehip ] ) take the form : @xmath390 where . stands for the lorentzian scalar product @xmath391 and @xmath392 . the string constraints ( [ conodos ] ) become for de sitter universe @xmath393 eqs.([desmo ] ) describe a non compact o(d,1 ) non - linear sigma model in two dimensions . in addition , the ( two dimensional ) energy - momentum tensor is required to vanish by the constraints eqs.([bincu ] ) . this system of non - linear partial differential equations can be reduced by choosing an appropriate basis for the string coordinates in the ( @xmath394)-dimensional minkowski space time @xmath383 to a noncompact toda model @xcite . these equations can be rewritten in the form of a chiral field model on the grassmanian @xmath395 . indeed , any element @xmath396 can be parametrized with a real vector @xmath397 of the unit pseudolength @xmath398 in terms of @xmath399 , the string equations ( [ desmo ] ) have the following form @xmath400 and the conformal constraints ( [ bincu ] ) become @xmath401 these conditions are equivalent to the existence of the representation ( [ g ] ) . equation ( [ geq ] ) is the compatibility condition for the following overdetermined linear system : @xmath404 where @xmath405 or , in terms of the vector @xmath397 @xmath406 @xmath407 eq . ( [ g ] ) can be easily inverted yielding @xmath408 in terms of the matrix @xmath181 : @xmath409 this condition is compatible with the above equations since the matrix function @xmath410 at the point @xmath412 satisfies the same equations as @xmath399 . thus the problem of constructing exact solutions of the string equations is reduced to finding compatible solutions of the linear equations ( [ psieq ] ) such that @xmath413 satisfies the constraints eqs.([gcond ] ) and ( [ gcond1 ] ) . in ref.@xcite the dressing method was applied as follows . we started from the exact ring - shaped string solution @xmath414 @xcite and we find the explicit solution @xmath415 of the associated linear system , where @xmath416 stands for the spectral parameter . then , we propose a new solution @xmath417 that differs from @xmath415 by a matrix rational in @xmath416 . notice that @xmath418 provides in general a new string solution . we then show that this rational matrix must have at least * four * poles , @xmath419 , as a consequence of the symmetries of the problem . the residues of these poles are shown to be one - dimensional projectors . we then prove that these projectors are formed by vectors which can all be expressed in terms of an arbitrary complex constant vector @xmath420 and the complex parameter @xmath421 . this result holds for arbitrary starting solutions @xmath414 . since we consider closed strings , we impose a @xmath87-periodicity on the string variable @xmath422 . this restricts @xmath421 to take discrete values that we succeed to express in terms of pythagorean numbers . in summary , our solutions depend on two arbitrary complex numbers contained in @xmath420 and two integers @xmath423 and @xmath424 . the counting of degrees of freedom is analogous to @xmath425 minkowski spacetime except that left and right modes are here mixed up in a non - linear and precise way . the vector @xmath420 somehow indicates the polarization of the string . the integers @xmath426 determine the string winding . they fix the way in which the string winds around the origin in the spatial dimensions ( here @xmath427 ) . our starting solution @xmath428 is a stable string winded @xmath429 times around the origin in de sitter space . the matrix multiplications involved in the computation of the final solution were done with the help of the computer program of symbolic calculation `` mathematica '' . the resulting solution @xmath430 is a complicated combination of trigonometric functions of @xmath422 and hyperbolic functions of @xmath21 . that is , these string solitonic solutions do not oscillate in time . this is a typical feature of string unstability @xcite - @xcite - @xcite . the new feature here is that strings ( even stable solutions ) do not oscillate neither for @xmath431 , nor for @xmath432 . we plot in figs . 1 - 7 the solutions for significative values of @xmath433 and @xmath434 in terms of the comoving coordinates @xmath435 @xmath436 the first feature to point out is that our solitonic solutions describe * multiple * ( here five or three ) strings , as it can be seen from the fact that for a given time @xmath23 we find several different values for @xmath21 . that is , @xmath21 is a * multivalued * function of @xmath23 for any fixed @xmath206 ( fig.1 - 2 ) . each branch of @xmath21 as a function of @xmath23 corresponds to a different string . this is a entirely new feature for strings in curved spacetime , with no analogue in flat spacetime where the time coordinate can always be chosen proportional to @xmath21 . in flat spacetime , multiple string solutions are described by multiple world - sheets . here , we have a * single * world - sheet describing several independent and simultaneous strings as a consequence of the coupling with the spacetime geometry . notice that we consider _ free _ strings . ( interactions among the strings as splitting or merging are not considered ) . five is the generic number of strings in our dressed solutions . the value five can be related to the fact that we are dressing a one - string solution ( @xmath414 ) with _ four _ poles . each pole adds here an unstable string . in order to describe the real physical evolution , we eliminated numerically @xmath437 from the solution and expressed the spatial comoving coordinates @xmath438 and @xmath439 in terms of @xmath23 and @xmath206 . we plot @xmath440 as a function of @xmath206 for different fixed values of @xmath23 in fig.3 - 4 . it is a sinusoidal - type function . besides the customary closed string period @xmath87 , another period appears which varies on @xmath21 . for small @xmath289 , @xmath441 has a convoluted shape while for larger @xmath26 ( here @xmath442 ) , it becomes a regular sinusoid . these behaviours reflect very clearly in the evolution of the spatial coordinates and shape of the string . the evolution of the five ( and three ) strings simultaneously described by our solution as a function of @xmath23 , for positive @xmath23 is shown in figs . one string is stable ( the 5th one ) . the other four are unstable . for the stable string , @xmath443 contracts in time precisely as @xmath444 , thus keeping the proper amplitude @xmath445 and proper size constant . for this stable string @xmath446 . ( @xmath447 = the horizon radius ) . for the other ( unstable ) strings , @xmath443 become very fast constant in time , the proper size expanding as the universe itself like @xmath448 . for these strings @xmath449 . these exact solutions display remarkably the asymptotic string behaviour found in refs.@xcite . in terms of the sinh - gordon description , this means that for the strings outside the horizon the sinh - gordon function @xmath377 is the same as the cosmic time @xmath23 up to a function of @xmath206 . more precisely , @xmath450 \right\ } + o(e^{-2ht } ) . \label{tcosa}\ ] ] here @xmath451 and @xmath452 are the @xmath438 and @xmath439 coordinates outside the horizon . for @xmath453 these strings are at the absolute _ minimum _ @xmath454 of the sinh - gordon potential with infinite size . the string inside the horizon ( stable string ) corresponds to the _ maximum _ of the potential , @xmath455 . @xmath456 is the only value in which the string can stay without being pushed down by the potential to @xmath457 and this also explains why only one stable string appears ( is not possible to put more than one string at the maximum of the potential without falling down ) . these features are _ generically _ exhibited by our one - soliton multistring solutions , independently of the particular initial state of the string ( fixed by @xmath458 and @xmath426 ) . for particular values of @xmath458 , the solution describes three strings , with symmetric shapes from @xmath459 , for instance like a rosette or a circle with festoons ( fig . 5 - 7 ) . the string solutions presented here trivially embedd on @xmath17-dimensional de sitter spacetime ( @xmath460 ) . it must be noticed that they exhibit the essential physics of strings in d - dimensional de sitter universe . moreover , the construction method used here works in any number of dimensions . h.j . de vega and n. snchez in `` string quantum gravity and the physics at the planck scale '' , proceedings of the erice workshop held in june 1992 . edited by n. snchez , world scientific , 1993 . pages 73 - 185 , and references given therein . * figure 3 : * @xmath465 for fixed @xmath23 for @xmath466 . three values of ht are displayed , corresponding to ht=0 ( full line ) , 1 ( dots ) , and 2 ( dashed line ) . for each ht , three curves are plotted , which correspond to the three strings . they are ordered with @xmath21 increasing . * figure 5 : * evolution as a function of cosmic time @xmath468 of the three strings , in the comoving coordinates @xmath469 , for @xmath470 . the comoving size of string ( 1 ) stays constant for @xmath471 , then decreases around @xmath472 , and stays constant again after @xmath473 . the invariant size of string ( 2 ) is constant for negative @xmath468 , then grows as the expansion factor for @xmath474 , and becomes identical to string ( 1 ) . the string ( 3 ) has a constant comoving size for @xmath475 , then collapses as @xmath476 for positive @xmath468 .

we investigate the structural properties of a two - dimensional system of ellipsoidal particles carrying a linear quadrupole moment in their center . these particles represent a simple model for a variety of uncharged , non - polar conjugated organic molecules . using optimization tools based on ideas of evolutionary algorithms , we first examine the ground state structures as we vary the aspect ratio of the particles and the pressure . interestingly , we find , besides the intuitively expected t - like configurations , a variety of complex structures , characterized with up to three different particle orientations . in an effort to explore the impact of thermal fluctuations , we perform constant - pressure molecular dynamics simulations within a range of rather low temperatures . we observe that ground state structures formed by particles with a large aspect ratio are in particular suited to withstand fluctuations up to rather high temperatures . our comprehensive investigations allow for a deeper understanding of molecular or colloidal monolayer arrangements under the influence of a typical electrostatic interaction on a coarse - grained level . authors who have contributed equally to the research

1512.01162

in recent years there is increasing interest in understanding the film morphologies of anisotropic , conjugated organic molecules at inorganic surfaces . a prime example are systems of optically active molecules at inorganic surfaces ; these so - called hybrid inorganic / organic systems ( hios ) represent a very promising material class in optoelectronics.@xcite typically , the corresponding organic molecules are strongly anisotropic in shape and are characterized by complex charge distributions , often dominated by a quadrupole moment . @xcite by manipulating the orientational structure of such organic layers it is possible to tune the efficiency of the charge carrier transport @xcite and thus to optimize the efficiency of the hybrid system . somewhat earlier , organic molecules at interfaces have attracted attention in the context of so - called langmuir monolayers , that is , two - dimensional ( 2d ) films of typically amphiphilic molecules constrained to a liquid - gas interface . @xcite again , the orientational and translational structure in such systems can be quite complex . from the theoretical side , a full microscopic treatment of an organic molecular layer is still challenging , and this holds particularly for hios ( where the substrate is typically patterned ) . on the level of atomistic molecular dynamics ( md ) simulations , one has to consider a large number of interactions such as atomic bonding , van der waals ( lennard - jones ) potentials , contributions due to bending or torsion of a molecule ( see e.g. ref . ) , and electrostatic interactions . even then , it often turns out that atomistic md calculations do not correctly predict the arrangement of several particles , one main reason being the insufficient treatment of polarizability effects . @xcite an example that demonstrates this issue is the quadrupolar molecule benzene , where a representation via atomic point charges @xcite or multipoles @xcite fails to reproduce the corresponding dimer configuration . this suggests to employ quantum chemical approaches to evaluate the electronic structure of the entire system , instead . however , at the moment , this level of complexity is computationally far too demanding , especially at finite temperatures . given these difficulties , the goal of the present paper is to understand _ generic _ aspects of the structural behavior of quadrupolar , anisotropic molecules at interfaces based on a _ coarse - grained _ model . specifically , we investigate a 2d many - particle system composed of ellipse - shaped particles with an embedded , axially symmetric quadrupole tensor ( higher - order multipoles are neglected ) . the quadrupole is oriented along one of the ellipsoidal axes as it is the case , e.g. , in benzene @xcite or naphtalene , whose quadrupole tensor nearly has axial symmetry . @xcite we further assume that the center of the particles is confined to a 2d plane , and that the molecules are allowed to rotate only within this plane . interestingly , 2d quadrupolar systems are , so far , rather unexplored . this contrasts the situation in the 3d case , where a considerable amount of pioneering work involving quadrupolar particles at finite temperatures @xcite and in the ground state @xcite is available . on the other hand , there are a number of simulation studies dealing with non - quadrupolar , shape - anisotropic particles in 2d , examples being hard ellipsoids , @xcite rectangles @xcite or hard spherocylinders.@xcite in particular , the density - driven nematic - isotropic transition of hard ellipsoids was already investigated in ref . and turned out being continuous or of first order depending on the aspect ratio . we also mention a density functional study @xcite of 2d systems of anisotropic particles exhibiting isotropic phases and phases with simultaneous orientational and translational order . thus , the important open question to be explored in the present article concerns the interplay of shape anisotropy and quadrupolar intermolecular interactions . to this end we employ two types of numerical calculations , that is , optimization techniques based on ideas of evolutionary algorithms ( ea ) and constant - pressure md simulations , revealing the system s behavior in the ground state and at finite temperatures . to facilitate the calculations , the non - electrostatic part of the interaction between two molecules is modeled via a purely repulsive , relatively stiff ( but not infinitely hard ) pair potential that foots on the ellipse - like contact distance suggested by berne and pechukas . @xcite the analytical form of this pair potential resembles an anisotropic soft - sphere potential , but has a higher exponent.@xcite this potential turned out to be suitable for both type of numerical calculations . van der waals interactions between the molecules are entirely neglected . although this may seem somewhat unrealistic for real molecules , there are several advantages of the resulting ( repulsive ) potential : it reduces the number of parameters , it allows ( as we will demonstrate ) for a transferability between results at different system parameters , and it makes it possible to establish a connection of the present study to investigations of anisotropic colloidal particles in confined geometry.@xcite the remainder of this article is organized as follows . in sec . [ sec : model ] we introduce the model and the pair interactions in detail . section [ sec : methods_data_analysis ] is devoted to the methods used to calculate equilibrium structures ( at @xmath1 and @xmath2 ) and corresponding data analysis methods . the resulting ground state structures at different pressures and aspect ratios are presented in sec . [ subsec : ground_state ] , and the corresponding finite - temperature results are discussed in sec . [ subsec : finite_temperatures ] . finally , we summarize our findings in sec . [ sec : conclusion ] . appendices a to c provide additional , complementary information . in our investigations we consider a two - dimensional system , embedded in the @xmath3-plane , where the anisotropic particles are only allowed to rotate around the @xmath4-axis . we use reduced units throughout , introducing parameters @xmath5 for the length scale , @xmath6 for the energy scale , and @xmath7 for the unit of mass . our particles are assumed to have an elliptic shape , characterized by the lengths of the two main axes , @xmath8 and @xmath9 ; the indices refer to the orientation of the corresponding axes relative to the linear quadrupolar moment to be introduced below ( see figure [ fig : model ] ) . defining a dimensionless shape anisotropy parameter , @xmath10 , we impose via @xmath11 that the surface area of our particles is independent of the actual value of @xmath10 . the isometric case of a disc is recovered for @xmath12 . the interaction between two - particles ( with indices @xmath13 and @xmath14 ) can be split into a short - ranged , anisotropic repulsion , @xmath15 , and a long - ranged potential , @xmath16 ; the latter one stems from the interaction of the linear quadrupoles that the particles carry . we define the distance @xmath17 between the centers of the particles and further @xmath18 . the @xmath19 and @xmath20 are the normalized orientation vectors of the respective linear quadrupolar moments . within our model the shape of the particle is the same for @xmath10 and @xmath21 , while the respective orientations of the embedded quadrupole moment are perpendicular for the cases @xmath10 and @xmath21 ( see bottom panels of figure [ fig : model ] ) . consequently , by choosing values of @xmath10 both smaller and larger than unity we are able to consider these two orientations of the quadrupolar moment within the same model . with the above vectors at hand we introduce @xmath22 for the short - range contribution of the inter - particle interaction ( subscript `` sr '' ) we use an anisotropic , repulsive potential : it has the simple functional form of an inverse power law ( ipl ) interaction , while its dependence on the connecting vector , @xmath23 , and on the orientations of the quadrupolar moments , @xmath24 and @xmath25 , is inspired by a gay - berne potential:@xcite @xmath26^{18 } ; \label{eqn : vsr}\end{aligned}\ ] ] here @xmath27 \right\}^{-1/2 } \label{eqn : sigma}\\ \chi & = & \frac{\kappa^2 - 1}{\kappa^2 + 1 } .\end{aligned}\ ] ] the electrostatic part of the inter - particle interaction is based on _ linear _ quadrupole moments , for which the quadrupole tensor , @xcite @xmath28 , becomes in its eigenbasis @xmath29 with @xmath30 being the strength of the moment . the interaction between two linear quadrupole moments can be written as @xcite @xmath31 , \label{eqn : vlr}\end{gathered}\ ] ] with @xmath32 , @xmath33 , and @xmath34 defined above and @xmath35 being the vacuum permittivity . even though the interaction decays slower with the distance ( i.e. , @xmath36 ) than , e.g. , van der waals interactions ( @xmath37 ) we can still calculate inter - particle energies with sufficient accuracy via a real - space lattice sum . here we have used a cutoff radius @xmath38 for both , electrostatic and repulsive interactions . close to unity ) , induced by the fact that neighboring particles tend to arrange in a mutually orthogonal orientation.,width=188 ] the interaction of two linear quadrupoles depends on their respective orientations . in contrast to dipoles , which tend to line up head - to - tail , isolated linear quadrupoles prefer the so - called t - configuration for @xmath10 close to unity ( see figure [ fig : t_configuration ] ) . a more detailed discussion of the impact of shape anisotropy on the two - particle arrangement with a minimal electrostatic energy is given in appendix [ sec : two - particle arrangement with minimal electrostatic energy ] . throughout , our numerical calculations were performed in the @xmath39-ensemble , where the relevant thermodynamic potential for vanishing temperature is the enthalpy , @xmath40 , @xmath41 here @xmath42 is the internal energy , @xmath43 is the pressure , and @xmath44 is the area occupied by the system . further , we introduce the temperature of the system , @xmath45 . based on the scales of length ( @xmath5 ) , mass ( @xmath7 ) , and energy ( @xmath6 ) , the quadrupole strength , @xmath46 , the pressure @xmath43 , the temperature @xmath45 and the time @xmath47 are expressed via their respective reduced units , @xmath48 , @xmath49 , @xmath50 ( with @xmath51 being the boltzmann constant ) and @xmath52 . in our approach we investigate in a _ first _ step the self - assembly scenarios of our particles at _ zero _ temperature ( i.e. , the ground state configurations ) . this information helps us to classify archetypical particle arrangements ( each characterized by a particular spatial and orientational order ) , which , in turn , help us to understand the strategy of the particles of how to assemble in the energetically most favorable manner . for these investigations we employ an optimization tool based on evolutionary algorithms , detailed below . in a _ second _ step we take advantage of these ground state configurations and use them as starting configurations in molecular dynamics ( md ) simulations , performed at small , but _ finite _ temperatures . these investigations provide information on the thermodynamic stability of the ordered phases . in our effort to identify ground state configurations of our systems , we use an optimization tool based on ideas of evolutionary algorithms ( eas ) . @xcite eas are heuristic approaches designed to search for global minima in high - dimensional search spaces and for problems that are characterized by rugged energy landscapes . in an effort to be compatible with the requirements of an @xmath39-ensemble , we introduce in our approach a unit cell of variable area and shape which creates ( together with its periodic images ) a system of infinite extent . in the desired configuration ( the so - called ground state configuration ) particles are located and oriented in this cell in such a way as to minimize the internal energy of the system , which at vanishing temperature is equivalent to the enthalpy . we initialize the algorithm by creating a set of configurations where particles are located in the cell at random positions and have random orientations . these arrangements are graded by their respective fitness value , a quantity that provides evidence on how suitable this configuration is to solve the optimization problem . since we are interested in finding ground state structures , a high fitness value of a particular configuration corresponds to a low value of the enthalpy per particle . then we iteratively use existing particle arrangements to create new ones by applying one of the two following operations : crossover and mutation . in the former one we first select two configurations where the choice is biased by high fitness values of the two configurations . characteristic features of both particle arrangements ( such as lattice vectors , particle positions , and particle orientations ) are then combined to form a new configuration . the mutation operation , on the other hand , introduces random changes to a randomly chosen configuration , such as moving or rotating an arbitrarily chosen particle or changing the lattice vectors . typically 2000 iterations are required for a particular state point until convergence towards the global minimum has been achieved . our implementation of eas is memetic , i.e. , we combine global and local search techniques : each time a new configuration has been created with one of the two above mentioned ea operations , we apply the l - bfgs - b @xcite algorithm which guides us to the nearest local minimum . this algorithm has been applied successfully for a broad spectrum of systems , both in two and three dimensions ( see , for instance , refs . ) . within the context of the present contribution , it should also be mentioned , that a suitable extension of the formalism is also able to cope with long - ranged interactions , as they are encountered in charged systems ( see refs . ) . for the present contribution we have performed with our ea approach computations for 211 evenly - spaced values of @xmath53 $ ] and for several combinations of @xmath54 and @xmath55 . due to computational limitations we have considered unit cells that contain up to twelve particles . in order to investigate structural changes of our system at finite temperature , we perform molecular dynamics ( md ) simulations at constant pressure and temperature for an ensemble of @xmath56 particles . to this end we use the berendsen weak - coupling scheme , @xcite where the classical newtonian equations of motion are supplemented by terms representing the coupling to a heat- and a pressure - bath . in the following , we first introduce the translational equations of motion proposed by berendsen _ _ , @xcite specializing to the case of a two - dimensional system . considering a particle with index @xmath13 ( @xmath57 one has @xmath58 \mathbf{b}_\alpha . \label{eqn : eqm_box_vectors}\end{aligned}\ ] ] where the vectors @xmath59 ( @xmath60 ) define the simulation box and its shape . in the above equations , the dot represents a time - derivative . further , @xmath61 and @xmath62 denote the position and the velocity of particle with index @xmath13 , respectively ; @xmath63 is the force on particle @xmath13 with mass @xmath7 , and @xmath64 is the compressibility . the pressure tensors , @xmath65 ( actual pressure ) and @xmath66 ( target pressure ) , are defined below . further , @xmath67 is the actual kinetic temperature , @xmath68 $ ] where the denominator @xmath69 represents the @xmath70 translational degrees of freedom minus two constraints due to momentum conservation in each spatial dimension . to ensure this conservation during the numerical integration , we set the total momentum to zero every 100 time steps . the ( time ) constants @xmath71 and @xmath72 determine the speed of relaxation of @xmath67 and @xmath73 towards their respective target values defined by the bath . specifically , the corresponding coupling equations read @xmath74 where the pressure tensors are defined as @xmath75 \\ & \text{and}\nonumber \\ & \underline{\mathbf{p}}= \frac{1}{2 } p\,\,\underline{\mathbf{1 } } . \label{eqn : pressure tensors}\end{aligned}\ ] ] here , @xmath76 denotes the dyadic product of two vectors , @xmath44 stands for the area of the simulation cell , @xmath77 is the unit - tensor , and @xmath78 is the force on particle @xmath13 exerted by particle @xmath14 . to describe the rotational motion of the particles , eqs . are supplemented with the following differential equations for the angular velocity , @xmath79 , of particle @xmath13 : @xmath80 here , @xmath81 and @xmath82 are the torque on particle @xmath13 and the corresponding moment of inertia of particle @xmath13 , respectively ; @xmath83 denotes the vector product . the kinetic temperature for the rotation is defined as @xmath84 . this temperature is controlled in analogy to eqs . . to solve the translational equations of motion , i.e. eqs . we use a modified leap - frog integrator as proposed in ref . . for the solution of the rotational equations of motion , eqs . , an analogous procedure is performed ( see ref . ) . to initialize the particle positions in our system , we use the unit cells obtained from the ground state calculations in the preceding investigations ( see subsec . [ subsec : ea ] ) . specifically , we take for each ( @xmath10-dependent ) ground state the respective unit cell and arrange copies of this cell such that the simulation cell has a minimal circumference . for this simulation cell ( which is now defined by the simulation box vectors @xmath85 and @xmath86 ) periodic boundary conditions are applied . the pair forces @xmath87 and torques @xmath88 are truncated at @xmath89 and are then shifted in order to make them vanish smoothly . the md calculations have been performed at several values of the reduced temperature @xmath90 , that is , @xmath91 . in our simulations we used a time increment @xmath92 ( corresponding at a macroscopic level to @xmath93 ) . an adequate choice of the time step is imposed by the mass of the particles , for which we have assumed the mass of benzene , @xmath94 . each simulation extends over 210 000 time steps . during the first 20 000 md steps , the target temperature is gradually increased from @xmath95 towards the respective target value . structural quantities are then extracted only during the final 10 000 time steps of the simulation . the values for the temperature- and pressure - coupling constants @xmath71 , @xmath96 and @xmath72 were chosen to be @xmath97 time steps . of course , the actual volume change of the box is also influenced by the compressibility ; its reduced , dimensionless counterpart , @xmath98 , is set to 0.01 . we assume the mass distribution within the particles to be homogeneous and thus obtain for the dimensionless moment of inertia @xmath99 . in order to quantify both the positional and the orientational order of the particles we introduce three different sets of order parameters : * the _ positional _ order can be quantified via two - dimensional bond - orientational order parameters ( boops ) , @xcite introducing , in addition , weight factors @xcite that are related to the side lengths of the voronoi polygon around each particle : @xmath100 here ( @xmath101 ) indicates that particle @xmath14 is a nearest neighbor of particle @xmath13 and @xmath102 is the length of the side of the voronoi polygon that separates the two particles . @xmath103 is the angle enclosed by the bond between particles @xmath13 and @xmath14 and the reference axis , which we choose as the @xmath104-axis . for the parameter @xmath105 we have chosen the values 4 and 6 , highlighting thus via @xmath106 or @xmath107 four- or six - fold symmetry , respectively . while boops are calculated for only the best configuration at vanishing temperature , averages over several configurations are taken in md simulations , indicated by the brackets in eq . and in the following . * an obvious candidate for quantifying the _ orientational _ order of the particles is the nematic order parameter @xmath108 , which is introduced via the tensor order parameter @xmath109 , defined as @xmath110 @xmath108 is then simply the positive eigenvalue of the tracless tensor @xmath109 , which can easily be calculated as @xmath111 the eigenvector associated with @xmath108 is called the director @xmath112 and indicates a preferred orientation in the configuration . note that @xmath108 is always positive , except in the case that all elements of @xmath109 vanish : then no preferred direction is present in the system . + a different way of how to measure the orientational order is via the parameter @xmath113 , defined as @xmath114 * in addition , a variety of order parameters combining _ positional _ and _ orientational _ order can be defined . an example of such an order parameter , which has been used in a previous contribution @xcite is @xmath115 concluding , we emphasize that orientational order refers to the orientation of the _ particles _ ; the orientation of the linear quadrupolar moment is then imposed by the value of @xmath10 : if @xmath116 , then the orientation of the moment is parallel to the main axis of the particles , while for @xmath117 these two orientations are perpendicular to each other . in our discussion of the ground state configurations we first present the different structural archetypes that we could identify with our ea - approach . we then present the diagram of states : it provides information on the @xmath10-ranges where the respective structures are the energetically most stable ones ; we further discuss how the thermodynamic properties and the order parameters of our systems vary over a representative range of @xmath10 . [ [ structures - with - one - preferred - orientation ] ] structures with one preferred orientation + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for @xmath10-values both considerably larger or smaller than unity ( i.e. , by a factor of @xmath118 ) , the particles tend to align parallel to each other in case of strong anisotropy : they accomplish this by forming tilted rows , characterized by a high density of particles along these lines ; these arrangements are denoted as _ parallel displaced_-configurations ( `` pd '' ; see figure [ fig : systems_quadrupolar_snap1 ] ) . since quadrupolar particles avoid a head - to - tail arrangement , neighboring rows repel each other , inducing thereby large gaps between these lanes ; the parallel offset between neighboring rows can be very sensitive to small changes in @xmath10 , leading to a slight modulation of the boops as functions of @xmath10 ( to be discussed below ) . an interesting special case of this structure is observed for a vanishing quadrupolar moment ( i.e. , @xmath119 ) , where particles form a distorted hexagonal lattice , denoted as the _ parallel_-configuration ( `` p '' ; see top left panel of figure [ fig : systems_quadrupolar_snap1 ] ) . [ [ structures - with - two - preferred - particle - orientations ] ] structures with two preferred particle orientations + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + here the preferred structural arrangements are throughout _ herringbone_-configurations ( `` hb '' ) where particles form alternating , parallel lanes , each of them being characterized by a specific particle orientation : in figure [ fig : systems_quadrupolar_snap2 ] particles pertaining to different lanes are colored red and blue , respectively . the relative orientation between neighboring particles depends in a highly sensitive manner on @xmath10 . apart from two special cases ( that are encountered for @xmath10-values very close to unity and which will be discussed below ) two different versions of the hb - configurations emerge from a closer analysis of the obtained structures : * for @xmath10-values somewhat closer to unity , we observe a rather dense structure , which we denote hb@xmath120-configuration ( see panels in the central column of figure [ fig : systems_quadrupolar_snap2 ] ) ; * for @xmath10-values that differ more strongly from unity , particles arrange in densely - populated rows of alternating orientation ( see panels in the right column of figure [ fig : systems_quadrupolar_snap2 ] ) . since neighboring rows repel each other , this structure has a lower overall density which we therefore denote as hb@xmath121 . the two special cases of the hb structure mentioned above are observed for @xmath10-values very close to unity where neighboring particles prefer strict mutual orthogonal orientations with respect to their nearest neighbors . ( i ) for large @xmath122- and small @xmath123-values , a perfect arrangement of mutually orthogonally oriented particles with an underlying square pattern can be observed , denoted as the _ square t_-configuration ( `` t@xmath124 '' ; see left panel of figure [ fig : systems_quadrupolar_snap2 ] ) . ( ii ) further , a closely related arrangement has been identified which is now based on an underlying hexagonal pattern ; it is denoted as the _ hexagonal t_-configuration ( `` t@xmath125 '' ; see bottom left panel of figure [ fig : systems_quadrupolar_snap2 ] ) ; for this particular configuration a mutually perfect orthogonal orientation of nearest neighbors can only be realized for @xmath126 . [ [ structures - with - three - preferred - particle - orientations ] ] structures with three preferred particle orientations + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + a very interesting phenomenon observed in our system is the occurrence of structures where particles arrange in three preferred orientations , characterized by a vanishing nematic order parameter @xmath108 . the _ spatial _ particle arrangement is here irrespective of the values of @xmath127 and @xmath123 reminiscent of a _ trihexagonal _ tiling , @xcite consisting of regular hexagons that are connected by triangles ; we thus denote it as the `` th''-configuration ( see figure [ fig : systems_quadrupolar_snap3 ] where also the hexagons and the triangles are highlighted ) . throughout , the relative angle between the orientations of neighboring particles is @xmath128 , reflected in the order parameter which assumes a value of @xmath129 ( see figure [ fig : ea_2_1 ] ) . we note that the occurrence of the th - configuration strongly depends on the choice of @xmath122 and @xmath123 ; in some cases , this particle arrangement is not observed at all ( see figure [ fig : ea_02_01 ] ) . in contrast , variants of the -configuration can almost always be observed in some range of @xmath10 ( see discussion below ) . [ [ more - complicated - structures ] ] more complicated structures + + + + + + + + + + + + + + + + + + + + + + + + + + + using the ea - approach we also identify more complicated structures , not conforming to the mechanisms described above . these structures require a larger number of particles per cell and turn out to be stable only within very small ranges of @xmath10 ( see subsec . [ subsubsec : diagram_states ] ) . while we observe several different variants of such particle configurations , they share a common feature , namely a mesh - like lattice , where chains of particles undulate back and forth ( see figure [ fig : systems_quadrupolar_snap4 ] ) . we denote them as _ branched _ structures ( `` b '' ) . we now discuss the diagram of states which summarizes the occurrence of the previously identified archetypes of ground state configurations of our system for selected values of @xmath122 and @xmath123 as we vary @xmath10 . since most of the interesting features of our systems can already be captured in the diagram of states obtained for @xmath130 and @xmath131 , we focus from now onwards on this particular set of parameters : we present data for the enthalpy and for a selection of appropriate order parameters introduced in subsec . [ subsec : order_parameters ] , that help to identify the respective structures . the filling fraction @xmath132 , which is also displayed in the following figures , is defined as @xmath133 . at this point we remind the reader that the _ shape _ ( and thus the orientation of the particles ) is invariant under the transformation @xmath134 while the relative orientations of the quadrupolar moments for the cases @xmath10 and @xmath21 are mutually orthogonal : with this symmetry in mind , we can easily disentangle the impact of @xmath122 and @xmath123 on the structure formation by comparing the relevant physical properties for the cases @xmath10 and @xmath21 . these properties are displayed for @xmath130 and @xmath131 in figure [ fig : ea_2_1 ] along with a color - coded , horizontal bar ( above this panel ) which indicates these @xmath10-ranges where the respective particle configurations are the energetically most favorable ones . on a qualitative level we observe that the enthalpy curve is continuous over the entire investigated @xmath10-range ; however , it shows kinks at particular values of the aspect ratio which provide a first evidence for the occurrence of discontinuous transitions between the ground state configurations . the specific enthalpy , @xmath135 , shows a pronounced local minimum at @xmath12 ( i.e. , for circular particle shapes ) : here the t - configuration is dominant , verified by the fact that in this @xmath10-region @xmath136 and @xmath137 . in addition , we observe @xmath138 and @xmath139 , indicating thereby a relative orthogonal orientation of neighboring particles . in contrast , for more anisometric particle shapes ( i.e. , small and large @xmath10-values ) , the enthalpy rapidly decays to rather small values ; actually @xmath40 tends to minus infinity for @xmath140 and @xmath141 as the quadrupoles start to overlap and the repulsive soft core shrinks as @xmath142 . for small and large @xmath10-values the formation of parallel rows is energetically most favorable and pd - configurations are observed . here the nematic order parameter @xmath108 is the appropriate quantity to characterize the emerging structure : indeed , @xmath108 assumes in these @xmath10-regions the value 1 , indicating that a single orientation prevails . in contrast , for the two intermediate @xmath10-ranges , where the enthalpy assumes local maxima the ground state configurations for @xmath10 and @xmath21 are distinctively different ; for these @xmath10-values we observe the formation of more complex structures , reflected by a rather intricate variation of the different order parameters with @xmath10 : ( i ) _ increasing _ first @xmath10 beyond the region where the t - configuration is stable , the system changes after a very narrow @xmath10-region where the hb@xmath143-structure occurs via a discontinuous _ structural _ transition ( identified by a jump in @xmath132 ) into the th - configuration ; the latter one is characterized by @xmath138 , @xmath144 and @xmath145 which provides evidence that the difference in orientation angles between nearest neighbors is @xmath128 . upon further increasing @xmath10 we pass again a very narrow interval where a branched structure is stable and we eventually reach again via a discontinuous structural transition the hb@xmath121-structure ; for this configuration none of the order parameters assume any characteristic value . eventually we identify for even larger @xmath10-values the aforementioned pd - structure . ( ii ) _ decreasing _ , on the other hand , the value of @xmath10 beyond the range where the t - structure is stable , we observe hb - configurations : first the one with the larger density ( i.e. , the hb@xmath143-structure ) and then the hb@xmath121-structure . the shape of @xmath113 ( see figure [ fig : ea_2_1 ] ) indicates that the transition between hb@xmath146 and hb@xmath121 ( and also the subsequent transition to the pd - structure ) is of first order , i.e. , at some point the relative orientation of neighboring particles is discontinuous . at this point it should be mentioned that @xmath132 and the order parameters do not necessarily behave in the same manner as the system changes from one structure to the other : some order parameters or @xmath132 may change abruptly , indicating a first order phase transition , while the other parameters change continuously . as an example we refer the reader to the transition hb@xmath121 @xmath147 pd for @xmath148 . another interesting feature emerges as we compare the curves of the order parameters shown in figure [ fig : ea_2_1 ] obtained for the parameters @xmath149 , with the corresponding data calculated for the parameter sets @xmath150 and @xmath151 , shown in figures [ fig : ea_02_01 ] and [ fig : ea_20_10 ] , respectively . the corresponding curves reveal striking similarities , suggesting that appropriate scaling relations of the order parameters via the values of the quadrupole moment and the pressure hold for the respective ground states . in contrast ( and interestingly ) , the enthalpy curves obtained for the different sets of data differ rather substantially in magnitude . we will discuss a possible background scenario of these observations in more detail in appendix [ sec : reduction of parameter space for hard particles ] . there we will show that indeed a scaling relation for the enthalpy of a systems of _ hard _ , quadrupolar particles can be derived ( see appendix [ sec : reduction of parameter space for hard particles : ground state ] ) . however , it seems that the softness of our particles remember that we consider in this contribution particles with a _ soft _ ( albeit rather steep ) core leads to a breakdown of this scaling law . this feature is presumably due to the fact that a simultaneous scaling of the quadrupole moment and of the pressure also induces a change of the density . to take into account this effect properly , we suggest in appendix [ sec : phenomenological enthalpy scaling for non - hard particles at ground state ] an empirical scaling law for the ground state enthalpy of a system of soft ellipsoidal particles and provide numerical evidence for its justification . in the current subsection , we analyze the md results that we have obtained for all considered aspect ratios @xmath10 and a set of reduced temperatures @xmath90 . as previously mentioned , the ground state configurations that we have obtained with the help of the ea algorithm for a variety of aspect ratios , serve now as initial configurations for the md simulations . corresponding snapshots for the most interesting and most representative configurations , obtained after equilibrating the system , are presented in figure [ fig : systems_quadrupolar_mdsnap ] . before starting the investigations , it is worth to briefly consider typical values of @xmath122 , @xmath123 in experimental systems . as an example , we consider the quadrupolar gay - berne model of benzene proposed by golubkov _ et al . _ we use their quadrupole strength , @xmath152 , and their spherical diameter , @xmath153 ( the latter value is based on the gay - berne contact distance ) . using the surface tension of water ( the 2d equivalent of pressure ) defined in ref . , @xmath154 at @xmath155 , we arrive at the ratio @xmath156 however , since the point quadrupole approximation is known to overestimate the interaction strength for narrow interparticle configurations @xcite due to the singularity in @xmath157 [ see eq . ] , we consider here a reduced value of @xmath158 instead of @xmath159 . with this choice we hope to cover not only benzene molecules , but also other quadrupolar molecules . by identifying @xmath131 with @xmath160 , we arrive at the following energy scale @xmath161 . in the following , we present md simulation results for the parameter pairs @xmath162 , and @xmath163 . we start with a detailed investigation of the structure at @xmath130 , @xmath131 and different temperatures . the md simulations reveal that the structures predicted for vanishing temperature remain stable also at low temperature ( @xmath164 , see panels in the second column of figure [ fig : systems_quadrupolar_mdsnap ] ) . as @xmath90 increases monotonously , defects start to form ( see panels in the third column of figure [ fig : systems_quadrupolar_mdsnap ] ) : the most common of these is a slight wave - like modulation of previously straight lines . finally , once the temperature has been raised above a certain threshold value ( see panels in the fourth column of figure [ fig : systems_quadrupolar_mdsnap ] ) , the crystalline order is rapidly lost . the corresponding transition temperatures depend strongly on @xmath10 : configurations with @xmath165 ( t@xmath166-configuration ) and @xmath10 far from unity ( pd - configuration ) turn out to be the most stable ones : in this situation , particles can approach very closely and exert thus a strongly attractive quadrupole - quadrupole interaction ; these ordered structures break up only at temperatures as high as @xmath167 and @xmath168 , respectively . in contrast , the rather complicated b - configurations melt already at temperatures as low as @xmath169 ( with @xmath170 ) . in order to investigate and to locate the transition of the system from the ordered to the disordered regime , we focus in the following on the reduced potential energy , @xmath171 , and the reduced system area , @xmath172 , ( in units of @xmath6 and @xmath5 , respectively ) . figure [ fig : upotentialvolume ] depicts @xmath173 and @xmath172 as functions of the reduced temperature @xmath90 for all considered values of the aspect ratio @xmath10 . upon increasing the temperature , @xmath171 progressively decreases in magnitude and finally approaches zero , reflecting the diminishing role of particle interactions . at the same time , the area @xmath172 increases . interestingly , we observe along this process that both @xmath174 and @xmath175 show discontinuous changes within very small temperature intervals for all @xmath10-values investigated ; these `` jumps '' are found for both quantities at approximately the same temperature . we attribute these observations to the occurrence of a first order phase transition . the temperatures that delimit these intervals are marked in figure [ fig : upotentialvolume ] by contour lines as functions of @xmath10 . only for @xmath176 no such discontinuity in @xmath177 could be resolved ; this fact can be related to the finite size of the temperature grid . intentionally , we have not evaluated the susceptibility , since it is well established that energy fluctuations are not correctly reproduced within the berendsen scheme . @xcite we now arrive at the discussion of the previously defined boops @xmath106 and @xmath107 [ see eq . ] ; they are displayed in figure [ fig : psi4psi6 ] . as already shown in figure [ fig : ea_2_1 ] , we find for all our ground state configurations intervals in @xmath10 where at least one of the two parameters does not vanish . considering now the temperature dependence of these quantities , we observe for all @xmath10-values investigated a discontinuous change of both @xmath106 and @xmath107 ( if not zero in the ground state ) at exactly the same temperatures where a discontinuous change in @xmath172 was observed . to analyze the structure of the system at finite temperatures on an even more quantitative level , we have calculated in addition various coefficients of the pair distribution function in an expansion in terms of rotational invariants . to be more specific , we use a two - dimensional version of the three - dimensional coefficient functions of the pair distribution function,@xcite i.e. @xmath178 where the @xmath179 are rotational invariants . here we focus on the coefficients @xmath180 , @xmath181 , and @xmath182 defined via the following two - dimensional rotational invariants : @xmath183 the function @xmath180 corresponds to the familiar pair correlation function . the other functions , @xmath182 and @xmath181 , describe , in addition , a local orientational order of the particles . similar to 3d systems @xcite we can interpret these coefficients as follows : @xmath181 provides information about the conditional probability density of a particle ( relative to the bulk probability density ) whose orientation axis is aligned in parallel ( positive value ) or orthogonal ( negative value ) to the orientation axis of a considered particle ; @xmath182 describes the conditional probability density of a particle ( again , relative to the bulk probability density ) positioned along or aside the axis of a considered particle . in the panels of figure [ fig : gxxxofrk21 ] , we present numerical results for these three functions for @xmath184 ( see also the corresponding snapshots shown in the panels in the bottom row of figure [ fig : systems_quadrupolar_mdsnap ] ) . for all three correlation functions the first peak is located at around @xmath185 : this position does not exactly mark the face - to - face alignment of the particles ( which occurs at @xmath186 ) and thus provides evidence for a slightly shifted parallel configuration , induced by the quadrupole ( see snapshots shown in the panels of the bottom row in figure [ fig : systems_quadrupolar_mdsnap ] ) . for @xmath168 , we observe a rapid decay of @xmath180 towards unity for large particle distances @xmath187 : the crystalline order has completely vanished , reflected by the missing peaks for larger @xmath187-values . since at low temperatures all particles are oriented in the same direction , @xmath181 is equivalent to @xmath180 ; however , as the temperature attains @xmath168 , we observe that for large @xmath187-values the orientational order is lost and thus @xmath181 vanishes already at @xmath188 . finally , @xmath182 oscillates around zero in the ordered phase for low temperatures ( @xmath189 ) due to the fact that @xmath182 is by definition not able to contribute to the overall particle density . for @xmath190 , @xmath182 completely vanishes for large @xmath187-values ( @xmath191 ) : obviously , a loss of orientational order occurs for @xmath190 . this temperature threshold , which can be interpreted as a melting temperature , agrees fairly well with the temperature where the discontinuous changes in @xmath177 and @xmath175 ( see figure [ fig : upotentialvolume ] ) and in the boops ( see figure [ fig : psi4psi6 ] ) were observed . concluding , we note that related investigations carried out for other @xmath10-values led to analogous conclusions about the corresponding melting temperature . panel ( a ) , @xmath181 panel ( b ) , and @xmath182 panel ( c ) for @xmath184 and various temperatures ( as labeled).,width=321 ] these observations motivate investigations on a _ melting curve _ that separates the ordered from the disordered phase as we increase the temperature . this line is displayed in figure [ fig : meltingdiagramp1qsqrt2 ] for all considered aspect ratios @xmath10 . we emphasize that these data represent only an estimate for the true , two - phase coexistence lines characterizing a first - order transition . for the latter , one would also expect the occurrence of a hysteresis , i.e. , the observation of two different curves , depending on whether the system is heated up or cooled down from a low or a high - temperature state , respectively . we briefly come back to this issue below . that separates the ordered phase ( at low temperatures ) from the disordered phase ( at high temperatures ) ; for details cf . text . data along on the line correspond to results obtained for disordered state points.,width=321 ] another interesting feature of the melting curve that can be observed is that it exhibits some degree of symmetry in shape when exchanging @xmath10 and @xmath21 ( see figure [ fig : meltingdiagramp1qsqrt2 ] ) . of course , we would not expect full symmetry since the electrostatic properties for the cases @xmath10 and @xmath21 are different ( see bottom panels in figure [ fig : model ] ) . from the data we can conclude that for particles with large eccentricities the melting occurs at higher temperatures than for @xmath10-values close to unity . given the large amount of numerical calculations required to construct the melting line in figure [ fig : meltingdiagramp1qsqrt2 ] , which was calculated for one particular set of parameters , @xmath192 , it would be obviously desirable and helpful to have a scaling relation at hand which allows to easily obtain ( or to extrapolate ) corresponding melting lines for other parameter sets . in appendix [ sec : reduction of parameter space in the hard limit at finite temperature ] we show that such a relation does indeed exist for _ hard _ particles . this relation states that the probability to encounter a microscopic configuration of the many - particle system in phase space is invariant under the simultaneous transformations @xmath193 , @xmath194 , @xmath195 , with @xmath196 being a scaling factor . however , in this contribution we consider _ soft _ particles ( even though characterized by a rather harsh repulsion , see eq . ) . nevertheless , as we discuss in appendix [ sec : scrutinizing the melting curve ] , it is also possible for the system at hand to provide via a suitably adapted scaling law a rough estimate of the location of the melting line at scaled parameters . finally and for the sake of completeness , we now discuss our investigations on the melting transition as obtained in a cooling process ( `` simulated annealing '' ) . the central question that we address is whether the ground state structures can be reproduced at least locally with md simulations starting from a disordered phase at higher temperatures . to this end we performed simulations at @xmath130 and several values of @xmath10 , initializing the system with random particle positions and orientations and cooling it down gradually . the initial pressure and temperature were set to @xmath197 and @xmath198 , respectively ; the initial box - shape is quadratic with a side length of @xmath199 . within the first @xmath200 md steps we linearly decreased the pressure and the temperature down to @xmath131 and @xmath201 , respectively . the simulations extended in total over @xmath202 md steps . for the other simulation parameters we refer the reader to sec . [ subsec : md ] . ( panel ( a ) ) for a system with @xmath203 at a temperature @xmath201 , pressure @xmath131 , and quadrupole strength @xmath130 after a melting ( snapshot in panel ( b ) ) and a cooling procedure ( snapshot in panel ( c ) ) . for the color code of the snapshots see inset in figure [ fig : systems_quadrupolar_mdsnap].,width=321 ] our data provide evidence that the predicted ground state could be obtained via this simulated annealing process only for very few state points . in general , the formation of ordered structures via such a process is hampered and delayed by frustration effects , especially for @xmath10-values far from unity , where particles encounter due to their elongated shapes difficulties to rotate . as an example we discuss the pair correlation function @xmath180 , obtained for @xmath203 ( i.e. , a value close to unity ) and depicted in figure [ fig : k08](a ) . we observe a coincidence in the peak positions of @xmath180 , but not in their heights . we interpret this deviation as an artefact of the crystallite structure appearing after cooling ( see figure [ fig : k08](b ) and ( c ) ) . this might be a consequence of the lack of long - range order in 2d systems . in summary , we have proposed a simple model that mimics the essential features of elongated , organic molecules without a net charge or dipole moment : it consists of soft ellipsoids with an embedded linear quadrupole moment . the preferred orthogonal arrangement of linear quadrupole moments in close proximity to each other represents an interesting contrast as compared to dipolar systems , where particles prefer parallel arrangements , thus often forming chains . the self - assembly scenarios of our system result from a competition between the shape anisotropy of the ellipsoids and the quadrupolar interactions . operating in the npt ensemble , we have investigated the system for numerous different sets of system parameters , @xmath122 and @xmath123 , and could identify different strategies of the system , depending on the shape anisotropy @xmath10 . while we always observed configurations of parallel rows of particles for @xmath10-values far from unity , the sequence of structures between these two limiting cases strongly depends on the competition between the short - ranged and the long - ranged , electrostatic contributions to the interaction . for @xmath165 we observe two different configurations with orthogonal particle arrangement , one of them based on a square lattice , the other closely related to a hexagonal lattice . intermediate ranges of @xmath10 both for @xmath204 as well as for @xmath205 are mostly dominated by two variations of the herringbone structure . in addition , we observe for selected @xmath10-values a non - trivial lattice , closely related to the trihexagonal tiling . in very small ranges of @xmath10 , more complicated , branched structures can emerge . however , these turn out to be in general rather unstable at finite temperatures , as shown in complementary md simulations : thus we speculate that they are metastable at vanishing temperature . for future work , it would be useful to extend our simple model in two directions : ( i ) the particles interaction should include a general quadrupole moment and ( ii ) the system could be confined to a slab geometry or be extended to full three dimensions . these extensions would nicely meet the considerable experimental interest in the self - assembly of complex organic molecules on surfaces . @xcite some of these systems show a remarkable variety of different structures , which can even be controlled via external parameters such as magnetic fields @xcite or light with different polarization , @xcite opening thereby the route to many interesting technological applications . @xcite the understanding of our simple model considered in the present contribution could serve as a suitable starting point for related investigations of the more complex molecules considered in these studies . extending the model step - by - step towards more complicated shapes and to more intricate effective interactions would help to understand the numerous competing effects . @xcite finally we note that so - called inverse patchy colloids ( ipcs ) , @xcite i.e. , colloids decorated with charged patches , represent due to their charge distribution a closely related system , as they also carry a linear quadrupole moment . ipcs have been observed to form similar structures as the ones encountered here both in simulations @xcite as well as in experiments this work was supported by the deutsche forschungsgemeinschaft ( dfg ) within the framework of the crc 951 ( project number a7 ) . ma and gk gratefully acknowledge financial support by the austrian science foundation ( fwf ) under project numbers p23910-n16 and f41 ( sfb vicom ) and by e - cam , an e - infrastructure centre of excellence for software , training , and consultancy in simulation and modelling funded by the european union ( proj . no . further , financial support from the bilateral projects phc - amadeus-2012 and 2015 ( under project numbers 26996uc and 33618yh ) , projekt amade ( under project numbers fr 10/2012 and fr 04/2015 ) , the german academic exchange service ( daad - under project number 10041371 ) , the bilateral projects phc - procope-2013 ( under project number 28400pd ) , and funding by investissement davenir labex palm ( grant anr-10-labx-0039 ) is acknowledged . we consider two quadrupolar particles at contact , i.e. , @xmath206 [ see eq . ] and optimize their respective and relative orientations by minimizing their electrostatic energies . in figure [ fig : angles ] we display the angles @xmath207 and @xmath208 , enclosed by the vector connecting the centers of the two particles and the orientations of the respective quadrupolar moments in the corresponding optimal configurations , as functions of @xmath10 ( for the definitions of the angles see also inset in figure [ fig : angles ] ) . obviously , the t - configuration is only stable for values of @xmath10 close to unity . for strong anisometry ( @xmath209 or @xmath210 ) the ground state configuration is parallel displaced ( pd ) . the goal of the present subsection is to derive scaling relations for the system parameters @xmath46 ( quadrupole strength ) and @xmath43 ( pressure ) under which the thermodynamic properties of the system ( and thus its ground state configurations ) remain unchanged . of course , these relations also hold for our reduced parameters , which we introduced at the end of sec . [ sec : model ] . we assume in the following a simplified version of our interaction potential , namely the _ quadrupolar hard - ellipse _ ( qhe ) model , defined by its interaction @xmath211 @xmath212 has been defined in eq . ( [ eqn : sigma ] ) . thus , contrary to our original system ( see sec . [ sec : model ] ) , the qhe model consists of _ impenetrable _ ellipsoidal particles , whereas the quadrupolar , long - range part of the interaction is identical to our original model , eq . . the potential energy , @xmath213 , of a specific configuration of the qhe system is defined via @xmath214 where @xmath215 and @xmath216 specify the microscopic configuration of the particles . at vanishing temperature , the particles are positioned and oriented in a non - degenerate ground state in an npt ensemble such that the enthalpy , @xmath40 , is minimal , that is @xmath217 .\end{aligned}\ ] ] the scaling properties of the qhe interaction potential ( via eq . ) , with the quadrupole strength @xmath46 induces a scaling property of the entire potential energy [ cf . eq . ] , i.e. , @xmath218 ; henceforward , @xmath196 is a simple scaling parameter . since the enthalpy @xmath40 is a linear combination of @xmath219 and @xmath43 , replacing @xmath46 by @xmath220 [ cf . . ] leads to the same ground state only if the pressure also scales with @xmath196 . thus , we can conclude that a particular ground state configuration of the qhe model ( specified by @xmath221 ) , remains invariant under the transformations @xmath222 and @xmath223 . we next consider the qhe system at @xmath2 . the phase space probability for the occurrence of a microscopic configuration ( specified by @xmath224 ) in a phase space volume of infinitely small extent in an npt - ensemble is given by @xmath225 \right\ } \mathrm{d}\mathbf{r}^n \,\mathrm{d}\mathbf{\hat{u}}^n \ , \mathrm{d}s_0 , \label{eqn : p_coords}\end{aligned}\ ] ] where @xmath226 is the corresponding partition function of the ensemble . it is obvious that eq . is invariant under the transformations @xmath227 , @xmath194 , and @xmath195 . since we know from eqs . and that the potential energy of non - overlapping , impenetrable particles is proportional to @xmath46 , i.e. , @xmath228 we obtain the simple rule , that @xmath229 is invariant under the scaling law @xmath193 , @xmath194 , and @xmath195 . thus , if we have a reference system ( specified by @xmath46 , @xmath43 , and @xmath45 ) at hand , we can simply calculate the properties of another system ( index `` new '' ) via the following procedure . given the size of the corresponding molecule , @xmath230 , its aspect ratio @xmath10 and the quadrupole strength , @xmath231 , it is thus possible to estimate the order - disorder transition temperature for the new system : we first calculate the reduced quadrupole strength via @xmath232 from which we obtain the scaling factor @xmath196 from @xmath233 . the order - disorder temperature and pressure of the new system then follow from the corresponding quantities of the reference system , @xmath234 and @xmath43 , via @xmath235 and @xmath236 . we now explore to which extent we can apply the parameter scaling relations derived for the qhe system to the actual system of soft particles investigated in the main part of this paper ( see sec . [ sec : model ] ) . the total potential energy for a specific configuration of soft quadrupolar ellipsoids is defined through @xmath237 as pointed out in the discussion of the ground states in sec . [ subsec : ground_state ] , the order parameters reveal as functions of @xmath10 strong similarities when one compares their values as obtained for the parameter sets @xmath238 and @xmath239 . this suggests that the soft particle system fulfills at @xmath1 a similar scaling relation as the qhe system . however , the corresponding enthalpy curves match only in terms of shape , but not in terms of magnitude . we thus introduce a _ phenomenological _ correction to the simple scaling law of the enthalpy as developed in appendix [ sec : reduction of parameter space for hard particles : ground state ] . specifically , we assume that a scaling of @xmath43 and @xmath46 with a factor @xmath196 results in a small rescaling of the ground state coordinates , i.e. @xmath240 , where @xmath241 attains values close to unity . we further assume that , when passing from one ground state , obtained for parameters @xmath242 and a filling fraction @xmath132 , to a new ground system ( index `` new '' ) , specified by parameters @xmath243 and a filling fraction @xmath244 , the rescaling factor @xmath241 is related to the change in density according to @xmath245 as a consequence , the ground state cell volume is scaled with @xmath246 . regarding the total potential energy , we assume that its scaling with respect to @xmath46 and the rescaling of the positions follows the same law in the qhe system [ see eqs . , and ] , yielding @xmath247 collecting all contributions , we obtain the following approximate expression for the enthalpy : @xmath248 p\ , s_0^{\rm gs } .\end{aligned}\ ] ] we check the applicability of this phenomenological scaling rule by calculating , as an example , enthalpies of a new system from the values of the enthalpies obtained at @xmath249 , @xmath250 and @xmath251 , @xmath197 . the corresponding values for the enthalpy as functions of @xmath10 are presented in figure [ fig : enthalpy18 ] . from these data we can conclude that eq . seems applicable for @xmath10-values close to unity where the corresponding structures possess a 4-fold symmetry ( t - configuration ) . for particles with larger aspect ratios , substantial deviations between the respective enthalpy curves occur . we now consider the question whether this rule can also be extended as an approximation to a system of soft particles . we are particularly interested in the melting curve , an example for which is shown in figure [ fig : meltingdiagramp1qsqrt2 ] pertaining to the parameters @xmath162 . to test the scaling rule , we have carried out the corresponding simulations for the case @xmath253 , i.e. @xmath254 at temperatures @xmath255 . the rescaled melting curve for the new system , specified by the latter parameter set is displayed in figure [ fig : meltingdiagram ] in the rescaled form together with the original melting curve , corresponding to @xmath162 . we observe that the new , rescaled melting curve is similar in shape to the original one , but the two sets of data do not coincide : the rescaled curve lies over the entire @xmath10-range above the reference curve ; we attribute this fact to the stronger molecular overlaps leading to an increased binding strength . specifically , we observe for all aspect ratios cohesion energies that are about three times higher for the parameter set @xmath254 at the ground state ; the factor three differs from the expected value of @xmath253 . in addition to the stronger overlaps , also the observed structural archetypes can differ between the two systems at the ground state . to be more specific we obtained for @xmath256 , a th - configuration for @xmath254 , whereas for @xmath162 a t - configuration is observed . we conclude that the scaling law put forward in sec . [ sec : reduction of parameter space in the hard limit at finite temperature ] only provides a rough estimate of the results obtained in actual calculations . that separates the ordered phase ( at low temperatures ) from the disordered phase ( at high temperatures ) . in addition , a rescaled melting curve ( using @xmath257 ) for @xmath258 is shown , which has been obtained by applying the scaling laws specified in the text.,width=321 ] 73ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) link:\doibase 10.1039/c004944c [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) \doibase http://dx.doi.org/10.1063/1.4907037 [ * * , ( ) ] @noop * * , ( ) http://www.ncbi.nlm.nih.gov/pubmed/10054722 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1002/jcc.20035 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1080/026782998207406 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1080/00268979200100331 [ * * , ( ) ] link:\doibase 10.1080/10587259908025635 [ * * , ( ) ] link:\doibase 10.1103/physreve.63.011706 [ * * , ( ) ] link:\doibase 10.1080/00268970802270067 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) http://www.ncbi.nlm.nih.gov/pubmed/9901252 [ * * , ( ) ] http://www.ncbi.nlm.nih.gov/pubmed/9904261 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1021/la901719 t [ * * , ( ) ] _ _ , http://katalog.ub.tuwien.ac.at/ac12656721[ph.d . thesis ] , ( ) , \doibase http://dx.doi.org/10.1063/1.441483 [ * * , ( ) ] @noop _ _ ( , ) \doibase http://dx.doi.org/10.1063/1.1901585 [ * * , ( ) ] link:\doibase 10.1137/0916069 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/revmodphys.60.161 [ * * , ( ) ] \doibase http://dx.doi.org/10.1063/1.4774084 [ * * , ( ) ] @noop * * , ( ) @noop _ _ ( , ) \doibase http://dx.doi.org/10.1063/1.555688 [ * * , ( ) ] link:\doibase 10.1021/j150665a034 [ * * , ( ) ] \doibase http://dx.doi.org/10.1063/1.4935063 [ * * , ( ) ] @noop _ _ http://www.amazon.com/exec/obidos/redirect?tag=citeulike07-20&path=asin/0198556454[__ ] , oxford science publications ( , ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1039/c4cc04189e [ * * , ( ) ] @noop * * , ( ) \doibase http://dx.doi.org/10.1016/j.cplett.2005.06.017 [ * * , ( ) ] link:\doibase 10.1021/la102788j [ * * , ( ) ] link:\doibase 10.1002/poc.1405 [ * * , ( ) ] link:\doibase 10.1021/am3010169 [ * * , ( ) ] link:\doibase 10.1021/jp301536z [ * * , ( ) ] http://stacks.iop.org/0953-8984/25/i=14/a=143202 [ * * , ( ) ] link:\doibase 10.1103/physrevb.61.16538 [ * * , ( ) ] http://stacks.iop.org/0268-1242/28/i=5/a=053001 [ * * , ( ) ] \doibase http://dx.doi.org/10.1063/1.4902824 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1021/nn401487 m [ * * , ( ) ] @noop * * , ( ) http://stacks.iop.org/0953-8984/27/i=23/a=234105 [ * * , ( ) ]

we extend the formulation of relativistic coulomb sum rules to account for the average effects of nuclear binding on the initial and final states of ejected nucleons . relativistic interactions are included by using a dirac representation adapted from a vector - scalar field theory . the scalar field reduces the effective nucleon mass and increases the relativistic effects of recoil and fermi motion . we consider two models for the off - shell behavior of the nuclear electromagnetic current , and demonstrate that the sum rule is accurate for applications to data over the interesting range of and three momentum . we further indicate that the form of the sum rule is sufficiently general to accommodate a broad class of off - shell form factor models.pacs numbers : 25.30.fj , 11.55.hx , 24.10.jv @=11 @=12 0.5 true in 0.5 true in submitted to _ physical review c _

nucl-th9412030

the subject of this paper is the coulomb sum rule ( csr ) for inelastic electron scattering @xcite , in a reformulation which extends its validity to relativistic momentum transfers @xmath2 , where @xmath3 is the three - momentum transfer to the target nucleus , and @xmath4 is the nucleon mass . the purpose of this work is to provide practical methods for the analysis of the longitudinal @xmath5 response of nuclear targets , at the higher energies now available at the continuous electron beam accelerator facility ( cebaf ) and other electron accelerators . in a previous paper @xcite , we derived a relativistic coulomb sum rule ( rcsr ) which incorporates the relativistic effects of nucleon recoil at large @xmath1 , as well as of fermi motion . in the present work we extend this theoretical approach to include the effects of nuclear interactions on the rcsr . in our recent article , we discussed in detail the assumptions under which a rcsr could be derived . the basic approximation is that only nucleon , i.e. , as opposed to antinucleon , degrees of freedom enter the coulomb response in the spacelike regime accessible by @xmath5 experiments . we call this the `` nucleons - only '' approximation ; it ignores effects of antinucleons , but includes fully the relativity of the nucleons . following conventional treatments , we adopt an impulse approximation which ignores explicit contributions from the exchange of charged mesons . we include nucleon anomalous moments and elastic form factors , which are important at higher energies . the final assumption is less conventional , but absolutely necessary . in order to derive a _ non - energy - weighted _ sum rule , it must be possible to factor _ all _ dependence on the photon energy @xmath6 from the current matrix element , otherwise dynamical effects enter into the coulomb response function and complicate the isolation of correlation effects . in this paper , we restrict our attention to form factor models which have factorable dependence on @xmath6 . we define the nuclear coulomb sum in terms of the coulomb response function @xmath7 and the proton electric form factor @xmath8 : @xmath9 where the lower limit @xmath10 excludes the quasielastic peak , and the upper limit @xmath1 restricts the integration to spacelike four - momenta . under the assumptions of ref . @xcite mentioned above , we obtain a rcsr which can be expressed in terms of one- and two - body contributions : @xmath11 where the one - body contribution is of the form @xmath12 @xmath13 includes two - body correlation information in momentum space , and @xmath14 is the uncorrelated two - body part which is related to the square of the nuclear elastic form factor , as discussed in ref . @xcite . in the one - body term ( [ ac ] ) , @xmath15 is the nucleon momentum distribution function for isospin projection @xmath16 and one spin projection , and @xmath17 is a kinematic factor which arises due to relativistic nucleon recoil and fermi motion in the target . in the nonrelativistic limit ( @xmath18 ) , we have @xmath19 and @xmath20 , which leads to @xmath21 ; then the sum rule ( [ ab ] ) is the result of ref . @xcite . the relativistic effects are all in the functions @xmath17 , representing recoil of the struck nucleon and fermi motion in the target ground state . in ref . @xcite we further showed that ( [ ac ] ) does not depend strongly on the details of @xmath15 , but only on the lowest momentum moments , e.g. , @xmath22 . this leads to a method of evaluating ( [ ac ] ) accurately in a weakly model - dependent manner , and in principle permits the extraction of the correlation function @xmath13 from the the experimentally measured coulomb sum ( [ aa ] ) , using ( [ ab ] ) . in the present paper , we investigate what changes are required in the sum rule when the coulomb sum @xmath23 is modified by relativistic two - body interactions . we study the effect of the average interaction in the nucleus using the mean field approach of quantum hadrodynamics , a relativistic field theory for nuclear physics @xcite . in particular , we consider qhd - i , which includes vector and scalar isoscalar mesons only . we then reformulate the coulomb sum rule of ref . @xcite so that results ( [ aa])([ac ] ) have a similar structure , but with modifications reflecting the mean - field effects of these relativistic interactions . these modifications have two main effects on the rcsr : first , there are kinematical effects resulting from mean - field interactions of the nucleons , which are represented in qhd - i by a reduction in the effective nucleon mass @xmath0 in the medium . consequences of a reduced effective nucleon mass @xmath0 for the coulomb sum have been considered previously in a fermi gas model @xcite . these calculations include interactions of both initial and final plane - wave nucleon states with the mean fields , through the effective mass @xmath0 , which can be interpreted as binding in the initial state , and final state interactions of the ejected nucleon with the nucleus . chinn , picklesimer and van orden @xcite have studied the effects of final state interactions on the coulomb response of a fermi gas , using more realistic interactions , and have seen similar effects to those seen due to @xmath24 . second , the electromagnetic coupling of the nucleon in medium may be modified by the mean fields , entering through the off - shell behavior of the nucleon elastic form factors . both modifications introduce a degree of model - dependence in the rcsr which is not present in the nonrelativistic formulation , nor in the relativistic formulation of ref . @xcite . we show how these features can be incorporated into the theory to allow the evaluation of the one - body contribution @xmath25 , and the subsequent extraction of the two - body correlation function @xmath13 from the measured coulomb response . this paper is organized as follows : in section 2 we introduce the basic formalism to include relativistic mean - field effects in the coulomb sum rule ( rcsr ) . we introduce two models ( f and g ) for the electromagnetic charge operator , and investigate the resulting behavior connected to different off - shell assumptions for the nucleon elastic form factors . in section 3 , we derive a modified version of the rcsr ( [ aa])([ac ] ) , concentrating on the explicit changes to the one - body term @xmath25 in off - shell models f and g. in section 4 , we illustrate the operation of the sum rule in a simple nuclear system : uniform nuclear matter treated in the mean - field approximation , with nuclear binding effects incorporated using qhd - i . we examine the sensitivity of the rcsr to @xmath0 and to the choice of off - shell models ( f and g ) , focussing on the convergence of the moment expansion in each case . we further demonstrate that the particular form of the rcsr given here is applicable to _ both _ models , and argue that the same form should also be valid for a broad class of form factor models . in section 5 , we draw conclusions , give guidelines for the application of the rcsr to data , and indicate important directions for future work . in this section , we review the formalism for electron scattering from nuclei , as it pertains to our development of relativistic coulomb sum rules . we first give some standard results of the plane - wave impulse approximation ( pwia ) , which is used for the analysis of @xmath5 experiments on nuclei , in a single - particle basis which accounts for the vector and scalar interactions of qhd - i . we then give two models for the nucleon form factors off shell , which will be used in the next section to illustrate the sensitivity of the sum rule to different off - shell assumptions . we begin with the differential cross section for the scattering of ultrarelativistic electrons from nuclear targets , which is commonly written in the form @xmath26,\label{ba}\ ] ] where @xmath27 is the four - momentum transferred from the electron to the nucleus via virtual photon exchange , and @xmath28 . the longitudinal contribution in ( [ ba ] ) has been expressed in terms of the coulomb response function @xmath29 where @xmath30 and @xmath31 denote initial . ] and final nuclear many - body states , respectively . in general , the electromagnetic current density operator @xmath32 may include contributions from both nucleons and charged mesons . meson exchange current ( mec ) contributions have been considered by schiavilla et . @xcite , for example , but are not included here . including only nucleons with electromagnetic form factors , the current density operator can be written in the form @xmath33 where @xmath34 is the ( schrdinger picture ) field operator for a point dirac nucleon , and @xmath35 represents the electromagnetic coupling at the @xmath36 vertex . there is also a sum over proton and neutron isospin projections , which will be suppressed until needed . any effects of charged mesons not included in the current operator ( [ bc ] ) may be interpreted as two - body effects , as discussed in ref . @xcite . to proceed to a relativistic sum rule , we shall expand the field operators @xmath34 of ( [ bc ] ) in a plane wave basis , i.e. , in a pwia . in ref . @xcite , we took this to be the free basis , i.e. , the momentum eigenstates of the free dirac equation . in the present paper , we shall instead expand the field operators in a basis of plane waves moving in the presence of uniform ( isoscalar ) lorentz scalar and vector potentials , such as would be generated by scalar and vector mesons in a relativistic field theory of nuclei , e.g. , qhd - i . with this modification , the theoretical development is formally similar to that of ref . the consequences of the change of basis show up after the nucleons - only approximation , in which antinucleon degrees of freedom are removed approximately from the spacelike coulomb response . in the following , nucleons - only will refer to nucleons in the presence of scalar and vector potentials in the target . the effect of this modification is to include efficiently the effects of these mean - field potentials in both the initial and final target states . we therefore use the plane - wave solutions of the dirac equation @xmath37 \psi(x)=0,\label{bd}\ ] ] where the scalar and vector potentials are written in the forms @xmath38 and @xmath39 , respectively , corresponding to the scalar and vector fields @xmath40 and @xmath41 , and the associated coupling constants . here , as in the mean - field solution for uniform nuclear matter at rest , @xmath42 , and @xmath43 and @xmath40 are constants . the solutions to ( [ bd ] ) are discussed in detail in ref . @xcite . in the case of uniform fields , the energy eigenvalues of ( [ bd ] ) take the simple form @xmath44 , where @xmath45 and the effective nucleon mass is defined @xmath46 . with the nucleons - only approximation , we need only consider positive - energy solutions to ( [ bd ] ) . these are plane wave solutions of momentum @xmath47 , which obey the equation @xmath48u_s({\bf p})=0.\label{be}\ ] ] explicit forms for the interacting solutions @xmath49 can be obtained directly from the free solutions , given in appendix a of ref . @xcite , by making the replacement @xmath50 . since the vector potential @xmath51 appears additively in the nucleon eigenenergy , it does not appear in ( [ be ] ) or its plane - wave spinor solutions . in ref . @xcite we argued that the coulomb response function ( [ bb ] ) for spacelike ( @xmath52 ) photon exchange is dominated by nucleon ( @xmath53 ) contributions to the current matrix elements , and that antinucleon ( @xmath54 ) and pair ( @xmath55 ) terms could be neglected . this is an exact result for a uniform free fermi gas , and leads to the nucleons - only approximation for interacting nuclear systems . the presence of a strong scalar field in the nucleus induces mixing of free @xmath56 and @xmath57 states , particularly for nucleons of high momentum , as in the final states of ( [ bb ] ) . however , transforming to the plane - wave basis formed from the solutions of ( [ be ] ) , hereafter referred to as the `` @xmath0-basis '' , removes this mixing by the potentials . the nucleons - only approximation is again adopted for interacting nuclei , but here refers to nucleons of mass @xmath0 . this use of the @xmath0-basis for final states implies substantial interaction of excited ( ejected ) nucleons before leaving the target , and is probably a better assumption for large nuclei than for small . then the electromagnetic current operator takes the form @xmath58 where @xmath59 and is a creation operator for a nucleon with spin projection @xmath60 and momentum @xmath47 . the energy denominators in ( [ bf ] ) reflect the normalization of the plane - wave spinors to @xmath61 particles / volume . the formal derivation of the rcsr now follows closely that of ref . @xcite , once we have discussed the form of the electromagnetic vertex operator @xmath62 . it is conventional to express the @xmath36 vertex operator in the form @xmath63 where @xmath64 is the nucleon anomalous magnetic moment , @xmath4 is the free nucleon mass , and @xmath65 and @xmath66 are the dirac and anomalous form factors , respectively . in general , @xmath65 and @xmath66 are scalar functions of @xmath67 , @xmath68 and @xmath1 . this form is sufficiently general for matrix elements between nucleon states ( antinucleons excluded ) as in the nucleons - only pwia . since the tensor @xmath69 is antisymmetric , only the three - momentum * q * enters explicitly the coulomb operator @xmath70 . any remaining dependence the photon energy @xmath6 enters @xmath70 _ only _ through the form factors @xmath65 and @xmath66 . for scattering from a free nucleon , the form factors @xmath65 and @xmath66 depend only on the scalar @xmath71 . to ensure the correct charge and magnetic moments for free nucleons , they are normalized at @xmath72 according to @xmath73 and @xmath74 . the form factors @xmath65 and @xmath66 are obtained from @xmath5 scattering data , usually in terms of the more convenient sachs electric and magnetic form factors @xmath75 and @xmath76 : @xmath77 where @xmath78 , and @xmath79 , i.e. , the photon four - momentum is determined by the four - momentum transfer at the @xmath80 vertex . for a single nucleon in free space , both energy and momentum are conserved at the @xmath36 vertex as well , and we have @xmath81 . see appendix a for an alternative form of ( 2.7 ) which is expressed in terms of the sachs form factors . we use the standard parameterization of the sachs form factors , along with the assumption @xmath82 , as discussed in ref . this choice has the convenient feature that all sachs form factors are proportional , which satisfies a condition assumed in our derivation of a _ non - energy - weighted _ sum rule in ref . @xcite . for interacting nucleons in a nucleus , one needs information about the form factors off their free mass shell , which is simply not known . this leaves considerable freedom to extrapolate off shell from the known on - shell forms . given no other information , a common approach has been to assume that the dependence on @xmath71 of the form factors @xmath83 and @xmath84 does not change off shell from that for a free nucleon . this implies , at least in part , a separation of the dynamics of the electromagnetic structure of the nucleon from that of the nucleus . however , even this assumption does not uniquely specify the off - shell form factors , because of the freedom to transform @xmath62 using the gordon identity to other operators with other form factors , each of which are equivalent on shell , but not off shell . this issue has been considered by de forest @xcite , and by chinn and picklesimer @xcite , who have given examples of possible choices of extrapolation , and studied the sensitivity of the response functions and of the coulomb sum to those choices . our method is similar in principle , but is based specifically on the gordon transformation which relates @xmath65 and @xmath66 to the free sachs form factors @xmath75 and @xmath76 , given in ( [ bh ] ) for free nucleons . we consider two possibilities : that the functional dependence on @xmath71 of the functions @xmath83 , @xmath84 is unchanged off shell ( model f ) , or alternatively , that the functional dependence on @xmath71 of the functions @xmath85 , @xmath86 is unchanged off shell ( model g ) . other choices are possible . for example , in principle , the vector field @xmath43 may also enter the current operator . we restrict our attention to models f and g , which illustrate the important issues and do not depend on @xmath43 . the first choice , model f , is the most common ; here it is simply assumed that the dirac and anomalous form factors @xmath83 and @xmath84 are unchanged in the nuclear environment , i.e. , that they are given by ( [ bh ] ) where both the sachs form factors and the kinematic variable @xmath78 are evaluated at the actual momentum transfer of the experiment , i.e. , at the momentum @xmath87 found at the electron vertex . this is in some sense a minimal assumption , in that the effects of the scalar field enter explicitly the current operator @xmath88 _ only _ through the nucleon field operators . in spite of its apparent simplicity , this model has the peculiar feature that the dirac and anomalous magnetic moments are treated differently in the nuclear medium . this follows since the dirac moment ( @xmath89 ) scales with the nucleon mass @xmath0 , as can be seen from the solution of ( [ bd ] ) for a nucleon at rest in the presence of a uniform magnetic field , while the anomalous moment ( @xmath90 ) is unchanged in the medium . this behavior is possible for a theory with point dirac nucleons dressed by charged meson fields , such as qhd - ii @xcite , since the anomalous moments in such a theory are typically included explicitly by hand , following a solution of the many - body problem using point nucleons . however , this behavior is not likely for a theory with internal nucleon electromagnetic structure , such as qcd , since there one expects that the dirac and anomalous magnetic moments may have similar origins , and therefore may respond in a similar manner to the nuclear environment . an alternative assumption , which we call model g , is that the sachs electric and magnetic form factors @xmath85 and @xmath86 are unchanged in the nuclear medium . to express @xmath62 in the form ( [ bg ] ) , the transformation ( [ bh ] ) is now performed in the @xmath0-basis : the off - shell forms of @xmath65 and @xmath66 are now given in terms of the free sachs form factors by @xmath91 where the new variable @xmath92/4{m^*}^2 $ ] . in contrast to model f described above , the off - shell choice ( [ bi ] ) leads to a total nucleon magnetic moment equal to @xmath93 . in particular , the effective dirac and anomalous magnetic moments behave similarly in the nuclear medium . ( note that the factor @xmath94 , appearing on the right - hand side of @xmath66 in ( [ bi ] ) , has the effect of replacing @xmath95 by @xmath96 in ( [ bg ] ) . ) also , since @xmath97 depends _ only _ on the momenta @xmath47 and @xmath98 , the photon energy @xmath6 now enters the form factors @xmath65 and @xmath66 _ only _ through the sachs form factors @xmath85 and @xmath86 . this can be seen by expressing @xmath70 in the alternate form given in ( a5 ) of appendix a. in the next section , we will see that model g leads to relativistic coulomb sum rule which is the direct analog of that derived in the free pwia , but with @xmath50 . the derivation of a rcsr based on the interacting pwia of section 2 is formally similar to that based on the free pwia of ref . we begin with the coulomb response function ( [ bb ] ) , and formally evaluate the spacelike coulomb sum ( [ aa ] ) by integration over the photon energy @xmath6 . as described in ref . @xcite , to arrive at a _ non - energy - weighted _ sum rule it must be possible to factor _ all _ dependence on @xmath6 ( which here enters through @xmath71 ) from the current matrix element in ( [ bb ] ) . this can be accomplished by requiring that the ratio @xmath99 be independent of @xmath6 , since the plane - wave spinors appearing in ( [ bc ] ) are functions only of the 3-momenta * p * and * q*. for a system of dirac protons , for which @xmath100 , this is satisfied trivially . in model g , it is satisfied by explicit construction , using the assumption of proportional sachs form factors . for more complicated off - shell models , such as model f , which do involve explicit dependence on @xmath6 , the derivation of a rcsr may require further assumptions . in this article , we will assume that the above condition is satisfied , and outline the remaining steps which lead to a rcsr . we next use closure to perform the sum over final states in the squared matrix element . in the nonrelativistic coulomb sum rule , in which the integration is over all @xmath6 , the use of closure here is exact . in the relativistic case , the use of closure over the spacelike states alone requires certain assumptions about the spacelike nuclear excitation spectrum , as discussed in ref . these arguments were based primarily on the example of a fermi gas . we make similar arguments here , and assume that the spacelike spectrum is saturated by the nucleons - only response , where now `` nucleons - only '' refers to positive - energy baryons of mass @xmath0 . after performing the sum over final states @xmath31 , the momentum - space anticommutation relations are used to separate one- and two - body terms , as in ( [ ab ] ) . the relativistic recoil function which appears in ( [ ac ] ) is of the form @xmath101 where the matrix element is defined by @xmath102 and @xmath103 represents a nucleon with momentum * p * , spin @xmath60 and isospin projection @xmath16 . the above expressions are identical in form to those obtained in the free pwia of ref . @xcite , but with the replacement @xmath50 . although the one - body term , evaluated using ( [ rjpq ] ) and ( [ jpq ] ) would account _ exactly _ for the relativistic effects of nucleon recoil and fermi motion , it requires knowledge of the nucleon momentum distribution @xmath15 in the target ground state , which is not normally known precisely . we therefore expand the recoil factor @xmath17 in powers of the nucleon momentum * p * , as in ref . @xcite . assuming spherical symmetry , the one - body term takes the form . ] @xmath104 where the momentum moments are defined @xmath105 and @xmath106 for @xmath107 , respectively . in ref . @xcite , we explained that only the first few moments are required to obtain an accurate result for the one - body term . in the next section , we will study how the convergence properties of this moment expansion depend on the parameter @xmath0 , and on the particular choice of off - shell model . we now give the precise forms of the recoil factor @xmath108 in models f and g. we first consider model g , in which we assume that the medium dependence of the nucleon electromagnetic coupling is described by the off - shell form factors ( [ bi ] ) . by construction the form factors @xmath65 and @xmath66 depend on @xmath71 only through their linear dependence on the free sachs form factors , which are assumed to be proportional . this ensures that the ratio @xmath109 is independent of the photon energy @xmath6 . with this condition satisfied , and since the baryon spinors are also independent of @xmath6 , the matrix element @xmath110 is a function only of three - momenta and the separation of one- and two - body terms in @xmath23 proceeds as described above , leading to ( [ ac ] ) . using the form factors ( [ bi ] ) in ( [ bg ] ) to evaluate the matrix element @xmath110 we obtain @xmath111 where we have defined the electric ( e ) and magnetic ( m ) recoil functions : @xmath112 and @xmath113 the functional forms of @xmath114 and @xmath115 can alternatively be obtained by writing @xmath35 of ( 2.7 ) in terms of @xmath75 and @xmath76 , as explained in appendix a. in ( [ rg ] ) we have also defined the nucleon electric charge @xmath116 ( in units of the proton charge ) and the total magnetic moment @xmath117 ( in nuclear magnetons ) : @xmath118 the functional form of the recoil function ( [ rg ] ) is identical to that obtained in the free pwia of ref . @xcite , but with @xmath50 . this was ensured by choosing to include the factor @xmath94 in ( [ bi ] ) . as described above , we expand @xmath17 in powers of the nucleon momentum * p * , which with ( [ ac ] ) , leads to an expression for the one - body term @xmath25 , as in ( [ momexp ] ) . the expansion coefficients are given in appendix b , and are identical to those obtained in the free pwia of ref . @xcite , but with @xmath50 . in this model , therefore , we can include the relativistic effects of a strong scalar field , in a straightforward extension of the rcsr derived in the free pwia . we now return to model f , in which we assume that the nucleon form factors are given by ( [ bh ] ) . now the photon energy @xmath6 enters not only through @xmath71 , which appears in the free sachs form factors , but also through @xmath78 , which enters in the transformation ( [ bh ] ) . this additional dependence on @xmath6 can _ not _ be removed by simply dividing by an overall factor @xmath119 , as in model g , without some further approximation . in order to proceed , we will first calculate the matrix element appearing in ( [ rjpq ] ) , using the vertex operator ( [ bg ] ) and the form factors ( [ bh ] ) . since division by @xmath119 is not sufficient to render the matrix element a function only of the momenta * p * and * q * , this function will not yet lead to the one - body term of the coulomb sum rule . however , it does serve as a useful starting point to illustrate the important issues . we obtain @xmath120 ^ 2\ r_e({\bf p},{\bf q } ) + \biggl[{{g_{m\sigma}^*}(q^2,\tilde\tau^ * ) \over g_{ep}(q^2)}\biggr]^2\ r_m({\bf p},{\bf q } ) , \label{rf}\ ] ] where we have emphasized explicitly the dependence on @xmath71 . the form of ( [ rf ] ) is similar to ( 3.5 ) for model g , but with `` effective '' sachs electric and magnetic form factors , defined by : @xmath121 which reduce to the free sachs form factors if @xmath122 . inserting @xmath83 , @xmath84 of ( [ bh ] ) into ( [ gsdef ] ) , we have @xmath123+\mu_\sigma\ \biggl[{\tau-{m^*\over m}\tilde \tau^*\over1+\tau}\biggr]\nonumber\\&&\label{gsofg}\\ { g_{m\sigma}^*(q^2,\tilde\tau^*)\over g_{ep}(q^2)}&= & \epsilon_\sigma\ \biggl[{1-{m^*\over m}\over1+\tau}\biggr ] + \mu_\sigma\ \biggl[{\tau+{m^*\over m}\over1+\tau}\biggr ] , \nonumber\end{aligned}\ ] ] which illustrates of `` mixing '' of ( free ) electric and magnetic contributions in this model . this mixing enters as a result of the reduced effective nucleon mass , and is distinct from the usual lorentz mixing which occurs for a moving particle . effective form factors of the sort ( 3.11 ) can also be seen in ref . @xcite , for example . to continue the derivation of a _ non - energy - weighted _ sum rule , we must make some assumption about the remaining @xmath6-dependence , which arises through the variable @xmath124 in ( 3.11 ) . a simple method is based on the excitation energy of a uniform fermi gas in qhd - i , which is expressible simply in terms of the three - momenta * p * and * q * : @xmath125 . we therefore make the replacement in ( 3.11 ) : @xmath126 to obtain energy - independent effective form factors . the factor @xmath127 arises directly from the definition of @xmath124 following ( 2.8 ) . with the replacement ( [ tau ] ) in ( 3.11 ) , the recoil function ( 3.9 ) is now independent of the photon energy @xmath6 , and the derivation of a non - energy - weighted coulomb sum rule for model f proceeds as for model g. at this point , it is possible to expand ( 3.9 ) about @xmath128 and obtain a moment expansion for @xmath25 , as in ( [ momexp ] ) . however , the resulting coefficients @xmath108 are much more complicated than those which arise from ( 3.5 ) in model g , due to the dependence on * p * of the effective form factors ( 3.11 ) . in addition , we will see that the essential physics of fermi motion enters ( [ rf ] ) only through the electric and magnetic recoil functions @xmath129 and @xmath130 , and can be ignored in the effective form factors themselves . it is therefore useful to consider an approximation scheme which will allow a more efficient evaluation of the one - body term : setting @xmath128 in ( 3.11 ) after applying ( 3.12 ) , we can write @xmath131 where we have defined the `` effective '' nucleon charges and magnetic moments : @xmath132 + \mu_\sigma\ \biggl[{(m^*-m)(e_{\bf q}^*-m^ * ) \over 2m^2+m^*(e_{\bf q}^*-m^*)}\biggr ] \nonumber\\ & & \label{emf}\\ \mu_\sigma^*({\bf q})&\equiv & { g_{m\sigma}^*(q^2,\tilde\tau^*)\over g_{ep}(q^2 ) } \biggr\vert_{{\bf p}=0}\nonumber\\ & & \nonumber\\ & = & \epsilon_\sigma\ \biggl[{2m(m - m^ * ) \over 2m^2+m^*(e_{\bf q}^*-m^*)}\biggr ] + \mu_\sigma\ \biggl[{2mm^*+m^*(e_{\bf q}^*-m^ * ) \over 2m^2+m^*(e_{\bf q}^*-m^*)}\biggr ] , \nonumber\end{aligned}\ ] ] following ( [ emg ] ) . we will refer to expression ( [ rfnew ] ) as the `` factored moment expansion . '' the effective form factors now enter ( [ rfnew ] ) as functions only of the momentum transfer * q * , but do not complicate its expansion in * p*. that this simplification is a valid approximation will be shown numerically in the next section . we now apply the sum rules derived in the previous section to a model nuclear system : uniform nuclear matter in the relativistic mean field approximation , as given by qhd - i . @xcite for this system the effects of nuclear structure on the coulomb sum @xmath23 enter through the effective nucleon mass @xmath0 and through the nuclear electromagnetic current , as we have discussed in section 3 . we are interested in the dependence of the coulomb sum on @xmath0 and on the choice of off - shell model ( f or g ) for the current . we then demonstrate how to apply the sum rule methods of section 3 to the model system , as if it were measured coulomb response data . the analysis will necessarily be model dependent , through the choice of @xmath0 and the off - shell model ( f or g ) , as well as through the moments @xmath133 of the nucleon momenta in the target . we shall examine the accuracy of the moment expansion for the `` best - case '' analysis , that is , for which the sum rule parameters match those of the assumed nuclear model . this will illustrate the convergence properties of the moment expansion in each model , and the validity of the factored moment expansion in model f. we begin by computing the coulomb sum @xmath23 for a uniform fermi gas of nucleons moving in uniform vector and scalar potentials . this has been studied previously @xcite . the result can be written in the form @xmath134 \ \sum_{ss^\prime}\ \bigl\vert j_{s^\prime s,\sigma}({\bf p},{\bf q } ) \bigr\vert^2,\label{rcsfg}\ ] ] where the matrix element @xmath110 is defined in ( [ jpq ] ) . we will evaluate ( [ rcsfg ] ) by numerical integration over @xmath67 and @xmath135 , the angle between * p * and * q*. the result is independent of @xmath43 , for reasons given following ( [ be ] ) . the factor @xmath136 restricts the sum over initial states to include only those states which are occupied in the nuclear ground state , and the factor @xmath137 $ ] ensures that final states for which @xmath138 are excluded . this pauli exclusion is the only source of two - particle correlations in this simple model . we shall consider two particular values of @xmath0 which arise in qhd - i , as well as @xmath122 . in the mean - field theory ( mft ) , static polarization of the baryon vacuum in the nuclear ground state ( due to the scalar field @xmath40 ) is ignored in the coupled field equations . in the relativistic hartree approximation ( rha ) , static vacuum polarization is included at the one - loop level . the particular values of @xmath0 are obtained from the self - consistent solutions for uniform nuclear matter at saturation . taking @xmath139 and using the input parameters @xmath140 and @xmath141 , leads to the values @xmath142 in mft , and @xmath143 in rha . although it may be argued that the model assumptions of mft are technically more consistent with the nucleons - only approximation , it may also be true that the resulting effective mass @xmath0 is too small due to the simplicity of the model , as compared to optical potential phenomenology ( see , e.g. , ref . @xcite ) . in the calculations to follow , we will show results using both the mft and rha values of @xmath0 to illustrate the kinds of effects which can be expected in this range . for applications to data , it is convenient to cast the rcsr in a slightly different form , so that the saturation of the sum rule is more apparent . as in ref . @xcite , we define the `` modified '' coulomb sum , which is not present in @xmath144 as defined in ref . @xcite . ] @xmath145 where the recoil `` correction '' factors are defined @xmath146\label{ridef}\ ] ] @xmath147 , \label{riidef}\ ] ] and so on for higher orders . includes terms through @xmath148 . ] the expansion coefficients @xmath149 are given in appendix b. the moments @xmath150 are defined in ( 3.4 ) . for applications of the rcsr to @xmath23 of ( [ rcsfg ] ) , we will assume the sharp distribution @xmath151 for the evaluation of the momentum moments . integrating this distribution over the fermi sphere of radius @xmath152 leads to the @xmath153 moment , given by @xmath154 this is consistent with the pwia and the evaluation of the rcsr on ( [ rcsfg ] ) . for applications to actual data , however , it would be preferable to use the most accurate values available for these moments , e.g. , using experimentally determined @xmath15 from @xmath155 . following ( [ ab ] ) , the above definitions lead to an _ approximate _ sum rule of the form @xmath156,\label{sapp}\ ] ] as discussed in section 5 of ref . the fermi gas model includes only pauli correlations , which vanish identically for @xmath157 . consequently , an application of the rcsr to ( [ rcsfg ] ) will be considered `` accurate '' if the modified coulomb sum @xmath158 tends to unity for momenta beyond the range of these correlations , i.e. , if @xmath159 for @xmath157 . this is frequently referred to in the literature as `` saturation '' of the sum rule . in the following examples , we take @xmath139 . to demonstrate the importance of the effects of a reduced effective mass @xmath0 , we first consider a system of dirac protons , for which @xmath160 and @xmath161 . in ref . @xcite , we showed for @xmath162 that the lowest - order sum rule , @xmath163 , is accurate to within @xmath164 for this case ( see figure 2 in ref . @xcite ) . in figure 1 , the dotted and dot - dashed curves show the coulomb sum @xmath23 calculated with @xmath162 and @xmath142 , respectively . in this example , we see a 1020% reduction in the coulomb sum over this momentum range , due to the reduced effective mass @xmath0 . this effect has been noted previously for lower momenta @xcite . the dashed and solid curves show the lowest order sum rule , @xmath163 , applied to @xmath165 in two different ways : in the dashed curve , the recoil factor @xmath166 has been evaluated at the free nucleon mass @xmath4 , while the solid curve shows a consistent application of the sum rule , in which @xmath166 has been evaluated at the reduced mass @xmath0 . the dashed curve demonstrates that if the effective nucleon mass is substantially reduced in the medium , then one must account for this in the application of the rcsr . the solid curve demonstrates that even when the effective mass is greatly reduced , e.g. , @xmath142 , the lowest - order sum rule @xmath163 applied to a system of dirac nucleons is accurate to within @xmath164 , _ if _ the appropriate value of @xmath0 is used . in figure 2 , we show the rcsr applied to a uniform system of nucleons with anomalous magnetic moments , for @xmath162 . ( here models f and g are identical . ) a similar figure was shown in ref . @xcite , but it is useful to review the main results here for comparison . the unmodified coulomb sum @xmath23 is shown by the dotted curve , and indicates an enhancement relative to the free dirac result . this is a consequence of the nucleon anomalous moments , as explained in ref . the lowest - order sum rule , @xmath163 , overshoots the `` correct '' result , i.e. , saturation , increasingly with higher momentum , and indicates the importance of fermi motion effects when anomalous moments are included . in ref . @xcite we explained that this is because the magnetic contribution is enhanced relative to the electric by a factor @xmath167 . this prevents the cancellation of electric and magnetic fermi motion effects , which occurs nearly exactly for dirac nucleons . the striking feature is that the @xmath168 result , @xmath169 is accurate to within @xmath170 over this momentum range . we also show the @xmath171 result , @xmath172 , which is accurate to better than @xmath173 over the same range . this we interpret as numerical evidence of convergence of the moment expansion in this model . the results shown here will be modified for @xmath24 , and will also depend on the choice of form factor model ( f or g ) . we are interested in the accuracy of sum rules designed to account for these modifications . we now consider @xmath24 . in figures 3a and 3b we show results for the rcsr in model g for rha and mft values of @xmath0 , respectively . as for figure 2 , the coulomb @xmath23 is enhanced relative to that for dirac nucleons , due to fermi motion effects , when anomalous moments are included . this model also has the rather peculiar feature that for @xmath174 , we have @xmath175 for @xmath176 . this value is coincidental ( see sec . 4.3 for further discussion ) . for still lower values of @xmath0 the coulomb sum can be significantly greater than unity . this is contrary to the common notion that relativistic effects necessarily suppress the coulomb sum relative to unity . note that the moment expansion converges more slowly for smaller @xmath0 , as can be seen by comparing @xmath169 in figures 3a and 3b . still , the rcsr to @xmath168 is reasonably accurate , even for @xmath177 . the very high accuracy of @xmath172 is interpreted as numerical evidence of convergence in this model . indeed , for @xmath176 , @xmath178 for the rha mass and @xmath179 for the mft mass , at @xmath180 . in figures 4a and 4b we show analogous results for a nuclear system under the off - shell assumption of model f. the coulomb sum @xmath23 behaves qualitatively differently in this model , with a minimum developing around @xmath181 as @xmath0 decreases to the mft value . the differences originate from the additional @xmath1-dependence in the effective form factors @xmath182 and @xmath183 , which are constants in model g. as in model g , @xmath163 is not adequate , and illustrates the importance of fermi motion effects when anomalous moments are included . also as for model g , @xmath169 , shown by the dashed curve , is reasonably accurate over this range of momenta and @xmath0 . the improvement in @xmath172 is interpreted as numerical evidence of convergence of the moment expansion . the most important feature of the sum rule results just presented for model f is the success of the factored moment expansion , which allows the use of the _ same _ expansion coefficients as in model g , i.e. , those given in appendix b. this approximation neglects only the fermi motion effects which enter through the ratios @xmath184 ^ 2 $ ] in ( 3.9 ) . to illustrate its accuracy , we have evaluated @xmath169 in a `` consistent '' moment expansion which correctly includes to @xmath168 additional fermi motion effects entering through these ratios . this result is shown by a dot - dashed curve , which is nearly identical to the `` factored '' result , shown by the dashed curve . these are small corrections on the scale of the other kinematical effects , lending support to the factored moment expansion in model f. in summary , we have shown that the relativistic effects of nucleon recoil , fermi motion and a reduced effective mass @xmath0 can be accounted for accurately in both models f and g. keeping only terms through @xmath168 gives @xmath185 accuracy , which is small on the scale of the dominant kinematical effects . keeping terms through @xmath171 gives @xmath164 accuracy . we emphasize that in both models f and g , the rcsr was evaluated using the _ same _ moment expansion coefficients ( those in appendix b ) , but in model f the effective charges and magnetic moments depend on @xmath1 and @xmath0 . this simplification requires using a factored moment expansion , in which fermi motion effects entering though the effective form factors @xmath186 , @xmath187 are ignored . in the cases studied this is an excellent approximation , as explained further below . one can understand the results just presented by looking individually at the electric and magnetic recoil functions @xmath114 and @xmath115 , defined in ( [ repq ] ) and ( [ rmpq ] ) , as functions of @xmath0 . we shall look at the variations of @xmath188 and @xmath189 of ( 3.14 ) in model f , compared to @xmath116 and @xmath117 in model g. we are especially interested in the factored moment expansion in model f. we shall examine the reasons for its success , and under what conditions such an approach may be expected to work for other off - shell assumptions . we argue that a factored moment expansion , like that adopted here , should be accurate in a large class of off - shell form factor models . in figures 5a and 5b , we show the electric and magnetic recoil functions : @xmath129 of ( [ repq ] ) and @xmath130 of ( [ rmpq ] ) , for two values of @xmath0 . in both figures , the solid curve represents the electric term @xmath190 , which dominates the coulomb response for a system of dirac nucleons . the magnetic term @xmath191 , as can be seen from the second equality in ( [ rmpq ] ) . thus the coulomb response for a nucleon at rest is purely electric , as noted in ref . the electric function @xmath190 decreases from unity at @xmath192 to its limiting value @xmath193 as @xmath194 . these limits are independent of @xmath0 , however , for smaller @xmath0 , @xmath190 decreases more rapidly with @xmath1 , since it is a monotonic function of @xmath195 . this increased suppression is easily accommodated by using an appropriate value for @xmath0 , as discussed for the dirac case in figure 1 . the corrections for fermi motion to @xmath114 and @xmath115 are also shown in figures 5a and 5b , and are denoted by @xmath196 and @xmath197 , respectively . these corrections were calculated by averaging @xmath129 and @xmath130 over the fermi sphere , taking @xmath140 . for a system of dirac nucleons , @xmath198 and @xmath199 enter the response with equal weighting . we noted in ref . @xcite for @xmath162 that this leads to a nearly exact cancellation over the entire momentum range , and ensures that the lowest - order sum rule , rcsr - i , is accurate to within @xmath164 . note that this cancellation persists to @xmath164 accuracy for @xmath24 , although _ both _ @xmath198 and @xmath199 nearly triple for @xmath177 , as in the mft example . for nucleons with anomalous magnetic moments , this near cancellation of @xmath196 and @xmath197 no longer occurs , since these function are multiplied by different factors : @xmath200 , @xmath201 of ( 3.8 ) for model g , or @xmath202 , @xmath203 of ( 3.14 ) for model f. in model g , @xmath197 is multiplied by the constant factor @xmath167 , while @xmath204 . for @xmath162 , this gives a 1020% increase in the coulomb sum @xmath23 , as was discussed in ref . @xcite . as @xmath205 decreases , the enhanced effect of @xmath197 increases further : for @xmath206 , the fermi motion correction increases with @xmath1 about as quickly as @xmath190 decreases , as shown in fig . this leads to the behavior of @xmath23 for this value of @xmath0 , shown in fig . 3b , and remarked on earlier , i.e. , @xmath207 . as @xmath0 is decreased to the mft value , this increase in @xmath199 effectively triples the importance of higher moments . this accounts for the decreased accuracy of the rcsr at @xmath168 , compared to that applied to free nucleons . model f has qualitatively similar behavior , as we saw in figs . 4a and 4b . here , however , @xmath23 is modified by the effective charges and moments ( 3.14 ) , which are shown in fig . 6 ( solid curves ) . for this case @xmath208 is somewhat smaller than @xmath209 and increases with @xmath1 over the range shown . this explains the smaller values of @xmath23 in figs . 4a and 4b than in figs . 3a and 3b , and the upturn with increasing @xmath1 in fig . we have seen that the convergence properties of the rcsr in model f are similar to those in model g , in spite of the use of a factored moment expansion that ignores certain fermi motion effects , which would enter through the effective form factors ( 3.11 ) . in order to understand this result , we first examine the effective form factors themselves . in figure 6 , we show these in the form in which they enter the recoil function @xmath210 in model f , i.e. , @xmath211 ^ 2 $ ] and @xmath212 ^ 2 $ ] . we have used ( 3.12 ) and @xmath142 in their evaluation . the dashed curves show the form factors at @xmath128 , and the solid curves were obtained by averaging over the fermi sphere with @xmath140 . ( note that the electric form factors tend to their free values in the limit @xmath213 , while the magnetic form factors tend to their free values in the limit @xmath194 . this can be seen from the definitions given in ( [ gsofg ] ) . ) we see that the fermi motion corrections to the ratios @xmath184 ^ 2 $ ] are small compared to the ratios themselves . thus terms @xmath168 and higher are small , especially in the magnetic form factors . ( terms @xmath214 vanish upon angle - integration with spherical symmetry , and therefore do not contribute to the results in figure 6 . ) to understand the success of the factored moment expansion , consider the second - order sum rule , @xmath169 . it omits two types of fermi corrections : 1 ) @xmath168 terms which come from a moment expansion of the effective form factors and multiply @xmath190 and @xmath215 , and 2 ) products of linear terms involving @xmath216 , one term from the effective form factors , and one from the recoil factors @xmath114 and @xmath115 . the first of these make at most @xmath185 correction to the electric contribution , and less than @xmath164 correction to the magnetic , as can be seen from figure 6 . these terms can therefore be neglected to a good approximation . linear terms do not enter the magnetic contribution , as can be seen from the @xmath217 factor in ( [ rmpq ] ) . linear terms have been omitted from the electric contribution , and could in principle have been appreciable . in fact , these terms taken together result in less than a @xmath218 error in the rcsr , as can be seen by comparing the dashed and dot - dashed curves for @xmath169 in figures 4a and 4b . thus the success of the factored moment expansion in model f can be attributed to the smallness of fermi corrections in the effective form factors themselves , and in part to the complete absence of linear terms in the magnetic contribution . we believe that the success of the factored moment expansion in model f may also occur for a much wider variety of off - shell models than we have considered here . the basis for this claim is that fermi corrections tend to be small in general , representing effects of order @xmath219 . the main exception is the magnetic correction , which is substantial because of two properties : @xmath220 , and @xmath221 . the magnetic contribution @xmath199 is small , but is then enhanced by roughly an order of magnitude by the anomalous magnetic moments . no such enhancement occurs for @xmath198 . thus in model f , the further corrections due to * p*-dependence in @xmath211 ^ 2 $ ] and @xmath212 ^ 2 $ ] are negligible . in other off - shell models , for which @xmath17 is expressible in the form ( 3.9 ) , we expect fermi motion effects in the corresponding effective form factors also to be small . we therefore propose that a factored moment expansion , like that employed here in model f , may also be accurate for the evaluation of fermi motion effects a wide variety of off - shell models . this allows one to evaluate the rcsr using the expansion coefficients given in appendix b. ( we include @xmath196 , although a small contribution in nucleon models with anomalous magnetic moments , to ensure the cancellation of fermi corrections for the dirac system , and to allow the `` consistent '' calculation of @xmath169 in figs . 5a and 5b . ) the main results of this paper can be summarized as follows : the effects of nuclear binding in the mean - field approximation enter the coulomb response function @xmath7 and the coulomb sum @xmath23 in a way that can be characterized by a reduced effective mass @xmath0 . these effects depend on the behavior of the electromagnetic current for the off - shell kinematics in the nuclear medium . the sensitivity of the coulomb sum @xmath23 to the choice of off - shell model has been discussed previously @xcite . we consider two illustrative models : f and g. we demonstrate in these models that the rcsr of ref . @xcite can be extended to account for relativistic binding effects , in addition to the purely kinematic effects of recoil and fermi motion treated in ref . the resulting rcsr is no longer model - independent , since one must make some assumptions about @xmath0 and the off - shell behavior of the form factors . however , to the extent that the dominant effects of the nuclear medium can be characterized by the simple parameter @xmath0 , the coulomb sum rule analysis can be applied to data , with the goal of looking for nuclear structure effects beyond the mean field , e.g. , two - body correlations . we emphasize that the form of the sum rule is sufficiently general to accommodate a broad class of off - shell form factor models , i.e. , those for which the recoil function @xmath17 can be expressed in the form ( 3.9 ) . the real change in transforming to the @xmath0 basis enters through the nucleons - only approximation , under which the sum rule is derived . strong potentials in relativistic models mix free @xmath57 components into the initial and final interacting nuclear states , and modify the spacelike response function . chinn , picklesimer and van orden @xcite have isolated the effects of this mixing of @xmath57 components , and show results qualitatively similar to those seen in fig . 4 . ( their form factors are closer to our model f than g. ) the transformation to an @xmath0 basis automatically incorporates these potential effects in a convenient representation , although the use of an effective mass independent of position or momentum may only approximate the physical situation . how might one apply this rcsr to experimental data ? first , the coulomb sum @xmath23 must be calculated from the measured coulomb response function , as in ( [ aa ] ) . then , one must make some specific assumptions about the off - shell behavior of the current operator , as discussed in section 3 . this should be cast in the @xmath222 form as in ( 3.10 ) and ( 3.11 ) for model f ( or @xmath223 for model g ) , with the substitution ( 3.12 ) to remove ( approximately ) any residual @xmath6-dependence from @xmath182 and @xmath183 . the choice of model is not restricted to those presented in sections 3.1 and 3.2 . with the factored moment expansion , the relativistic recoil function ( 3.1 ) takes the form ( 3.13 ) and the ratios @xmath224 are evaluated at @xmath128 , as in ( 3.14 ) . the recoil factor @xmath225 is then evaluated in a moment expansion , e.g. , to @xmath168 as in ( 4.4 ) , with @xmath226 fixed by other experimental information , or by a model as in ( 4.5 ) , or as a free parameter . the expansion coefficients @xmath108 are given in appendix b , and are functions of the parameter @xmath0 . the modified coulomb sum @xmath158 is obtained by forming the ratio ( 4.2 ) of the experimentally determined numerator @xmath23 , and the ( model dependent ) recoil denominator @xmath227 . the expected behavior of @xmath23 with increasing @xmath1 is that it will approach the one - body term @xmath25 of ( 1.3 ) , assuming that both @xmath13 and @xmath228 in ( 1.2 ) , as @xmath194 . the modified sum @xmath158 will then approach unity _ if _ the assumed form of the current is correct , _ and if _ the effective mass @xmath0 is appropriately chosen . should that be the case for a given set of data , it would be reasonable to assume that the recoil functions have been correctly chosen . the sum rule ( 1.3 ) can then be used to investigate the two - nucleon correlation function in the ratio form is related to the standard nonrelativistic correlation function . ( see eqs . ( 5.8 ) and ( 5.16 ) in ref . an interesting application of the rcsr is the constraint of the viable off - shell form factor models by analyzing the experimentally determined coulomb sum @xmath23 in different off - shell models and comparing to @xmath231 beyond the expected range of correlations . a further remark about the analysis of data with the rcsr seems appropriate . it has become customary for the experimental coulomb response data to be integrated in a modified form of ( 1.1 ) , as suggested by deforest @xcite , in which the proton electric form factor @xmath119 is replaced by @xmath232 , with @xmath78 ( see , e.g. , refs . we explained in ref . @xcite why this procedure will not lead to a _ non - energy - weighted _ sum rule : to obtain a sum rule of the form ( 1.2 ) or ( 1.5 ) , the extra @xmath6-dependence should not be introduced into the definition of the coulomb sum . the kinematic effects included in the recoil factors @xmath225 depend on the 3-momentum transfer * q * , rather than on the invariant @xmath71 . it is this property which preserves the sum rule in passing from ( 1.2 ) to ( 5.1 ) . the use of the @xmath0-basis for both initial and final states , with the same value of @xmath0 , implicitly assumes that the nucleus is large enough that the nucleon kinematics in the final state are sensitive to the scalar field . for smaller nuclei , it may be necessary to account for the effects of finite size . this could possibly be accomplished through a rederivation of the rcsr in a hartree representation based on bound , localized nuclear states . another issue is the momentum - dependence of the mean - field potentials , as discussed in refs . @xcite , for example . since the rcsr has been derived in momentum space , such a modification is relatively straightforward . as a first approximation , which is consistent with our conclusion that fermi motion effects tend to be small , it seems reasonable to use @xmath233 for initial states and @xmath234 for final states . this prescription preserves the basic structure of the rcsr , in that no further dependence on @xmath47 is introduced . this would require the initial- and final - state masses to be treated distinctly , however , and would complicate the form of the coefficients in appendix b. we are currently investigating the effects of virtual @xmath55 pairs ( of mass @xmath0 ) on the rcsr , as they enter intermediate excited states in the random phase approximation . these were considered previously by horowitz @xcite , and appear to be significant . we are also interested in how energy dependent terms , which can enter in off - shell models where a substitution of the form ( 3.12 ) is not appropriate , affect the rcsr . this research was supported in part by the u.s . department of energy under grant no . de - fg02 - 88er40425 with the university of rochester . the authors would also like to thank the high energy physics group for use of the vax computer . in model g , we use the on - shell forms of @xmath85 and @xmath86 , given in ( [ aae ] ) . in model f , we use @xmath186 and @xmath187 , given in ( 3.11 ) . expression ( a5 ) makes clear the origin of the functional forms in ( 3.6 ) and ( 3.7 ) . in this appendix , we give the coefficients @xmath245 , for even powers through @xmath171 in the moment expansion . these are written as they appear in model g , i.e. , in terms of the usual nucleon charge ( @xmath116 ) and magnetic moment ( @xmath117 ) : @xmath248\nonumber\\ & & \label{abc}\\ & & + \,\mu_\sigma^2\ \bigl[-64{e_{\bf q}^*}^9 - 24{e_{\bf q}^*}^8m^ * + 16{e_{\bf q}^*}^7{m^*}^2\nonumber\\ & & \nonumber\\ & & \qquad-16{e_{\bf q}^*}^6{m^*}^3 -48{e_{\bf q}^*}^5{m^*}^4 - 24{e_{\bf q}^*}^4{m^*}^5\bigr ] . \biggr]\nonumber\end{aligned}\ ] ] k. w. mcvoy and l. van hove , phys . rev . * 125 * , 1034 ( 1962 ) . fetter and j.d . walecka , _ quantum theory of many - particle systems , _ ( mcgraw - hill , new york , 1971 ) . ferre and d.s . koltun , phys . c * 49 * , 1961 ( 1994 ) . serot and j.d . walecka , adv . * 16 * , 1 ( 1986 ) . t. matsui , phys . lett . b * 132 * , 260 ( 1983 ) . g. do dang , m. lhuillier , nguyen van giai and j. w. van orden , phys . c * 35 * , 1637 ( 1987 ) . chinn , a. picklesimer and j.w . van orden , phys . c * 40 * , 790 ( 1989 ) . chinn , a. picklesimer and j.w . van orden , phys . c * 40 * , 1159 ( 1989 ) . r. schiavilla , r.b . wiringa and j. carlson , phys . * 70 * , 3856 ( 1993 ) . t. de forest , jr . , * a392 * , 232 ( 1983 ) . chinn and a. picklesimer , _ nuovo cim . a _ * 105 * , 1149 ( 1992 ) . s.j . wallace , ann . nucl . & part . , * 37 * , 267 ( 1987 ) . t. de forest , jr . * a414 * , 347 ( 1984 ) . chen _ et . al . , _ phys . * 66 * , 1283 ( 1991 ) . meziani _ et . al . , _ phys . lett . * 69 * , 41 ( 1992 ) . g. do dang and pham van thieu , phys . c * 28 * , 1845 ( 1983 ) . h. kim , c.j . horowitz and m.r . frank , indiana university nuclear theory center , preprint no . 94 - 11 , submitted to phys . c ( 1994 ) . c. j. horowitz , phys . b * 208 * , 8 ( 1988 ) . barnes , phys . * 1 * , 166 ( 1962 ) . hand , d.g . miller and r. wilson , rev . mod . phys . * 35 * , 335 ( 1963 ) . scadron , _ advanced quantum theory , _ ( springer - verlag , new york , 1979 ) . ( see problem 5.8c , pg . 352 . ) 0.3 true in * figure captions * 0.2 true in fig . 1 coulomb sum @xmath23 for a uniform system of dirac protons , at two values of @xmath0 . two different evaluations of the lowest - order rcsr are shown as indicated.fig . 2 . rcsr evaluated for a uniform system of nucleons with anomalous magnetic moments , for @xmath162 . the coulomb sum @xmath23 in ( 4.1 ) and @xmath158 in ( 4.2 ) are shown as indicated.fig . same as fig . 2 , but for @xmath24 in model g. results are shown for ( a ) @xmath143 and ( b ) @xmath142 . same as fig . 3 , but in model f. note that @xmath169 is calculated in two ways : the `` factored '' ( dashed ) and `` consistent '' ( similar dot - dashed ) moment expansions.fig . 5 . electric and magnetic recoil functions , @xmath129 and @xmath130 , in a convenient separation . results are shown for ( a ) @xmath162 and ( b ) @xmath142 . the functions @xmath190 ( solid ) , @xmath196 ( dashed ) and @xmath197 ( dot - dashed ) are shown ; @xmath191 . ( note the scale changes for @xmath198 and @xmath199.)fig . effective form factors ( squared ) in model f , for @xmath142 . solid curves are obtained by averaging ( 3.11 ) with ( 3.12 ) over the fermi sphere , and include fermi corrections to all orders . dashed curves are @xmath202 and @xmath203 of ( 3.14 ) , as used in the `` factored '' moment expansion .

we consider a metallic wire coupled to two metallic electrodes via two junctions placed nearby . a bias voltage applied to one of such junctions alters the electron distribution function in the wire in the vicinity of another junction thus modifying both its noise and the coulomb blockade correction to its conductance . we evaluate such interaction corrections to both local and non - local conductances demonstrating non - trivial coulomb anomalies in the system under consideration . experiments on non - local electron transport with coulomb effects can be conveniently used to test inelastic electron relaxation in metallic conductors at low temperatures .

1110.2666

a direct relation between shot noise and coulomb blockade of electron transport in mesoscopic conductors is well known . in normal conductors this relation was established theoretically @xcite and subsequently confirmed experimentally @xcite . later the same ideas were extended to subgap electron transport in normal - superconducting ( ns ) hybrids @xcite . the latter results appear to provide an adequate interpretation for experimental observations @xcite of coulomb effects in such systems . while all the above developments concern local electron transport and shot noise , the question arises if there also exists any general relation between non - locally correlated shot noise in multi - terminal conductors and coulomb effects on non - local electron transport in such systems . an important example is provided by three - terminal nsn structures which have recently received a great deal of attention in both experiments @xcite and theory @xcite in connection with the phenomenon of crossed andreev reflection . the latter phenomenon yields non - trivial behavior of the non - local subgap conductance in such structures . further interesting features emerge if one takes into account electron - electron interactions . one can observe , for example , the sign change of the non - local conductance caused either by the influence of the electromagnetic modes propagating along the wire @xcite , or by positive cross - correlations in non - local current noise @xcite . furthermore , positive cross - correlations in shot noise are directly linked to coulomb ani - blockade of non - local electron transport @xcite . thus , a general relation between cross - correlated shot noise and coulomb effects in non - local subgap electron transport in nsn systems turns out to be much richer than that in the local case @xcite . in this paper we will address the impact of electron - electron interactions on non - local effects in normal metallic structures depicted in fig . [ fig_two_junctions ] . non - local properties of such systems turn out to be very sensitive to inelastic processes . at low temperatures such processes in metallic conductors usually become rather weak and electrons can propagate at long distances , typically of order microns , without suffering any significant energy changes . hence , provided voltage bias is applied to a mesoscopic conductor , its electron distribution function @xmath0 may substantially deviate from its equilibrium value universally defined by the fermi function @xmath1 . for example , low temperature distribution function @xmath0 may take the characteristic double - step form in comparatively short metallic wires attached to two big reservoirs with different electrostatic potentials @xcite . connecting the electrodes 3 and 4 via the two junctions with resistances @xmath2 , fano factors @xmath3 and capacitances @xmath4 . ( b ) equivalent electric circuit of the system depicted in panel ( a ) . , width=283 ] further interesting effects emerge if one takes into account an interplay between non - equilibrium effects and electron - electron interactions . consider , e.g. , a tunnel junction between two metallic leads . provided the the junction resistance significantly exceeds that of the leads , the effect of coulomb interaction can be modeled by introducing interactions between electrons and some linear electro - magnetic environment @xcite . in this case the strength of coulomb interaction is characterized by an effective impedance of the environment and the current across the tunnel junction reads @xmath5 p(e_l - e_r - ev ) \nonumber\\ & & - \big[1-f_l(e_l)\big]f_r(e_r)p(-e_l+e_r+ev ) \big\ } , \label{pofe}\end{aligned}\ ] ] where @xmath6 are the electron distribution functions in the left and right electrodes and @xmath7 is the probability to excite a photon with energy @xmath8 due to interaction between the junction and the environment . provided the environment has a non - zero impedance and both distribution functions @xmath9 and @xmath10 are close to the fermi function , eq . ( [ pofe ] ) yields the well known zero - bias anomaly on the i - v curve , i.e. the coulomb blockade dip in the differential conductance @xmath11 in the limit of low voltages @xcite . furthermore , should at least one of the distribution functions deviate from the equilibrium one , the i - v curve can receive further significant modifications . for instance , if one distribution function takes the double step form @xcite , it follows immediately from eq . ( [ pofe ] ) that the coulomb blockade dip in the conductance should split into two separate dips . these dips can be and have been @xcite detected experimentally thus offering a possibility to investigate non - equilibrium effects with the aid of small capacitance tunnel junctions as it was demonstrated , e.g. , by experimental analysis of the impact of magnetic impurities on inelastic relaxation of electrons in normal metals @xcite . despite clear advantages and simplicity of eq . ( [ pofe ] ) , it might not always be convenient to employ in order to analyze combined effects of non - equilibrium and coulomb interaction in metallic conductors . indeed , the applicability of eq . ( [ pofe ] ) is restricted to junctions with very low barrier transmissions , i.e. the effect of higher transmissions can not be correctly accounted for my means of this equation . the latter effect might be important , in particular if one needs to evaluate the non - local conductance . in addition , the function @xmath7 is usually evaluated under the assumption of thermodynamic equilibrium in electromagnetic environment , which effectively implies equilibrium electron distributions in both leads . if , however , the electron subsystem is driven out of equilibrium , self - consistent evaluation of @xmath7 might become a non - trivial problem . furthermore , the function @xmath7 would in general be difficult to evaluate for effectively non - linear electromagnetic environments . the above complications are avoided within the kinetic equation analysis presented below . this approach only requires resistances of metallic leads to remain smaller than the quantum resistance unit @xmath12 . within the same theoretical framework it allows to evaluate both non - local shot noise and the effect of electron - electron interactions on non - local electron transport in normal metallic conductors as well as to describe a non - trivial interplay between coulomb effects and inelastic processes in such structures the paper in organized as follows . in sec . [ sec : model ] we outline our model and define the hamiltonian of our system . in sec . [ sec : noise ] we analyze non - local correlated shot noise in the system under consideration . in sec . [ sec : current ] we extend this analysis taking into account electron - electron interactions and demonstrating direct relation between shot noise and interaction effects in non - local electron transport . a brief summary of our key observations is contained in sec . [ sec : summary ] . some technical details are relegated to appendices . in appendix [ sec : derivation ] we outline key steps of our derivation of the kinetic equation employed in our analysis . necessary details of our solution of this kinetic equation are displayed in appendix [ sec : solution ] . in this paper we will consider the system depicted in fig . [ fig_two_junctions ] . it consists of a metallic wire of length @xmath13 connected to two leads 1 and 2 by two small area junctions located at @xmath14 , @xmath15 and two bulk reservoirs 3 and 4 at @xmath16 and @xmath17 ( @xmath18 is the coordinate along the wire ) . the system depicted in fig . [ fig_two_junctions ] is described by the hamiltonian @xmath19 where @xmath20 are the hamiltonians of the normal metals , @xmath21 is the hamiltonian of the wire and @xmath22 \label{ht}\end{aligned}\ ] ] are tunneling hamiltonians describing transfer of electrons across the contacts with area @xmath23 and tunneling amplitude @xmath24 . here and below @xmath25 stands for the electron mass , @xmath26 is the chemical potential , the index @xmath27 labels the spin projection , the potential @xmath28 accounts for disorder inside the wire and @xmath29 represents the scalar potential . the transmissions of the conducting channels of the junctions are related to the matrix elements of the tunnel amplitudes @xmath30 between the states belonging to the same conducting channel as follows @xmath31 where @xmath32 ( @xmath15 ) is the density of states in the corresponding terminal and @xmath33 is the density of states inside the wire . the barrier resistances @xmath34 and @xmath35 and their fano factors @xmath36 and @xmath37 are expressed in a standard way as @xmath38 a voltage bias , respectively @xmath39 and @xmath40 , can be applied to all four metallic terminals 1,2,3 and 4 . in the setup of fig . [ fig_two_junctions ] one of the junctions , e.g. the junction 2 , may be viewed as an injector , which drives electron distribution function in the wire out of equilibrium . the junction 1 may then be used as a detector for experimental investigation of nonequilibrium effects . one of the ways to observe such effects is to study the non - local differential conductance @xmath41 of our system . clearly , in such kind of experiments the distance between the junctions should not exceed an effective electron inelastic relaxation length @xmath42 which sets the scale for non - equilibrium effects in the wire at a given temperature . thus , the setup of fig . [ fig_two_junctions ] may be used to directly measure @xmath43 . finally we note that the above particular system geometry is chosen merely for the sake of definiteness . the key steps of our subsequent analysis and the results obtained from it remain applicable to a much broader class of systems than that depicted in fig . [ fig_two_junctions ] . e.g. , the wire may be replaced by a metallic lead of any shape , and ultimately all geometry specific details can be absorbed in few elements of the conductance matrix . we begin with the analysis of shot noise employing the so - called boltzmann - langevin technique @xcite based on a kinetic equation for the electron distribution function @xmath44 . low frequency cross - correlated shot noise in multi - terminal metallic structures has already been studied before , see , e.g. , ref . . here we will briefly rederive and somewhat extend the corresponding results in order to illustrate the basic idea of the approach in a relatively simple case . in the next section we will extend this approach in order to include electron - electron interactions where more involved calculations will be necessary . the boltzmann - langevin kinetic equation accounts for current noise produced by the junctions 1 and 2 and has the form @xmath45 here @xmath46 and @xmath33 are respectively the electron diffusion constant and the electron density of states at the fermi energy inside the wire . we also introduced electrostatic potentials of the leads @xmath47 and @xmath48 in the vicinity of the junctions 1 and 2 , @xmath49 where the resistances of the leads @xmath50 are defined in fig . [ fig_two_junctions]b , and @xmath51 in the inelastic relaxation time . note that here we are not going to discuss physical mechanisms dominating the process of electron energy relaxation at low temperatures and simply treat @xmath52 as a phenomenological parameter . the potential @xmath53 should be determined self - consistently from the equation @xmath54=0 , \label{self - consist}\end{aligned}\ ] ] which directly follows from the charge neutrality condition inside the normal metal . this charge neutrality condition in metals is a direct consequence of strong coulomb interaction between electrons as well as between electrons and lattice ions . integrating eq . ( [ kinetic_0 ] ) over energy we obtain @xmath55 note that inelastic relaxation time @xmath52 drops out from this equation . the stochastic variables @xmath56 and @xmath57 in eqs . ( [ kinetic_0 ] ) and ( [ diffusion_v ] ) account for low frequency fluctuations of the current carried by electrons with energy @xmath8 through the junctions 1 and 2 respectively . the corresponding correlators read@xcite @xmath58 \nonumber\\ & & + \ , \beta_j\left[1-f(t_1,e_1,\bm{x}_j)\right]f_f(e_1-ew_j ) \nonumber\\ & & + \,(1-\beta_j)f(t_1,e_1,\bm{x}_j)\left[1-f(t_1,e_1,\bm{x}_j)\right ] \nonumber\\ & & + \,(1-\beta_j)f_f(e_1-ew_j)\left[1-f_f(e_1-ew_j)\right ] \big\}. \label{corr_eta_0}\end{aligned}\ ] ] finally , no fluctuations occur at fully open contacts between the wire and the terminals 3 and 4 . these contacts are accounted for by the boundary conditions @xmath59 note that in the eq . ( [ kinetic_0 ] ) we have neglected the internal current noise generated in the wire @xcite . in order to justify this approximation , in what follows we will assume @xmath60 i.e. we will assume the junction resistances to be much higher than the resistances of the metallic leads and the wire ( see fig . [ fig_two_junctions]b for the definition of the resistances ) . thus , the task at hand is to solve eqs . ( [ kinetic_0 ] ) , ( [ diffusion_v ] ) supplemented by eqs . ( [ corr_eta_0 ] ) , ( [ bou ] ) and to evaluate the current noise in our system . as we already discussed above , the form of the distribution function inside the wire may essentially depend on the relation between its size @xmath13 and the inelastic relaxation length @xmath43 . yet another relevant parameter to be compared with @xmath43 is the distance between the two junctions @xmath61 . provided inelastic relaxation is very strong , @xmath62 , the inelastic term in eq . ( [ kinetic_0 ] ) plays the dominant role and the electron distribution function @xmath63 in the wire remains close to the fermi function @xmath64 with the voltage @xmath65 to be derived from eq . ( [ diffusion_v ] ) . in the opposite weak relaxation limit @xmath66 the inelastic collision integral in eq . ( [ kinetic_0 ] ) can be neglected . of interest is also the intermediate limit of a long wire @xmath67 but relatively weak relaxation @xmath68 . we begin our analysis by defining the currents @xmath69 and @xmath70 across junctions 1 and 2 : @xmath71+\delta\tilde i_j , \label{current_0}\end{aligned}\ ] ] where @xmath72 is the fluctuating current in the @xmath73-th junction . in the limit of full inelastic relaxation , @xmath74 , the distribution function in the wire has the equilibrium form , and with the aid of eq . ( [ corr_eta_0 ] ) we derive the zero frequency spectral noise power @xmath75 , @xmath76 where @xmath77 are voltage drops across the junctions . under the condition ( [ resistances ] ) one finds @xmath78 naturally , eq . ( [ s_bare ] ) just coincides with the noise power for a perfectly voltage biased junction @xcite . let us now consider the limit @xmath79 . in this case , according to eq . ( [ kinetic_0 ] ) the electron distribution function @xmath80 deviates from the equilibrium form and fluctuates . hence , the total current noise should acquire an additional contribution . in order to proceed let us establish the relation between the distribution functions @xmath80 and the stochastic variables @xmath81 . this goal can be achieved with the aid of the diffuson @xmath82 , which is defined as a solution of the diffusion equation @xmath83 with boundary conditions @xmath84 the physical meaning of the diffuson @xmath82 is well known : it defines the probability for an electron injected into the wire at the point @xmath85 to reach the point @xmath86 during the time @xmath87 . we also define the fourier transformed diffuson @xmath88 the solution of eq . ( [ kinetic_0 ] ) can be expressed in the form @xmath89 . \label{distribution_1}\end{aligned}\ ] ] this general expression gets simplified in the limit @xmath90 and provided current fluctuations can be neglected , i.e. @xmath91 . in this case the electric potential @xmath65 does not depend on time and slowly varies in space . then one can approximately replace @xmath92 by @xmath93 . afterwards , employing the properties of the diffuson , one finds @xmath94f_f\left(e - ev(\bm{x})\right ) \nonumber\\ & & + \,\frac{\tilde{\cal d}(0,\bm{x},\bm{x}_1)}{2e^2\nu_0r_1}f_f(e - ew_1 ) + \frac{\tilde{\cal d}(0,\bm{x},\bm{x}_2)}{2e^2\nu_0r_2}f_f(e - ew_2 ) . \label{distribution_2}\end{aligned}\ ] ] the non - equilibrium distribution function in this regime has three steps , see also fig . [ fig_fofe ] . the first one comes from the distribution function of the isolated wire @xmath93 , while the other two steps , @xmath95 , originate from the junctions . since the diffuson @xmath96 decays at distances @xmath97 , the distribution function acquires its equilibrium form far away from the junctions . . the solid line shows @xmath98 given by eq . ( [ distribution_2 ] ) which is applicable at intermediate values of inelastic relaxation . in this regime the distribution function has three steps . the dashed line corresponds to elastic limit in which the distribution function @xmath98 is defined in eq . ( [ distribution_elastic ] ) . the system parameters are : @xmath99 mk , @xmath100 , @xmath101 , @xmath102 @xmath26v , @xmath103 @xmath26v , @xmath104 @xmath26v , @xmath105 @xmath26v , @xmath106.,width=302 ] the currents @xmath69 and @xmath70 can be evaluated with the aid of eqs . ( [ self - consist ] ) and ( [ diffusion_v ] ) . they read @xmath107 here @xmath108 and @xmath109 define respectively local and non - local conductances of our structure : @xmath110 where @xmath111 is the solution of the diffusion equation ( [ diffusion ] ) with @xmath112 . one can equivalently write these conductances in the from @xmath113 here we again assumed that @xmath114 and @xmath115 . let us emphasize that the results ( [ gij ] ) , ( [ gij_r ] ) were derived from eq . ( [ diffusion_v ] ) and , hence , are not sensitive to inelastic relaxation at all . besides that , general expressions ( [ gij ] ) are not restricted to the wire geometry and remain valid for any shape of the leads . finally , the noise terms @xmath116 appearing in eq . ( [ current ] ) read @xmath117 , \label{xi1 } \\ \delta i_2 & = & \int de dt ' \left[-\frac{{\cal d}_0(t - t',\bm{x}_2,\bm{x}_1)}{2e^2\nu_0r_2}\eta_1(t',e ) \right . \nonumber\\ & & \left . + \,\left(\delta(t - t')-\frac{{\cal d}_0(t - t',\bm{x}_2,\bm{x}_2)}{2e^2\nu_0r_2}\right)\eta_2(t',e)\right ] . \label{xi2}\end{aligned}\ ] ] clearly , they differ from the bare noise terms @xmath118 since the contributions coming from electrons diffusing from one junction to the other or returning back to the same junction are also taken into account . we are now in position to evaluate the zero frequency noise power matrix @xmath119 with the aid of eqs . ( [ corr_eta_0 ] ) and ( [ xi1],[xi2 ] ) we express the noise power for the first junction as @xmath120 \nonumber\\ & & + \ , \beta_1 \left[1-f(e,\bm{x}_1)\right]f_f(e - ew_1 ) \nonumber\\ & & + \ , ( 1-\beta_1 ) f(e,\bm{x}_1)\left[1-f(e,\bm{x}_1)\right ] \nonumber\\ & & + \ , ( 1-\beta_1 ) \left[1-f_f(e - ew_1)\right]f_f(e - ew_1 ) \big\ } \nonumber\\ & & + \ , r_2 g_{\rm nl}^2 \int de \big\ { \beta_2 f(e,\bm{x}_2)\left[1-f_f(e - ew_2)\right ] \nonumber\\ & & + \ , \beta_2\left[1-f(e,\bm{x}_2)\right]f_f(e - ew_2 ) \nonumber\\ & & + \ , ( 1-\beta_2 ) f(e,\bm{x}_2)\left[1-f(e,\bm{x}_2)\right ] \nonumber\\ & & + \ , ( 1-\beta_2 ) \left[1-f_f(e - ew_2)\right]f_f(e - ew_2 ) \big\}. \label{noise11_0}\end{aligned}\ ] ] substituting the distribution function ( [ distribution_2 ] ) into this expression , assuming @xmath121 , defining the function @xmath122 and under the condition ( [ resistances ] ) , we arrive at the final result @xmath123 \nonumber\\ & & + \ , \beta_1\,\tilde g_{\rm nl}\,w(v_1-v_2 ) + ( 1-\beta_1)\ , g_{\rm nl } w(v_2 ) , \label{s11_final } \\ s_{12 } & = & -g_{\rm nl } \big [ \beta_1w(v_1 ) % \nonumber\\ & & + \,\beta_2 w(v_2)+(2-\beta_1-\beta_2)2t\big ] . \label{s12_final}\end{aligned}\ ] ] noise power for the second junction @xmath124 is defined by eq . ( [ s11_final ] ) with interchanged indices @xmath125 . here we have introduced the effective non - local conductance @xmath126 which , in contrast to @xmath109 , is suppressed by inelastic relaxation . one has @xmath127 if the distance between the junctions exceeds @xmath43 , and @xmath128 if @xmath129 . the first line of eq . ( [ s11_final ] ) just coincides with the standard expression for the shot noise of a mesoscopic conductor with the fano factor @xmath36 , while the second and third lines provide the corrections induced in the first junction by the second one . the origin of these corrections is simple : voltage bias applied to the second junction yields modifications in the electron distribution function in the vicinity of the first junction ( cf . ( [ distribution_1 ] ) ) thus changing its current noise . now we turn to the regime of a short wire , @xmath130 , where inelastic relaxation can be fully ignored . accordingly in eq . ( [ kinetic_0 ] ) we set @xmath131 and repeat the above calculation in this limit . as a result , the distribution function in the wire acquires the four step shape @xmath132 \nonumber\\ & & \times\ , \left [ \frac{r_2}{r}f_f(e - ev_3)+\frac{r_1}{r}f_f(e - ev_4 ) \right ] \nonumber\\ & & + \,\frac{\tilde{\cal d}_0(0,\bm{x},\bm{x}_1)}{2e^2\nu_0r_1}f_f(e - ew_1 ) \nonumber\\ & & + \,\frac{\tilde{\cal d}_0(0,\bm{x},\bm{x}_2)}{2e^2\nu_0r_2}f_f(e - ew_2 ) . \label{distribution_elastic}\end{aligned}\ ] ] this function is also illustrated in fig . [ fig_fofe ] . here we introduced the total resistance of the wire @xmath133 and assumed @xmath134 . the noise in the limit ( [ resistances ] ) , @xmath134 and @xmath135 becomes @xmath136 \nonumber\\ & & + \,\tilde g_{\rm nl } w(w_1-w_2 ) , \label{s11_elastic } \\ s_{12 } & = & -g_{\rm nl}\bigg[\frac{r_2}{r}w(w_1-v_3 ) + \frac{r_1}{r}w(w_1-v_4 ) \nonumber\\ & & + \,\frac{r_2}{r}w(w_2-v_3 ) + \frac{r_1}{r}w(w_2-v_4)\bigg ] . \label{s12_elastic}\end{aligned}\ ] ] comparing these expressions with eqs . ( [ s11_final ] ) , ( [ s12_final ] ) , we observe that they coincide either provided @xmath137 or in the large bias limit @xmath138 . otherwise , every function @xmath139 entering the result in the limit of strong relaxation splits up into two functions in the limit @xmath130 . until now we have ignored interaction effects and restricted our consideration to low frequency current fluctuations . below we will account for electron - electron interactions and evaluate the interaction correction to the conductance matrix of our system . extending the arguments @xcite , we will demonstrate a close relation between coulomb blockade of non - local electron transport and shot noise in the system under consideration . for this purpose it will be necessary to go beyond the low frequency limit and allow for arbitrary ( not necessarily slow ) fluctuations of voltages @xmath140 across the junctions . in this regime the time and energy dependent electron distribution function in the wire @xmath44 becomes ill - defined due to quantum mechanical uncertainty principle . this problem can be cured by employing the keldysh green function of electrons @xmath141 , \label{g_keldysh}\end{aligned}\ ] ] which fully describes electron dynamics at arbitrarily high frequencies . applying the fourier transformation ( [ g_keldysh ] ) to the kinetic equation ( [ kinetic_0 ] ) we cast it to the form @xcite @xmath142}}{\sinh\pi t(t_1-t_2 ) } \nonumber\\ & & -\ , \frac{\delta(\bm{x}-\bm{x}_1)}{2e^2\nu_0r_1}\left(g-\frac{-it e^{-i[\phi_1(t_1)-\phi_1(t_2)]}}{\sinh\pi t(t_1-t_2)}\right ) \nonumber\\ & & -\ , \frac{\delta(\bm{x}-\bm{x}_2)}{2e^2\nu_0r_2}\left(g-\frac{-it e^{-i[\phi_2(t_1)-\phi_2(t_2)]}}{\sinh\pi t(t_1-t_2)}\right ) \nonumber\\ & & -\ , \frac{\delta(\bm{x}-\bm{x}_1)}{e\nu_0}\eta_1(t_1,t_2 ) - \frac{\delta(\bm{x}-\bm{x}_2)}{e\nu_0}\eta_2(t_1,t_2 ) . \label{kinetic}\end{aligned}\ ] ] here the stochastic variables @xmath143 , which now also depend on two times , are correlated as follows @xmath144}}{\sinh\pi t(t_3-t_2 ) } \nonumber\\ & & + \ , \frac{-it e^{-i[\phi_j(t_1)-\phi_j(t_4)]}}{\sinh\pi t(t_1-t_4)}g(t_3,t_2 ) \bigg ) \nonumber\\ & & -\,(1-\beta_j)\bigg ( g(t_1,t_4)g(t_3,t_2 ) \nonumber\\ & & + \ , \frac{-it e^{-i[\phi_j(t_1)-\phi_j(t_4)]}}{\sinh\pi t(t_1-t_4)}\frac{-it e^{-i[\phi_j(t_3)-\phi_j(t_2)]}}{\sinh\pi t(t_3-t_2 ) } \bigg ) \bigg ] . \label{corr_eta}\end{aligned}\ ] ] in eqs . ( [ kinetic ] ) and ( [ corr_eta ] ) we defined the fluctuating phases of the leads @xmath145 as well as the phase @xmath146 where @xmath147 is the electric potential inside the wire which fluctuates both in time and in space and includes interaction effects . note that fully quantum mechanical description of interaction effects in metallic conductors generally involves two ( rather than one ) quantum fluctuating phase fields @xmath148 and @xmath149 ( defined on the two branches of the keldysh contour ) appearing after the standard hubbard - stratonovich decoupling of the coulomb term in the hamiltonian @xcite . provided interaction effects are sufficiently small ( as is the case here , see below ) one can effectively eliminate one of these fields , @xmath150 , and retain only the `` center - of - mass '' field @xmath151 . the derivation of the kinetic equation ( [ kinetic ] ) in the tunnel limit @xmath135 is presented in the appendix [ sec : model ] . rigorous derivation of the kinetic equation ( [ kinetic ] ) based on the non - linear @xmath152model as well as its applicability conditions can be found in ref . . now we turn to the expression for the current through the first junction @xmath69 . in order to derive this expression it is necessary to solve the kinetic equation ( [ kinetic ] ) . technical details of this procedure are presented in appendix [ sec : solution ] . here we directly proceed to the corresponding results . let us first consider the limit of strong inelastic relaxation , @xmath153 , and assume that the wire potential varies in space slowly enough , @xmath90 . in this case the current through the first junction acquires the form @xmath154\right ] \nonumber\\ & & -\ , \tilde g_{\rm nl}\frac{\beta_1}{e}\int_0^\infty dt \frac{\pi t^2}{\sinh^2\pi tt } k_{11}(t ) \sin[e(v_1-v_2)t ] \nonumber\\ & & + \ , \frac{\beta_1 } { e}\int dt'dt '' k_{12}(t''-t')\frac{{\cal d}(t',\bm{x}_1,\bm{x}_2)}{2e^2\nu_0r_1r_2 } \nonumber\\ & & \times\ , \frac{\pi t^2}{\sinh^2\pi tt''}\sin[(v_1-v_2)t '' ] \nonumber\\ & & -\,g_{\rm nl } \left [ v_2 - \frac{\beta_2 } { e}\int_0^\infty dt \frac{\pi t^2}{\sinh^2\pi tt } k_{22}(t)\sin[ev_2 t ] \right ] \nonumber\\ & & + \ , \frac{1-\beta_1}{e } \int dt'dt''k_{12}(t''-t ' ) \frac{{\cal d}(t',\bm{x}_1,\bm{x}_2)}{2e^2\nu_0r_1r_2 } \nonumber\\ & & \times\ , \frac{\pi t^2}{\sinh^2\pi tt''}\sin[ev_2 t '' ] . \label{current_final}\end{aligned}\ ] ] here we have defined the response functions @xmath155 which characterize the response of voltage fluctuations in the junction @xmath156 on the current noise of the junction @xmath73 . the corresponding impedance matrix @xmath157 is defined in appendix b , see eq . ( [ impedance ] ) . as before , here the voltage drops @xmath158 and @xmath159 are defined in eq . ( [ v1v2 ] ) . repeating now the same calculation in the elastic limit @xmath130 , we obtain @xmath160+\frac{r_1}{r}\sin[e(w_1-v_4)t]\right)\bigg ] \nonumber\\ & & -\ , g_{\rm nl } \bigg [ v_2 - \frac{\beta_2 } { e}\int_0^\infty dt \frac{\pi t^2}{\sinh^2\pi tt } k_{22}(t ) \nonumber\\ & & \times\ , \left(\frac{r_2}{r}\sin[e(w_2-v_3)t]+\frac{r_1}{r}\sin[e(w_2-v_4)t]\right ) \bigg ] \nonumber\\ & & -\ , \tilde g_{\rm nl}\frac{\beta_1}{e}\int_0^\infty dt \frac{\pi t^2}{\sinh^2\pi tt } k_{11}(t ) \sin[e(w_1-w_2)t ] \nonumber\\ & & + \,\frac{\beta_1 } { e}\int dt'dt '' k_{12}(t''-t ' ) \frac{\pi t^2}{\sinh^2\pi tt''}\frac{{\cal d}(t',\bm{x}_1,\bm{x}_2)}{2e^2\nu_0r_1r_2 } \nonumber\\ & & \times\ , \sin[(w_1-w_2)t '' ] \nonumber\\ & & + \ , \frac{1-\beta_1}{e } \int dt'dt''k_{12}(t''-t ' ) \frac{\pi t^2}{\sinh^2\pi tt '' } \frac{{\cal d}(t',\bm{x}_1,\bm{x}_2)}{2e^2\nu_0r_1r_2 } \nonumber\\ & & \times\ , \left ( \frac{r_2}{r}\sin[e(w_2-v_3)t'']+\frac{r_1}{r}\sin[e(w_2-v_4)t '' ] \right ) , \label{current_elastic}\end{aligned}\ ] ] where the lead potentials @xmath161 are defined in eq . ( [ w1w2 ] ) . ( [ current_final ] ) , ( [ current_elastic ] ) represent the central results of this paper which fully determines the leading coulomb corrections to the conductance matrix of our structure in both relevant limits of strong and weak inelastic relaxation . these results also allow to demonstrate a close relation between shot noise and interaction effects , which is now extended to include non - local electron transport . for example , the first line of eq . ( [ current_final ] ) describes the standard `` local '' coulomb anomaly caused by charging effects and related to local shot noise @xcite . the next three lines in eq . ( [ current_final ] ) contain terms depending on the voltage difference @xmath162 and describing non - local effects . their origin can be traced back to the corresponding contribution to the shot noise in the first junction , cf . the second line in eq . ( [ s11_final ] ) . finally , the contribution in the last three lines in eq . ( [ current_final ] ) depends only on the voltage @xmath159 and emerges from the last term of eq . ( [ current_dc ] ) @xmath163 . in the same way one can establish the correspondence between various terms in the expressions for the current ( [ current_elastic ] ) and noise ( [ s11_elastic ] ) , ( [ s12_elastic ] ) in the elastic limit . perhaps we should also add that the above results remain applicable to a much broader class of systems than that depicted in fig . [ fig_two_junctions ] . e.g. , the wire may be replaced by a metallic lead of any shape , and ultimately all geometry specific details can be absorbed in few elements of the conductance matrix . it is interesting to compare the results ( [ current_final],[current_elastic ] ) with the predictions of the @xmath164theory ( [ pofe ] ) . employing the usual definition of the @xmath7 function@xcite @xmath165 \frac{[\cos\omega t-1]\coth\frac{\omega}{2t}+i\sin\omega t}{\omega } , \nonumber \end{aligned}\ ] ] and combining it with the solution of the kinetic equation ( [ kinetic_0 ] ) one can evaluate the current ( [ pofe ] ) in the limit of low resistances of the leads @xmath166 . comparing the result with the eqs . ( [ current_final],[current_elastic ] ) in the tunnel limit @xmath135 , one observes that the @xmath7 approach reproduces the contributions containing the local response functions @xmath167 , while the corrections @xmath168 are missing . one can further verify that the latter corrections originate from the cross - correlation of the junction shot noises which are ignored in the formula ( [ pofe2 ] ) . ( [ di1dv1 ] ) and ( [ di1dv2 ] ) respectively . the system parameters are : @xmath99 mk , @xmath169 ns , @xmath170 , @xmath171 , @xmath135 , @xmath172 ms , @xmath173 ms . the curves at @xmath174 in the top panel and at @xmath175 in the bottom panel are shown in real scale , other curves are shifted vertically for clarity . local differential conductance @xmath176 exhibits a small dip at @xmath177 . non - local conductance @xmath178 shows a much more pronounced peak at @xmath179 . , width=264 ] in order to further specify our results it is necessary to make certain assumptions about the form of the kernels @xmath180 . for typical experimental setups and at sufficiently low voltages and temperature it is reasonable to adopt the following approximation for the elements of the admittance matrix of the environment @xmath181 ( see eq . ( [ admittance ] ) for their precise definition ) : @xmath182 , @xmath183 , @xmath184 , where @xmath185 and @xmath186 are effective shunt resistances . these resistances can roughly be estimated as @xmath187 in practice , the shunt resistances may deviate from these simple estimates due to impedance dispersion in metallic wires at high frequencies @xcite . further assuming that @xmath109 is small as compared to @xmath188 one finds @xmath189 and @xmath190 where @xmath191 is the charge relaxation time which for simplicity is taken equal for both junctions . this simplification is by no means restrictive since in our final result @xmath192 appears only under the logarithm as an effective cutoff parameter . under these conditions the current in the limit @xmath153 ( [ current_final ] ) can be evaluated analytically and takes the form @xmath193 % \nonumber\\ & & -g_{\rm nl}\left [ v_2 - \frac{4\pi \beta_2 t}{eg_2 } f_i(v_2)\right ] \nonumber\\ & & -\ , \tilde g_{\rm nl } \frac{4\pi\beta_1 t}{eg_1 } f_i(v_1-v_2),\end{aligned}\ ] ] where we defined the dimensionless conductances of the environment @xmath194 , @xmath195 and the dimensionless function @xmath196.\end{aligned}\ ] ] here @xmath197 stands for the digamma function . both local and non - local differential conductances read @xmath198 % \nonumber\\ & & - \tilde g_{\rm nl } \frac{2\beta_1}{g_1 } f(v_1-v_2 ) , \label{di1dv1 } \\ \frac{\partial i_1}{\partial v_2 } = -g_{\rm nl}\left[1-\frac{2\beta_2}{g_2 } f(v_2)\right ] % \nonumber\\ & & + \tilde g_{\rm nl } \frac{2\beta_1}{g_1 } f(v_1-v_2 ) , \label{di1dv2}\end{aligned}\ ] ] where we introduced another function @xmath199 . \label{f}\end{aligned}\ ] ] , eqs . ( [ di1dw1 ] ) and ( [ di1dw2 ] ) respectively . the system parameters are the same as in fig . the voltage values are : @xmath200 @xmath26v , @xmath201 @xmath26v , @xmath202 @xmath26v in panel ( a ) and @xmath203 @xmath26v in panel ( b ) . , width=264 ] in the elastic limit @xmath130 we find @xmath204 \nonumber\\ -\,g_{\rm nl}\bigg [ v_2 - \frac{4\pi \beta_2 t } { eg_2}\left ( \frac{r_2}{r}f_i(w_2-v_3)+\frac{r_1}{r}f_i(w_2-v_4 ) \right ) \bigg ] \nonumber\\ -\ , \tilde g_{\rm nl } \frac{4\pi\beta_1 t}{eg_1 } f_i(w_1-w_2).\hspace{4.3 cm } \nonumber\end{aligned}\ ] ] accordingly , local and non - local differential conductances acquire the form @xmath205 \nonumber\\ & & - \tilde g_{\rm nl } \frac{2\beta_1}{g_1 } f(v_1-v_2 ) , \label{di1dw1 } \\ \frac{\partial i_1}{\partial w_2 } & = & -g_{\rm nl}\left[1-\frac{2\beta_2}{g_2 } \left(\frac{r_2}{r}f(w_2-v_3)+\frac{r_1}{r}f(w_2-v_4)\right ) \right ] \nonumber\\ & & + \tilde g_{\rm nl } \frac{2\beta_1}{g_1 } f(w_1-w_2 ) . \label{di1dw2}\end{aligned}\ ] ] local differential conductance ( [ di1dv1 ] ) of a long wire with @xmath153 is plotted in fig . [ fig_didv1]a . for a chosen set of parameters it is weakly affected by the second junction , although a small dip at @xmath177 is observed . in contrast , non - local differential conductance @xmath178 is very sensitive to @xmath158 and has two peaks centered , respectively , at @xmath206 and @xmath179 , see fig . [ fig_didv1]b . [ fig_didv2]a shows local differential conductance of a short wire , @xmath130 , in which the electron distribution function does not relax . we observe that the conductance @xmath207 given by eq . ( [ di1dw1 ] ) has three dips centered respectively at @xmath208 . likewise , non - local conductance @xmath209 defined in eq . ( [ di1dw2 ] ) shows peaks at @xmath210 , see fig . [ fig_didv2]b . comparing figs . [ fig_didv1 ] and [ fig_didv2 ] we observe that the dip in @xmath176 ( the peak in @xmath178 ) occuring for strong inelastic electron relaxation splits into two dips ( peaks ) in the weak relaxation limit . let us briefly summarize our key observations . we have demonstrated that coulomb blockade corrections to both local and non - local conductances in metallic conductors may change significantly provided the electron distribution function in at least one of the leads is driven out of equilibrium . provided the conductor length @xmath13 is shorter than the inelastic relaxation length @xmath43 , at low temperatures and under non - zero voltage bias the electron distribution function acquires a characteristic double step form and the coulomb dip in the differential conductance splits into two dips . this effect disappears provided inelastic relaxation becomes strong @xmath211 . if two leads are attached to a metallic wire as it is shown in fig . 1 , the electron distribution function in the vicinity of one junction may also be driven out of equilibrium provided electrons are injected through the second junction and do not relax their energies at distances shorter than the distance between these two junctions . in this situation additional coulomb dip in the differential conductance appears . the latter configuration with two junctions also allows to study coulomb blockade of non - local electron transport in the presence of non - equilibrium . it turns out that in this case an interplay between coulomb and non - equilibrium effects yields more pronounced peaks in the non - local differential conductance , see figs . [ fig_didv1]b , [ fig_didv2]b . this observation indicates that experiments on non - local electron transport in the presence of coulomb effects can be conveniently used to test inelastic electron relaxation in metallic conductors at low temperatures , as it has already been demonstrated , e.g. , in experiments @xcite . the analysis developed here applies in the weak coulomb blockade regime implying that either the resistances of metallic leads should be much smaller than the quantum resistance unit @xmath12 , or the temperature should exceed charging energies of the barriers . in this regime there exists a transparent relation between shot noise and interaction effects in the electron transport @xcite . here we extended this fundamental relation to the non - local case , demonstrating that negative cross - correlations in shot noise are directly linked to coulomb suppression of non - local conductance . this is in contrast to nsn structures where coulomb anti - blockade of non - local conductance may occur being related to positive cross - correlations in shot noise induced by crossed andreev reflection . let us briefly discuss the main steps of our derivation of the kinetic equation ( [ kinetic ] ) . we start by defining the electron keldysh green function @xmath212 this function obeys the equation @xmath213 here we have defined the four - dimensional vectors @xmath214 . below we will stick to the diffusive limit in which case the electron distribution function remains isotropic . then applying the standard quasiclassical technique @xcite , we equalize the coordinates , @xmath215 , and make the replacement @xmath216 we further note that the operators @xmath217 and @xmath218 in the vicinity of the barriers are not independent . they are related to each other via the scattering matrices of the barrier . consider for simplicity the tunneling limit @xmath219 in which case the corresponding transmission amplitudes in eq . ( [ ttr ] ) read @xmath220 then we obtain @xmath221 here the superscript @xmath222 labels incoming waves unaffected by the barriers . in the tunneling limit considered here it suffices to identify the `` incoming '' operators with the full ones . substituting the above expressions into eq . ( [ kinetic_gf ] ) , performing the replacement ( [ repl ] ) and setting @xmath223 , after adding the phenomenological term describing inelastic relaxation we arrive at eq . ( [ kinetic ] ) for the function @xmath224 without noise terms . the pre - factors in front of the terms on the right hand side of eq . ( [ kinetic ] ) are fixed by the requirement that in the absence of interactions the currents across the barriers have the standard ohmic form @xmath225 . the noise terms may be derived if one employs eq . ( [ kinetic_gf ] ) for non - averaged operator green function @xmath226 the noise operator @xmath227 is then defined as follows @xmath228 evaluating the symmetrized correlator of two such operators , @xmath229 one can verify that it coincides with the correlator ( [ corr_eta ] ) in the tunneling limit @xmath230 . the pre - factors in front of the noise terms in the eq . ( [ kinetic ] ) are again determined by comparison with the noises of the junctions in the known non - interacting limit . the noise variable @xmath231 is defined analogously . the operators @xmath232 may be treated as classical fluctuating functions in the spirit of the @xmath152model and path integral formulation @xcite . we finally note that the kinetic equation ( [ kinetic ] ) can also be derived beyond the tunneling limit , i.e. for @xmath233 . however , in this general case the corresponding analysis turns rather complicated since the full scattering matrices of the barriers should be employed in eq . ( [ relation ] ) . without going into such complicated algebra here we refer the reader to ref . where a general and rigorous derivation of the kinetic equation ( [ kinetic ] ) has been carried out . in order to solve eq . ( [ kinetic ] ) we make use of the same procedure as in sec . [ sec : noise ] . with the aid of eq . ( [ g_keldysh ] ) the expression for the current ( [ current_0 ] ) can be rewritten as @xmath234}}{\sinh\pi t(t - t ' ) } \right ) \nonumber\\ & & + \ , c_j\dot v_j+2\pi\eta_j(t , t ) , \label{current_1}\end{aligned}\ ] ] where we added displacement currents recharging the capacitors @xmath235 . the solution of eq . ( [ kinetic ] ) takes the form @xmath236}}{\sinh\pi t(t_1-t_2 ) } \int dt ' d^3\bm{x}'\ , \nonumber\\ & & \frac{{\cal d}\left(\frac{t_1+t_2}{2}-t',\bm{x},\bm{x}'\right)}{\tau_{\rm in } } \ , \nonumber\\ & & + \ , \frac{-it e^{-i\left[\phi(t_1,\bm{x})-\phi(t_2,\bm{x})\right]}}{\sinh \pi t(t_1-t_2)}\int dt ' \nonumber\\ & & \times\ , \left[\frac{{\cal d}\left(\frac{t_1+t_2}{2}-t',\bm{x},\bm{x}_1\right)}{2e^2\nu_0r_1 } e^{-i\left[\varphi_1\left(t'+\frac{t_1-t_2}{2}\right)-\varphi_1\left(t'-\frac{t_1-t_2}{2}\right)\right ] } \right . \nonumber\\ & & \left . + \ , \frac{{\cal d}\left(\frac{t_1+t_2}{2}-t',\bm{x},\bm{x}_2\right)}{2e^2\nu_0r_2 } e^{-i\left[\varphi_2\left(t'+\frac{t_1-t_2}{2}\right)-\varphi_2\left(t'-\frac{t_1-t_2}{2}\right)\right ] } \right ] \nonumber\\ & & -\ , e^{-i\left[\phi(t_1,\bm{x})-\phi(t_2,\bm{x})\right ] } \int dt ' \nonumber\\ & & \times\ , \bigg [ \frac{{\cal d}\left(\frac{t_1+t_2}{2}-t',\bm{x},\bm{x}_1\right)}{e\nu_0 } e^{i\left[\phi\left(t'+\frac{t_1-t_2}{2},\bm{x}_1\right)-\phi\left(t'-\frac{t_1-t_2}{2},\bm{x}_1\right)\right ] } \nonumber\\ & & \times\ , \eta_1\left(t'+\frac{t_1-t_2}{2},t'-\frac{t_1-t_2}{2}\right ) \nonumber\\ & & + \,\frac{{\cal d}\left(\frac{t_1+t_2}{2}-t',\bm{x},\bm{x}_2\right)}{e\nu_0 } e^{i\left[\phi\left(t'+\frac{t_1-t_2}{2},\bm{x}_2\right)-\phi\left(t'-\frac{t_1-t_2}{2},\bm{x}_2\right)\right ] } \nonumber\\ & & \times\ , \eta_2\left(t'+\frac{t_1-t_2}{2},t'-\frac{t_1-t_2}{2}\right ) \bigg ] , \label{g0}\end{aligned}\ ] ] where we defined @xmath237 . here we have already assumed that inelastic relaxation is strong , @xmath130 , and that the the wire potential varies in space slowly enough , @xmath90 . combining eqs . ( [ current_1 ] ) and ( [ g0 ] ) we evaluate the instantaneous current value in the first junction @xmath238 + c_1\dot v_1+\delta i_1(t ) , \label{current_2}\end{aligned}\ ] ] where the noise term @xmath239 is defined in eq . ( [ xi1 ] ) with the following replacement @xmath240 . averaging the expression for the current ( [ current_2 ] ) , ( [ xi1 ] ) over time we arrive at the following current - voltage characteristics @xmath241 it is important to emphasize that here the average values @xmath242 differ from zero due to the presence of fluctuating phases which account for interaction effects . in order to evaluate these averages it is convenient to split the time - dependent phases into regular and fluctuating parts , @xmath243 where the potentials @xmath244 are defined in eq . ( [ v1v2 ] ) . in what follows we will assume that interaction effects remain sufficiently weak , which is the case provided either the resistances of metallic wires are much smaller than the quantum resistance unit , @xmath245 , or the temperature is sufficiently high , @xmath246 . in either case phase fluctuations remain small , @xmath247 , and the average @xmath248 can be expressed in the form @xmath249 note that fluctuating phases @xmath250 , in turn , depend on the stochastic variables @xmath251 . in order to establish this dependence we will make use of fourier transformed eq . ( [ current_1 ] ) which yields @xmath252 here we introduced the fourier transform of the fluctuating currents @xmath253 and used the relation @xmath254 . the conductances @xmath255 , @xmath256 and @xmath257 are again defined in eqs . ( [ gij ] ) where one should now substitute @xmath258 , i.e. these conductances are expressed via fourier transformed diffusons at a frequency @xmath259 . from the equivalent circuit of fig . 1b we can also define the fluctuating currents @xmath260 where @xmath181 is the admittance matrix of our structure . the off - diagonal elements @xmath261 are responsible for cross - correlations between the junctions , which may be caused , e.g. , by capacitive coupling between the leads 1 and 2 . excluding the currents @xmath262 from the above equations we obtain @xmath263 where the kernels @xmath180 read @xmath155 with @xmath157 being an effective impedance matrix @xmath264 and @xmath265 combining eqs . ( [ av1 ] ) and ( [ 31 ] ) , we obtain @xmath266 due to causality the variable @xmath267 can only depend on the phases @xmath268 taken at earlier times ( i.e. at @xmath269 ) , while the function @xmath270 differs from zero only for @xmath271 . hence , the variable @xmath272 is independent of @xmath268 , and eq . ( [ av2 ] ) can be rewritten in the form @xmath273 here the correlator @xmath274 is defined in eq . ( [ corr_eta ] ) with the function @xmath275 set by eq . ( [ g0 ] ) with omitted noise terms , i.e. with @xmath276 . the average value @xmath277 is derived in exactly the same manner . golubev and a.d . zaikin , phys . lett . * 86 * , 4887 ( 2001 ) ; phys . b * 69 * , 075318 ( 2004 ) . a. levy yeyati , a. martin - rodero , d. esteve , and c. urbina , phys . 87 * , 046802 ( 2001 ) . c. altimiras , u. gennser , a. cavanna , d. mailly , and f. pierre , phys . * 99 * , 256805 ( 2007 ) . a.v . galaktionov and a.d . zaikin , phys . b * 80 * , 174527 ( 2009 ) . bollinger , a. rogachev , and a. bezryadin , europhys . * 76 * , 505 ( 2006 ) . d. beckmann , h.b . weber , and h. v. lhneysen , phys . * 93 * , 197003 ( 2004 ) ; d. beckmann and h. v. lhneysen , appl . phys . a * 89 * , 603 ( 2007 ) . s. russo , m. kroug , t.m . klapwijk , and a.f . morpurgo , phys . . lett . * 95 * , 027002 ( 2005 ) . p. cadden - zimansky and v. chandrasekhar , phys . lett . * 97 * , 237003 ( 2006 ) ; p. cadden - zimansky , z. jiang , and v. chandrasekhar , new j. phys . * 9 * , 116 ( 2007 ) . a. kleine , a. baumgartner , j. trbovic , c. schnenberger , europhys . lett . * 87 * , 27011 ( 2009 ) ; a. kleine , a. baumgartner , j. trbovic , d.s . golubev , a.d . zaikin , and c. schnenberger , nanotechnology * 21 * , 274002 ( 2010 ) . g. falci , d. feinberg , and f.w.j . hekking , europhys . lett . * 54 * , 255 ( 2001 ) ; m.s . kalenkov and a.d . zaikin , phys . b * 75 * , 172503 ( 2007 ) ; d.s . golubev , m.s . kalenkov , and a.d . zaikin , phys . lett . * 103 * , 067006 ( 2009 ) and further references therein . a. levy yeyati , f.s . bergeret , a. martin - rodero , and t.m . klawijk , nat . phys . * 3 * , 455 ( 2007 ) . d.s . golubev and a.d . zaikin , phys . b * 82 * , 134508 ( 2010 ) . coulomb anti - blockade of local electron transport in superconducting quantum point contacts was also discussed in a different context by a. levy yeyati , j.c . cuevas , and a. martin - rodero , phys . lett . * 95 * , 056804 ( 2005 ) . h. pothier , s. gueron , n.o . birge , d. esteve , and m.h . devoret , phys . * 79 * , 3490 ( 1997 ) . g. schn and a.d . zaikin , phys . rep . * 198 * , 237 ( 1990 ) . ingold and yu.v . nazarov ( 1992 ) , in _ single charge tunneling _ , edited by h. grabert and m. h. devoret , nato asi , ser . b , vol . 294 , pp . 21 - 107 ( plenum , new york , 1992 ) . s.v.panyukov and a.d.zaikin . j. low temp . phys . * 73 * , 1 ( 1988 ) . a. anthore , f. pierre , h. pothier , and d. esteve , phys . rev . lett . * 90 * , 076806 ( 2003 ) . b. huard , a. anthore , n.o . birge , h. pothier , and d. esteve , phys . lett . * 95 * , 036802 ( 2005 ) ; b. huard , a. anthore , f. pierre , h. pothier , n.o . birge , d. esteve , solid state comm . * 131 * , 599 ( 2004 ) . sukhorukov and d. loss , phys . lett . * 80 * , 4959 ( 1998 ) ; phys . b * 59 * , 13054 ( 1999 ) . ya.m . blanter and m. bttiker , phys . rep . * 336 * , 1 ( 2000 ) . d.s . golubev and a.d . zaikin , phys . b * 59 * , 9195 ( 1999 ) ; physica b * 255 * , 164 ( 1998 ) . d.s . golubev and a.d . zaikin , phys . b * 70 * , 165423 ( 2004 ) . p. wahlgren , p. delsing , t. claeson , and d.b . haviland phys . b * 57 * , 2375 ( 1998 ) ; j.s . penttil , . parts , p. j. hakonen , m.a . paalanen , and e.b . sonin , phys . b * 61 * , 10890 ( 2000 ) . j. rammer and h. smith , rev . phys . * 58 * , 323 ( 1986 ) . a. kamenev and a. andreev , phys . b * 60 * , 2218 ( 1999 ) .

we present interplanetary network localization information for 218 gamma - ray bursts in the 3rd batse catalog , obtained by analyzing the arrival times of these bursts at the _ ulysses and _ compton gamma - ray observatory ( cgro ) spacecraft . for any given burst observed by these two spacecraft , arrival time analysis ( or `` triangulation '' ) results in an annulus of possible arrival directions whose half - width varies between 7 arcseconds and 32 arcminutes , depending on the intensity and time history of the burst , and the distance of the _ ulysses spacecraft from earth . this annulus generally intersects the batse error circle , resulting in an average reduction of the error box area of a factor of 30 . _ _ _

astro-ph9811473

a knowledge of the precise positions of cosmic gamma ray bursts ( grbs ) is important for many studies . for example , recent observations of the optical counterparts to bursts have provided evidence that at least two of them are at cosmological distances ( metzger et al . 1997 , kulkarni et al . 1998 ) and that , by implication , most or all of them may also be . one optical counterpart has been associated with a galaxy , and the majority of the models for the energy release in bursts have long invoked neutron star - neutron star or neutron star- black hole mergers ( e.g. paczynski 1991 ) . if this is correct , grbs should originate in or close to galaxies , and correlations between burst positions and the large scale structure of the universe might be expected . in this context , it has been debated whether the positions of bursts are correlated with those of abell clusters ( kolatt & piran 1996 ; hurley et al . 1997a ; struble & rood 1997 ) and radio - quiet quasars ( hurley et al . 1997b ; schartel , andernach , and greiner , 1997 ) . in both cases , increased precision in the location accuracy leads to more stringent tests . similarly , precise positions greatly improve searches for quiescent counterparts to grbs in other wavelength ranges ( e.g. tokanai et al . 1997 ) , and can help confirm proposed associations between grbs and fading counterparts ( hurley et al . finally , burst recurrence from a single source is excluded in most cosmological models . such recurrence would reveal itself through small - scale clustering in the angular distribution of burst locations , and the sensitivity of the distribution function to repeating sources is greater when the burst location accuracy is improved ( kippen , hurley , & pendleton 1998 ) . we present here a catalog of 218 gamma ray bursts from the detector aboard the _ ulysses spacecraft , which were also observed by the burst and transient source experiment ( batse ) , and cataloged in the 3rd batse catalog ( meegan et al . 1996 ) . in the present catalog , their positions have been improved in accuracy by arrival time analysis , or `` triangulation '' . detection of an event by these two widely separated instruments leads to a narrow annulus which intersects the batse error circle and reduces its area by a factor of @xmath0 30 . such annulus / error circle intersections have played important roles in the examples cited above . _ the _ ulysses grb detector ( hurley et al . 1992 ) is one instrument in the 3rd interplanetary network of burst detectors . it consists of two 3 mm thick hemispherical csi scintillators with a projected area of about 20 @xmath1 in any direction . the detector is mounted on a magnetometer boom far from the body of the spacecraft , and therefore has a practically unobstructed view of the full sky . because the _ ulysses mission is in interplanetary space , the instrument also benefits from an exceptionally stable background . the grb detector operates continuously , and over 97% of the data are recovered . _ _ a unique feature of the mission is the fact that it is in heliocentric orbit with aphelion @xmath0 5 au . thus it reaches an apogee of @xmath0 6 au or 3000 light - seconds , and during portions of the orbit , the grb experiment is farther than any burst detector has ever been from earth , resulting in greatly improved location accuracies for gamma - ray bursts . the grb instrument observes bursts in both a triggered and a real - time mode . the triggered mode is initiated when the number of counts in one of two possible time intervals exceeds a preset value . because the background is stable , this number is defined in terms of an absolute number of counts , rather than a number of standard deviations above the background . the two time intervals may be selected among values from 0.125 to 4 s. each of the two time intervals is associated with an independently selectable energy window , whose lower and upper thresholds lie between @xmath0 15 and 150 kev . the data collected in the triggered mode consist of energy spectra , and either 16 s of 8 ms count rates , or 64 s of 32 ms count rates , in the @xmath0 25 to 150 kev energy range ; @xmath0 1.8 and 7.4 s of the data stream record the time histories prior to trigger , respectively . following a trigger , data are read out for @xmath0 40 m , during which time the experiment can not re - trigger . an example is shown in figure 1 . independent of the trigger , a continuous stream of real time data is transmitted . this consists of count rates taken with 0.25 , 0.5 , 1 , or 2 s time resolution . the resolution depends on the spacecraft telemetry rate , and the energy range is @xmath0 25 to 150 kev , as for the trigger . data are compressed in this mode . an example is shown in figure 2 . the real time data serve many purposes . for example , numerous gamma - ray bursts which are not intense enough to trigger the detector can be reliably identified in the real time data , as we demonstrate below . bursts arriving during trigger readout may also be identified . finally , bursts whose duration exceeds the duration of the triggered memory can be fully recorded in this mode . batse consists of eight detector modules situated at the corners of the _ compton gamma - ray observatory spacecraft . each contains a large area detector ( lad ) , a 50.8 cm diameter by 1.27 cm thick nai scintillator . the trigger algorithm examines the count rates of two or more detectors in the 50 - 300 kev energy range , over 64 , 256 , and 1024 ms intervals . the experiment is described in more detail in meegan et al . ( 1996 ) . for most of the events in this catalog , we have utilized the discsc and preb data types summed over two , three , or four detectors . this gives a 0.064 s resolution time history from 2 seconds prior to the trigger to 240 s or more after the trigger . we select discriminator channels 1 and 2 , corresponding to the energy range 25 - 100 kev , to match the _ ulysses energy range as closely as possible . rarely , this data type is not available ; in that case we use mer , tts , tte , or 1 s resolution data . as batse is in low earth orbit , the time history data are occasionally interrupted by data gaps and earth occultation . _ _ every cosmic burst detected by batse is systematically searched for in the _ ulysses data as soon as the batse data are available for it . this is done by using the approximate arrival direction from batse and the position of the _ ulysses spacecraft to calculate a range of possible arrival times at _ ulysses . the calculation assumes generous uncertainties in the arrival direction to assure that the crossing window always includes the burst . typical window lengths are 300 - 500 s ( e.g. , figure 2 ) . in cases where the batse positions undergo substantial revision , a revised search window is calculated and the search is redone . _ _ _ to carry out the search , the _ ulysses real time count rates are extracted for the crossing window and plotted . the data are searched both automatically and by eye . in the automatic procedure , the background count rate is typically calculated over ten time intervals , and a @xmath2 increase is searched for over 2 or more consecutive time intervals . the search by eye confirms the presence of the increase , and in some cases , identifies increases that were not detected in the automatic procedure due , for example , to a long rise time . if the burst is intense enough to trigger the _ ulysses grb detector , burst detection is usually immediately apparent in both the real time and the triggered data streams , with one noteworthy exception . very short bursts ( durations @xmath3 0.25 s ) can trigger , but remain difficult to identify in the lower time resolution real time data . conversely , long bursts which do not reach large peak intensities can be identified in the real time data , but may not trigger regardless of their time - integrated fluxes . _ _ when a grb arrives at two spacecraft with a delay @xmath4 t , it may be localized to an annulus whose half - angle @xmath5 with respect to the vector joining the two spacecraft is given by cos@xmath5=c@xmath4t / d , where c is the speed of light and d is the distance between the two spacecraft . ( this assumes that the burst is a plane wave , i.e. that its distance is much greater than d. ) the annulus width d@xmath5 , and thus one dimension of the resulting error box , is c @xmath6t)/dsin@xmath5 where @xmath6 t ) is the uncertainty in the time delay . this term has two components , one statistical and one systematic . the first is associated with the uncertainty in comparing the time histories of a burst recorded by two different instruments ; the second is due to the uncertainty in the clock calibrations of the two spacecraft . a third source of error is the uncertainty in the knowledge of the spacecraft positions . we discuss each in turn . in the simple case of identical detectors with identical backgrounds , the most probable delay @xmath4 t and its error @xmath6 t ) may be estimated as follows . denote the two background - subtracted time histories by x@xmath7 , y@xmath7 , i=1, ... m , that is , x@xmath8 is the number of counts in the first time bin for detector x , and y@xmath8 the number of counts in the first time bin for detector y , and so on . assume that the background can be estimated with arbitrary accuracy , and further that each value of x@xmath7 and y@xmath7 is an independent normal random variable with variance x@xmath7 and y@xmath7 respectively . for simplicity of notation , assume that the two time histories recorded by detectors x and y are perfectly aligned when time bin 1 of detector x corresponds to time bin 1 of detector y. then for this perfect alignment , corresponding to zero lag between the two time histories , the quantity @xmath9 is a random variable with mean zero and variance unity for every i. thus @xmath10 is distributed as @xmath11 with n degrees of freedom . here the sum is performed over some number n of time bins , where n @xmath12 m. now allow the lag to vary slightly about zero , so that time bin i of detector x corresponds to time bin j of detector y , and form the random variable @xmath13 where the lag ij is now variable . the quantity @xmath14 where the summation is over the intervals which x and y have in common , is similarly distributed as @xmath11 . in practice we do not know _ a priori the lag at which the best correlation will occur , so we test various lags ij until r@xmath15 reaches a minimum , r@xmath16 . then the lag or time delay corresponding to a 3@xmath17 equivalent confidence level is r@xmath15=r@xmath16 + 9 . ( all annuli in this catalog have widths corresponding to 3@xmath17 equivalent confidence . ) _ although this simple example describes the essential features of the method , there are numerous complicating factors in practice . for batse , the background varies over the orbit , and its value may not always be known accurately ; the assumption of a normal distribution may also be violated when cyg x-1 is in the field of view . the batse and _ ulysses detectors have different areas , burst and background count rates , and time resolutions . we treat this by first , normalizing the batse background- subtracted time history to the _ ulysses one so that two time histories have the same total number of counts . we also generate a batse time history with identical time resolution to the _ ulysses one by scaling the counts in the original batse time bins . these adjustments complicate the above expressions considerably . however , through monte - carlo simulations , as well as time history comparisons using a completely different method ( laros et al . 1998 ) , we have verified that the procedure gives correct results . in addition to calculating the r@xmath15 statistic , we also compute the correlation coefficient as a function of lag , namely @xmath18 where the summations are over the intervals that x and y have in common , and find its maximum value . the purpose of this calculation is to assure that the r@xmath15 statistic has indeed identified the best lag ; we require that the maximum @xmath19 and the minimum r@xmath15 correspond to identical lags , or at worst , that the difference between the lags be much less than the 3@xmath17 confidence interval . finally , parabolic fits are done to both statistics to identify the best lag and the confidence intervals . an example is shown in figures 3 and 4 . _ _ _ the second uncertainty , due to the clock calibrations , is in principle much smaller than the 3@xmath17 confidence interval . the clock on the _ compton gamma - ray observatory is accurate to 100 @xmath20s , and this accuracy is verified through pulsar timing . onboard software increases the uncertainty in the batse trigger times to @xmath0 1 ms . the clock on the _ ulysses spacecraft is calibrated during each daily tracking pass , and the clock drift is extrapolated until the next pass to obtain the correct time ; the calibration is verified at six month intervals by sending commands to the grb experiment at precisely known times . the exact procedure is described in hurley ( 1994 ) . due to spacecraft buffering of the commands , the result is that _ if it can be assumed that no errors larger than 125 ms are present , _ then the clock can be shown to be accurate to several milliseconds . for the purposes of triangulation , however , we take the extremely conservative approach that no 3 @xmath17 cross - correlation uncertainty is less than 125 ms . larger timing errors have been identified in the past for both missions . the _ cgro clock was found to have a consistent 2 s offset in the early part of the mission , and random timing errors between 0.25 and several hundred seconds occurred in certain types of _ ulysses data . in all cases the data were reprocessed to obtain the results in this paper . _ _ _ _ _ _ finally , the uncertainties in the spacecraft ephemerides are truly negligible . the _ ulysses range is known to @xmath0 10 km , and the right ascension and declination of the spacecraft are accurate to @xmath0 0.00003 degrees . the gro range is known to better than 50 km , and the right ascension and declination to better than 0.01 degrees . the error that these uncertainties produce in the knowledge of the interspacecraft vector ( i.e. , the center of the annulus ) is of order 0.0005 degrees , and therefore usually two orders of magnitude smaller than the timing uncertainties . as a conservative measure , this uncertainty is added linearly to the uncertainties in the annulus parameters caused by the timing . it typically increases the annulus width by several tenths of an arcsecond . _ the radius of each annulus and its right ascension and declination are transformed to a heliocentric frame . this is equivalent to an aberration correction ( maximum value @xmath0 20 arcseconds ) , and involves two adjustments . the first is to the spacecraft time . the signal from a distant spacecraft may travel several thousand seconds on its way to earth . depending on the earth s velocity vector relative to the spacecraft , the relative motion of the earth and the spacecraft may not be negligible during this period . `` geocentric '' clock times are related to the position of the earth when the signal is _ received from the spacecraft . `` heliocentric '' clock times , i.e. , times related to the position of the earth when the signal was _ transmitted , must be used . the difference between the two may be as large as several hundred milliseconds , and is often comparable to or greater than the statistical uncertainty of the time history comparison . the second adjustment is to the inter - spacecraft vector . two observers moving with the earth and with the sun see different vectors ( i.e. , annulus centers ) as a result of their motions during the burst s transit time between the two spacecraft . this correction , too , may not be negligible . the result of triangulating the 1993 january 31 burst is shown in figure 5 . _ _ although there have never been more than three widely separated spacecraft in the 3rd ipn , which would have provided redundant determinations and therefore a verification of burst positions , there have been numerous instances where a burst source with a known position was triangulated with _ ulysses , batse , and other spacecraft . these include the soft gamma repeater sgr1806 - 20 and the bursting pulsar gro j1744 - 28 ( hurley et al . 1996 , 1998a ) , the 1997 january 11 burst localized by bepposax ( galama et al . 1997 ) , the 1997 february 28 burst , with both a bepposax location and an optical transient ( hurley et al . 1997c ) , and numerous other bepposax bursts ( e.g. hurley et al . in each case the triangulated position was in good agreement with the known position of the source . _ we have used several selection criteria for the bursts in this catalog . the first is that the burst must have been detected by _ ulysses and batse . because of the very different detector sizes , all the bursts satisfying this criterion are observed to be rather intense by batse , and therefore detection by batse is unambiguous . however , the weaker events ( as observed by _ ulysses in the real time mode ) may be difficult to identify . in these cases , we rely on several supplementary semi - empirical criteria . one is that the r@xmath16 statistic , divided by the number of degrees of freedom , be less than unity . another is that the maximum value of the correlation coefficient @xmath21 be greater than 0.5 . yet another is that the ratio of _ ulysses counts to batse counts lie in the range 0.019 @xmath22 0.009 . it might seem unusual that the average value of this ratio is 0.019 , whereas the ratio of the the _ ulysses projected area in any direction to a single batse lad area is closer to 0.01 , and the batse counts are derived from at least two lad s in this procedure . several factors account for this . first , the _ ulysses energy range is wider than batse s : 25 - 150 kev _ vs. 25 - 100 kev . second , the _ ulysses lower energy threshold actually allows photons with energies as low as 15 kev to be counted , albeit with lower efficiency . third , the two _ ulysses hemispherical detectors are thin , so photons which do not interact in one hemisphere may interact in the other , increasing the effective area . finally , the batse count rates are not corrected for dead time , which is known to affect the brighter bursts ( which are in general among the bursts we observe ) . _ _ _ _ _ _ _ _ this set of criteria is not an absolute one . to a good approximation , every gamma - ray burst is different from every other one . in addition , data gaps , earth occultation , telemetry noise , unusual background behavior , and different data modes complicate some analyses . in some cases , bursts satisfying all the criteria above are rejected because they are simply too weak to allow reliable triangulation . in other cases , bursts which do not satisfy one or more criteria may be accepted . a final criterion for inclusion in this catalog is that the burst must not have been observed by a third interplanetary spacecraft . the _ ulysses / batse bursts also observed by _ mars observer have been published in laros et al . the _ ulysses /batse bursts also observed by _ pioneer venus orbiter are in press ( laros et al . 1998 ) . over the period covered by the 3b catalog , _ ulysses observed 346 confirmed cosmic gamma - ray bursts ( i.e. , bursts that were observed by at least one other spacecraft ) . of these , 274 were observed by _ ulysses and batse and in some cases , other spacecraft as well . since batse observed 1122 bursts during this period , _ ulysses observes approximately one out of every 4 batse bursts . of the 274 , 64 were observed by _ ulysses in triggered mode , and 210 in real time mode . finally , 220 were observed by _ ulysses and batse , but not by a third interplanetary spacecraft ; two of these could not be triangulated according to the criteria explained above . _ _ _ _ _ the histogram of figure 6 shows the distribution of annulus half - widths for the 218 bursts localized . the smallest is about 7 , the largest 32 , and the average is 4.5 . 158 of the annuli , or 72% , intersect the batse 1 @xmath17 error circles , whose radii are defined by @xmath23 , where @xmath24 is the systematic error , 1.6@xmath25 , and @xmath26 is the statistical error . figure 5 shows one example . this is less than the number which would be predicted ( 87% ) . an analysis of a preliminary version of the ipn catalog describes several more complicated batse error models that are consistent with the batse - ipn separations ( briggs et al . a more extensive analysis , utilizing the ulysses supplement to the 4b catalog , is in progress ( briggs et al . one quantity of interest is how close the annulus passes to the center of the error circle . let @xmath27 be the right ascension and declination of the center of a batse error circle , and let @xmath28 be the right ascension , declination , and radius of an annulus . then the minimum distance between the error circle and the annulus is given by @xmath29 . a histogram of the distribution of minimum distances between the annuli and the centers of the batse error circles is given in figure 7 . in general , the annuli obtained by triangulations are small circles on the celestial sphere , so their curvature , even across a relatively small batse error circle , is not always negligible , and a simple , four - sided error box can not be defined . this curvature can be seen in figure 5 . for this reason , we do not cite the intersection points of the annulus with the error circle . however , we give here the formulas for finding these points for those cases where it may be useful . let @xmath30 and @xmath28 be the right ascension , declination , and radii of the two small circles . let @xmath31 be the right ascension and declination of the intersection points . then @xmath32 where a=@xmath33 , b=@xmath34 , and c=@xmath35 , for each of the two values of the declination @xmath36 the corresponding right ascension @xmath37 is given by @xmath38 figure 8 shows the batse peak fluxes and fluences for 162 of the 218 bursts with flux and fluence entries in the 3b catalog . there is a slight tendency for some of the smaller fluence events to have relatively large peak fluxes over a time interval comparable to that of the _ ulysses real time data , which helps to explain why these bursts were detected by _ ulysses . to estimate the completeness of this catalog , we have calculated the ratios of the number of bursts detected by _ ulysses to the number detected by batse above several flux and fluence thresholds starting with the weakest events in the batse catalog ; we have used the 25 - 100 kev fluences in erg @xmath39 and the peak 25 - 300 kev fluxes over 256 ms in photons @xmath39 s@xmath40 . since events detected by a third spacecraft are not included in the present catalog , the numbers given represent lower limits to the completeness . table 1 summarizes the results . _ _ _ cc 1.8 x 10@xmath41 erg @xmath39 & .25 10@xmath42 erg @xmath39 & .49 10@xmath43 erg @xmath39 & .74 .27 photons @xmath39 s@xmath40 & .25 1 photon @xmath39 s@xmath40 & .34 10 photons @xmath39 s@xmath40 & .65 finally , we have analyzed these annuli for intersections which might provide evidence of repeating bursts from a single source . although there was a large , but statistically insignificant number of two annulus intersections , we found only one case where a three - way intersection almost occurred ( batse bursts 451 , 2286 , and 2619 ) ; the data are therefore consistent with no burst repetition . the ten columns in table 1 give : 1 ) the date of the burst , in ddmmyy format , 2 ) the universal time of the burst at earth , 3 ) the batse number for the burst , 4 ) the batse right ascension of the center of the error circle ( j2000 ) , in degrees , 5 ) the batse declination of the center of the error circle ( j2000 ) , in degrees , 6 ) the total 1 @xmath17 statistical batse error circle radius , in degrees , ( the approximate total 1@xmath17 radius is obtained by adding 1.6@xmath25 in quadrature , but see briggs et al . 1998a , b for an improved error model ) , 7 ) the right ascension of the center of the ipn ( batse/_ulysses ) annulus , epoch j2000 , corrected to the heliocentric frame , in degrees , 8) the declination of the center of the ipn ( batse/_ulysses ) annulus , epoch j2000 , corrected to the heliocentric frame , in degrees , 9 ) the angular radius of the ipn ( batse/_ulysses ) annulus , corrected to the heliocentric frame , in degrees , and 10 ) the half width of the ipn ( batse/_ulysses ) annulus , in degrees ; the 3 @xmath17 confidence annulus is given by r@xmath44 @xmath22 @xmath36 r@xmath45 . _ _ _ _ this list supercedes previous ones which have been circulated in the past , but not published . the changes to it include : 1 ) the use of final time history and spacecraft ephemeris data ; earlier lists often used preliminary and predict data , with correspondingly larger timing uncertainties , 2 ) transformation of annuli to heliocentric coordinates ; earlier lists often gave geocentric coordinates , with correspondingly larger uncertainties in the annulus widths , 3 ) removal of weak events according to the criteria in section 5 , 4 ) reprocessing of data to eliminate known clock errors , and 5 ) careful examination of bursts whose batse time histories were truncated by data gaps and earth occultation . figures 9 and 10 compare the batse error circles with the ipn annulus / error circle intersections for the bursts in this catalog . to generate this plot , it was assumed that all annuli pass through the centers of their corresponding batse error circles . we stress that while the batse data are for the bursts listed in the 3b catalog , the locations and error circle radii are in fact the updated ones which will appear in the 4b catalog . they have been taken from the latest online catalog , and are given here for convenience only . finally , we add that 150 _ ulysses /batse bursts were detected over the period between the end of the 3b catalog and the end of the 4b catalog . analysis of these events has been completed and is being published in conjunction with the 4b catalog . _ cccccccccc 250491 & 00:37:46 & 109 & 91.29 & -22.77 & 1.02 & 115.140 & 23.171 & 51.784 & 0.051 290491 & 03:11:50 & 121 & 174.75 & 14.33 & 1.36 & 116.426 & 22.919 & 59.157 & 0.136 020591 & 22:37:23 & 142 & 46.22 & -52.71 & 0.62 & 297.624 & -22.677 & 87.126 & 0.023 030591 & 07:04:13 & 143 & 87.45 & 38.74 & 0.87 & 117.731 & 22.655 & 30.360 & 0.009 110591 & 02:11:48 & 179 & 266.48 & 58.30 & 1.66 & 300.193 & -22.137 & 83.666 & 0.018 230591 & 19:03:25 & 222 & 106.99 & 0.36 & 1.28 & 124.202 & 21.228 & 26.847 & 0.051 010691 & 19:22:14 & 249 & 310.12 & 32.34 & 0.17 & 307.035 & -20.537 & 52.942 & 0.008 020691 & 22:55:01 & 257 & 142.76 & 53.99 & 2.06 & 127.387 & 20.448 & 34.526 & 0.163 260691 & 07:15:13 & 444 & 133.38 & 7.72 & 1.27 & 134.557 & 18.501 & 10.164 & 0.020 300691 & 07:37:02 & 469 & 304.70 & 35.13 & 0.91 & 315.764 & -18.148 & 52.439 & 0.085 210791 & 19:30:12 & 563 & 315.52 & -14.29 & 1.31 & 322.022 & -16.198 & 6.769 & 0.436 090891 & 06:15:00 & 659 & 195.43 & 19.37 & 1.85 & 147.088 & 14.483 & 50.860 & 0.092 090891 & 19:35:45 & 660 & 281.96 & 64.74 & 2.01 & 327.238 & -14.431 & 85.747 & 0.016 050991 & 23:48:55 & 761 & 360.00 & -81.05 & 0.82 & 334.075 & -11.942 & 70.787 & 0.017 070991 & 01:05:23 & 764 & 136.55 & -28.23 & 1.30 & 154.322 & 11.849 & 43.573 & 0.131 260991 & 18:25:57 & 824 & 346.07 & 21.36 & 2.70 & 338.688 & -10.171 & 32.155 & 0.294 270991 & 23:26:53 & 829 & 50.16 & -39.65 & 0.34 & 338.937 & -10.074 & 68.870 & 0.036 300991 & 11:55:30 & 841 & 131.80 & -20.62 & 1.27 & 159.447 & 9.874 & 39.201 & 0.013 061091 & 09:01:54 & 871 & 33.30 & -45.44 & 3.38 & 340.603 & -9.419 & 50.544 & 0.087 221091 & 04:13:58 & 914 & 123.91 & 70.34 & 1.03 & 163.395 & 8.312 & 68.583 & 0.017 261091 & 13:07:01 & 938 & 74.68 & -22.27 & 2.40 & 344.080 & -8.038 & 80.644 & 0.068 061191 & 03:42:01 & 1008 & 343.03 & -35.16 & 0.83 & 345.570 & -7.445 & 29.223 & 0.023 231191 & 08:15:33 & 1114 & 22.81 & 30.34 & 1.14 & 347.386 & -6.730 & 49.421 & 0.015 271191 & 04:22:08 & 1122 & 269.14 & 49.73 & 0.38 & 347.680 & -6.617 & 88.466 & 0.005 071291 & 09:50:45 & 1150 & 307.79 & 29.30 & 1.59 & 348.248 & -6.408 & 53.876 & 0.059 091291 & 18:35:58 & 1157 & 259.98 & -45.08 & 0.40 & 348.333 & -6.380 & 82.576 & 0.003 171291 & 08:22:06 & 1190 & 6.96 & 22.67 & 1.02 & 348.480 & -6.340 & 35.915 & 0.032 211291 & 22:03:26 & 1200 & 215.82 & -42.72 & 1.98 & 168.475 & 6.356 & 61.764 & 0.111 241291 & 11:02:58 & 1212 & 45.43 & -10.79 & 4.68 & 348.444 & -6.377 & 57.290 & 0.117 271291 & 01:05:53 & 1235 & 350.44 & -68.25 & 0.67 & 348.389 & -6.408 & 61.852 & 0.064 100192 & 09:17:58 & 1288 & 58.08 & -20.80 & 0.42 & 347.700 & -6.748 & 69.190 & 0.018 210292 & 06:14:32 & 1425 & 184.14 & 48.48 & 0.65 & 161.368 & 8.446 & 42.142 & 0.008 240292 & 02:43:39 & 1432 & 169.65 & -66.78 & 2.73 & 160.726 & 8.515 & 75.175 & 0.539 270292 & 06:00:51 & 1443 & 68.36 & 66.84 & 1.33 & 160.016 & 8.591 & 83.353 & 0.005 070392 & 00:18:04 & 1467 & 355.57 & -45.15 & 1.16 & 338.049 & -8.795 & 39.316 & 0.066 150392 & 04:19:27 & 1484 & 320.58 & -20.89 & 0.46 & 336.280 & -8.960 & 16.873 & 0.013 070492 & 12:03:40 & 1543 & 341.44 & -10.68 & 2.04 & 332.036 & -9.187 & 9.927 & 0.437 080492 & 12:05:56 & 1545 & 138.96 & 40.83 & 0.58 & 151.890 & 9.190 & 34.380 & 0.022 290492 & 18:59:32 & 1571 & 179.08 & -14.33 & 2.85 & 149.602 & 8.963 & 39.318 & 0.091 020592 & 06:02:59 & 1577 & 33.05 & -17.73 & 1.04 & 329.440 & -8.908 & 60.781 & 0.026 130592 & 16:52:39 & 1606 & 211.16 & -44.77 & 0.65 & 148.956 & 8.591 & 74.250 & 0.005 300592 & 22:59:55 & 1630 & 162.68 & 18.13 & 1.67 & 148.993 & 7.901 & 17.618 & 0.070 170692 & 05:26:51 & 1652 & 37.66 & 77.51 & 0.51 & 149.807 & 6.977 & 89.064 & 0.012 220692 & 07:05:05 & 1663 & 162.10 & 47.17 & 0.31 & 150.169 & 6.667 & 41.336 & 0.005 220692 & 12:54:21 & 1664 & 346.81 & 10.01 & 0.55 & 330.187 & -6.649 & 18.892 & 0.018 270692 & 13:02:36 & 1676 & 166.02 & -2.58 & 0.65 & 150.593 & 6.327 & 23.504 & 0.017 010792 & 19:38:19 & 1683 & 314.64 & 33.44 & 0.63 & 330.973 & -6.033 & 42.560 & 0.017 140792 & 20:23:26 & 1700 & 140.36 & -44.64 & 3.28 & 152.298 & 5.079 & 50.262 & 0.099 180792 & 14:40:36 & 1708 & 22.44 & -3.98 & 1.14 & 332.718 & -4.782 & 49.196 & 0.032 180792 & 21:32:43 & 1709 & 296.39 & -55.98 & 0.48 & 332.751 & -4.760 & 58.948 & 0.004 210792 & 18:17:48 & 1717 & 37.55 & 33.49 & 0.73 & 333.083 & -4.530 & 72.485 & 0.033 230792 & 01:00:34 & 1721 & 85.79 & 7.78 & 0.48 & 153.234 & 4.427 & 68.971 & 0.008 010892 & 01:16:40 & 1733 & 157.86 & 68.90 & 1.12 & 154.338 & 3.667 & 64.462 & 0.021 130892 & 03:33:48 & 1807 & 206.22 & -28.34 & 1.75 & 155.907 & 2.581 & 63.921 & 0.051 140892 & 06:09:53 & 1815 & 258.68 & -44.63 & 0.79 & 336.053 & -2.478 & 78.143 & 0.006 240892 & 10:53:03 & 1872 & 228.59 & -51.79 & 0.33 & 157.427 & 1.505 & 76.788 & 0.003 300892 & 01:45:18 & 1883 & 258.53 & -73.99 & 0.54 & 338.192 & -0.949 & 86.267 & 0.010 020992 & 00:29:05 & 1886 & 279.15 & -20.70 & 2.29 & 338.594 & -0.651 & 61.607 & 0.015 030992 & 00:49:17 & 1888 & 61.97 & -68.37 & 0.75 & 338.732 & -0.549 & 88.165 & 0.010 250992 & 21:45:16 & 1956 & 160.26 & 67.38 & 1.34 & 161.773 & -1.864 & 68.233 & 0.020 031092 & 06:35:40 & 1974 & 150.32 & -14.71 & 0.55 & 162.694 & -2.674 & 19.765 & 0.021 091092 & 06:41:11 & 1983 & 120.41 & -29.10 & 0.39 & 163.411 & -3.345 & 46.240 & 0.003 151092 & 01:36:32 & 1989 & 123.04 & 41.69 & 0.95 & 164.066 & -3.998 & 57.677 & 0.040 221092 & 15:20:59 & 1997 & 253.78 & -12.42 & 0.41 & 164.862 & -4.861 & 88.377 & 0.002 291092 & 12:37:56 & 2018 & 34.14 & -1.57 & 1.67 & 345.514 & 5.651 & 50.169 & 0.045 301092 & 01:16:31 & 2020 & 86.16 & 6.86 & 1.75 & 165.561 & -5.710 & 79.293 & 0.042 181192 & 22:11:44 & 2061 & 350.45 & 50.32 & 1.07 & 346.930 & 7.990 & 41.232 & 0.128 231192 & 06:18:31 & 2067 & 334.24 & -53.25 & 0.72 & 347.111 & 8.481 & 64.897 & 0.008 031292 & 21:44:22 & 2074 & 233.62 & 21.03 & 0.74 & 167.337 & -9.664 & 68.339 & 0.119 061292 & 18:24:19 & 2080 & 176.48 & 45.72 & 0.77 & 167.341 & -9.975 & 51.667 & 0.029 071292 & 16:00:48 & 2083 & 306.33 & -42.93 & 0.44 & 347.336 & 10.074 & 62.709 & 0.003 091292 & 02:40:56 & 2089 & 172.98 & 61.45 & 0.51 & 167.325 & -10.228 & 68.392 & 0.048 091292 & 11:35:38 & 2090 & 332.97 & -52.02 & 0.68 & 347.320 & 10.269 & 63.995 & 0.009 181292 & 02:30:02 & 2102 & 355.62 & -55.31 & 1.76 & 347.091 & 11.168 & 66.903 & 0.076 271292 & 12:44:08 & 2106 & 206.33 & -25.65 & 1.04 & 166.535 & -12.089 & 39.698 & 0.113 301292 & 08:57:51 & 2110 & 19.52 & 17.21 & 0.33 & 346.301 & 12.353 & 32.939 & 0.112 301292 & 14:05:46 & 2111 & 150.54 & 28.63 & 1.71 & 166.282 & -12.370 & 45.110 & 0.086 060193 & 15:37:39 & 2121 & 6.08 & 2.00 & 0.53 & 345.556 & 12.984 & 21.048 & 0.021 060193 & 19:57:15 & 2122 & 124.79 & -32.62 & 0.68 & 165.536 & -12.996 & 42.302 & 0.149 080193 & 02:28:55 & 2123 & 94.62 & -30.44 & 0.99 & 165.383 & -13.100 & 65.116 & 0.069 120193 & 03:47:31 & 2127 & 226.32 & 27.74 & 1.22 & 164.853 & -13.415 & 72.911 & 0.021 120193 & 15:18:02 & 2128 & 210.10 & 51.92 & 1.02 & 164.786 & -13.451 & 76.849 & 0.066 160193 & 02:47:02 & 2136 & 246.99 & 15.91 & 1.02 & 164.275 & -13.700 & 88.748 & 0.026 270193 & 16:14:25 & 2149 & 310.95 & -39.63 & 1.16 & 342.242 & 14.382 & 60.557 & 0.155 310193 & 18:57:12 & 2151 & 182.04 & -8.24 & 2.16 & 161.406 & -14.563 & 22.249 & 0.009 010293 & 16:41:55 & 2156 & 333.53 & 41.15 & 1.26 & 341.214 & 14.601 & 27.869 & 0.023 050293 & 10:10:31 & 2166 & 278.00 & -45.74 & 5.98 & 340.403 & 14.729 & 83.936 & 0.277 170293 & 14:44:38 & 2197 & 117.78 & 0.98 & 2.27 & 157.517 & -14.945 & 41.809 & 0.234 020393 & 21:33:17 & 2213 & 90.33 & 5.43 & 0.62 & 154.166 & -14.821 & 66.384 & 0.016 090393 & 03:07:50 & 2228 & 325.41 & 51.73 & 0.72 & 332.605 & 14.648 & 36.594 & 0.012 100393 & 07:19:20 & 2232 & 284.35 & -52.39 & 0.89 & 332.316 & 14.607 & 79.362 & 0.007 180393 & 12:29:53 & 2255 & 150.48 & -18.85 & 1.09 & 150.361 & -14.267 & 5.323 & 0.320 290393 & 03:09:25 & 2273 & 179.25 & -8.65 & 0.94 & 148.090 & -13.723 & 35.582 & 0.010 310393 & 03:11:16 & 2276 & 43.00 & 47.27 & 2.03 & 327.701 & 13.613 & 77.070 & 0.078 050493 & 08:54:49 & 2286 & 198.18 & -8.80 & 0.55 & 146.754 & -13.315 & 52.136 & 0.005 060493 & 07:16:25 & 2289 & 299.81 & 29.01 & 0.51 & 326.595 & 13.263 & 31.383 & 0.011 090493 & 04:50:10 & 2295 & 295.83 & 43.18 & 2.04 & 326.125 & 13.098 & 37.466 & 0.366 090493 & 21:25:40 & 2297 & 176.91 & -39.75 & 1.25 & 146.019 & -13.056 & 39.155 & 0.088 140493 & 02:27:15 & 2303 & 311.42 & 9.92 & 0.91 & 325.406 & 12.819 & 14.314 & 0.061 180493 & 14:38:18 & 2307 & 311.65 & 21.37 & 1.27 & 324.831 & 12.570 & 14.596 & 0.415 240493 & 19:19:00 & 2315 & 54.22 & 41.60 & 1.67 & 324.179 & 12.245 & 86.680 & 0.134 250493 & 10:17:29 & 2316 & 18.00 & -35.19 & 0.31 & 324.122 & 12.214 & 70.285 & 0.034 260493 & 12:40:32 & 2319 & 67.82 & 10.13 & 1.10 & 144.025 & -12.159 & 75.515 & 0.028 280493 & 01:07:12 & 2320 & 129.85 & -55.45 & 1.68 & 143.900 & -12.085 & 46.202 & 0.005 300493 & 15:01:18 & 2321 & 106.33 & 20.97 & 0.36 & 143.709 & -11.967 & 50.803 & 0.009 020593 & 13:52:52 & 2323 & 254.72 & -46.53 & 0.78 & 323.580 & 11.882 & 87.103 & 0.039 060593 & 14:52:51 & 2329 & 64.45 & -5.97 & 0.24 & 143.364 & -11.717 & 77.827 & 0.004 180593 & 03:20:25 & 2346 & 349.17 & -37.51 & 1.04 & 323.086 & 11.354 & 53.387 & 0.019 230593 & 19:40:53 & 2350 & 14.80 & 4.21 & 2.64 & 323.121 & 11.238 & 44.153 & 0.124 310593 & 15:53:38 & 2362 & 340.04 & 13.25 & 2.16 & 323.341 & 11.152 & 15.895 & 0.044 090693 & 10:07:24 & 2383 & 208.73 & -16.38 & 1.62 & 143.799 & -11.158 & 64.377 & 0.132 140693 & 03:40:30 & 2393 & 168.14 & 23.95 & 1.15 & 144.132 & -11.210 & 42.800 & 0.008 050793 & 12:39:08 & 2429 & 283.08 & -39.24 & 0.99 & 326.244 & 11.861 & 65.410 & 0.124 060793 & 08:09:10 & 2432 & 210.09 & -1.36 & 1.31 & 146.342 & -11.898 & 63.498 & 0.045 080793 & 05:12:38 & 2433 & 86.55 & -69.02 & 6.25 & 146.572 & -11.989 & 62.867 & 0.076 090793 & 06:14:22 & 2436 & 194.41 & -46.73 & 0.67 & 146.700 & -12.042 & 54.525 & 0.028 140793 & 16:12:53 & 2446 & 257.70 & 23.70 & 0.97 & 327.399 & 12.344 & 68.720 & 0.048 210793 & 01:17:22 & 2452 & 78.66 & -33.15 & 1.52 & 148.275 & -12.751 & 66.262 & 0.085 210793 & 12:45:35 & 2453 & 298.58 & -40.79 & 1.79 & 328.342 & 12.785 & 61.079 & 0.101 270793 & 08:08:29 & 2466 & 159.99 & -23.89 & 1.28 & 149.185 & -13.209 & 16.546 & 0.136 310793 & 03:16:31 & 2475 & 116.53 & 45.49 & 2.15 & 149.754 & -13.514 & 66.345 & 0.036 050893 & 12:39:28 & 2482 & 334.55 & -15.60 & 0.83 & 330.584 & 13.983 & 30.305 & 0.208 090893 & 05:11:04 & 2486 & 37.97 & -74.63 & 1.79 & 151.166 & -14.323 & 80.857 & 0.088 220893 & 14:44:51 & 2500 & 213.09 & -0.68 & 2.28 & 153.348 & -15.724 & 65.872 & 0.232 310893 & 09:16:05 & 2507 & 223.29 & -3.63 & 0.47 & 154.815 & -16.774 & 67.536 & 0.007 030993 & 00:19:29 & 2511 & 88.67 & -10.42 & 4.20 & 155.258 & -17.109 & 64.672 & 0.109 050993 & 03:26:30 & 2514 & 310.95 & -1.22 & 0.63 & 335.615 & 17.389 & 28.629 & 0.006 100993 & 12:12:08 & 2522 & 167.59 & -66.69 & 0.54 & 156.521 & -18.114 & 48.810 & 0.044 140993 & 02:49:11 & 2530 & 34.26 & 34.29 & 1.53 & 337.128 & 18.628 & 50.198 & 0.173 160993 & 20:19:23 & 2533 & 280.71 & 65.39 & 0.22 & 337.586 & 19.026 & 57.392 & 0.013 220993 & 06:24:46 & 2537 & 271.20 & 55.89 & 0.45 & 338.488 & 19.847 & 67.999 & 0.003 270993 & 04:18:07 & 2542 & 108.61 & -9.52 & 1.62 & 159.296 & -20.624 & 52.084 & 0.115 061093 & 21:31:50 & 2566 & 66.31 & 65.28 & 1.70 & 340.850 & 22.262 & 70.461 & 0.049 081093 & 07:18:14 & 2570 & 42.98 & 21.37 & 0.85 & 341.069 & 22.510 & 58.202 & 0.082 081093 & 11:09:10 & 2571 & 165.40 & 33.93 & 0.86 & 161.094 & -22.537 & 60.104 & 0.027 141093 & 17:01:01 & 2586 & 272.50 & 9.34 & 0.50 & 342.042 & 23.672 & 67.956 & 0.010 171093 & 01:17:49 & 2590 & 181.44 & -83.11 & 0.65 & 162.389 & -24.110 & 58.979 & 0.003 191093 & 18:22:34 & 2593 & 242.92 & -18.70 & 1.66 & 162.779 & -24.628 & 73.476 & 0.025 241093 & 13:29:09 & 2603 & 137.63 & -8.49 & 0.53 & 163.447 & -25.567 & 30.574 & 0.083 261093 & 11:35:57 & 2606 & 51.36 & -10.56 & 0.48 & 343.704 & 25.955 & 75.482 & 0.087 301093 & 08:08:42 & 2609 & 326.41 & 58.24 & 0.55 & 344.202 & 26.745 & 31.863 & 0.058 311093 & 04:06:35 & 2611 & 325.10 & 62.73 & 0.32 & 344.306 & 26.918 & 37.073 & 0.005 031193 & 16:25:42 & 2617 & 87.88 & 65.05 & 0.31 & 344.732 & 27.657 & 72.278 & 0.008 061193 & 20:32:04 & 2619 & 196.75 & -3.13 & 1.11 & 165.094 & -28.337 & 40.406 & 0.232 121193 & 18:45:53 & 2628 & 311.51 & 55.95 & 0.53 & 345.711 & 29.654 & 34.291 & 0.010 261193 & 04:46:46 & 2660 & 157.28 & 70.08 & 3.02 & 346.750 & 32.795 & 80.134 & 0.109 261193 & 19:31:39 & 2661 & 3.29 & -17.40 & 0.22 & 346.785 & 32.945 & 50.498 & 0.004 041293 & 09:48:10 & 2676 & 250.15 & 38.23 & 0.18 & 347.083 & 34.824 & 76.909 & 0.004 051293 & 14:59:51 & 2679 & 15.56 & 66.49 & 2.22 & 347.108 & 35.133 & 38.364 & 0.018 081293 & 04:37:19 & 2682 & 198.73 & -29.93 & 0.56 & 167.142 & -35.783 & 34.384 & 0.050 081293 & 09:34:40 & 2683 & 234.11 & -40.29 & 3.06 & 167.143 & -35.836 & 51.960 & 0.143 211293 & 02:09:50 & 2700 & 91.02 & -43.17 & 0.50 & 166.799 & -39.143 & 56.073 & 0.036 221293 & 21:21:33 & 2703 & 192.52 & 28.65 & 0.66 & 166.675 & -39.621 & 73.261 & 0.251 251293 & 23:20:12 & 2709 & 4.85 & 9.87 & 0.79 & 346.415 & 40.440 & 35.141 & 0.212 261293 & 20:11:09 & 2711 & 205.95 & 21.63 & 2.82 & 166.329 & -40.669 & 74.178 & 0.127 010194 & 22:42:00 & 2729 & 189.27 & 62.84 & 3.47 & 345.578 & 42.289 & 78.226 & 0.064 020194 & 02:45:52 & 2732 & 10.90 & 45.11 & 1.27 & 345.554 & 42.335 & 18.951 & 0.195 030194 & 22:12:56 & 2736 & 267.90 & 7.14 & 0.65 & 345.274 & 42.810 & 76.721 & 0.016 130194 & 16:47:54 & 2756 & 209.39 & -23.16 & 9.65 & 163.289 & -45.332 & 42.651 & 0.049 190194 & 15:44:40 & 2773 & 131.41 & 39.29 & 1.39 & 161.654 & -46.793 & 89.606 & 0.051 280194 & 16:50:58 & 2790 & 224.02 & -15.22 & 0.96 & 158.510 & -48.816 & 64.647 & 0.059 290194 & 10:43:17 & 2793 & 127.46 & 28.59 & 0.82 & 158.217 & -48.971 & 81.612 & 0.072 010294 & 07:40:40 & 2794 & 268.61 & 22.37 & 1.67 & 337.027 & 49.542 & 59.277 & 0.125 030294 & 15:46:56 & 2797 & 184.16 & -40.95 & 0.58 & 156.002 & -49.973 & 20.512 & 0.052 060294 & 00:08:37 & 2798 & 144.20 & -59.96 & 0.15 & 154.922 & -50.380 & 13.687 & 0.039 100294 & 19:13:16 & 2812 & 152.29 & 82.13 & 0.66 & 332.566 & 51.121 & 48.298 & 0.007 170294 & 23:02:42 & 2831 & 29.07 & 4.55 & 0.69 & 328.739 & 51.947 & 69.334 & 0.005 180294 & 19:32:28 & 2833 & 349.71 & -19.69 & 0.35 & 328.264 & 52.023 & 73.939 & 0.010 280294 & 11:29:00 & 2852 & 127.98 & -12.36 & 0.44 & 142.713 & -52.495 & 42.173 & 0.012 010394 & 20:10:37 & 2855 & 103.51 & 64.35 & 0.35 & 321.919 & 52.509 & 59.400 & 0.029 020394 & 05:08:31 & 2856 & 12.49 & -23.97 & 0.21 & 321.700 & 52.508 & 88.572 & 0.005 120394 & 11:28:22 & 2877 & 221.50 & -56.17 & 0.61 & 135.895 & -52.124 & 48.964 & 0.052 190394 & 23:57:20 & 2889 & 68.74 & -12.60 & 0.39 & 132.014 & -51.403 & 62.552 & 0.090 210394 & 22:05:07 & 2890 & 160.33 & 3.10 & 0.85 & 131.093 & -51.168 & 58.918 & 0.233 230394 & 22:04:38 & 2891 & 213.89 & 8.37 & 0.85 & 310.173 & 50.907 & 85.032 & 0.016 280394 & 12:05:34 & 2894 & 97.18 & -32.35 & 0.58 & 128.216 & -50.236 & 26.692 & 0.028 290394 & 05:49:26 & 2895 & 237.51 & -68.89 & 0.50 & 127.917 & -50.121 & 51.112 & 0.006 290394 & 18:15:37 & 2897 & 204.91 & -10.33 & 1.00 & 127.712 & -50.040 & 69.434 & 0.026 300394 & 20:44:39 & 2899 & 193.80 & -33.63 & 1.72 & 127.286 & -49.861 & 48.714 & 0.320 060494 & 01:10:54 & 2913 & 225.44 & -39.10 & 1.24 & 125.133 & -48.806 & 69.435 & 0.043 100494 & 15:45:02 & 2919 & 254.89 & 30.80 & 0.93 & 303.788 & 47.974 & 42.261 & 0.047 120494 & 01:40:31 & 2922 & 217.17 & -38.35 & 1.12 & 123.422 & -47.710 & 58.846 & 0.303 140494 & 16:46:25 & 2929 & 181.26 & -26.15 & 0.50 & 122.794 & -47.222 & 49.154 & 0.026 190494 & 19:10:58 & 2940 & 359.86 & -47.95 & 0.58 & 121.768 & -46.270 & 73.314 & 0.062 290494 & 00:43:53 & 2953 & 35.10 & -57.52 & 0.32 & 120.524 & -44.594 & 51.727 & 0.006 030594 & 05:07:44 & 2958 & 161.00 & 10.06 & 0.60 & 120.195 & -43.873 & 64.039 & 0.057 120594 & 17:37:50 & 2974 & 232.97 & 55.88 & 2.65 & 299.935 & 42.376 & 44.482 & 0.007 200594 & 00:21:38 & 2984 & 323.50 & 6.93 & 0.88 & 300.146 & 41.389 & 40.428 & 0.019 260594 & 10:55:26 & 2993 & 80.41 & -30.39 & 0.96 & 120.590 & -40.650 & 33.571 & 0.347 260594 & 20:20:05 & 2994 & 131.85 & 34.20 & 1.69 & 120.622 & -40.611 & 76.690 & 0.012 290594 & 03:17:02 & 2998 & 163.64 & -25.33 & 3.45 & 120.834 & -40.382 & 35.002 & 0.334 290594 & 16:20:36 & 3002 & 66.19 & -23.75 & 3.30 & 120.889 & -40.330 & 47.443 & 0.021 290594 & 21:18:56 & 3003 & 214.84 & 58.93 & 0.46 & 300.908 & 40.314 & 56.125 & 0.083 090694 & 16:29:06 & 3024 & 284.14 & 23.34 & 1.01 & 302.269 & 39.513 & 18.866 & 0.118 190694 & 21:31:20 & 3035 & 299.05 & -29.89 & 0.34 & 303.996 & 39.141 & 69.575 & 0.026 220694 & 13:18:16 & 3039 & 299.58 & -31.24 & 0.82 & 304.505 & 39.108 & 72.495 & 0.012 230694 & 18:46:23 & 3042 & 108.37 & 75.78 & 0.46 & 304.747 & 39.101 & 65.666 & 0.013 010794 & 21:46:38 & 3055 & 145.18 & -6.43 & 0.81 & 126.479 & -39.199 & 37.139 & 0.157 020794 & 09:28:11 & 3056 & 29.73 & -23.53 & 0.57 & 126.586 & -39.214 & 78.876 & 0.116 030794 & 04:40:46 & 3057 & 131.50 & 27.39 & 1.25 & 126.771 & -39.239 & 67.508 & 0.005 040794 & 23:32:07 & 3060 & 212.49 & 47.25 & 0.64 & 307.181 & 39.305 & 67.688 & 0.082 060794 & 19:24:33 & 3062 & 265.11 & 26.40 & 1.86 & 307.614 & 39.384 & 36.694 & 0.287 080794 & 20:42:06 & 3067 & 301.58 & 24.66 & 1.02 & 308.112 & 39.490 & 17.427 & 0.022 140794 & 12:54:21 & 3075 & 333.58 & -42.10 & 1.59 & 309.555 & 39.862 & 85.495 & 0.067 170794 & 03:24:29 & 3087 & 109.80 & 12.93 & 0.71 & 130.246 & -40.076 & 54.328 & 0.006 280794 & 02:55:32 & 3099 & 42.57 & -49.08 & 0.76 & 133.354 & -41.280 & 60.269 & 0.045 280794 & 14:06:40 & 3100 & 291.22 & 4.45 & 1.29 & 313.491 & 41.344 & 44.189 & 0.181 280794 & 23:58:54 & 3101 & 85.49 & -39.87 & 1.44 & 133.620 & -41.397 & 36.263 & 0.066 060894 & 04:38:48 & 3108 & 276.21 & -31.33 & 10.69 & 316.177 & 42.651 & 86.962 & 0.079 060894 & 09:32:56 & 3109 & 242.38 & 13.65 & 1.22 & 316.242 & 42.686 & 69.105 & 0.126 100894 & 02:22:42 & 3115 & 212.01 & -17.47 & 0.53 & 137.468 & -43.347 & 67.696 & 0.009 120894 & 00:27:42 & 3119 & 340.49 & -46.12 & 2.23 & 138.120 & -43.715 & 86.692 & 0.263 170894 & 08:40:15 & 3128 & 280.84 & 3.47 & 0.40 & 319.994 & 44.823 & 53.907 & 0.010 210894 & 21:51:27 & 3131 & 294.69 & 43.48 & 1.77 & 321.670 & 45.864 & 16.142 & 0.191 260894 & 20:55:20 & 3138 & 93.59 & -34.63 & 0.45 & 143.593 & -47.103 & 37.845 & 0.013 300894 & 08:59:56 & 3143 & 193.66 & 25.66 & 1.58 & 145.011 & -48.048 & 89.313 & 0.042 100994 & 19:31:24 & 3163 & 332.01 & -34.11 & 0.94 & 330.099 & 51.525 & 87.032 & 0.065 150994 & 06:52:19 & 3168 & 219.42 & 47.70 & 4.53 & 332.318 & 53.052 & 62.836 & 0.393 support for the _ ulysses grb experiment is provided by jpl contract 958056 . joint analysis of _ ulysses and batse data is supported by nasa grant nag 5 - 1560 . ck acknowledges support from nasa grant nag5 - 2560 . _ _ briggs , m. , pendleton , g. , brainerd , j. , connaughton , v. , kippen , r. m. , meegan , c. , and hurley , k. 1998a , in gamma - ray bursts : 4th huntsville symposium , eds . c. meegan , r. preece , and t. koshut , ( aip : new york ) , p. 104 briggs , m. , et al . 1998b , , in press galama , t. , et al . 1997 , , 486 , l5 hurley , k. 1994 , in gamma - ray bursts - second workshop , eds . j. fishman , j. brainerd , and k. hurley , aip conference proceedings 307 ( aip : new york ) , 682 hurley , k. , et al . 1992 , astron . ser . , 92(2 ) , 401 hurley , k. , et al . 1996 , , 28(4 ) , 1408 hurley , k. , et al . 1997a , , 479 , l113 hurley , k. , et al . 1997b , proc . 25th icrc , durban , south africa , og 2.2.6 , 45 hurley , k. , et al . 1997c , , 485 , l1 hurley , k. , et al . 1998a , in gamma - ray bursts : 4th huntsville symposium , eds . c. meegan , r. preece , and t. koshut , ( aip : new york ) , p. 50 hurley , k. , et al . 1998b , in the active x - ray sky : results from bepposax and rossi - xte , eds . l. scarsi , h. bradt , p. giommi , and f. fiore , nucl . b ( proceedings supplement ) , 69(1 - 3 ) , p.660 kippen , m. , hurley , k. , and pendleton , g. 1998 , in gamma - ray bursts : 4th huntsville symposium , eds . c. meegan , r. preece , and t. koshut , ( aip : new york ) , p. 114 kolatt , t. , and piran , t. 1996 , , 467 , l41 kulkarni , s. , et al . 1998 , , 393 , 35 laros , j. , et al . , 1997 , , 110 , 157 laros , j. , et al . , 1998 , , accepted meegan , c. , et al . , 1996 , , 106 , 45 metzger , m. , et al . , 1997 , , 387 , 878 paczynski , b. 1991 , acta astronomica , 41 , 257 schartel , n. , andernach , h. , and greiner , j. 1997 , , 323 , 659 struble , m. , and rood , h. 1997 , , 490 , 109 tokanai , f. , et al . 1997 , , 49 , 207